CHAPTER ONE
'No,' said Sally-Anne McAllister dazedly.
'No, please, no,' and she struggled fiercely against the arms which held her -- a man's, she noted, and that was enough to start her struggling even harder.
She would not be held by a man ever again.
No, not at all, and then, even in her confused state, her mind shied away from the reasons for her distaste, and she found herself saying even through her pain and shock, 'I will not think about that, I will not,' and so saying she stopped struggling and sank back into oblivion once more .
The next time she returned to consciousness she discovered that the whole right side of her face was numb, and that was all she registered.
The memory of being held in a man's hard arms had disappeared.
Her eyes opened; she was on her back.
Above her she saw a ceiling, grey and white, a plaster rose from which depended a gas-light inside a glass globe, engraved with roses.
She heard voices which at first made little sense, could not, for the moment, think where she might be or even who she was.
'Mama,' she said, her voice a thread, speech strangely difficult.
'She's coming round at last.'
It was a man's voice, educated, a pleasant if cold baritone.
An earlier memory returned.
Was he the man who had held her?
She did not want him to hold her again.
She tried to sit up, but was pushed gently back, by a woman's hand this time.
'Oh, dear God, Dr Neil.
Thanks be to Him she's conscious again.
I thought she would never recover,' said a woman's voice this time.
'The second blow was a cruel one.'
'Stupid,' said the man's cold voice.
'It was stupid of her to try to intervene between Jem Higgins and his Poll.'
Sally-Anne opened her eyes, tried to sit up -- a mistake that, everything reeled around her again; but memory had returned and she knew why she was in this room...and even why she had reached there, but not how.
And how dared the man standing between herself and the light so that she could not see him clearly speak so harshly of what she had done?
It might indeed be stupid to try to stop a man from attacking a helpless woman...but...
'Someone,' she announced, her voice suddenly strong again, 'someone has to try to prevent poor women from being beaten by great strong brutes.'
'Oh, yes,' said the man drily, 'I could not agree with you more.
But not young girls who could barely defend themselves against a schoolboy, let alone the professional bruiser Jem Higgins once was.
How came you here, anyway?'
And from what she could see of him, which was not much, he was giving her the coldest of stares.
Her memory returned fully...
She remembered quite clearly what she had been doing scarce ten minutes ago.
Sally-Anne McAllister -- although that was not her real name -- walked along Vetch Street in London's East End late on a hot afternoon in the early summer of 1903.
Not that summer did anything for Vetch Street; dust motes hung in the warm air and the sunlight was pitiless, starkly revealing the cracks in the broken pavement, the decaying brickwork of the small terrace houses, and the larger tenements which stood among them, and the rottenness of the wood in the unpainted doors and window-frames.
Here and there a larger house stood, once the home of some magnate now long gone, broken and rotten, a warren where different families lived in every room.
In all her short life Sally-Anne had never before encountered the squalor which she had seen in the few days since she had arrived in these poor streets in the hinterland between London Docks and Stepney.
Although she was not aware of it she was, in her shabby white cotton blouse and her dark green skirt, with her blue-black and glossy curls drawn up and knotted simply on the top of her head, her face glowing and vital, the only touch of colour in the grimy street.
She carried a small basket in her hand containing food for her frugal evening meal, although the rancid smells around her were strong enough to diminish any desire to eat -- she had not yet grown sufficiently accustomed to them to ignore them.
Later she was to think how little she was prepared for the simple events which were to change her life completely -- indeed she would have said that so much had happened to her already that any further incidents must be minor, a judgement which could not have been more faulty.
She had been living in Crow Court off Vetch Street for nearly a week, and the narrow lives of the people among whom she found herself appalled her.
What perhaps struck her the most was that, despite all, many of them appeared to be happy, while she, Sally-Anne, who could if she wished command the most luxurious life a woman could dream of, was most desperately unhappy, and had no idea of how she might become otherwise.
Her thoughts, which as usual these days were depressing, were disturbed by the noise of a fight, a fracas on the corner where Vetch Street met Millstone Lane.
A man, a large man, was beating a woman, a little woman who seemed scarcely more than a child, and was trying to drag her into one of the tenements which lined the opposite side of the road.
A group of ragged children and some idle women were watching him with amusement rather than disapproval, half applauding him with their cruel laughter, and although one woman boldly cried, 'Shame,' most seemed to be enjoying the unequal struggle.
The little woman suddenly broke away from him, ran across the ill-paved road towards her, tripped on the broken pavement, fell on to her knees before Sally-Anne, and, face wild, looked up at her imploringly, wailing, 'Help me, missis, please help me.
He'll kill me yet,' in a thin broken voice.
Sally-Anne, hampered by her basket, tried to pull herself free.
The woman stank of neglect, her clothes were torn and filthy, and tears had made twin furrows down her face.
Pity rose in her.
And anger -- anger at the man, at all men.
He had loomed up before her, had arrived to claim the woman again, pulling her away from Sally-Anne, his own face twisted with rage.
He cuffed the woman on the head, attempting to drag her back into the doorway.
Sally-Anne was almost choking between rage and fear.
She put her basket down on the pavement, held on to the woman with one arm, pulled on the man's huge arm with the other, said to him firmly, if more than a little fearfully, 'No, you are not to touch her again.
Leave her alone.'
This came out with more bravado than Sally-Anne really felt, and had she seen him more clearly before she intervened she might not have said anything at all.
The man was a bear, unshaven, face mottled purple, his eyes yellow and feral, his teeth broken, but his body, huge and strong, running to fat.
He paused to stare at Sally-Anne, face ugly.
'An' who might you be, to tell me what to do wi' mine?'
'Never mind who I might be,' said Sally-Anne firmly, trying to swallow her fear.
'Just stop what you're doing or I'll set the police on you.'
She might as well have saved her breath.
The feral eyes glared cruelly at her.
'Leave go, missis, or it'll be the worse for you.'
It was impossible for Sally-Anne to obey even if she had wished to.
The woman had sunk to the ground to avoid her tormentor, had clasped Sally-Anne around her knees, and was shrieking up at her, 'Oh, help me, help me, do.'
She cried and wailed into Sally-Anne's skirt so pitifully that Sally-Anne's own fear of the brute before her was lost in sympathy for his wretched victim.
The watchers were now bellowing encouragement to each of the players in a game which had taken a new turn.
They hallooed and shouted.
Windows were thrown open, heads appeared.
The unequal tug of war continued, and something kept the man from actually striking Sally-Anne, although he rained blows on the woman, who was now giving vent to a low, keening moan.
Emboldened, Sally-Anne cried to the watchers, 'Fetch a policeman.
He could arrest this man for assault.'
Guffaws greeted her.
'Not likely, missis; it's only Jem keeping his Poll in order.
No case for the Peelers.'
And their cruel laughter was that of any mindless mob finding entertainment in violence.
By now Jem had almost succeeded in prising Poll free from Sally-Anne, with the result that Poll's wailing went up an octave.
Almost dragged free, her hands had dropped around Sally-Anne's ankles, nearly bringing her, too, to the ground.
Looking up, eyes wild, Poll made one last supplication to Sally-Anne.
'Oh, don't let him take me, missis, please, don't let him take me.
He'll kill me this time, for sure.'
'No,' said Sally-Anne, breathless.
She fell on to her knees, held on to the woman, whose face was now on her shoulder.
She looked up at Jem bravely.
'No, I won't.
You, Jem, stop this at once.
Shame on you for hitting a woman.
I shall certainly set the police on you if you continue to go on as you are doing.'
This reasoned and ridiculous plea had no effect at all on Jem, other than to inflame him to further violence.
He bent down, thrust his unshaven face and stinking breath into Sally-Anne's, put his great hands under her armpits and hauled her to her feet by main force, Poll still clinging desperately to her.
'You will have it, then, damn you,' he growled, and he struck not Poll, but Sally-Anne, hard in the face.
She saw the blow coming, but hampered by Poll could take no avoiding action other than to turn her head slightly on receiving it.
All her senses stunned, she realised that he was about to strike her again, and this time, when he did so, oblivion took her, man, woman and jeering watchers all disappearing into the vacant dark...
And now, here she was, wherever here was, with a woman, short and comfortable, who possessed a kind face, and a man who was none of these things, judging by the coolly impersonal way in which he spoke to her.
She could see little of him, only his tall body, lean against the light.
'How did I get here?' she asked, looking around a shabby but pleasant room, feeling so frail that she might have been made of china, china broken into a thousand pieces.
She remembered Jem's first blow, but not his second.
'Dr Neil carried you in,' said the woman, who was now holding Sally-Anne's right hand with her own left hand, while gently wiping Sally-Anne's poor bruised face with a damp cloth held in her right.
Sally-Anne's returning senses told her that she was lying on an old-fashioned sofa with a high back at one end, and no sides.
The man standing before her had earlier been on his knees on the opposite side from the woman.
Painfully, looking up at him, she saw a good strong profile, sandy hair tipped with gold from the sun coming in through the lace-curtained window.
'Dr Neil?' she said questioningly.
'Dr Neil Cochrane,' he replied brusquely.
'What on earth possessed you to get involved with Jem Higgins?
Suicidal, if I may say so.
You have got off lightly, even if your face will be swollen and painful for some time.'
He looked hard at Sally-Anne, ignored her shabby clothes and shoes, looked instead at her hands and face, and all the signs of good care and feeding about her, so different from most of the women who lived around Vetch Street.
He was about to ask her what in the world she was doing in the East End when he was reproached by an angry Sally-Anne.
Oh, she hated all men, did she not?
Particularly officious, domineering ones.
She had not pushed into, or invited, the fight; Poll had involved her.
But she would not defend herself, by no means, and she answered him in tones as brusque and hard as his own.
'I could not see such a large man mistreating such a poor little woman.'
'I can only say, Miss...? that if you are to defend every maltreated woman in London's East End you will spend all your days in such labour, to no purpose.
The law does not like, or support, those who come between husband and wife.'
'Oh, indeed?' said Sally-Anne furiously to this supporter of wife-beaters -- and if only he would move so that she could see him, and not continue to stand between herself and the light --'How typically male.
Only a man could say such cruel things,' and then, belatedly, remembering her manners, which her mama had so often told her she was always forgetting, 'But I must thank you for taking me in.
And my name is Sally-Anne...
McAllister,' and she almost tripped over the last bit, but hoped he had not noticed.
She did notice that the nice little woman had gently squeezed her hand when Dr Neil was being at his most dictatorial over men's right to beat their wives.
Whether Dr Neil had noticed her hesitation or not, he said to the little woman -- ignoring both Sally-Anne's anger over his cruelty, and her thanks -- which just went to show, she thought crossly, that it was no use trying to placate the unpleasant creatures --'Matey, you will make us all a hot pot of tea, strong and sweet.
It will restore Miss...
McAllister a little.
She has received quite an unpleasant shock.'
'Thank you, no,' said Sally-Anne impatiently; she needed no condescension from him or any man.
Unpleasant shock, indeed.
She hardly knew which was worst, Jem with his blatant nastiness, or this man defending it -- implicitly, anyway.
'I will be on my way.
Drink your tea yourself.
You seem to need it more than I do.'
And how would Mama react to that ?
Oh, Sally-Anne, do you never think of what you are saying before you say it?-- all in the most sorrowful manner, no doubt.
She swung her legs over the side of the sofa, which reduced her defiance to mere childishness, alas, since her head began to swim dreadfully, and also caused him to say impatiently -- he really ought to meet Mama, at least he would agree with her --'Oh, do be quiet for a moment, madam.
You are in no condition to go any where.
Miss Mates will make us all some tea, give you some sweet biscuits, and only when I am fully satisfied that you are fit again shall I allow you to leave.
While you are here, in my home, you are my patient.
Perhaps some sweet tea will restore your temper a little-- or are you always so pointlessly combative?'
Pointlessly combative!
Only her spinning head prevented her from rising and leaving at once.
She stared fierily at him.
What was he doing in the East End?
His voice was educated, his clothing good, if a little careless -- a tweed suit, and a cream shirt, well cut, if his collar was a little frayed.
She could still see only his profile, and when he moved away, into the room's shadows, she could not even see that.
He seemed to take it for granted that everyone would do what he told them.
Miss Mates had already sped to do his bidding, giving Sally-Anne one last sympathetic squeeze of the hand before she left her.
Even in her dazed and painful state his imperious, if not to say imperial manner amused as well as annoyed her.
Few people ever expected Sally-Anne to do as she was told.
It was usually a vain hope.
She had been independent since childhood and hoped to remain so.
Thoroughly spoiled, her mama had said sadly, the last time she had seen her.
But suddenly the thought of her mama was painful to her, and to push it away she leaned back again, said, amusement plain in her voice, which, although she did not know it, surprised Neil Cochrane a little, 'Oh, I will be good, because in my present condition I cannot be anything else, I fear.'
She could almost feel her absent mother's approval of her belated politeness.
Dr Neil turned away from inspecting his own bookshelves, said, apropos of what she could not imagine, 'You are not English, I perceive, Miss McAllister.'
'You are quick, Dr Neil,' she said, a little surprised in her turn.
Her East Coast American accent deceived most people, it was so like, and yet in some ways so unlike, that of the society in which she had been living for the last six months.
'No, I am American, from the East Coast, but I have been living in England for some time.'
Neil Cochrane moved forward, sat down in an armchair facing her, and said, 'I am happy to see that you are prepared to rest a little, after all.'
But what he said had little effect on Sally-Anne compared with the sight of his face at last.
She had already noticed that when he walked he limped more than a little, favouring his left leg, but his face had taken a greater hurt.
His left profile, first glimpsed, was that of a handsome man, but the right side of his face was a ruin.
It was marred by a puckered scar which had destroyed his cheek, beginning just above the right-hand corner of his mouth, but, fortunately for him, narrowly missed his eye to disappear into his hair, twisting and distorting his whole face.
Neil Cochrane's mouth tightened when for the first time Sally-Anne saw his face plain, and reacted by controlling her own, so that the shock which she had received was only momentarily shown.
He gave no other sign of distress, merely turned his head to greet Miss Mates, saying coolly, 'You are prompt, Matey, even without your kitchen help.'
Miss Mates was carrying a black lacquered tray with a delicately flowered china tea-set on it -- Wedgwood, Sally-Anne noted -- and a silver teapot.
The elegantly shabby room in which she had found a haven was an oasis in the East End's desert, which could be glimpsed through the small bow window.
Sally-Anne took the tea which Miss Mates offered, and drank it gratefully.
Full awareness was returning to her, her head had ceased to swim, and all that was left of her recent encounter was the pain in her face, and her consequent anger at every member of the male sex.
'What happened to Poll?' she asked, determined not to be silenced by Dr Neil Cochrane, and not to allow the ruin of his looks to create any pity for him.
She did, though, wonder what dreadful accident had left him marked for life, and lame into the bargain.
She tried to eat a small biscuit, grimaced a little at the consequent pain.
This made her think suddenly of the pain which Neil Cochrane must have endured, so that she felt pity for him, after all.
'What do you think?' said Dr Neil, watching her.
He had already noted in her favour that she did not avoid looking him full in the face as many did; nor did she flinch or stare when she did.
'Jem took her home, as he intended to.'
He forbore to add that Poll might have earned a few more blows as the result of Sally-Anne's intervention, to prevent her from involving anyone else in future.
'And beat her again, no doubt,' said Sally-Anne acidly.
'No doubt.'
'You are annoyingly cool about it, sir.'
'Dr Neil, or Dr Cochrane, if you please,' he said, remaining determinedly cool -- to reproach her, no doubt, for what he considered misplaced pity.
'No sirs.
We cannot dance up and down about poor Poll; it would not help her.
That does not mean that I approve of what Jem does.'
'But the police --' began Sally-Anne.
'Oh, the police do not care about, or interfere in, domestic matters.
She is his wife, whether married in church, or what we call over here his common-law wife.
That is, she lives with him, and that, I fear, is the end of it.'
'You fear!' she flashed at him.
'The wrongs of women do not concern you?'
'Oh, everyone's wrongs concern me,' he said, but his voice was as cool as ever.
'But I do not flatter myself that I can do anything about them.
Except, of course, where they directly impinge on me, that is .
You are a suffragette, I take it?'
Sally-Anne could not tell whether he said this critically or not.
His cold, impassive manner had not changed when he spoke.
'What decent woman would not be,' said Sally-Anne, waving her biscuit at him, 'given the way in which society treats us?
It is not only East End bruisers who mistreat their women, by no means.'
She was fast recovering, Dr Neil saw; the pretty, wilful, if swollen and bruised face was vital, the springing blue-black curls were a sign of vigour and health.
Miss Sally-Anne McAllister had always been well fed and well cared for, he noted professionally.
She was also a young person well accustomed to having her own way -- there was no doubt of that.
'And what is this decent, well-educated woman doing wandering around London's East End?' he asked.
Before Sally-Anne could reply -- and fortunately for her, for she needed to consider her answer carefully -- there was a knocking at the door, and Dr Neil was required there for a moment.
In his temporary absence, Miss Mates refreshed Sally-Anne's cup again, said gently to her, 'Do not mind Dr Neil's manner, my dear.
He is the kindest of men beneath his brusque exterior.
His patients swear by him.
Do not let him frighten you.'
'Oh, he does not do that ,' said Sally-Anne cheerfully; all her normal brio had returned, and she was not going to allow Dr Neil to put her down, and when he came back she gave him her most dazzling smile, and prepared to do battle with him, whenever battle was necessary.
Kind he might be, although it seemed improbable from what she had so far seen of him, but Sally-Anne wanted nothing from men, neither kindness...nor love...nor anything.
CHAPTER TWO
But when Dr Neil Cochrane returned to the business of Miss Sally-Anne McAllister who had so strangely arrived in his home he was as firmly pressing of her as he was before.
Miss Mates had brought in Sally-Anne's basket after Dr Neil had intervened to save her, and driven off a Jem Higgins who was by then a little fearful at having laid Sally-Anne so low.
Dr Neil had seen her few poor items of food decently arranged upon a napkin in the bottom of the basket, and there seemed little doubt that she was actually living in the district -- although why he could not imagine.
He sat down by Sally-Anne, his ruined cheek slightly averted, poured himself another cup of tea and began to question her again.
Sally-Anne stirred restlessly.
He might have saved her from Jem, but did that give him the right to such a ruthless inquisition?
She demurred a little when he said, picking up his teacup, and looking at her over the top of it, 'You never answered my question, Miss McAllister.
What exactly are you doing here?
I see by your basket and your dress that you are domiciled in the neighbourhood.'
Domiciled in the neighbourhood, indeed!
What a pompous way of putting it.
Amusement at that made her tone light.
'If you must know, Dr Neil, and I suppose I owe you that, I am looking for work.'
'Work?' he said, raising his eyebrows and looking at her as though she had said drilling for oil or prospecting for gold.
'Yes, work,' she said sharply.
'One must eat.
Money is needed to buy food.
One is given money for work.
What is so surprising about that?'
Dr Neil could have given her several answers, beginning with the beautiful hands lying in her lap which had visibly never done a stroke of work in their owner's life, but said instead, 'A strange place for a young lady like yourself to look for it.'
'Indeed not,' replied Sally-Anne, with all the fiery determination which her own family knew so well.
'It is hard for an unqualified and unapprenticed female to find any work in London, except in the East End where, I am reliably informed, there is casual labour aplenty, and unskilled work, too.'
'Yes, I know that,' he said, smiling a little at her charming vehemence.
'But how does a young American lady --' and he stressed the last word slightly '-- come to be alone and seeking work in London?'
How indeed? thought Sally-Anne ruefully.
Her fertile and inventive imagination came to her aid.
She was frequently appalled at her own ability to lie convincingly, thought that perhaps she got it from Papa, one of the world's great tricksters -- not knowing that on occasion her apparently innocent mama had the gift as well.
She had no alternative but to tell him a series of absolute whoppers, as her younger brother, Rob, would have said.
The poser being, of course, that she could not tell him either the true reason she was here, or even who she really was.
She improvised wildly and skilfully, dabbing at her eyes while she spoke with a rather grubby handkerchief.
'My papa died after he lost all our money in the last great depression.
My...mama...had long gone.'
Even an unscrupulous Sally-Anne could not quite kill her mama off completely -- if he thought that her mama had bolted, well, so be it.
She gave a half-sob, added, 'I had to find work.
An old acquaintance of the family needed a governess-companion for her little girl during a prolonged visit to Europe.
She came from New York State, too -- her husband was an executive in an automobile works there -- and I thought my troubles were over.'
She heaved a great sigh of which she was privately rather proud, thinking that it added a touch of verisimilitude to the sad tale she was spinning.
Later she was to wish that she had neither been so deceitful, nor so colourful, but at the time excitement and expediency combined carried her away.
'Unfortunately my employer's husband...'
And she cast her eyes modestly downwards.
'Yes?' prompted Dr Neil with a grin -- for some reason he did not believe a word of this farrago -- to hear her reply in a low voice,
'He took more than a proper liking to me, and when I indicated that I was not interested in his advances he had me cast off...without a reference, so that I cannot find suitable employment.'
Sally-Anne was particularly pleased with the last bit, and saw by Miss Mates's sympathetic expression that she was swallowing every word.
She was not quite so sure of Dr Neil, but she ploughed steadily on.
'I was left with very little money, came here to live in Crow Court, because it is cheap.
I need employment to live, and in the hope that I can save my passage home again.
I would do anything to earn a wage, however small -- be a servant, even.
I would not mind being trained, and I am not afraid of hard work.'
'A servant?' said Miss Mates, coming over and taking Sally-Anne's hand.
'You poor child,' and Sally-Anne swallowed a little at this, feeling ashamed of telling such a kind-hearted creature such whoppers.
'But the work is so hard, for so little.
And you are so old.'
'Old?' said Sally-Anne a little indignantly, to Dr Neil's amusement as he watched the two women.
'Why, I am not yet twenty-one.'
'Oh, but one starts training housemaids and cooks at twelve and thirteen,' said Miss Mates gently, but at the same time she was thinking of her own problems in finding and keeping satisfactory girls, and Miss McAllister looked both strong and willing.
A little high-spirited, perhaps, but hard work and long hours would soon cool her down.
'I learn very quickly,' said Sally-Anne submissively.
'And I really do need to earn my passage home.
I meant it when I said that I would do anything.
Perhaps you might know where I could usefully apply?'
Was it her imagination or was the man opposite her regarding her warily?
Since Miss Mates had intervened he had said little, although he had had enough to say before that, goodness knew, thought Sally-Anne briskly.
Strange how handsome he looked with the shadow concealing his ruined cheek a little.
What a pity for him.
She thought that perhaps her manner was somewhat too bold for one who so recently had been only a governess, and now wanted to be even less.
She smiled and bent her head.
She must remember not to be her usual confidently aggressive self if she was offering to take up such a subordinate post.
'Are you serious, Miss McAllister?'
Dr Neil asked.
'About becoming a servant, I mean.
You do know how hard the work is, I hope?'
'Oh, yes!' exclaimed Sally-Anne eagerly.
'But I do not intend to do it forever, and I am very strong.
Papa says that most horses would envy me in that line!'
This frankly offered statement brought an almost unwilling laugh from him.
'Did he, indeed?
Let us put Papa's notions to the test, then.
I know that Matey here wants a maid of all work, and that the last two were highly unsatisfactory.
You could hardly be worse, I suppose.'
Sally-Anne bridled a little at this.
She liked to think that she was very efficient in all she did.
From what she had seen of maid's work it was not very difficult.
An idiot could do it.
What she did not know was that it was not so much that the work was difficult, but that there was so very much of it, and all tiring.
Like most young ladies of her class she had no idea at all of the effort which went into making her own charmed life easy.
She was not so much spoilt as totally unaware.
'I'm sure that I should be satisfactory,' she said stiffly.
'You could put me on probation, I suppose.'
'I have no intention of doing anything else,' said Dr Neil briskly.
He was now all employer, coolly assessing Sally-Anne as a prospective slave about the house.
'I should be taking you on at a week's notice.
Your wage would be five shillings and all found, one evening and one afternoon off a week -- you can arrange all that with Miss Mates.
To church on Sunday, with Miss Mates, of course.
Oh, and no followers.
I want no policemen hanging about the kitchen.'
He thought that if Miss McAllister was not serious that would be sure to put her off.
He decided to inform her of her duties, said casually, watching her as he spoke, 'Your day would begin at six o'clock, finish at Miss Mates's bedtime between half-past ten and eleven, and you will have the attic bedroom.
Miss Mates will show you to it.'
Sally-Anne was scarlet.
Policemen followers, indeed!
And her hours!
She had genuinely possessed no idea of how long servants worked, and for how little.
She suspected that the good doctor might be laying it on a little thick, but she would not be put off, and said in her stiffest voice, 'Yes, I understand all that, and I am prepared to work hard,' and she added for good measure, 'You need not worry about followers.
I truly despise all men.'
This came out in Sally-Anne's best manner.
'So very glad to hear it,' said Dr Neil Cochrane brightly.
'Now we all know where we are.
You will not be flirting at the kitchen door with the milkman and the coalman, and Matey will not be chasing you for doing so.'
He was no fool.
Despite being an East End doctor he also knew the great and wide world outside.
There was something vaguely odd about Sally-Anne McAllister and the farrago to which she had treated them.
He did not doubt that she was an American, but as to her story -- well, he was not sure.
But he was a man who liked challenges and this wilful -- child -- for despite her boasted nearly twenty-one years, to Neil Cochrane she seemed little more -- appeared to offer one.
And Matey badly needed help; indeed, she was looking at him with something like approval.
She said to Sally-Anne, 'Well, I do need a maid of all work, that's true, but I never thought of hiring a lady...'
And her voice trailed off as she surveyed Sally-Anne in all her proud and pampered beauty.
Did she really understand what she was taking on?
'I said I was willing to work and to learn,' said Sally-Anne, exasperated by all this  havering .
Goodness, how the British went on.
A good American would have had her in the kitchen with a flue-brush, or making pastry, by now.
'I'm an excellent cook,' she announced proudly.
'Mama made sure of that.'
'Well, that's certainly something,' said Miss Mates, overwhelmed by such artless eagerness, so foreign to any maid of all work she had encountered before.
'I could start immediately,' Sally-Anne roared on, striking while the iron was hot, just as Papa always advised -- one maxim which she had not disdained to learn from him.
'Because I am running out of money,' she added, and she crossed her fingers beneath the light blanket which Miss Mates had thrown over her earlier and asked God to forgive her for such dreadful lies, and so many of them.
She brought her hands up, clasped them together, said impulsively, 'Oh, do say yes, Dr Neil, sir, Miss Mates -- help a poor, lonely orphan,' and this last flew out without her even thinking, and certainly dissolved Miss Mates, even if Dr C, as she privately thought of him, looked a little sardonic.
Miss Mates, indeed, kissed her on the cheek, said, 'Oh, you poor thing,' which made Sally-Anne feel a little mean, even if it was the last copper-bottomed clincher -- that was her Uncle Orrin this time, not Papa -- which got her the job.
For good measure, she added, a trifle pathetically, 'I am a good girl, really truly, even if I can't give you any references, and I will work hard, I promise.'
'Oh, we shan't give you any opportunity to be anything else, Miss McAllister.
You're sure you want her, Matey?' drawled Dr Neil, still sardonic, but his use of Miss Mates's nickname was the final sign that he was prepared to take her on.
But for Miss Mates the 'really truly' was the icing on the cake, even if she had more doubts about Sally-Anne's stamina than her missing references.
'Perhaps you could start this evening,' she offered.
'If Dr Neil says that you are fit.'
Sally-Anne tried to stand up, found that she was still somewhat rocky, looked at Dr Neil, and said doubtfully, 'I don't think that I am quite fully recovered yet.
I might be able to fetch my things from Crow Court where I am rooming a little later on, but as to work...'
'Oh, yes, work,' said Dr Neil in his rather hateful way, as though every word she said was dubious in itself, and, even though they were, Sally-Anne could not help feeling righteously indignant.
'I would not want to slave-drive you as early as this actual moment.'
And he looked pointedly at his watch, a fine gold hunter which he pulled out of his pocket and cocked an eye at.
'Five o'clock,' he announced.
'I should think you might be ready at six-thirty,' and now the quizzical eye was cocked at Sally-Anne.
'Surgery at five-fifteen,' he said.
'Matey will look after you, and perhaps later on we can collect your traps from Crow Court.
I take it that you will not need a pantechnicon?'
'You take it correctly,' said Sally-Anne, a little peevish, despite her pleasure at having found employment so easily, and in an area where she could observe and record the life going on around her.
She must not forget the other reason why she had wished to find work here, and exile from her old life was not the main one for her presence in Vetch Street.
Miss Mates, who had removed the tea things, came bustling back, said to Dr Neil, 'Mind you are ready for surgery,' and brought him a plain black coat which he put on instead of the rather sporty Harris tweed one which he was wearing.
'You see, McAllister,' he said gravely, 'were it not for Matey I should never look presentable.
I tend to forget what a proper doctor should wear.
That's why it's so important that she has efficient help.'
He strolled off to the surgery, Matey still fussing about him.
She came back, sat by the fireside, and said to Sally-Anne, 'The surgery's at the side of the house.
I suppose you've seen it when you've walked by.'
She sighed.
'He works too hard, but there, you can't tell him anything.
He's always gone his own way.
I've made another pot of tea for us.
It will do your head the world of good.
No restorative like tea.'
I wonder the British don't bath in it as well, thought Sally-Anne naughtily, but she was grateful for the extra cup, and for the sandwiches which Miss Mates brought in with it for them both.
'To tide us over.
We usually have a bit of supper when Dr Neil's day is ended.
Not that it ever really ends.
Many's the night he's been called out in the small hours...'
And Matey heaved another great sigh.
She spoke of him, Sally-Anne thought, as though he were still only twelve years old.
She giggled internally at this, visualising tall and haughty Dr Neil wearing the black knickerbockers, funny jacket and cap of an English schoolboy.
The smile this brought to her face pleased the old nurse.
'Beginning to feel a bit better, are you?' she said kindly.
She leaned forward, pulled the wrapper over Sally-Anne up to her chin.
'Have a bit of a doze, my dear.
You've had a hard day, and by the sound of it not an easy life.
Yes, have a little nap; it'll set you up for a day's work tomorrow.
I'll leave you to go and wash the pots.'
She tiptoed out, leaving Sally-Anne to fall into a light sleep -- but not before she felt a little mean, as she would have said in the USA, about deceiving the good old lady -- and Dr Neil Cochrane, of course.
Although she didn't feel so badly about him .
He was only a man when all was said and done, and men -- well, they deserved to be deceived.
But it was not a great deceit, after all, and perhaps one day she might be able to explain it to them, and apologise, but for the present matters could not have arranged themselves much better.
What a pity about Dr Neil's ruined face -- and his suspicious manner, were her last conscious thoughts before sleep took her.
CHAPTER THREE
Really, the worst thing about being a kitchenmaid-cum-parlourmaid-cum-everything else, Sally-Anne ruefully considered, was all the scrubbing and the grate-cleaning which had to be done every day.
It was not that she minded the hard work; it was what the hard work did to her poor hands.
The scrubbing was the nastiest, she thought despairingly, bad though blacking the grates, particularly the kitchen range, was.
She was down on her knees, finishing off the doorstep; an apron made from some coarse sacking was protecting her plain black dress, her legs were clad in thick black wool stockings and her feet were shod in heavy, much mended shoes -- second-hand into the bargain, as was most of her servant's clothing.
And what a ritual the doorstep was.
First the scrubbing, and then the almost religious whitening not only of the step, but their share of the pavement before the step.
It had to be done every day, too.
What was worse, within minutes people would be walking all over her back-breaking work -- it did not bear thinking of.
She straightened up, and carried the pail, the soap, the brush and the swab, as Miss Mates called the cloth, to the outdoor drain at the back of the house.
And she shuddered at the chore the whole thing would be in the winter -- if she lasted that long -- seeing that it was such a drag in the mildness of summer.
Her soft hands were so unaccustomed to such hard work that it became -- temporarily, Miss Mates said, until they hardened -- difficult for her to help with the fine sewing, and the darning and repairing needed in the little house.
Which was a pity, Miss Mates also said, looking approvingly at Sally-Anne's neat and careful stitchery, done before her hands became ruined.
'Oh, Mama always made sure that I learned to do the practical stuff, as well as fine embroidery and canvas work,' Sally-Anne had said.
'She always said that bread and butter was needed as well as cake.'
'Your mama sounds like a sensible woman,' said Miss Mates.
'A pity that you had to lose her and earn your own living.'
'Needs must,' lied Sally-Anne, who was beginning to wonder whether she would ever be able to tell the whole truth again.
That was the worst thing about duplicity, she was beginning to find.
That and having to watch what you were saying.
Keep it simple was a good motto.
She returned to the present, replaced the cleaned pail, the soap, the brush and the swab beside the pump in the lean-to outhouse at the back -- it stood in a small soot-encrusted yard where a few fearless London sparrows frequently gathered -- and returned to the house to clean the surgery.
She had already thrown yesterday's damp tea-leaves on to the parlour carpet, and when she had finished in the surgery would take a small, hard hand-brush and, down on her knees yet again, would painstakingly sweep them all up into a dustpan, to remove the previous day's dirt and dust, Miss Mates said, and purify the air.
Sally-Anne liked the surgery.
Small and cramped although it was, it reminded her a little of Papa's office back home: a room where a man obviously worked.
She briefly envied Dr Neil and Papa -- so easy for a man to do useful and purposeful work.
On the other hand, she conceded, someone had to do the kind of menial tasks she was at present condemned to -- but why was it always women who did them?
And for such a poor reward, too.
She knew what Papa would say to that .
'Now, Sally-Anne, men do menial work, a lot of it, and harsher often, than women's.
They go down the mines, herd cattle, tend sewers -- you must admit that.'
She knew what Papa would say because he had said it a few years ago when she had begun to reproach him and all men for their oppression of women.
He had not denied it -- he had explained it instead, which was worse, she had once said angrily.
Her eyes filled with tears.
Oh, she did miss them so, both Mama and Papa.
But she would not crawl home for comfort, defeated.
Twice defeated now, and the second one so bitter...and so irrevocable.
Humpty Dumpty could never be put back again...
No, she would not think of that -- no, never that, and she began to wield her feather duster with such angry vigour that Dr Neil, coming in, was amused at her bright energy which even the ever-lasting drudgery which she performed could not dim.
'You are supposed to be suppressing dust, McAllister, not raising it and spreading it about,' he said cheerfully to her back.
Sally-Anne could not get used to being known so abruptly by her surname.
She had never queried such a thing before -- indeed, had taken servants, their duties, and how one addressed them for granted, after the fashion of all wealthy and pampered girls, she now supposed.
She would never do so again.
She stopped, turned to face Dr Neil -- as everyone, she found, called him, even his poor patients as well as Miss Mates -- and said cheekily, 'Oh, come, Dr Neil.
If I am dull and slow I am reproached, and now it appears that being vigorous does not answer, either.'
Dr Neil put his head on one side and said gravely, 'You must learn not to be impudent to the Master, McAllister -- it really will not do,' but she knew by his manner that he was not serious, and she thought again what a pity it was that the ugly scar so spoiled and distorted his face.
And he minded that it did; Sally-Anne knew that.
She was not so young and giddy that she had no understanding of the desires and problems of others.
Besides, living as she now was in a small house on top of two other people, and Dr Neil's surgery boy, and seeing at close hand the wretched people who frequented his surgery, was giving her insights into the motivations and behaviour of the people around her because she had to consider them, whereas always before, in the past, other people had been compelled to consider her .
What pity she had possessed for others had been diffuse and impersonal, for women as a mass rather than for individual women such as poor, defeated Poll whom she saw daily, the marks of Jem's fists on her face.
Dr Neil was now seated at his desk and she lingered a little.
For some reason, even though he was often sharp and rather short with her, however kind he was to his poor patients, she had begun to like being with him.
Perhaps it was because he did not defer to her, flatter her, praise her beauty and her charm, admire her ready wit, as all the men she had known before had done, when what they really liked and deferred to was the knowledge of her father's immense fortune and the certainty that she was sure to inherit a great part of it.
Miss Mates worshipped him.
She had been his nurse when he was a little boy, she had told Sally-Anne when they were preparing dinner for him one evening.
Sally-Anne remembered the carrots falling into small red-gold coins as she cut them while Matey talked.
'He was a soldier, you know,' she had said, podding peas so rapidly that Sally-Anne was full of awe for such expertise, pea-podding being difficult, she had discovered.
'Sent into the army at eighteen, as younger sons always are.
And then, when he was doing so well, he was badly hurt in that nasty Africa.'
Africa was always nasty to Miss Mates because it had maimed and nearly killed her treasure.
'When he came home he had to leave the army, you see.
He was ill for a long time.
He still limps, as I suppose you have noticed.'
Yes, Sally-Anne had noticed.
'Well, he said that he wanted to do something useful with his life, seeing that it had been given back to him when he hadn't expected it.
Africa had changed him, you see, and he still only in his early twenties.
He said he wanted to be a doctor, and his papa, Sir Hanley, agreed in the end to him becoming one -- not a thing for a gentleman to be, he said, but Dr Neil insisted, even when Sir Hanley said that he didn't want an apothecary for a son.'
He had called Neil a damned snivelling apothecary, Miss Mates remembered, but she had never used, or repeated, such language.
'So, here we are,' Miss Mates had finished.
'He said that he didn't want to be a fashionable doctor.
He owed his life to his men, he said, and he had been horrified to discover the conditions in which they lived at home, particularly in the towns -- he being a gentleman, you see, and always living in the country in comfort.
Goodness, child, do hurry up.
You work carefully, but you must learn not to be so slow.'
'I shall be better with more practice,' Sally-Anne had said humbly.
She had thought that she was doing rather well, but that appeared to be a fatal thought to have.
Whenever she felt that it nearly always turned out that she was wrong.
Another lesson she was learning.
'Pay-day today,' said Dr Neil suddenly, watching her whisk the duster over the windows and the drapes -- she must remember to call them curtains while she was in England.
'So it is,' said Sally-Anne, struck.
Friday had come round so quickly again -- such a contrast to her old life when the week had often seemed boringly endless.
'Might as well pay you now,' he said, and pulled a little tin box out of one of his desk drawers, opened it, and began to count out her meagre wage.
Why was it, Sally-Anne thought when she went forward to take it, that this pitiful sum seemed more precious to her than her huge allowance which she could claim any day she wished?
Perhaps it was because the shining pennies were a reward for her hard work, some recompense for her sore knees and the work-reddened hands which took the coins from Dr Neil.
Her pleasure was written so plainly on her face that Dr Neil wondered again about the true nature of his new and strange housemaid.
Looking into her great eyes, their blue so dark that it was almost black, he was uncomfortably aware that having Miss Sally-Anne McAllister in the house was a most disturbing influence on a man who had not only denied himself sexually for some years, but who had rarely mixed with young women at all since he had left the army.
He had been a bit of a gay dog before that, he remembered, enjoying himself freely like all young men of his age and class.
Certainly he had not been in such close proximity to a pretty young girl for as long as he could remember.
He had forgotten how pleasing a woman's soft voice was, the faintly suggestive noise her skirts made, and the delightfully pleasant scent which accompanied McAllister everywhere, so different from Matey's sensible Lifebuoy and carbolic.
Was the scent essentially McAllister, or was it some subtle perfume she chose to wear?
As he had passed her in the little hall this morning a tendril of her silky black hair had brushed his ruined cheek, and the smell of it had been so distracting that he had feared for his composure.
Good God, what was he coming to, that his previously monastic life should be tottering, if ever so slightly under the impact of a pretty minx who had invaded his chaste home?
And a pretty minx who was an accomplished liar, and who doubtless had some ulterior motive, though God knew what, in choosing to live in a milieu so unlike her usual habitat.
There was no doubt that, disguise it as she might, McAllister reeked of wealth and privilege.
And he, Neil Cochrane, really did know about such Birds of Paradise, for had he not been loved and betrayed by one, and learned his bitter lesson -- never to love or trust a woman again?
To check his wandering thoughts he started to ask her to resume her labours elsewhere and finish the surgery when he had completed his own tasks in it -- Only to find that she was hopping gently from foot to foot, and was bursting to ask him a question.
Dr Neil sighed.
He decided that temptation would have to be removed later rather than sooner.
He could not snub the child so openly.
He mentally called her a child to avoid admitting that she was very much a woman.
Sally-Anne's question was, for her, a very important one.
It was one which she had been longing to ask him ever since the night on which he had come home so late, just before she had turned off all the gas-lamps.
He had arrived grimly silent, his face old and drawn, the great scar livid and prominent.
He had told a worried Miss Mates that his patient had just died having her twelfth child.
The child had survived, but it was deformed.
Sally-Anne had been horrified.
Twelve children!
Even she knew that there was no need for this...that there were ways and means.
Single girls of good family were not supposed to know of such things, but Sally-Anne, if not an active suffragette, knew about Miss Annie Besant, and those who said that women should not be tied to endless childbearing and that there were practical ways of avoiding it.
'Dr Neil,' she said bravely, her dusting finished.
She must ask him, she must, however forward he might think her.
'McAllister?' he said, looking up, surprised, to impress on her that maid servants, even American ones, were not supposed to quiz the Master when he was at work -- should not even originate a conversation with him.
'May I ask you a question, please?'
'Yes,' he said, his mouth quirking a little, despite himself.
McAllister's perfect manners, running in tandem with her fiery, impulsive nature, made a mix which had begun to amuse him, and added to her disturbing charm.
'That woman...the one who had twelve children.'
She trod on the word in her horror.
'You're a doctor.
Why don't you tell them?'
Somehow, in her mixed anxiety and embarrassment, this was coming out all wrong.
Perhaps Mama was right.
She should think more before she spoke.
'Tell them what?' he said, genuinely puzzled, admiring against his will the pretty, ardent face opposite him.
'Tell them...it's not necessary...to have twelve children.
That there are ways.
You could...instruct them.'
Sally-Anne was suddenly rosy red at her daring, but she had to say it; it needed to be said.
Women were dying .
Dr Neil came down to earth with a bump.
In all his thirty-three years no single woman had ever addressed him after such a fashion, or spoken of such a delicate matter.
'It's not my business, McAllister,' he said stiffly.
'Nor yours, either.'
'Not your business?
I suppose it's your business to watch them die having their deformed babies!
How dare you be so cool?'
Dr Neil was not cool at all beneath his impassive exterior.
He had grieved over the dead woman, and Sally-Anne had put her work-reddened finger on something which had frequently troubled him in his slum practice.
'Come now, McAllister.
You are talking about matters between a husband and a wife.'
He had never thought to find himself discussing such a thing with a single young woman, but he had heard of American girls' frankness, and he supposed that this was a sample of it.
'But even you, a man, must agree that the whole business is so unequal.
What chance has any woman, rich or poor, against the tyranny which her husband exercises?
As a doctor, you know that such suffering is not necessary.
Is it not your duty to do something about it, to help such victims?
If not you, then who?'
Sally-Anne was by now in full flow, and when she paused for breath Dr Neil said, as drily as he could, 'I will only say to you what I told you on the day when you arrived here: it is useless to take the world's burdens on your shoulders.
It is not a simple matter of informing people what to do as you seem to think.
No, wait,' he continued, when she opened her mouth to speak again.
'Think.
Since you seem to know whereof you speak I will only say that you must know that the burden of restraint by means of the...mechanism...employed will, of necessity, fall upon the man.
Think back to Jem Higgins and ask yourself of what use it would be for me to preach your message to him.'
It was just like arguing with Papa, thought Sally-Anne with some disgust.
He was so reasoned, and there was so much truth in what he said.
But against that was the suffering endured by women which she saw all about her in London's East End.
Not simply the women dying of too many children, but the women who were compelled to sell their bodies because that was all that they had to sell.
She might have said no more, except that, unluckily for him, Dr Neil felt impelled to continue.
'You see,' he said kindly, 'it is one thing for someone like yourself who has always had an easy life -- up to now, that is -- to speak glibly about suffering women, and think that there is an easy way out --'
He was violently interrupted by a Sally-Anne almost incandescent with rage.
Not know how a woman could suffer!
How dared he?
He knew nothing, nothing, of what Sally-Anne had had to suffer at the hands of one of the monsters who controlled...what were his weaselling words?...the mechanism.
How would he, a man, know anything?
And why should she expect him to change a system which benefited men so greatly?
'Oh, yes,' she said, waving her feather duster at him with such violence that the end tickled his nose.
'Do nothing.
That is the way, I see.
Why not do something?
Publicise, make the ability to control birth available to the poor women in the East End as well as to their rich sisters in the West End.
It may be indelicate of me, but I think that women will only be truly free when someone invents a...mechanism...which women can control.
That would be even better and more useful to poor women than the vote, desirable though that might be.'
Dr Neil could not but admire her.
However mistaken he thought that she might be, and that her vision of life was based on a charming naivete which took little account of the cruel realities of existence, it was, to him, admirable that she should care about such things, and in such a practical way when all was said and done.
The simple demand for the vote, the be-all and end-all of most suffragettes, was truly seen by McAllister as a minor step compared with relieving poor women's social and economic disabilities, a measure which had little to do with suffrage.
He gave a half-smile which was immediately interpreted by Sally-Anne as patronising contempt.
'Oh,' she said, 'I see that for all your fine words you are much like the rest,'and there were tears in her eyes when she turned away from him.
She should have known better.
Despite the cool way in which he had always spoken to her she had thought him different from other men.
For some reason which she could not understand a feeling of desolation swept over her.
She groped blindly for the door-knob, her glossy head bent.
Long suffering had made Dr Neil sensitive to the feelings of others.
He stood up, walked round his desk, put a hand on Sally-Anne's arm, and said gently, 'McAllister, look at me.'
His voice was so unexpectedly kind that Sally-Anne's anger drained away.
She looked up at him, mouth quivering a little.
'I did not mean to patronise you,' he said.
'Even if I thought that you were mistaken.'
He paused.
'It is hard enough for me to run my practice here, seeing that I am that strange animal, a gentleman.
If I started trying to come between husband and wife, and that is how it would be seen, I could hardly help anyone at all.
I must have these people's trust, the men's as well as the women's.'
Sally-Anne nodded mutely.
Perhaps he was not just another male ogre, after all.
She would like to think so -- even if he did defend the status quo, just like Papa.
Dr Neil put out a hand.
He suddenly did not want to lose her respect.
Woman she might be, but she was displaying a genuine compassion rarely seen among rich young girls.
'Come,' he offered, 'shake hands on it, McAllister.'
Something else occurred to Sally-Anne.
Almost for the first time in her life she checked her wilful and impulsive self, even questioned a little a previous action or speech.
'Dr Neil,' she said, and her manner was almost shy, 'I hope you do not think the less of me for raising such a delicate matter.'
Goodness, she thought, is this really me speaking?
And she looked down at her work-worn shoes, half afraid to meet his eyes.
Dr Neil looked down at the bent blue-black head, and some idea of the enormous concession she was making to him struck home.
'No, indeed,' he answered gravely, although he had been a little shocked if truth were told.
'Not at all.
On the contrary, I think it admirable that a gentlewoman should think seriously of such matters.'
And suddenly he, too, thought, Is this really I, Neil Cochrane speaking?
What can be coming over me?
For despite having an open mind in many ways, his attitude to women and their problems had always been the conventional one of the young aristocrat he had once been.
He held McAllister's hand for a little longer than propriety might demand, and the pair of them stood for a moment, rapt, until the surgery boy, who ran the doctor's errands on his bike, delivered prescriptions and generally did a great deal of donkey work, knocked on the door before he came in, saying excitedly and importantly, 'You're wanted, Dr Neil, sir.
Carrie Jackson in Vincent's Buildings is having her tenth and her ma has sent for you.
She says it's urgent.
Things going wrong.'
Her tenth!
Sally-Anne was indignant all over again.
Dr Neil dropped her hand smartly, picked up his bags, and was out of the room in a flash.
Miss Mates, who had followed young Eddie into the surgery, looked sadly after her departing treasure.
'He never thinks of himself these days.
Only lives for his work, and now he's bound to miss breakfast.
What he would do if I weren't here to look after him, I'm sure I don't know.'
The surgery seemed empty without Dr Neil Cochrane in it.
On impulse Sally-Anne went to the kitchen and picked up a jam jar which she had earlier filled with sweet peas brought by a grateful patient who owned a little garden.
She carefully carried it through, put it on his window-sill.
She hoped that he would not find the jam jar too utilitarian, but he appeared to possess few vases.
After all, despite his backward views about the wrongs of women, he really did work very hard for his poor patients!
CHAPTER FOUR
'Oh, shoot,' said Sally-Anne disgustedly when the candle flickered and almost went out -- again.
She was in her tiny attic bedroom, sitting up in bed, propped up against a hard and lumpy pillow, sitting on a hard and lumpy mattress.
Opposite her was a small washstand with a coarse crockery toilet set on a fake marble top.
A roller towel hung on the door.
Beside the washstand was a closet -- cupboard, the English said, containing her few clothes.
Side by side hung her coarse morning working outfits, made out of casement cloth, which she had had to buy herself, much of it second-hand to save money, and two afternoon ones -- a white cotton shirtwaister, black skirt, lacy cap with streamers and lacy white pinafore.
Her afternoon shoes, slightly better than her morning ones, were ranged neatly beneath them.
By her bed was a chest of drawers which contained her underwear -- several pairs of cheap directoire knickers, vests, petticoats, and corsets, plus black stockings of wool, cotton, and one precious silk pair which she had brought from her old life.
She was busily engaged in writing in a penny exercise book.
It was already eleven-thirty at night.
Through the gap in the curtains, also casement cloth, a romantic moon shone down on her, but Sally-Anne had little interest in romance or the moon.
She was bone-weary from a day's back-breaking labour, but it was essential that she finish what she was writing, for she had a deadline to reach and that deadline was the day after tomorrow, her half-day, and she had barely begun her task.
The candle flickered again, so badly that it almost flickered out.
There was nothing for it.
She would have to go downstairs to find and fetch a new one from the store cupboard in the kitchen.
She threw back the covers, slipped a light shawl around her shoulders above her coarse calico nightgown, thrust her bare feet into her felt slippers, and crossed the room, avoiding the small oak bureau by the door, which she opened cautiously.
The whole house was quiet and she told herself that she must be careful not to awaken the two other sleeping inhabitants.
Outside, for once, the East End was quiet, too.
She crept downstairs, holding the green enamel candlestick high above her head so that she didn't lose her footing on the narrow wooden stairs.
The stair carpet didn't begin until the floor below the attic -- another discovery she had made about the lifestyle of servants.
She would have to pass through the parlour to reach the kitchen, an arrangement which had shocked the pampered girl she had been, but it was obviously designed to conserve space in the small house.
Even so, Dr Neil's modest home was larger and better appointed than most around Vetch Street, and was vast compared with the one stifling room she had briefly occupied in Crow Court.
The hall at the bottom of the stairs was a tiny square, a door on one side opened to what had been designed as the best parlour, but was now converted to a waiting-room for the surgery, which was a lean-to structure at the back.
The door opposite gave access to the parlour where Dr Neil and Miss Mates lived and ate.
Sally-Anne's own preserve was the kitchen, where she ate either at the kitchen table, or, when that was full, at a small card table which was folded up and put away when she had finished.
A wooden Windsor chair, with a hard cushion, was provided for her when she was allowed to rest, which wasn't often, given the quantity of work which keeping even a small house clean necessitated.
She pushed the parlour door open and tiptoed in.
It was dark and quiet.
The candle she carried gave one last flicker -- and expired.
'Oh, shoot,' she whispered again, trying to avoid bumping into the large oval dining table which stood in the window.
Someone, something, moved in the dark shadows by the empty fire-grate, filled in summer with a copper jug stuffed full of artificial flowers.
'Who's there?' said a blurred voice.
'Matey?'
Sally-Anne jumped and said falteringly, 'It's only me -- McAllister.
Come for a new candle.'
The someone, who was, of course, Dr Neil, struck a Swan Vesta to light the oil-lamp which always stood on a side-table where he usually kept the book which he was currently reading.
He had a small but good library, much of it kept on shelves on the first landing.
The lamp's dim yellow light showed him to be seated, or rather slumped, in his big armchair.
His tie was pulled loose, his shirt unbuttoned, and his hair was tousled.
His face, too, was also blurred, only the scar on it was more livid and sharper than usual.
The reason for his blurred face and voice stood on an occasional table before him -- a whisky bottle and a shot glass.
She had never seen either him, or Miss Mates, take a drop of alcohol before in the weeks which she had already spent in Vetch Street.
'McAllister,' he said.
'And why do you want a candle at this hour?
You should be asleep.'
'So should you be,' said Sally-Anne, greatly daring.
'Touche ,' he said lazily, not at all put out.
'But you haven't given me an answer.'
'I don't like being up three flights of stairs at night without a candle.
This one,' she said, setting the candlestick down on the table.
'isn't satisfactory.
As you must have seen, it died on me a moment ago.'
'Fair enough,' Dr Neil replied, and by the careful way he spoke he had drunk quite a lot of the whisky from the half-empty bottle.
Sally-Anne had seen drunken and half-drunken men before in Washington and London.
American legislators were not noted for their abstemiousness, and nor were the denizens of London Society.
'If you will excuse me,' she said politely -- no single woman ought to be talking alone with a man at nearly midnight --'I will collect a new candle from the kitchen and retire again.'
'No, I will not excuse you, McAllister,' was his answer to that, made with a kind of growling good nature.
'I require entertainment, and it is a good servant's part to do the Master's bidding.
Sit down, McAllister, and entertain the Master.'
He had made no move to rise.
All his polite gentlemanliness which he particularly observed with his poor patients, Matey and herself, was quite missing.
On the other hand, it did not seem likely that he was prepared to do anything improper, such as jump on her.
Sally-Anne was very conscious that she was wearing only a nightgown and a light shawl, and the fear which she sometimes felt these days in the company of men, and had felt ever since -- no, forget that -- was threatening to overwhelm her.
Nevertheless she thought it best to humour him and sat down on the side of the table away from him and put her clasped hands on its polished top.
She took them off again when she remembered that it would be she who would have to Ronuk it again if Matey saw its shine marred.
'How shall I entertain you, Dr Neil?' she enquired, and then regretted what she had said -- it might bear the wrong meaning.
But he made no double-edged comment in return, simply said, 'Tell me of America, McAllister.
Of your old home.
I have never visited America and probably never shall.'
Well, that seemed innocent enough and it was probably best to humour a drunken man.
She had overheard Papa say that once to Mama about a particularly notorious senator whom he was compelled to entertain.
'My home?'
She thought for a moment, then decided to tell him of the little house in Washington where she and Mama had lived before Mama married Papa -- She wondered briefly what Dr Neil would have made of that story.
'We lived for a time in Washington DC,' she said slowly.
'We weren't rich, only comfortable.'
And that part, at least, was true.
'We had a frame house with a garden around it, quite small.
There was a picket fence, and a small gate.
I have seen nothing quite like it in England.
We were not in central Washington, you understand, but in a new suburb.
Mama worked as an aide to a senator...'
She knew immediately that she should not have said that.
Dr Neil picked her up immediately, saying, 'And your papa, McAllister -- he did not object to your mama working?
What did he do?'
Sally-Anne's fertile mind provided an answer which had a kind of truth in it, or at least made a passing gesture in that direction.
'Oh, Papa was an accountant then.
But after he inherited a little money and set up in his own business Mama stopped working, and then she had my little brothers.'
And that piece of undeniable truth was, perhaps, a mistake, too, for Dr Neil was not so drunk that he did not ask, without a pause, 'And what happened to them, McAllister, when you lost your parents and your fortune?'
This was a bit of a poser, but Sally-Anne, never at a loss, said, thinking of kind Uncle Orrin who would surely look after her and her three younger brothers if anything happened to Papa and Mama, 'An uncle took them in, but he said that he wasn't prepared to keep a great girl, and that I must fend for myself, and find employment, which I have done, ever since.'
'So, you didn't take to drink, McAllister, when you were disappointed in life and love.'
And his voice had a note of self-mockery in it which surprised her.
'Women usually don't,' she said.
'That's a man's privilege.'
'Oh,touche again,' he riposted, the laugh in his voice genuine this time.
'You have a sharp and perceptive tongue, McAllister.
Are many American girls like you?
If so, you all reproach me with your cheerful resilience.'
'I suppose,' said Sally-Anne.
She had never thought herself as part of a mass called American girls.
She was Sally-Anne...
McAllister.
Dr Neil picked up his glass, filled it from the bottle, waved the bottle at her, and asked, 'Do liberated American girls ever drink spirits, McAllister?
Am I being inhospitable?'
'Some do.'
Sally-Anne gave the matter as grave a consideration as though he were asking her her opinion of the latest exhibition of Japanese art, or Beerbohm Tree's newest play.
'But not the kind that I am likely to know.'
Dr Neil gave a crack of laughter at that.
'And that should teach me not to ask ridiculous and impertinent questions,' he remarked, and his normal cheerfulness seemed to have returned, which was a relief.
Sally-Anne thought that Dr Neil might be a man, and therefore to be hated, but he was a good and caring doctor, and really shouldn't be abusing himself with alcohol.
Papa said that it was a good servant, but a bad master, and he never joined in temperance rant.
'Should you be drinking so much?' she asked, greatly daring again.
'Probably not, McAllister, probably not.
But tell me, can you think of circumstances where one might get drunk to reduce pain?
Either physical pain -- or that induced by unwanted memories?'
Oh, he had struck home harder than he knew.
Sally-Anne had one pain, one memory of which she dared not even think for fear that she would lose all command of herself, one memory which she always pushed away when it tried to attack her.
She pushed it away now.
Would spirits dim that pain?
Had he a pain like that?
If she joined Dr Neil and his bottle, would she feel better?
Dr Neil had seen McAllister's face change even as he spoke so carelessly to her.
He knew that in some way he had hurt her.
He regretted it.
The man who had endured great mental and physical pain always, when in command of himself, tried to avoid inflicting it on others.
But the demon which had driven him to drink that night, after months of abstinence, had him in its thrall.
Perhaps -- no, not perhaps, but because McAllister, with all her youthful ebullience and charm, was in his house, she had revived something in him which he did not want to feel and he had called up the demon to assuage it -- no, to kill it.
'So,' he said, when she did not answer him, and the beautiful mouth quivered, ever so slightly, 'even pretty little McAllister has her secrets.
Not so young and green, after all.'
'Oh,' said Sally-Anne, rising, 'you are hateful like this.
You will not respect yourself in the morning.
And I don't think that you own me after my day is over, Dr Cochrane.
Pray excuse me.
I will return to my room, and dispense with the candle.'
Her own memories were so strong that she feared that she would burst into hysterics before him, and that would never do.
She had vowed never to give way to that.
The lonely dark was preferable to staying to be taunted.
She made for the door, but as perforce she had to pass him he put out a hand and caught her by the wrist.
Not hard, not tightly, but gently, a warm, almost loving clasp, but when he spoke his words were far from loving -- they were jeering, even.
'Oh, come, McAllister.
What's your game, eh?
Tell me that.'
'My game?' echoed Sally-Anne, her heart suddenly bumping now that he was holding her, acutely aware of how little she was wearing.
Besides, she had been caught like that by a man once before.
'Yes, your game, McAllister.
You are playing a game, are you not?
What are you doing here?
Is it a bet?'
'A bet?'
Scalding anger at him consumed her.
'Of course it's not a bet!
I...need the work.'
'You do?'
His voice was hatefully mocking, and whether it was the drink talking, or his resentment of all women because of what one beautiful woman had done to him, Dr Neil did not know.
'Now, why don't I believe you, McAllister?
Entertain me even more.
Tell me the truth for once.
That would make you a pearl among women, and no mistake.'
Sally-Anne tried to wrench her wrist away, but to no avail.
His grip tightened.
'No, indeed.
You are not to go.
The Master will not have it.
We may face the black night together for a time.
You may tell me why a young lady who, whatever she says, has never done any work in her life before, has come to the East End to find it.'
'I told you,' said Sally-Anne, exasperated by his probing, annoyed that he had seen through her whoppers.
'And,' she added a little triumphantly, 'if my performance as a parlourmaid doesn't satisfy, then have the goodness to dismiss me, not bully me in the middle of the night.'
One word in this impressive little speech was unfortunate.
'Ah, yes, performance,' he murmured, taking another great gulp of whisky from the glass in his left hand, still holding her tightly with his right.
'What a good word, McAllister.
I have never, in all my life, known such a diligent, hard-working, uncomplaining maid as you are.
As good as a play.
Real maids are quite different, moaning and wailing and kissing the local bobby between the dustbin and the outhouse.
I shouldn't complain, I suppose.
Matey and I get the benefit of your...performance.
No, don't pull away.
The Master commands you to stay.
Don't you know that a really submissive maid soon learns to please the Master in every way...every way, McAllister?
You take me, I'm sure.
I most desperately need entertaining, as you can see.'
He was playing with her, teasing her.
He had no real intention of assaulting her, however great the temptation which she presented to him.
But he felt that McAllister had to pay something back for all the fairy-tales which she had told Matey and himself, and which Matey had so gullibly swallowed.
He was not to know that a real fear was beginning to overwhelm Sally-Anne.
Oh, it could not happen again.
God could not be so unkind.
She began to tremble, tried to compose herself, to appeal to the coolly aloof Dr Neil of the day, not the drunken midnight man slouching in the great armchair.
She tried to control her voice, and was pleased that it was as steadily calm as she could have hoped.
'Please release me, Dr Cochrane.
I am sure that you do not really wish to frighten me.
I know that drink makes men...irresponsible.
I am not so young and green that I am unaware of that.
And if you let me go I will forgive you for the way in which you have just spoken to me.'
Dr Neil, thus so firmly rebuked, closed his eyes.
He heard the calm voice, but could feel the trembling body which gave it the lie.
He released the small hand, and said, his voice suddenly pleading, 'Don't go, McAllister.
I didn't mean to frighten you.
I...would rather not be alone.
Stay, if only for a little while.'
'And you should not ask me that, either,' said Sally-Anne, head erect, carriage proud, refusing to be won over.
'We should not be talking alone, down here, in the middle of the night.'
And then, free of his grip, she recovered her courage and her gallant spirit to say to him, 'And you really ought to stop drinking.
At once!
What would Miss Mates think if she found us here like this?'
'The worst, I suppose,' said Dr Neil, somewhat wryly.
'You recall me to common sense and my duties, McAllister.
A parlourmaid beyond compare.
You are right to rebuke the Master when he forgets himself so.'
He sat up, corked the bottle, picked up his glass, pushed it away, cocked his head on one side, and said in a pathetic, slightly injured voice, as though he were the victim, not Sally-Anne, 'There -- will that do, McAllister?
I promise not to offend again, although I cannot promise that I will keep the promise.'
Sally-Anne struggled to repress the laughter which suddenly swept over her -- relief after tension, she supposed.
'Oh, you really are too ridiculous.
You owe it to yourself to behave properly, not to me.'
'A female Solomon,' said Neil, sighing, and then he gave her the most charming grin, and through the ruin of his face she suddenly saw what Matey must have known -- the handsome soldier-boy he had once been.
'And even female Solomons must have their proper rest.
Goodnight, McAllister; take the lamp.
I can find my way up in the dark, or fetch a candle from the kitchen cupboard for myself.
Here --' And he handed her the lamp, which had been designed to be portable.
'Be careful on the stairs, mind.
I should not like my playing the Good Samaritan to end in a flaming holocaust.'
Sally-Anne could not demur.
She took the lamp from him.
At least she would be able to finish her article now.
'Goodnight, then,' she said, 'and thank you.'
She could not prevent herself from saying in a doubtful voice, 'You will be careful when you go to bed, won't you?'
Dr Neil laughed a little at that.
'I may be overset,' he said, his voice and manner that of the daylight man again, 'but I am still quite capable,'and he gave her his charming smile.
She took the memory of it upstairs to bed with her, but all the time that she wrote she could see him sitting there as he had been when he had first lit the lamp, his face full of an old pain.
Daylight brought the prosaic world back again, and a Dr Neil who was exactly like the man she had always known -- it was as though she had imagined the improper advances of the night hours.
'You're dreamy today, McAllister,' said Miss Mates accusingly, making Sally-Anne scrub the kitchen floor all over again, the first effort not being deemed sufficiently satisfactory.
The fact that she was basically a kind soul did not mean that she was soft on Sally-Anne or herself.
'You must try a little harder.
We are having company this afternoon.
The teapot, milk jug, sugar basin and tongs will require cleaning with Goddard's powder before midday, so that all will be proper, and the visitors will not be able to say that this house is a slum inside, even if there is one outside.'
'Who are we expecting, then?' said Sally-Anne, who, like all good servants, was beginning to identify herself with her employers, and was one reason for her surprising success as a maid of all work.
'Mrs Teresa Darrell,' said Matey, making a little of a face.
'She is a kind of cousin of Dr Neil's.
She was married to Dr Neil's captain who was killed in nasty Africa, as poor Dr Neil nearly was.'
She paused.
She did not feel that she could tell McAllister that Mrs Darrell had designs on Dr Neil, and that she, Matey, was not very happy about the idea that Mrs Darrell might become Mrs Cochrane.
To begin with, Mrs Darrell did not approve of Matey.
She thought that it was Matey who had encouraged Dr Neil to become a doctor and take up an East End practice.
'Not at all the thing for a Cochrane to do,' she had said more than once.
'After all, he is still poor Stair's heir, poor Stair never having married.'
Stair was Alastair, Neil's older and only brother, and was usually called 'poor Stair' because he had drunk, gambled and wenched away the small remains of the Cochrane estate, most of which his father had already dissipated before him.
Matey thought unkindly that Mrs Darrell rather liked the idea of becoming Lady Cochrane when Stair went to his last rest, which at the rate he was living might not be long.
Sally-Anne knew that Neil was Stair's brother; she had met him, and had not liked him much.
He had been a friend...of...him.
Neil's origins made his presence in the East End even stranger -- and he had the impertinence to quiz her for her presence, when his own was just as odd.
Dressing for the afternoon, she wondered what Mrs Darrell was like, and whether she was worth all the tohu-bohu of preparing a slap-up tea, and being formally 'At Home', which had resulted in even more work for McAllister, as Sally-Anne was increasingly beginning to think herself when she was being a maid of all work.
Prompt at four o'clock, Dr Neil, complainingly dressed in his best suit at Matey's request --'Do I have to?' he had said, rolling his eyes at McAllister -- Matey in a bottle-green velvet gown which McAllister had not seen before, McAllister in her best sateen skirt, white blouse, and small lacy cap with streamers, were all sitting in the parlour when the doorbell rang.
At least Dr Neil and Miss Mates were sitting; McAllister was standing at the kitchen door, and moved sedately forward to answer the urgent bell.
Not promptly enough, apparently, for the bell rang again, peremptorily.
Mrs Darrell was in her middle thirties, was tall and statuesque and had been a bit of a beauty in her day.
She was not, McAllister thought, very tastefully turned out, being inclined to the upholstered look, with a remarkable array of feathers in her hat above a maroon walking dress, trimmed with black.
Her hat was nearly extinguished by the feathers.
She had brought along her companion, Norton, a thin, harassed-looking woman, whose drab grey attire set off her employer's brilliance.
McAllister took their hats, boas, parasols and wraps into the waiting-room, there being nowhere else to take them, after showing the visitors in.
'Well, my deeah Neil, you are looking positively deevy.
So much bettah,' said Mrs Darrell in what McAllister recognised as an imitation of the society shriek affected by those around His Majesty King Edward VII.
She kissed his cheek, managing to avoid the scarred one by some contortion recognised by both her victim and McAllister, who had taken an instant dislike to her.
Why, she could not think.
'Do sit down, Tess,' requested Dr Neil.
'Yes, I am feeling better these days.
I must say that you look in health.'
Doubtless so rosy because of all the rouge on top of papier poudre, thought McAllister nastily.
She had taken up her post by the kitchen door, standing there, hands neatly folded over her spotless apron.
'I see that you have a new gel, Matey,' remarked Mrs Darrell.
'Bit elderly, ain't she, for a maid?'
Dr Neil avoided looking at a bridling McAllister, particularly when Mrs Darrell went on to say, still in the same shriek, 'Hope she's honest, Neil.
The last chit I hired not only got herself -- well, you know, with the coachman, but made off with some of my best lace when I turned her away.
Oh, servants, servants!'
'Just as well I have so few of them,' said Dr Neil.
He still dared not look in McAllister's direction, for Teresa Darrell, frequently calling on Norton for support, continued her diatribe about the servant question.
I really should not be surprised by this, thought McAllister, fuming.
As I well know, the servant question is a constant subject for conversation in every drawing-room, so why do I resent it so much now?
Because I have spent the whole day since six o'clock this morning working, and working hard, and this...painted maypole has done nothing all day long and will do nothing tomorrow, except complain about those who do work.
Does she think that I have no feelings, or have been struck deaf because I stand by the kitchen door, silent?
'Tea, McAllister,' said Matey, to stem Teresa Darrell's flow.
This, diverted, now turned into a recital about every distant relative Neil Cochrane possessed, all of whom, apparently, had only one wish -- to see him back in polite society again.
She had not finished exhorting Dr Neil about this when McAllister, who could hear every word in the kitchen, returned with the tea-tray, staggering under its weight.
By the time that she had balanced the tray on a small table, the silver teapot carefully placed where Matey could preside over it, Mrs Darrell had embarked on an attack on Dr Neil and the profession which he had taken up.
It was all done so genteelly that it set McAllister's teeth on edge.
She was busy handing around cups and saucers, damask napkins, silver knives and cake forks, setting out the cake-stand, circulating plates of cucumber sandwiches, little patties and everything else considered appropriate for an afternoon tea in 1903.
'And I have no appetite, no appetite at all,' shrieked Teresa Darrell at Dr Neil, and consuming her fourth sandwich while she spoke.
'Now, if you set up a decent practice, somewhere in Belgravia, instead of this...slum, you could come and treat me.
Think of the fortune such a charming creature as yourself could make.
No one would mind calling in a doctor of your social standing.'
'Now, Tess, I have told you often enough that I have no wish to set up in Belgravia, Harley Street, or Wimpole Street,' said Dr Neil quietly, refusing the sandwiches which a servile McAllister was handing him.
'There are plenty of medical men to do that, and few to practise here, where the need is so great.'
'Oh, but you are a saint, Neil.
We all know that.
He is a saint, is he not, Norton, and Matey, my dear?
But you should think of yourself a little; a halo is all very well...'
And she burst into tinkling and self-approving laughter.
And what sort of halo does a drunken man at midnight, threatening to assault his parlourmaid, wear? thought McAllister nastily.
And how could Dr Neil, who was such a man of sense, tolerate such a...nodcock?
Teresa Darrell waved for the cake-stand.
McAllister took it round to her, at the point where she began again on the necessity of Dr Neil to rejoin polite society.
This went on for some time, with Dr Neil fencing politely, Matey looking grim, and Norton sighing in counterpoint to la Darrell.
The climax to the whole tasteless display came when tea ended, and McAllister was required to take everything away.
Balancing the tray, to which Mrs Darrell had kindly added the plate containing those cream cakes which she had not managed to devour, she heard her say to Dr Neil, 'You really ought to go into society again, my dear.
You must surely have recovered from your shyness over your unfortunate injury.
After all, we all know what you look like by now!'
Rage, pure and delightful, exploded inside McAllister.
She could look neither at Dr Neil, nor at the tasteless harpy who was tormenting him with her tactlessness.
The rage incited her to the most positive and extreme action.
A small tuffet stood between herself, Mrs Darrell and the kitchen door.
With the most artfully devised deliberation she managed not to avoid it, tripped spectacularly in such a fashion that she fell forward, the contents of the tea-tray, cups, saucers, dregs of tea, milk, sugar and cream cakes, all cascading neatly into Mrs Darrell's lap, with McAllister herself, purple in the face as a consequence of stifling a dreadful desire to laugh, landing gracefully on her knees at the good lady's feet.
For a moment the noise was indescribable, much of it contributed by McAllister, who set up a keening cry, and, in endeavouring to make matters better by dabbing at the debris on Mrs Darrell's lap with a damask napkin, made them worse.
Inside McAllister, who was enjoying her own performance as a clumsy hoyden, a wicked devil was laughing itself stupid, until she felt strong hands under her armpits and Dr Neil hoisted her clear of her victim.
He set her on her feet, and said in a voice which she hardly recognised, 'You will go immediately to the kitchen, McAllister, to await further instructions, but only after you have apologised to Mrs Darrell, at once!'
And his voice rose on the last two words.
Dr Neil knew perfectly well that what had been done was no accident.
He had seen McAllister's face just before her trip and something had alerted him, even before everything had landed in Teresa Darrell's lap.
'No,' said Mrs Darrell violently to McAllister, who was dipping a curtsy, and beginning a stammering apology, but only because Dr Neil had asked her, not because she was sorry for what she had done.
No, indeed.
It was no more and no less than such a creature deserved.
'No, you wicked gel.
I don't want your apologies.
You have quite ruined my gown.
I demand that you turn her away, Neil.
You surely cannot wish to keep such a clumsy thing.'
'No,' said McAllister in her turn, face white, and trembling as though to lose her post would be the tragedy which it would have been to the servant she was pretending to be.
But she did not wish to lose her place, hard though her life as a servant was; such an outcome would be a failure which she could not endure.
'It was an accident; please don't turn me away.'
And the slight wail in her voice was genuine, as Dr Neil could tell.
'Go to the kitchen, McAllister!' he commanded.
'Matey, you must help Tess and Miss Norton to clean Tess's ruined gown.
I am truly sorry, my dear cousin, but the girl is new, not yet trained.'
'From what I have seen of her, she never will be trained,' snapped Teresa Darrell.
'An insolent little piece.
You ought to turn her away.
But there, you were always soft-hearted.'
And she suffered Matey to lead her upstairs to try to repair the ravages which McAllister had wrought.
Left alone, Dr Neil walked to the kitchen to find a strange scene.
McAllister sat in the Windsor chair, her head bent, crying.
It was the very last thing he had expected to discover given McAllister's fiery and impetuous nature.
He had fully intended to tell her exactly what he thought of her disgraceful conduct, for he had no doubt at all that what had been done had been done deliberately.
But the sight of her in tears disarmed him in the strangest way.
Crying women had always annoyed him, and he briefly wondered why McAllister's tears should affect him so differently.
She was crying with an almost fierce abandon.
One might have thought her heart was broken.
'McAllister,' he said gently.
'Why did you do it?'
McAllister looked up.
She did not know what had come over her.
Always before, after similar wickednesses, she had felt almost gleeful triumph, but this time, although she was not truly sorry for what she had done to Mrs Darrell, she felt something like remorse.
And to be turned away as well!
'Do what?' she sobbed.
'I tripped.'
'Tell the truth, McAllister,' said Dr Neil, still gentle.
'I saw you, immediately before you fell.
I know that you tripped, but you tripped deliberately.
Why?'
McAllister dropped her head.
Why did she feel so strange, so...ashamed?
Almost as though she had let Dr Neil down, rather than defend him, as she had meant to, by punishing a tactless fool who was hurting him.
Why did she mind his being hurt so much?
For once, she must tell him the truth, never mind that he had behaved so badly last night.
'Yes,' she said into her sodden handkerchief.
'It was deliberate.'
'But why?' he repeated, genuinely puzzled.
He was hardened to Teresa Darrell's tactlessness.
She had been practising it on him for years.
'I didn't like her,' said McAllister stiffly.
'She deserved it.'
And then, with a flash of her usual impetuous spirit, 'She was horrible to you, not once but again and again.'
'She doesn't mean to be,' said Dr Neil perceptively.
'She thinks that she is helping and encouraging me.'
'That makes it worse, not better,' burst out McAllister.
'What a fool!'
Dr Neil sighed.
Like McAllister, he wanted to know what was coming over him.
He wanted to comfort naughty McAllister, not punish her.
'It was a very unkind thing to do, McAllister,' he said.
'Was it right to criticise her for being, as you thought, unkind, and then be even more unkind yourself?
Teresa Darrell is a most unhappy creature who lost her young husband shortly after their honeymoon, and I am one of the few people left who can remember and mourn him.'
McAllister began to cry again, dreadfully.
It was as though all that had happened to her in the past year was suddenly before her, and all her own shortcomings into the bargain had landed on her in a heap, and were destroying her.
She had been almost unnaturally brave for so long, and now it was as though she had been given licence to cry over everything.
Worse, Dr Neil was being so kind, when reason told her that no one would blame him if he turned her away for what she had done.
It was plain that he was not going to.
Suddenly, as much to his own surprise as McAllister's, Dr Neil went on his knees beside her.
He pulled out his own spotlessly white handkerchief, which McAllister had laundered and ironed earlier that week, and said urgently, 'Oh, do stop crying, McAllister.
It is not like you at all.
I am not going to ask Matey to turn you away.
Suppose I suggest to you that we stop part of your wages to help pay for the damage to the china and to Mrs Darrell's dignity, and you promise to be good in future?'
'Oh, dear,' said McAllister tragically.
'I have promised that so often in the past, and it has never answered yet!'
Dr Neil gave a shout of laughter.
'Come, that's better!' he said.
'You remind me of myself last night.
We are a sorry pair, are we not?'
McAllister stopped crying, said shyly, 'Not promising to keep the promise, you mean?'
She wielded his handkerchief with her customary vigour.
'Something like that.
Now I must go to say goodbye to Tess Darrell, and apologise for half-trained servant girls.
You will behave yourself in future?'
There was such an appeal in his voice that McAllister felt really ashamed, and then when he had left her, refusing to take his hanky back, saying, 'You need it more than I do,' she also felt something else -- a kind of purging, a relief.
She knew that she would probably be naughty again, because that was the way she was, but never again would she be quite so prone to take such instant action.
And she also knew that Dr Neil was well aware of why she had done what she did, and she knew something else -- he was not only sorry for McAllister, but was even sorrier for Mrs Darrell, as well, a widow who had lost her husband in the most cruel circumstances, and for whom he knew that he would never offer, despite his pity for her.
And that was why she had done a wrong thing, and that her thoughtlessness was selfish -- a new idea for the spoilt and pampered girl McAllister had so recently been.
But had it not been to rid herself of her aura of wealth and privilege which had created her feeling that she was the darling of the gods -- although the same gods knew how brutally they had treated her -- that she had come to the East End to work, and to live as though she really needed to, and to survive on the pittance which she had earned, without bolting back again to luxury and comfort?
She would be more careful in future, consider others a little more.
She could not, must not fail in her self-appointed task so soon after beginning it.
CHAPTER FIVE
'Up West?' said Matey, handing Sally-Anne the china from the small dresser in the kitchen, which had been scrupulously washed that morning and was now being replaced by Sally-Anne who was standing on a small stool to reach the top shelf.
'You are thinking of going there this afternoon?'
'Yes,' said Sally-Anne.
It was her afternoon off in about twenty minutes.
'I thought I might like to window-shop.
I can't afford anything this week, of course.'
And she cast her eyes down modestly at this reference to her pay being docked for yesterday afternoon's fiasco.
'I suppose you could take the bus,' allowed Matey.
It ran from the end of the road, pulled by a rather weary horse which had a habit of stopping suddenly to the dismay of its passengers.
As usual Sally-Anne was not quite telling the truth.
Her real destination was Fleet Street and the offices of the weekly Clarion Cry , but she could hardly inform Dr Neil or Matey of that interesting fact -- it didn't quite fit the sad picture which she had drawn of the friendless, abandoned orphan without employment -- not that the money which she earned from the Clarion would have kept her.
Dr Neil had been kind and friendly to her when she had served him breakfast.
Despite himself he had been touched by her fierce defence of him, and although he frequently entertained Mrs Darrell he only did so because she was the widow of his dead friend.
He was quite aware of her marital designs on him, but she was very much part of the life which he had rejected, even if she refused to recognise that fact, and tried to reclaim him for it.
Eating his bacon and eggs, 'Sunny-side up,' Sally-Anne had once said when she put his plate before him, drinking his coffee -- he preferred it to everlasting tea, he said -- he looked much better than he had done the day before; no midnight drinking, Sally-Anne thought.
Mental transference of some sort must have taken place because, putting down his napkin and rising, he had said to her when she had begun to remove his breakfast china, 'McAllister.'
'Dr Neil?' she said, turning and bobbing at him like a proper servant, a manoeuvre which amused him, so that his lips twitched at the unlikely sight -- it was so much at odds with her determined personality.
'About yesterday -- and the night before...'
He hesitated, then gave her his unexpected smile again.
'If you keep to your promise -- to behave yourself in future when I have guests -- then I shall try to keep to my promise.
Fair's fair, after all.'
'Oh, how splendid!' exclaimed Sally-Anne warmly, and with Matey safe in the kitchen, unable to hear, added, 'No more whisky at midnight!'
'And no more tilting tea-trays,' he said.
Above all things he wanted to see McAllister's bright and bubbling face, and she had been unwontedly shy and subdued this morning.
She rewarded him with such a beaming smile that he took the memory of it into the surgery with him, where it stayed all morning, brightening the day for him.
His unlikely American maid-of-all-work was beginning to occupy his thoughts more and more...
Sally-Anne took the horse bus.
She was wearing her servant's clothing -- a bottle-green cotton skirt, a white cotton shirtwaister, a light black shawl, black stockings and heavy shoes.
She was carrying a small tapestry bag which contained her rather crumpled manuscript.
But she did not ride on the bus all the way up West, leaving it instead where its route crossed Fleet Street.
She walked briskly by St Paul's Cathedral, along Ludgate Hill, to reach her destination.
The Clarion's offices were small and dingy, but despite this, and the fact that it only appeared once a week, the magazine had a large and growing circulation because of the crusading character of its editor, J. D. O'Connor, who had been a leading journalist with the Morning Post before he had struck out on his own, financed by money left to him by his land-owning father.
Sally-Anne had met him at a party; he had long been settled in England, and he had been impressed and amused by her fiery conversation and her obvious intelligence.
He knew that her mother had been a well-known journalist on the Washington Gazette before her marriage to Jared Tunstall, Sally-Anne's father and a Yankee millionaire, who as the result of his fame as one of America's so-called robber barons, and his friendship with King Edward VII and his circle, was nothing if not notorious.
Half seriously, half as a joke he had asked her to write a short piece for him on what she had termed in conversation with him 'wasted women'.
By this she meant society women who were compelled by their roles as wives and hostesses to live empty and meaningless lives.
He had thought that that would be the last he would hear of her, but a few days later she had sent him her first piece, and he had been so impressed by it that he had printed it and asked her to write more for him.
'A weekly column,' he had suggested, and when some weeks later, looking pale and rather ill, she had visited him and suggested a series of articles about the stunted lives of women and children in the East End, to be written from the inside, she had said, not the outside, he had rapidly agreed.
Her first piece, written before she had joined Neil Cochrane's household, from the viewpoint of a privileged outsider, had been even better than he could have hoped, and had caused favourable comment.
To preserve her anonymity they had agreed on the pseudonym Vesta, and if at first he had thought a few days in the East End might dampen her enthusiasm he had been proved wrong.
Sally-Anne approached the cubbyhole giving admission to the building with some trepidation, rightly so when the commissionaire had sniffed at her, and demanded to know why she sought admission -- her clothing and general appearance were not those worn by other visitors to the editor.
She produced O'Connor's card, with the words, 'This lady is to be admitted to my office on demand', written and signed by the great man himself, and mounted some uncarpeted wooden stairs to his somewhat grimy sanctum.
O'Connor was in his shirt-sleeves; he rapidly assumed his jacket and pulled out a chair for her, first wiping it with a none too clean handkerchief.
'Pray sit, Miss Tunstall.
I trust I see you well?
You have something for me?'
'Two pieces,' said Sally-Anne, sitting down, grateful that her clothes were utilitarian, rather than the exquisite gowns which she had formerly worn.
'One is on the general treatment of poor women, the other is more particular.
It is called, 'An East End Doctor's Surgery', and tells of his morning with his patients, their problems, and the difficulties of giving adequate medical care to those who can barely afford it -- even when the doctor is good-hearted -- as this one plainly is.'
She handed the two pieces over, and O'Connor began to read them rapidly, looking up at her occasionally while he did so.
He had already recognised that she was wearing poor clothing, and it was also obvious that she was taking her work seriously enough to sacrifice her privileged lifestyle and live among those of whom she wrote.
He was filled with sudden admiration for a professionalism which he had not suspected she possessed.
He put the papers down when he had finished them and looked at her long and steadily.
He had noticed that she did not identify the doctor, calling him Doctor X. Sally-Anne experienced a sudden dismay.
Oh, no, never say that he did not like them, after all her hard work and the writing of them in the small hours after her demanding duties as a dogsbody had already tired her!
She clasped her hands together, and cast a look of entreaty at him -- which he correctly interpreted.
'Have no fear,' he said gently.
'Your work is excellent, written from the heart, but skilfully so.
I could do with you on my permanent staff -- but I think that these articles will hit home, cause a stir, such a stir as Henry Mayhew created some time ago when he, too, wrote of the cabined and confined lives of the capital's poor.
No, it is not these articles which cause me concern, but you.'
'Me?' said Sally-Anne inelegantly, then added, 'I don't understand.'
'Miss Tunstall,' he said.
'Does the American Embassy, does the ambassador, know where you are living -- and, more to the point, what you are doing?'
Sally-Anne flushed a little, and said, 'Uncle Orrin?
No, I thought that...'
She paused.
'That he might stop you,' said O'Connor, amusement on his face.
'Yes, I'm sure that he would have done -- the East End is not the safest place for a beautiful and rich girl of good family to live.'
'Oh, I'm not a fool,' said Sally-Anne spiritedly.
'No one knows that I am that -- rich and of good family, I mean.'
O'Connor looked at the glowing and vital face opposite him, and thought wryly that Miss Sally-Anne Tunstall, society beauty and heiress, carried her pedigree on her face and in her carriage.
'You are,' he said, 'presumably living in this doctor's household.
By your appearance you are posing as a servant.
You must find the work hard -- which makes the writing of these articles even more commendable.
At least let me know of your address, even if you feel you cannot tell the Ambassador, your uncle, or his wife where you are.
Then, if I hear concern expressed about you -- that you have disappeared...'
Sally-Anne thought all this a great to-do about nothing.
'Oh, no,' she said confidently.
'Yes, I will give you my address, but they think that I am touring with my friend Laura Parslow and her family.
Laura and I arranged everything between us.
She thinks it all very romantic -- she does not know what hard work being a servant is -- and nor did I, for that matter.
I shall write you a piece on that, too.
It is quite unbelievable.
Even the kindest of souls, and the housekeeper for whom I work is certainly that, have no idea of the magnitude of the demands they make on one.
It's a good thing I'm very strong -- otherwise I might have dropped dead, like an over-worked horse, after a week of it.
I shall never feel the same about cosy coal fires again; I have heaved too many buckets about for that.
You have no idea of how much coal is needed to keep a small grate going so that one might cook.'
This heartfelt confession had its effect on her hearer.
It's a good thing I'm not a womaniser, he thought, or else Miss T would be in danger; she is so attractive in her joie de vivre -- and when that is added to beauty!
I wonder what the good doctor and his housekeeper make of her -- and what story she has concocted to explain her presence there?
But he said nothing of this, merely noted down her address, sighed a little at it, and made his latest recruit promise to visit him once a week with her piece.
'Without fail,' he said sternly.
'And if you are too tired to write you must come to see me in any case, and then I shall know you are safe.
I have a good mind to ask you to serve full-time on my reporting staff, but I am selfish enough to want you to go on writing for me these exposes of low life, particularly the way in which such misery afflicts women.
You are doing your sex a service, you know,' he finished.
That did it!
That just did it!
Sally-Anne could have jumped up and down for joy.
She thought of all the wretched women in the East End, and aye, in the West End too, not excluding her own maltreated self, and thought that at whatever cost she would carry out the task she had set herself -- to find out how the under-privileged lived, and to strike a blow for suffering womanhood while she did so.
Waiting at the stop for the horse bus to take her back to Vetch Street, Sally-Anne heard a voice she knew.
After her interview with J. D. O'Connor she had mitigated her whoppers to Matey and Dr Neil by moving on to the West End, where she walked along Oxford Street, entering Mr Gordon Selfridge's store, gazing as raptly at its wonders as though she were truly the poor girl whom she pretended to be, the whole place seeming quite different now that she no longer had her papa's bottomless purse at her command.
But she did make two purchases from the hat and the dress departments with the money which J. D. O'Conner had given her for the two articles which she had written for him.
She turned to see the maid-of-all-work from St Jude's Rectory, whom she had met after church last Sunday, staring at her.
Rose Bailey had a sharply pretty face with the knowing expression of the East End urchin which she had so recently been.
'Why!
If it ain't McAllister, from the doctor's.
Bin up West?'
Rose's cockney accent was so strong that Sally-Anne could hardly understand her.
Rose, for her part, thought that McAllister talked funny, and when Sally-Anne had told her that she came from the United States of America she had stared at her as though she had said that she came from the moon.
'Yes,' Sally-Anne said, 'I've been window-shopping.
I had my pay cut this week for being naughty at a tea party.'
'Garn,' said Rose, staring at her, and revising her opinion that Sally-Anne was a bit prim and proper.
'Yer never!
I've bought some nice ribbons, from a barrer off Oxford Street -- cheap, they were.'
And she fished a small parcel from her pocket, and showed Sally-Anne her treasure just as the horse bus finally groaned up -- or neighed up, thought Sally-Anne irreverently.
She had never thought that she would ever travel on such a thing, although she could not tell Rose that.
'Like to come wiv me next week?' offered Rose, the bus slowly bringing them nearer home with each plodding step.
'Could do wiv a pal, since my friend Lottie went into service up West.
Call for me at the rectory, why don't yer?
We could 'ave tea at Ma's, and then go up West in the evenin' fer a bit of all right.'
And she closed one eye in a grotesque wink, leaving Sally-Anne to guess what the bit of all right was -- a visit to a music hall or a theatre, she assumed -- wrongly, for she still had a lot to learn about the ways of the aliens among whom she lived.
Innocent that she was, for all the sophistication conferred on her by the great world in which she had once lived, she readily agreed to Rose's suggestion, and when they had parted Rose took care to remind her of her promise.
She thought that McAllister would be an added attraction for any nobs they might meet -- she was so much prettier than Lottie had been, and knew how to wear her clothes, looking better in her drab outfit than Lottie did in her furs and feathers.
She began to look forward to the following Wednesday -- Who knew where they might end up?
Sally-Anne, feeling that with Rose's offer of friendship she was entering more and more into her new life, was humming 'Marching through Georgia', and had just reached Atlanta, when she let herself into Dr Neil's.
The house was quiet, no sign of Matey's bustling presence, and when she walked into the kitchen, taking off her shabby black hat, pulling the hat pins out of her abundant hair, she found Dr Neil, sitting at the kitchen table, his tea before him, quite alone.
He cocked a sardonic eyebrow at her.
'Well, well, if it isn't McAllister!
Two wonders to consider: first of all it's your afternoon off, and secondly I was debating whether or not you might have tired of your games here and retired back to the place from whence you sprung.'
'Oh, indeed,' riposted Sally-Anne, eyes flashing and giving off all the danger signs which her family would have recognised.
'Now, why should you think any of that?
I can surely do what I like on my day off, not be reproached if I choose to come home early because I feel tired, and secondly, as I said before, I am not playing games.
I take my work here very seriously indeed .'
'So there...' said Dr Neil wickedly.
'I think your answer really needed that at the end to make it truly effective.
Or perhaps a tongue stuck out would be even more so.
Learning bad habits down East, are we?'
'If I am,' said Sally-Anne, glaring at him, 'it must be catching.
You seemed to have learned plenty, I must say -- baiting poor servant girls being one of them.'
'Ah, but we're not a poor servant girl, really, are we?' said Dr Neil, his head on one side again, the scar presented to her; it was a sign that despite his mockery he had accepted her, for with those of whom he was unsure, or not used to, he always attempted to shield them from its sight.
'So all that piff-paff won't serve, will it?
And if you really mean to be a good servant you should have noticed that the Master's tea is cold, and the pot needs -- refreshing, I believe, is the Yankee word.'
'I am off-duty,' said Sally-Anne  awefully , 'and I am bound to tell you that although I am here I am not here in the sense of waiting on you.'
'Not just waiting on me, McAllister,' he answered -- he seemed to have an answer for everything, she thought.
'I was about to suggest that we eat tea together, and, that being so, you would not mind a fresh pot.
However --'
'Well, that's an artful way to get me to wait on you,' interrupted Sally-Anne frankly, 'and I suppose, since you are so cunning, and I am dying for a drink, you might have your way -- perhaps a bonus on my wages might be an idea, seeing that I shall be doing this out of hours .'
Dr Neil thought that the way she occasionally trod on phrases when she spoke was charming rather than irritating, given that it enlivened and illuminated her forthright nature.
He was forthright back.
'Spoken like a true Yankee,' he said approvingly.
'Always on the chase for the Almighty dollar.'
Sally-Anne's laugh was robust in the extreme.
'The Almighty dollar!
Land sakes!
You are talking of a few pennies' pittance added to my weekly wage.
Papa would think that an odd way to describe the Almighty dollar.
And, come to that, why shouldn't I bargain for doing something extra?
You'd dock my pay soon enough if I didn't do everything you and Matey ordered me to do in my waking hours.'
She had forgotten that her papa was supposed to be dead, and had spoken of him in the present tense, a fact not wasted on Dr Neil, who made no comment, but asked politely, 'And does all this stockmarket bargaining mean that you are going to make me a cup of tea, or not?
I see that I must do a little bargaining on my own behalf.
I am hoping that you will observe that I have fetched in enough buckets of coal this afternoon to spare you having to carry them in tomorrow.
Now, what do you say to that, McAllister?
Quits?'
Sally-Anne had finished putting her things down, and was feeling enormously hungry as well as thirsty -- Papa had once said that she had an appetite as indecent as his own, and the wonder was that neither of them ever put any weight on!
'Not surprising,' her mama had said gently, 'seeing that neither of you ever sits still for a moment, nor ever stops talking, either.'
She would have been surprised to learn that observant Dr Neil was thinking the same of her and her abundant energy.
'Quits,' she said cheerfully, and carried the copper kettle to the brass tap let into the front of the water tank which sat in the big iron stove alongside the kitchen fire, and ran the hot water into it before she set it not on the fire, but on a small gas ring in the corner, lighting the gas with a match from a box of Swan Vestas.
After that, watched by her amused employer she emptied the pot of cold tea, scrupulously dried it, and set it to warm on the hotplates beside the fire, then placed a tea-caddy, ornamented with the features of King Edward VII and Queen Alexandra, a souvenir of their coronation, on the table, ready to spoon the required amount into the pot.
'So, our Yankee maid has learned the ritual of tea-making,' Dr Neil said with a naughty grin.
'How could I not,' said Sally-Anne, 'with Matey hard on my trail?
You may tell her when she returns how well I have learned my lesson.'
'Oh, yes,' said Dr Neil, watching her lay a place for herself, putting out the china, cutting fresh bread, buttering it from a clay crock, and spooning home-made jam lavishly on it.
'Eat like that, McAllister, and you will grow fat.'
'Not so long as I am working for you.'
And her blue-black eyes on him were merry.
'I shall grow thin rather, with exercise, and all the knee-bending I perform when I clean the carpets and scrub the floors.'
The kettle began to boil, and she completed the ritual of tea-making, finally standing the pot on a porcelain plate with Windsor Castle painted on it, mounted on sterling-silver trivets, and crowning it with a tea cosy knitted by Matey back in the days when she had lived in the Cochranes country house, now almost derelict since Stair's expensive debauchery had impoverished the estate.
Sally-Anne sat down, and began to eat the bread and butter, home-made scones and Victoria sponge cake as though she had not eaten for weeks.
'You did not treat yourself to a snack in an ABC or Lyons, then,' said Neil, gratefully drinking his own fresh cup of tea.
'No, indeed,' said Sally-Anne, mouth now full of currant bun.
'My pay does not stretch to that, especially after the moths got at it yesterday.'
'But you are making up for that today,' said Dr Neil gravely, his mouth twitching, particularly when she looked at him, a question on her pretty face, her mouth being too full for further speech.
'By eating so much of your and Matey's baking,' he explained.
'Can we afford such an appetite, I wonder?'
Sally-Anne was too busy eating to want to spar with him further.
It occurred to her suddenly that she had rarely enjoyed herself so much as she was now doing, seated in Dr Neil's kitchen, dressed like a skivvy and eating a huge meal, rather than languidly nibbling at it, as she would have done at one of Aunt Nella's 'At Homes'.
And how come Neil Cochrane was amusing her so easily without even trying?
He was now leaning back in his chair, having completed his own frugal meal, contemplating her while she polished off the final remains of her own large one.
'And tonight, McAllister,' he said, at last.
'What are you going to do tonight?'
'Why,' she said sweetly, rising and picking up her clothes, and her bag, now empty of the articles which she had given J. D. O'Connor, 'tonight I am going to turn in early, read and rest -- such a change after all I have done since this time last week.
And you,' she added, greatly daring -- but, since the Master was quizzing her so relentlessly, why should she not quiz the Master?
'What will you be doing tonight, after surgery?'
'You forget, McAllister,' he said lazily, 'no surgery on a Wednesday.
And that is why I am wondering whether you would favour me with a game of chess before retiring to your boudoir?'
'Some boudoir,' she replied, inelegant again.
'Are you serious?'
'I am always serious about chess, McAllister,' he replied, eyes mocking her.
'It is the only way to treat it.
Do you play?'
'Papa taught me,' she said.
'But I am not his equal.
Few people are -- at chess, or anything else.'
'Does that mean that you will give me a game?'
His voice was sardonic again -- he had noted the present tense when she had spoken of her poor 'dead' papa!
Sally-Anne nodded at him, to hear him say, 'Right -- leave the pots, McAllister.
Your duty tonight is not to wash up, but to allow the Master to slaughter you on the black and white field -- unless you play a game as wicked as Papa's -- in which case you may finish off what the Africans began on me!'
Well, he was certainly in a good mood, and no sign so far as she could see that he had been drinking what Matey called 'his nasty whisky', so she prepared herself to play him.
It was some time since she and Papa had played one of their fierce games; he never gave quarter to anyone, but she thought that she could remember what he had taught her.
'You are sure I am up to it?' she remarked a little naughtily; it would help if he underrated her -- another of Papa's precepts she had found worth following.
'If you're not,' said Neil grandly, 'then the Master will teach you.
What are servant girls for, but being taught by the Master?'
He took her arm, and the contact, for both of them, was startling.
A mixture of fear, at being touched by a man at all-- and something else, quite different -- swept through Sally-Anne so that she began to tremble as he walked her through to the parlour.
For Dr Neil the feel of her, so soft, the scent of her hair, and of McAllister herself, was so wild and sweet and struck him with such force that his breathing grew a little ragged and his body reminded him of how long it had been since he had satisfied it.
Unwanted emotions overcame him, too, not helped by her trembling, so much at odds with her usual fierce, unflinching pride.
It surprised him a little; he would not have thought her so easily affected by a man's touch.
Sternly, he told his body to behave itself, and his face reflected the power of his feelings, so that Sally-Anne, looking up at him, was a little daunted by his expression -- goodness, was that the effect the prospect of playing a game of chess had on him?-- and resigned herself to a grim half-hour.
But it wasn't like that at all.
Dr Neil sat her down in a chair opposite to his, and drew out the permanently set-up board on which she had frequently seen him working out problems.
His expression lightening a little once he had let go of her, he pulled a penny from his pocket, and said cheerfully, 'We'll toss for white.
Agreed, McAllister?
No favours for the fair sex.
I always play to win.'
Sally-Anne was delighted by his changed manner; he always looked so charming when he smiled -- even the scar seemed to disappear a little, and as she was always ready for fun herself -- a trait she shared with her papa -- and the game looked like being fun, she said, eyes shining, 'Oh, I play to win, too, but a good servant always does what the Master commands,' and she cast her eyes down in the manner of a stage domestic registering submission.
'At last, some proper respect!'
Dr Neil tossed the penny up, she called 'heads'-- and won.
'So,' he said, leaning back in his big armchair, his expression enigmatic, 'let battle commence.'
Oh, it was a battle, and no mistake, thought Sally-Anne; it was almost like playing against Papa -- only Dr Neil played more slowly; but he was just as cunning.
As she looked across at him during the game to see him, chin propped on his hand, his whole face a mask of concentration, but paradoxically, at the same time, as relaxed and self-forgetful as she had ever known him, it was plain how handsome he must once have been.
Living so close to him, she understood; she was beginning to see not the disfiguring scar and limp, but instead the essential Neil, who was so kind to herself -- for all his teasing of her -- and to Matey and his poor patients.
Neither of them spoke, and if Dr Neil was a little amused to see how playing a highly competitive game silenced McAllister, so that the whole of her vital spirit was directed towards winning, he did not betray it.
She had been well taught, and after making the odd minor mistake during the opening moves she soon recovered herself, and began to play with a concentration and a ferocity which was also a surprise.
He wondered a little at what the unknown papa who had taught her so well was like, if this was the kind of game he had favoured.
Time was magic, Sally-Anne thought; it stretched or contracted according to one's pleasure.
In one sense the game took no time at all, in another it seemed to last forever, and there in the quiet room, opposite the quiet man, she began to feel a kind of peace.
The clock above the fireplace chimed out its message, but it was lost on them.
Once, his bishop slaughtered in an exchange of pieces, Dr Neil looked up at her and said, 'It's a good thing that the Master didn't underestimate you overmuch, McAllister.
You play a wicked game -- if reckless.'
It was his only comment after an hour's play, and she gave him her flashing smile, remarking, 'Dare all,' Papa used to say, 'and the game will reward you -- always provided that you cover your back.''
This brought one of Neil's cracks of laughter and the comment, 'A wily gent, I see,' and then silence took over again, and lasted for a long time, until the front door opened, and Matey returned from the sewing circle at St Jude's Church Hall, and stared at them, sitting there in the dusk, hunched over the board, rapt.
She did not speak -- she knew better -- but walked into the kitchen, thinking furiously, among other things, that her treasure had not looked so happy and fulfilled for years as he did sitting opposite to his American servant girl -- and what could that mean, and did she approve?
She was not sure.
'Cocoa?' she offered, coming in a little time later, carrying a small tray with three enamel mugs of steaming beverage on it, and a plate of Marie biscuits.
She could see that the game was drawing to a close -- nearly all the pieces on the board had disappeared -- and although she had no idea who was winning she thought that the gladiators needed sustenance.
Sally-Anne sighed, yawned and stretched, not at all ladylike, and said joyously, 'Stalemate!
What a happy conclusion.
No one wins, and no one loses.
Oh, I did enjoy that!'
'Spoken like a true woman,' said Dr Neil, staring at the board, empty save for a few pieces clustered round a pair of kings in one corner.
'I cannot agree with you.
We men like to win, and it is quite improper for the servant to be so impudent to the Master as not to allow him to win.'
'Papa always said that only fools did that,' proclaimed Sally-Anne cheerfully.
'Throw a game, I mean.'
She took the proffered mug from Matey's hand.
'Unless you did it to gain some advantage, of course.'
'From all you say of Papa,' announced Dr Neil naughtily, 'I'm astonished to learn that he went bankrupt.
That sounds like a fate reserved for his competitors.'
Oh, dear, thought Sally-Anne, there I go again.
I really must watch what I say -- he has a mind like a knife; but aloud she said, doing her best to look distressed, 'Oh, he had a run of bad luck --' an explanation that hardly seemed to convince her hearer.
All three of them drank their cocoa with pleasure.
Neil was astonished at how pleasant the evening had been; he had seldom enjoyed a game of chess more, and, thinking this, said a little slyly, 'Another game, another night, perhaps?'
Sally-Anne, her heart thudding in the strangest way when he asked her this, as though he had said something much more intimate, and then suddenly understanding by the ambiguous way in which he had spoken that he had offered her other games than chess, and his wicked expression betraying that he had seen her confusion, flushed, and he added softly, so that Matey could not hear him, 'Come, McAllister, give me your answer -- you surely wish to please the Master in every way possible,' the last bit in a fake American accent so bad that she laughed out loud.
'Oh, I will play you at chess any time you like,' she informed him demurely.
'Pity that,' he murmured.
'There are other games master and servant can play, you know.'
'Yes, I do know,' she said, rising to collect her things from the kitchen where she had left them, 'but I have no intention of obliging the Master further, either now, or in the future.'
He watched her go, his smile crooked, heard her say goodnight to Matey -- her manners were always impeccable -- then followed her into the kitchen to pick up her candlestick and light the candle for her; it had grown dark while they had drunk their cocoa and eaten their biscuits.
'At least let me be your linkboy,' he said.
'I will escort you up the first flight of stairs, but I shall observe the proprieties by leaving you to look after yourself on the second landing.'
Sally-Anne saw Matey's face, half approving, half disapproving of Dr Neil's boldness, and said in reply, 'Thank you, sir,' in a neat parody of every servant who had ever lived.
And they set off, Dr Neil behind her, holding the candle high, so that it threw strange shadows on the walls and ceilings.
There was a small old table on the second landing.
He put the candle-holder down on it, then turned towards her, and before she could stop him he gave her a kiss so gentle that it hardly registered, and she stared at him, wide-eyed, put her hand to where he had saluted her, and said huskily, 'No,' although what she meant by that neither he nor she knew.
'Goodnight,' he said.
'Sweet dreams, McAllister.
Stalemates in life are not as satisfactory as those on the chessboard, you know.'
And then he was gone down the stairs, running lightly, for all his limp, and the last which she saw of him was the sandy-blond head before he turned the corner of the first landing.
CHAPTER SIX
'Hello, Rose,' said Sally-Anne briskly.
She was at the rectory kitchen door a week later, meeting Rose as she had promised, for tea with her ma and a trip up West in the evening.
She had been reviewing the previous week on her walk to the rectory.
It had been hectic, to say the least.
Dr Neil had been busy with a small epidemic of low fever which had fortunately missed Vetch Street, and Matey had decided to do some summer cleaning, spring cleaning having been missed out, due to the  deficiencies of Sally-Anne's predecessor.
Such a carry-on Sally-Anne had never seen, and all of it exhausting.
What with hauling carpets in and out, cleaning the wallpaper with a loose dough -- a trick Sally-Anne found oddly satisfying, if messy -- washing every curtain, sheet, blanket, cushion-cover, piece of crockery, floorboard, door and window-frame, ceiling -- you name it, Sally-Ann cleaned it -- by Sunday she felt quite stunned and had gone to sleep during the rector's sermon, only to be prodded awake by Matey, to a smiling Dr Neil's amusement.
'Heathen, are you, McAllister?' he had murmured during her fervent singing of 'All things bright and beautiful'.
'Which Amerindian tribe do you belong to, Huron or Apache?
And is the Great Manitou the God you worship?'
Which had made it difficult for her to stifle her irreverent giggles, particularly with Matey's severe stare on her.
Afterwards they had gone out into the brilliant sunshine of mid-June, the English summer being fine for a change, and Matey had introduced her to her other, lesser treasure, the curate Mr Julian Sands.
She called Sally-Anne, 'My new help,' which sounded more dignified than maid-of-all-work, somehow, McAllister had allowed.
Mr Sands, a pale, shy young man, had taken one look at Mr Cochrane's vivid servant girl and had fallen promptly into a deep and worshipful love for her, a fact written on his ingenuous face.
He had even felt compelled to shake hands with a girl whose like he had never seen before.
He held her hand a little longer than he should have done, and had stammered at her, 'I -- I have a cousin in Florida, Miss McAllister; I wonder if you have run across her?'
Sally-Anne had to stifle more giggles -- by his conversation Mr Sands thought that the USA was roughly the size of Wales.
Nevertheless she was her usual polite society self with Mr Sands -- she had learned in her papa's home and at the embassy how to put people at their ease -- and all in all the sight of his skivvy doing the gracious afforded Dr Neil another bout of inward sardonic mirth, however it delighted young Mr Sands, who thought what a splendid creature she was, and a great pity she was only a servant -- she had so much presence.
He had invited her to join the ladies' sewing circle, and even suggested that she might like to attend the Tuesday evening Bible readings which he ran -- he was sure that 'dear Miss Mates' would release her for the hour and a half the readings usually took.
'Oh, I'm sure we could let you out for that, McAllister,' Dr Neil had said, only for Sally-Anne to murmur modestly that since she was already well treated for time off she could not possibly ask for more.
'What a hard working treasure you are, to be sure,' Dr Neil had said, walking his two women back to Vetch Street, tipping his soft hat to those who were happy to greet the young doctor.
'Not at all,' said Sally-Anne, 'and before you plead for me with Matey let me tell you that you will have your tea salted and your eggs sugared every breakfast if you do any such thing.
If you are so enthusiastic about Bible readings, why not attend yourself?'
'Ah, but I am not a pretty young skivvy,' said Dr Neil wickedly, 'and dear Mr Sands does not wish to see me walk through the door.
What a conquest you have made, McAllister, and think what a triumph it would be for a maid-of-all-work to become a curate's wife.
You could hardly have dreamed of such a possibility.
It beggars belief.'
Sally-Anne could see that the good doctor, as she had naughtily begun to call him, was going to take a great deal of delight in mercilessly teasing her about Mr Sands.
'It would beggar belief even more for me to attend a ladies' sewing circle, or a Bible reading for that matter,'she replied, hardly able to keep amusement out of her voice.
It occurred to her that in any other circumstances the only description for what she and Dr Neil constantly did was flirting.
She looked sideways at him; the ruined cheek was away from her, and this morning he was hardly limping.
His crooked smile was very much in evidence and Matey could have told her that since her arrival Dr Neil had been happier than she had seen him for a long time -- there had been fewer backslidings towards the 'nasty whisky' since McAllister had appeared in his life to provide him with such rich amusement.
To her astonishment, Sally-Anne, who for two dreadful months had hardly been able to bring herself to be near any man, however young and apparently innocuous, wanted to stroke the corner of his mouth -- more, she wanted to smooth the scar away -- or, rather, since that was impossible, to run her hand down it and tell him that it did not matter, such a thing could only disturb those who were themselves already disturbed!
Whatever was coming over her, to think any such thing?
Especially of a man whose extreme scepticism about herself and her doings was daily expressed.
She shook herself inwardly, said severely, Control yourself, Sally-Anne Tunstall; remember what happened when you had such soft thoughts about a man before, and the sudden dreadful memory this evoked hit her so hard that she stopped dead in her tracks, gave a stifled wail, and went so white that Dr Neil, hearing her, and looking at her, saw that her pallor was so extreme that he thought her on the verge of fainting.
'Good God!' he exclaimed.
'What is it, McAllister?'
And he went to take her by the arm, for he feared that she might fall -- only for an even worse horror to grip Sally-Anne, so that she pushed him violently away, quite unable to control herself, stammering, 'No -- no...'
The bright day was ruined, and the peace that the last few weeks had given her shattered.
For one agonised moment she thought that Dr Neil was going to insist, to touch her again, and she could not bear that -- no, no, not that!
Such a strange thought when a moment ago she had fantasised so cheerfully of touching him !
It was a thought that made her breathing grow even shorter, and blackness was on her, and everything disappeared, to reappear a few moments later.
She was now seated on the ground, propped up against the wall of one of the grimy tenements they had been passing; someone had brought water, and Dr Neil was gently wiping her face with his damp handkerchief.
A look of relief passed over his own face when, the dreadful malaise gone, she croaked at him, 'Oh, dear, what happened?'
'You fainted,' said Dr Neil, looking at her keenly.
'Do you,' he said -- and he was suddenly hesitant --'know of any reason why you should?'
He was being very much the doctor, Sally-Anne noted, even through her distress, which was growing less by the minute as she regained the self-control which she had so suddenly and disastrously lost.
She must not let that happen again, she must not, and with Dr Neil looking at her so sharply -- and why was that?-- she was just able to turn her thoughts away from the unthinkable into whose pit she had just fallen.
And then it suddenly struck her what he must be thinking, and she remembered Aunt Nella fainting like this in the early days of her pregnancy with Cousin Lawrence.
Did he think she was hiding in the East End because of that ?
Something between distress and amusement struck her this time.
Of all things she knew that that was not true.
She tried to stand up, only to find that he would not allow her to do so, and Matey was saying, sharply to her, 'Sit child, sit.'
'No,' Sally-Anne cried vehemently.
'It is not what you think, Dr Neil.
No, not at all, no,never .'
And the 'never' came out so wildly that he was surprised again, and, his eyes still watchful, he said, 'I thought not, but you must see, McAllister, I had to ask you -- for your sake.'
Her answer was as cool as she could make it, with her head held high, even though holding it so steadily erect made her feel dizzy.
'I think it must have been coming from the cold of the church into the warmth, or, perhaps --' and she made a wan attempt at humour --'it might have been the thought of attending the ladies'sewing circle which overcame me!'
She could tell that Dr Neil was looking at her most sceptically, although he was touching her so gently that the black fear which she had felt before she had fainted did not return -- and pooh to his suspicions!
He could have no idea at all of what was really wrong with her; Sally-Anne was sure of that.
'Well,' he said finally, helping her to walk with Matey's assistance, 'I have heard of the ladies'sewing circle accused of many things, but never of bringing on a swoon.
Resisting them, rather, is the usual belief.'
He was being kind, she could tell, and not probing further, so she rewarded him with a rather watery smile, and said sententiously, 'There's always a first time for everything.'
And what on earth Papa would say of that banal piece of wisdom she could not imagine -- and why was Papa, whom she had parted from in a high old anger, so much in her thoughts these days?
Dr Neil registered how withdrawn she was, and once they were back home hot, sugary tea was poured down her in a constant stream, and then, despite the warmth of the day, Matey put a shawl around her shoulders and made her sit near the fire.
Normally Sally-Anne would have rejected such molly-coddling, but she did feel dreadfully cold, and said so to Dr Neil who said quietly, 'It is the usual consequence of shock, McAllister, and for some reason you were suddenly very shocked.
You would not care to talk about it with Matey -- if you do not feel you could confide in me?'
Sally-Anne tossed her head at this, and repeated that it was all nothing --'And really you should not make such a fuss over so little'-- but all the same she was happy to let Matey help her up to bed; she felt strangely weak, and the thought of Sunday lunch and washing up, and all the work to be done before the day was over, made her feel worse than ever.
She sat up, cheeks flaming.
'Oh, I do hope they don't think I did it to get out of doing my chores; that would be too bad.'
But neither Dr Neil nor Matey thought that, and when Matey said hesitantly, 'You don't think it's the obvious...?'
Dr Neil shook his head, and said firmly.
'No, she said not, and although I think that virtually everything she has told us about herself is a pack of lies I don't think she's lying about that.'
Sally-Anne did not know of this conversation, of course, and by the next day she was her usual bouncing, energetic self, and it being Monday not only was it wash-day -- oh, dear -- but summer cleaning also began again, with everything being put back, and the whole place was shining and sweet-smelling, so that , Wednesday having come round, McAllister had sunk down into Dr Neil's big armchair with a great, 'Phew, I'm glad that's all over!' amusing Matey, whose soft spot for the girl who seemed to be making Dr Neil more reconciled to his lot grew greater every day.
How it would all end Matey could not imagine.
She was nearly as sceptical as Dr Neil about the reason for McAllister's presence in their life, but she was such a good, hard-working creature, despite her decorative appearance, that Matey could not quarrel with her.
And when, after lunch, she came downstairs in her new outfit, bought from Selfridge's last week with the money which J. D. O'Connor had paid her for her articles, and with her next two articles in her bag, ready to be handed in to the great man himself before she returned to the rectory to pick up Rose Bailey, whose time off did not begin until four-thirty, both Dr Neil and Matey thought that she looked enchanting.
Her frock, which had cost far less than one of her hats in her old life, was a short-sleeved model in navy blue, with a sailor's collar in navy blue and white silk, and had anchors embroidered around the hem of her skirt.
With it she wore an imitation of a sailor's hat -- if one could imagine a sailor's hat with a broad brim and long navy blue streamers -- and she looked extremely desirable in her cheap finery.
Dr Neil felt quite a pang.
No one like himself, a partial cripple, with a badly scarred body -- his limp was paining him this week -- and with a ruined face into the bargain could hope to aspire to such a Bird of Paradise; he could only imagine what she looked like in the clothes she normally wore.
He wondered again what her game was in Vetch Street, and what rich young fellow was waiting for her, back in her real life.
He was sure that there was a rich young fellow -- no one who looked and behaved like McAllister could be without one.
'Will I do?' she asked Matey, in her charming, half-joking manner, for Matey to reply in like vein,
'Delightful, my dear.
Have a good day, and be careful, mind.'
'Oh, I will,' said McAllister gaily; every day she spent in Vetch Street she felt more and more like McAllister, Dr Neil's unconventional parlourmaid, and less and less like Miss Sally-Anne Tunstall, American beauty and heiress.
Miss Tunstall had not possessed work-reddened hands and knees.
And if, at first, she had had to remind herself to be careful not to be found out, deceit was becoming second nature to her.
Yesterday she had gone to the dingy little newsagent at the corner of the street to pay the paper bill, and to buy Matey a writing pad and envelopes, when she had seen on the counter a pile of postcards depicting society beauties.
To her horror the top one was of herself.
It had been taken when she had been presented to King Edward VII and Queen Alexandra earlier in the year.
The presentation had been quite an unnecessary proceeding, as she and Uncle Orrin Tunstall had agreed, seeing that she had met and known both the King and Queen since she was ten years old and had first visited England with her mama and papa.
Her papa, Jared Tunstall, the senior senator for Nevada, was a personal friend of the King, and Sally-Anne had first met him when he was the Prince of Wales.
The Prince had loved children, had been very kind to them, and Sally-Anne had been delighted to see him gravely encourage two four-year-old boys, the twin sons of an old friend, to race pieces of bread, butter and jam up and down his expensively trousered leg.
When Sally-Anne, dolled up in ostrich feathers and pearls, had gravely made her elaborate curtsy to him at Court, he had put out his hand to lift her, and after she had replied, 'My thanks, Your Majesty,' he had twinkled at her, using the famous charm which had somehow made one ignore his years and his obesity, and said,
'Oh, I hope I shall always be Kingy to you, Miss Sally-Anne,' making it clear that he included her in the charmed circle of his friends.
While she was standing in the grubby little shop, this incongruous memory had flashed into her head at the sight of her face.
How the printers had got hold of her photograph she did not know, but they had, and now it was being sold all over London, along with ones of Lillie Langtry and other noted belles.
Beneath the photograph was printed not her name, but a sobriquet, a common practice -- Mrs Langtry was always know as 'The Jersey Lily'.
Sally-Anne had been christened 'An American Princess' soon after her arrival in London, and the name had stuck.
She thought that the photograph neither flattered, nor was very like her, but suppose Rose, or, worse still, Dr Neil or Matey saw it?
Rose had entered the newsagent at that very moment, and Sally-Anne, acting with characteristic speed, had swept up her photograph and one of Mrs Langtry, and, putting the second one on top, she had handed them over to buy them with her hard-earned pennies.
Her friend -- for she and Rose had become quite close -- had sniffed at Mrs Langtry, and said kindly, 'I 'ear as 'ow she's a right fat cow these days,' and fortunately had not asked to look at the other one she had bought.
Sally-Anne had already checked that this was the only one of herself in the pile.
Now, ready to leave for her afternoon and evening's fun she thought it unlikely that she would meet anyone she knew in Fleet Street, or even up West with Rose, and anyway it was a chance which she was prepared to take, for she intended her little excursion to be the basis of yet another article.
Her only regret was that Dr Neil did not buy the Clarion Cry so that so far the only piece of her work which she had seen in print was the one which she had written before she had arrived in the East End.
Dr Neil wanted to tell her to be careful, that she really had little idea of how hard and cruel the world was outside the privileged fastness in which she had previously lived, but he contented himself with saying instead, 'Have a good time, McAllister.
Don't miss the last bus back.'
He was intending to go up West himself that evening, but hardly expected to meet McAllister when he did so, seeing that, for once , he was going out with his disreputable brother Stair, and God knew where they would end up -- nowhere respectable, that was for sure .
So troubled was he for a girl who, after all, was of age, high-spirited and clever, that after she had gone, and Matey had answered reflectively when he had asked her if she knew where McAllister was going, 'I don't know.
With Rose Bailey from the rectory, I believe,' he felt compelled to reply, a little worried,
'With Rose?
Are you sure?'
He had met Rose, recognised what she was, and guessed why she was going up West, and what for, and he was suddenly certain that while McAllister might be a member of a sophisticated society she knew little of the world that Rose Bailey lived in.
'You ought not to have let her go with Rose,' he said reproachfully.
'Whyever not?' said Matey, looking up from her mending, surprised at the urgency in his voice.
What could he say?
Matey's knowledge, too, was circumscribed, and she had no idea of the bold way in which Rose looked at him, and the unmistakable invitation in her eyes.
He did not like to think of McAllister alone with her, but that was stupid -- McAllister was no concern of his, and surely did not need a battered doctor to look out for her.
But all the same he wished that McAllister were not going up West with Rose that night, and wondered what was coming over him that such a fly-trap could fill his waking -- and his sleeping -- thoughts!
Rose was waiting for McAllister at the rectory.
She was wearing a bright red dress, shiny black boots, and despite the heat of the day a black feather boa to match the huge feathers in her large picture hat.
In street clothes she looked very flash, not at all as she did in her neat black and white servant's uniform.
She had run to the door after seeing McAllister walk up the front path, and greeted her with a wide grin, brought on by the sight of McAllister's sailor outfit.
'Cor,' she shrieked, 'yer'll 'ave the mashers after yer tonight, an'no mistake!'
Which surprised Sally-Anne a little, mashers not having figured in her plans.
The two girls, so near in age, but so different in their experience of life, walked briskly through the mean and dirty streets.
On one corner they were hooted after by a group of young men wearing brown bowlers tipped over their eyes, and then they grew particularly bold, following them along the cracked pavement, Rose turned and screamed after them, 'Does yer muvver know yer art?'
A witticism which convulsed herself and the recipients of it.
'Give as good as yer get, is my motter,' she informed McAllister.
'No p'int in takin' up wiv them as 'as no blunt.'
Rose's ma lived with five of her children on the ground floor of one of the tenement buildings about three-quarters of a mile from Vetch Street.
'Ere we are,' said Rose, and pushed McAllister through an unpainted door, along a grimy passage into a smelly room facing an interior courtyard.
The table was already set for high tea, with ham, bread and butter and a very yellow cake.
There was no cow's milk provided -- it went off so quickly in hot weather, with nowhere cool to store it -- and a tin of Nestle's condensed milk stood beside a jam jar full of sugar.
A saucepan full of water was hissing on a tiny range, a miniature of the one in Matey's kitchen.
Rose and her ma exchanged smacking kisses, and 'Ere's McAllister from the doctor's, ma, come to 'ave tea wiv us,' Rose announced.
Ma Bailey wiped her hand on her dirty apron and extended it to Sally-Anne.
'McAllister?' she said.
'Ain't she got annuver nime?'
'Sally-Anne, please,' said the Tunstall heiress.
'McAllister reminds me of scrubbing floors and cleaning ranges.'
This was not strictly true.
McAllister, she felt, was reserved for Dr Neil.
'McAl -- Sally-Anne,' said Rose, reading her ma's stare correctly, 'is a Yankee; that's why she talks so funny.
Don't mind that.
She's a bit of all right.'
One thing, thought Sally-Anne, looking around her, after Ma Bailey had set her down in one of the few chairs the room possessed, was that no one here had ever done any summer cleaning, nor spring, autumn nor winter cleaning either.
There was a bed in one corner still unmade, although it was five o'clock in the afternoon.
Three or four dirty children arrived for tea; she and Rose were the only clean persons in the kitchen -- for so Sally-Anne supposed the room -- and while Ma, Rose and Sally-Anne sat down at table the children all stood, and wolfed the food down mannerlessly, fighting among themselves while they did so.
The ham, which everyone ate avidly, and which Sally-Anne forced down, was off and the butter was so rancid that she almost choked over it.
As for the tea!
Matey had already told her that most of the tea sold in East End grocers was collected from the leavings of the West End, dried, repacked and sold again.
Drinking her tea at Rose's ma's, Sally-Anne could well believe it.
Afterwards she helped to clear the table and to carry the pots through into a lean-to kitchen, the room in which they had eaten being the best parlour.
'Don't bovver wiv 'em,' said Rose, but,
'At least let me wash them up,' replied Sally-Anne determinedly, and, Rose giving way, she discovered that the water had to be fetched from a tap in the inside court, there being none piped into any of the surrounding buildings, and the privy was a common one, also standing in the yard.
Outside, running the water into the misshapen saucepan, Sally-Anne found the stench indescribable, so much so that she began a desperate burst of coughing to prevent herself from vomiting.
She was about to commiserate with Rose on her unfortunate circumstances when Rose proudly said, 'We're ever so lucky to be'ere Sally-Anne, wiv the 'ole grarnd floor to ourselves -- not like them upstairs what only 'ave one room.'
'Them upstairs' were obviously of an inferior race, and their habits were disgusting, Rose said.
Sally-Anne also made the mistake of assuming that Rose's ma was a widow, only for Rose to stare at her, burst out laughing and say, 'Ow, there weren't no Mr Bailey, and me dad 'opped it long ago and 'is name weren't Bailey; he was only 'im as lived wiv 'er.'
''Er' was apparently Rose's ma, and it seemed likely from Rose's hints and sniffs that all the children had different dads, which, looking at Ma Bailey, fat, defeated and unlovely, Sally-Anne found difficult to believe.
How could any man want to lay a finger on her, never mind father her children?
The youngest of whom, she discovered, was only six weeks old, and was brought from the bedroom in an elderly bassinet to be fed, not mother's milk, but some patent milk powder made up with the dubious water from the outside tap.
It was one thing for Miss Sally-Anne Tunstall, pampered beauty, to contemplate the horrors of poverty from the relative comfort of Vetch Street, but to see it in practice, that was quite another thing.
All the loose women Sally-Anne had previously met in both English and American society had been particularly lovely and tempting creatures, not gross and depressed objects who looked like unmade beds -- as Rose's ma did.
Just before six o'clock they finally left for the West with Rose's ma's warnings ringing in their ears.
'You be careful, mind, our Rose, and you too, miss.'
Rose's ma, even more than her daughter, had appreciated what a strange creature had arrived in their meagre lives.
How fortunate I am, thought Sally-Anne, running with Rose to catch the horse bus.
How can people live like that?
And are there those, as Rose suggests, who live even more straitened lives?
She shuddered -- for if there were she really must find them out and write of them, or her articles could not be considered authentic.
She did not reflect that only a few weeks ago she would have thought it impossible that she could have lived the life of a servant, let alone bear eating and drinking in the den in which Rose and her ma lived, but she did know one thing -- she would never take her comfortable life for granted again.
And what she did not know was that she had even more lessons to learn -- and that some of them might not be pleasant.
CHAPTER SEVEN
Neil Cochrane was not enjoying himself.
Dressed as he rarely was these days, in evening togs and carrying an opera hat, a white silk scarf around his neck, he was, at ten o'clock at night, completely sober, although the group he was with, headed by his brother Stair, and Havelock, usually known as Havvie, Marquess of Blaine, heir of the Duke of Innescourt, whose stag party night this was, were all three-parts drunk.
Stair, who usually ignored Neil since his brother had given up his privileged life and become an East End doctor, occasionally felt that he had some sort of duty to him; he was, after all, his heir.
His idea of duty took the form of inviting Neil to join him in the nightly round of enjoyment which made up Stair's life.
Neil refused most of these invitations, but had accepted this last one -- the temptation to indulge himself, since he had stopped his midnight drinking, was suddenly great.
But the sight of their debauchery stifled his own, and the knowledge that they were going to end the evening at Madam Rachel's select brothel, just off the Haymarket, far from exciting him, disgusted him.
Havvie Blaine disgusted him most of all.
He was marrying one of America's richest heiresses, who was, Stair had drunkenly informed him, 'As ugly as a pug dog,' solely for her money, and was proposing to bed Madame Rachel herself as a preliminary to his marriage at St Margaret's, Westminster, the following morning, the biggest society wedding of the year.
For very form's sake, and because, after all, Stair was his brother, he had stayed with the party through one act of a musical comedy at the Gaiety Theatre where they had made so much noise that their departure at the first interval must have pleased the audience which they had left behind, had gone to Quaggers -- Quaglino's -- to dine -- which meant drink -- in a private room, and were now on their way to crown their evening's pleasure by 'Pushing the boat out for Havvie', Stair's witticism.
For whatever reason, Neil, who had almost looked forward to relieving his body's urgent demands -- he seemed to have been continent for years, and McAllister's arrival in Vetch Street had most inconveniently revived his dormant senses -- found that he was disgusted with himself most of all.
Stair thrust an arm around him, laid his head on Neil's shoulder, and said, 'Let me treat you to a good 'un at Rachel's, Neil,' which completed the destruction of any desire Neil might have had to treat himself.
How he was to avoid patronising Madame Rachel's girls without offending Stair he could not think.
They were by now in Piccadilly Circus, which was as bright as day, and were surrounded by the crowds streaming from the theatres, cafes and dives which populated the area, painted ladies of a certain character being prominent among them -- as well as the enthusiastic amateurs who had come up from the East End to make a few pennies, or even be given supper, as a price for their favours.
It was only because it was so rare that Stair ever troubled with him at all these days that Neil felt compelled to go along with him, willy-nilly.
He had a brief memory of McAllister's bright face, earlier that day, and the thought of enjoying himself with one of Madame Rachel's girls did not attract.
Salvation came suddenly.
Another crowd of gilded aristocrats and gentlemen, also out on the town and also half cut, recognising Stair's party, and Havvie Blaine in particular, decided to enliven his celebrations by staging an impromptu rugger game.
Forming a scrum, they attacked Stair's party, and one young fellow who had brought along a hunting horn blew blasts on it to encourage his fellows.
The whole group, locked together, whirled around.
Stair, turning to see what was happening, was borne away from Neil, who was himself carried down the Haymarket by a shouting throng -- the rugger match had triggered off an impromptu riot as well.
It's an ill wind that blows nobody any good, thought Neil platitudinously, as a large bruiser pushed him against one of the pillars of the Haymarket Theatre -- I might get my nose broken, but at least my virtue looks like remaining intact!
He dodged away and took the opportunity to walk rapidly in the opposite direction from the one in which Stair's lot might be, towards Pall Mall, passing the narrow courts and alleys which led from the brilliantly lit main street.
Halfway down the Haymarket, now quite a distance from Piccadilly Circus, Neil paused, breathless, and afterwards wondered what kind god had arranged for him to be where he was.
To his left from the dark street came a frightened female voice, shouting in a despairing manner, 'No, please, no!'
Neil, who never thought of himself as in any way a Galahad, or a knight errant in shining armour, saving virgins under attack, nevertheless felt compelled to investigate, particularly when the cry came again, more despairing than ever.
He ran the short distance down the side-street as fast as his gammy leg would let him.
He saw that a tall young man, dressed in the uniform of a lower-class masher -- check trousers, brown bowler, shiny brown boots, and huge moustache -- had a girl pinned against the wall of one of the office buildings, and was attempting to kiss her -- or worse.
The girl was trying to fight him off, was kicking his shins vigorously, trying to hold her face away from his, and was giving vent to muffled shouts, which were, of necessity, growing feebler and feebler as her strength ran out.
It was quite apparent that her resistance was not token, but real.
The chivalry towards women which gentleman were supposed to possess -- but often didn't -- was a very real part of Neil's nature, and, having seen what was happening, he could not stand by, or walk away, telling himself that it was no business of his.
'Let her go,' he said, hoping that the masher would not see what a useless cripple he was.
He was carrying Stair's silver-topped cane, given to him to look after because in his drunken state Stair could no longer remember to keep it with him.
The masher turned his attention from the girl to Neil and moved towards him menacingly.
Neil responded by lifting the cane high, ready to defend himself from a man who he saw was bigger, and probably stronger, than himself.
The masher dodged round Neil, lifted his fists, and said in what Neil could tell was by no means a confident manner, 'Throw that away, and put up yer dukes like a man!'
The only sensible response to this, Neil decided ruefully, was to keep hold of the cane, and continue to threaten the bully with it, and for a very short space of time they circled around one another-- neither man, it soon became plain, really wishing to do more than threaten.
The masher's intended victim was not quite so indecisive; far from taking the opportunity to run off, leaving St George to fight the dragon on his own, she had taken the opportunity once her attacker's attention had been diverted from her to remove her right shoe, and she now proceeded to attack him from the rear, pounding him first about the head and shoulders with the shoe, and then with her handbag shrieking, 'Take that, you cowardly bully, and that,' leaving him open to any attack Neil might care to make.
Assailed from both sides, the masher decided that the game was not worth the candle, and, dodging by Neil, to the latter's great relief, he scuttled off towards the Haymarket.
Neil dropped his cane, and turned towards the victim, now transformed into her attacker's Nemesis.
Her hair down and around her shoulders, her dress torn, her right shoe in one hand, and her bag in the other, the light of battle in her eye, McAllister stared at her saviour, Dr Neil, and he stared back at her.
'Good God, you idiot,' he said romantically, 'what on earth are you doing here, and where is Rose?
Have you taken leave of your senses to end up being attacked in a back alley off the Haymarket of all places, by a man you don't know?'
'Oh,' said McAllister furiously, remembering something which she did not like to remember, 'you mean it would have been perfectly in order for him to attack me if he'd known me?'
All the anger she felt for the wretch who had assaulted her was transferred to Dr Neil who had arrived to save her.
'I might have expected such an answer from you, McAllister; it fits in with the general picture,' said Dr Neil angrily, picking up his cane.
'I suppose that you will be gracious  enough first to tell me how you arrived here, and secondly to allow me to take you home before you end up in another alley.
Thanks I don't expect.'
Enormous relief at being saved, and delayed shock, hit Sally-Anne together.
For one dreadful moment she had thought that God was going to play yet another ghastly trick on her, and one which she had not invited -- any more than she had invited her previous one.
'Oh,' she almost wailed, 'I am being so thoughtless.
Not to thank you, I mean -- and Rose -- oh, dear...' and her face crumpled, but the steely resolution which lay beneath her outward charm prevented her from giving way to the hysterics which she knew would take her over if she allowed herself to show any weakness.
Compunction struck Dr Neil.
After all, the poor child was probably still in shock compounded of shame at being so discovered, and relief at her salvation.
'Let me help you to put your shoe back on,' he said gently, 'and I must thank you for coming to my rescue.
The princess certainly helped her saviour tonight.'
Princess!
The irony of the description struck Sally-Anne hard.
She allowed Dr Neil to sit her down on a doorstep and help her to replace her shoe, and he guarded her while she tidied her hair, and pulled her torn dress together.
'I didn't ask for this,' she said.
'I...think I had better tell you what happened; I wouldn't like you to entertain any wrong thoughts.'
She looked up at him, the enormous blue-black eyes wide and shadowed by her recent experience.
She was wise enough to understand, now that her fear was leaving her, that it was Dr Neil's very real concern which had made him so sharp with her when he had found out who it was he had rescued.
'As for Rose...' she shivered, and Dr Neil, seeing her face change, said abruptly,
'When did you last eat, McAllister?'
Sally-Anne stared at him, surprised by this strange request.
'At Rose's ma's,' she answered.
'We had tea there, before we came up West.'
'I thought so,' said Dr Neil kindly.
'You are suffering from both shock -- and, knowing your appetite, from hunger as well.'
His eyes twinkled at her as he added, 'Would you allow me to buy you supper before I take you home?'
'Oh, yes, please,' said Sally-Anne fervently, realising that yes, hunger was succeeding shock, and the thought of food and drink was a pleasant one.
By now they had walked some distance along the Haymarket, and were on the edge of Piccadilly Circus where Dr Neil was relieved to see that Stair's party had disappeared and the crowds there had grown less.
He turned to Sally-Anne who was clinging to his arm, and said, 'I know a small chop-house near here, and when you have eaten you may tell me what happened, and how you came to be on your own, under attack.'
He hesitated.
'I blame myself a little.
You know nothing of the world in which you have come to live, and I should have warned you about Rose, only I did not know that you were going out with her until after you had gone, and then, of course, it was too late.'
Both of them realised that he had abandoned all pretence that Sally-Anne was an ordinary young woman come to work in Vetch Street, but neither of them pursued the matter, Dr Neil from delicacy, and Sally-Anne because she could not tell him the real truth about herself -- he would undoubtedly immediately send her back to the embassy, and she did not want that at all-- it would be failure.
Only when they were sitting down in the chop-house and Dr Neil had ordered them soup and rolls, followed by lamb chops with seasonal vegetables, and a glass of red wine each --'Good for your shattered nerves,' he said gravely -- and they were waiting for the soup to arrive did Sally-Anne have time to look about her.
She had never been in such a place before, and she saw at once that most of the men had girls with them who were certainly not their wives, or the kind of girl one took home to mother.
The whole day had been a strange one for a young lady like herself who had never been allowed to go out on her own, had been carefully looked after and protected at all times from the impact of the world in which most people lived.
The soup arrived, a rich broth, and rolls and butter and a carafe of wine as well.
Sally-Anne attacked her food with vigour, and when she had drunk her soup, buttered and eaten her roll, Dr Neil prompted her gently, 'Talk to me about it, McAllister.
It will make you feel better.'
'Oh,' said Sally-Anne, wiping her mouth with her napkin, 'it was all such a surprise, the evening started out so well.
'We went to the theatre where Marie Lloyd was singing.
I had never been to such a show before -- not at all the thing that well-bred young ladies are supposed to attend -- but she made me laugh...
We went upstairs -- Rose called it the gods, because we were so near to heaven, I suppose, and the stage was so far away.'
Her expression was wistful, reminiscent, as she recalled how, once they had gone up endless stairs and were seated on the hard wooden benches, an attendant had come along and pushed them all closer and closer together so that as many spectators as possible could be packed in.
At first she had disliked this, but Rose's amusement, and the joking of those around them, had infected her, too, and she had found herself giggling and laughing with the rest as they were shuffled along.
'And then,' she said, 'after the show was over, and we were outside again, Rose said, 'What about a bit of supper?'
And, oh, how stupid I was.'
She remembered how she had looked at Rose in some surprise, and said, 'Yes, that would be swell, but I haven't enough money with me to pay for supper.'
Rose had put her finger by her nose with a look of infinite cunning, and replied, 'Nah, Sally-Anne, yer ain't as green as that, surely.
We don't need to pay.
Watch.'
By now they were in Piccadilly Circus, anonymous among the largest crowd Sally-Anne had ever been in without some sort of companion to protect her, or to look out for her.
A rich young girl of good family was seldom, if ever, on her own.
Rose had moved back a little to stand in front of one of the shop windows, scanning the crowd with knowing eyes.
Two young men, superior artisans probably, out for a night on the town, dressed in an imitation of their betters, their brown bowlers tipped over their foreheads, their eyes searching the crowd for likely partners for a bit of fun, had attracted Rose's attention, and she had made sure that she attracted theirs.
Sally-Anne never knew precisely how she did it, but the next moment the shorter one was speaking to Rose, and the taller one had his wandering eye on a shrinking Sally-Anne, who was too confused to do more than understand that the Mashers were agreeing to buy supper for Rose and herself -- in exchange for what?
But she was suddenly sure that she knew!
'No,' she said, 'I want to go home,'only for Rose to ignore her plea, and say,
'Come on Sally-Anne; never say die!'
And before Sally-Anne could do anything about it the tall young man was tucking her arm in his, and walking her briskly along at such a pace that she was horrified to discover that Rose and her companion were gone, lost behind them as fresh crowds emerged from yet more emptying theatres and dance halls.
Panic-stricken, she tried to pull her arm out of his, but her companion ignored her struggles, saying, 'Come on, ducks, no need to bovver ourselves with supper; let's 'ave our bit of fun first,' and pulled her along into the alley where Dr Neil had found her.
Their lamb chops had arrived as she finished telling him this sad story.
'I had no idea,' she informed him earnestly, 'no idea at all what Rose had in mind, or I would never have gone with her.
She was going to sell herself -- she has sold herself by now-- and would have sold me -- for supper.
Just for supper!'
Dr Neil decided that McAllister's education in life needed adding to a little.
'Ah, but it isn't just for supper for Rose, is it?
Rose could never afford for herself the sort of supper her young man will provide for her, and she's fly enough not to get herself dragged into an alley and done out of it.
She probably managed to make him pay her a little more before she agreed to do what he wanted -- and altogether earned for herself in one evening more than she does in several days of being a parlourmaid.
'Your partner probably recognised you as the innocent you are -- and tried to take advantage of the fact.
Eat up your lamb chops, McAllister, before they grow cold, there's a good girl.
I don't like to waste good money, or good food either -- and I promise to respect your virtue when you've finished.'
Oh, it was good food, Sally-Anne thought gratefully, tucking into it, beginning to enjoy herself now that her ordeal was over, and was happy to sit there with Dr Neil who was also enjoying his supper, by his expression.
He poured her another glass of wine from the carafe which stood between them on the table.
She thought that he looked nicer than ever in his evening dress, and sounded nicer now that he had dropped his brusque, cold manner with her again.
'You didn't recognise what Rose was?' he asked her.
She shook her head vigorously.
'No.
I was foolish, I suppose.'
It was dreadful to her to think that Rose was an amateur tart earning such a pittance in her legitimate work that she was ready to sell herself for so little.
She should have realised the truth when she had met her mother and seen her wretched home.
Another awful thought struck her.
How many of the young men and women of her own world were doing exactly the same as Rose?
And the fact that they were doing it for a great deal more money, like Havvie Blaine, rather than for supper and a few pence, didn't make it any better.
She had finished her lamb chops, and they were really top-hole, or she was so hungry that anything would have seemed like ambrosia, the food of the gods.
She looked across at Dr Neil; the overhead light was making a golden aureole around his sandy head, and she thought again that even with his dreadful scar he was an extremely attractive man.
He intercepted her candid gaze, said gravely, 'Some apple pie, McAllister?
Surely that monstrous appetite isn't yet satisfied?'
'Oh, yes, please!' returned Sally-Anne, and then caught her breath again, when he turned his head, presenting her with his beautiful, undamaged profile -- he must have been an absolute stunner before he was disfigured.
Not only did he order the apple pie and cream, but he also poured her yet another glass of wine -- somehow she seemed to have drunk the other already!
'Should I really?' she said, looking at the wine glass doubtfully, speaking as much to dispel the strange feeling that was coming over her every time she looked at him as to reassure herself over the wine.
It was a feeling which she had never had before, not even with Terry or Havvie, both of whom she had thought herself in love with.
This feeling was quite different; in some strange way it seemed to reach into her very soul.
She wanted to protect him, and for him to protect her -- it was as though her identity was bound up with his.
And yet it was physical as well.
She wanted to lean over and touch him, and, fearful though she was of men, and she shrank at the thought, she wanted him to touch her -- gently, of course.
Surely she could not be stupid enough to be falling in love, be bowled over by a man's charm again, and so strongly this time?
The mere idea made her feel faintly sick, as well as excited, but not sick enough to refuse the apple pie and cream when it came.
And Dr Neil, staring at her wine glass as though it were a bomb, said, 'Trust me McAllister; alcohol will do you nothing but good after your adventures tonight -- you will sleep well, and I have already given you my word that I will behave myself.'
So he had, and Sally-Anne knew that, above all, Dr Neil was a man of his word.
She ate her apple pie, which was excellent, and finished off her glass of wine, only for Dr Neil to look at her quizzically and say, 'Cheese and biscuits, McAllister?
Surely you will not refuse cheese and biscuits?'
'Oh, I must,' she said gaily, but she accepted the offer of coffee, and, waiting for it to arrive, said, 'Now it is my turn to quiz you.
What is Dr Neil doing in the West End, dressed up like a toff?'
A word which she had learned from Rose.
Dr Neil leaned back in his chair and appreciated his maid's pretty face.
She was quite the most attractive thing in the cafe now that the colour had come back to her face and she had regained her normal cheerful manner.
More than one man had looked enviously at him.
'Oh, my brother Stair -- I am sure that Matey has told you about Stair -- invited me out tonight to the stag party of one of his friends; I used to know him slightly.
He is marrying a rich American heiress tomorrow and is spending his last night on the town as a free man.
I am sorry for the poor girl if her bridegroom chooses to behave as he did tonight, and, knowing him well, I cannot imagine that he will be any better behaved when he is married to her than he was before.'
'A rich American heiress,' said Sally-Anne thoughtfully.
'I wonder if I have heard of her?'
'Perhaps,' said Neil.
'Her name is Maybelle Foy -- Stair says that she is not very pretty, but is enormously rich.'
For once, Dr Neil, engrossed in drinking his coffee, did not notice that McAllister had gone very still.
She picked up her cup, and said in a voice as cool as she could make it, 'Yes, I have heard of Miss Foy, and what you say of her is true.
What lucky man is marrying her?'
She thought that she knew who the lucky man might be, and it took all her strength of mind not to betray the dreadful emotions which merely thinking of him aroused in her.
'Havvie Blaine, old Innescourt's heir.
That family is as poor as church mice, and Stair told me that Havvie had already managed to snare one Yankee heiress, but she got away, and now he has caught another -- not quite so rich, but rich enough.
They will be restoring Dunblaine Castle -- unless Havvie spends the lot on loose women, drink and gambling -- begging your pardon for being frank, McAllister.
The Havvie Blaines of this world ought to wear a sign warning decent women off -- and he's as handsome as the devil, which doesn't help.'
So he ought, thought Sally-Anne, and so he is, and, desperate to change the conversation, to steer it away from dangerous ground, she said, as brightly as she could, through numb lips, 'Shan't we miss the last horse bus home if we don't leave soon?' thinking how fortunate it was that Stair had not told Dr Neil the other heiress's identity -- Sally-Anne Tunstall might have been a dead give-away; she really ought to have changed her Christian name.
Dr Neil, who had just called for the bill, pulled out his watch, and said cheerfully, 'Oh, we have already missed it, McAllister.
If you can trust yourself to me in a cab we can travel back to Vetch Street in style, and if Matey is up try to think of an explanation that will not involve Rose, or men who trap young girls in alleys, or how I, fortunately for you, became separated from Stair and rescued you.
We don't want her thinking that you and I made a secret assignation in Piccadilly Circus.
No, by good fortune we met, after you were accidentally parted from Rose, and I saw you chastely home.
Oh, and by the by, I want you to promise me that you will not go out on the rampage with Rose again.'
'No, indeed,' said Sally-Anne with a shudder.
'I can think of nothing that I would like less.'
She took the arm he offered her, and they left the cafe together.
Somehow Dr Neil's touch did not seem to affect her as badly as that of most men, even though in the cab home sitting so near to him nearly brought on the kind of faintness which she had felt on the walk home from church.
She managed to control herself and fortunately Dr Neil was as quiet as she was, and when they reached Vetch Street Matey had already gone to bed.
Dr Neil lit the gas-lamp in the tiny hall and the one in the parlour and fetched their candles for them.
'Good night, McAllister,' he said.
'If your evening began well and had a bad middle, I hope you enjoyed the end.'
As once before he leaned over and gave her cheek a gentle, passionless kiss.
It was almost as though he recognised that she would not -- no, could not -- accept more.
Sally-Anne took the kiss and the last sight of him standing in the parlour in his beautiful evening clothes up to bed with her, and agreed with him -- for really the end of the evening and their happy supper together had been the best thing of all.
CHAPTER EIGHT
'Surprisingly I enjoyed going to the ladies' sewing circle, with Matey,'wrote Sally-Anne in her journal.
It was five-thirty in the morning and she was already dressed for the day.
She had recently found that she was waking with the dawn -- or even slightly earlier -- and she used the opportunity either to keep up with her journal or to rough out a column for J. D. O'Connor who grew increasingly pleased with her work.
She closed her book; her papa had given it to her before she left for England and she blushed a little on remembering how short she had been with him when he had handed it over to her.
Somehow, since being at Vetch Street, working so hard, and living a life so different from her old one, many events in her past had taken on a different colour.
She rose from her chair at the little antique bureau which stood in the attic window space, bent down and fumbled for the catch of the secret drawer which she had found there.
She had recorded in her journal everything which had happened to her since she had arrived in London and she wanted no prying eyes to read what she had written.
Not that she thought Matey, or Dr Neil, would come poking around her room, but she was changing rapidly from the trusting and innocent girl she had once been to someone more cautious -- she, Sally-Anne, more cautious!
How pleased Mama would be!-- and she no longer took things for granted as she once had.
Six o'clock found her downstairs at the kitchen grate, readying the little house for the day.
She could hear stirrings upstairs and assumed that Matey, who often rose early when she could not sleep late, was on her way down to start her chores alongside McAllister.
Matey might not work so hard as McAllister did, or perform such menial tasks, like scrubbing the kitchen floor and whitening the front doorstep -- which was next on McAllister's list of duties -- but she did her share and was never idle, even in her spare time -- hence her membership of the ladies' sewing circle and her encouragement of her housemaid, McAllister, to accompany her to it.
Only, it wasn't Matey who walked into the kitchen as McAllister began her blackleading, but Dr Neil, yawning and fastening the buttons of a rather natty grey coat which she had never seen before.
'Good morning, McAllister,' he said; his manners to all those society considered his inferiors were always punctilious.
'Would you leave off what you are doing for a moment?
The Master would like a cup of coffee.'
McAllister put down her brush, placed her hands on her hips in the manner of every recalcitrant servant and said aggressively, with the hint of a twinkle in her eye, 'And what would Miss Mates say if I neglect my early morning duties?'
Neil sat down at the table and placed his newspaper on its gaudy oilcloth cover.
'Never mind what Matey might say --' and he was almost short with her '-- I'll square her for you.
It's the Master speaking, McAllister.
Jump to it!'
And he picked the paper up, not at first to read it, preferring to look over its top at a bustling McAllister instead.
She was so neat and quick in all her movements, and Matey had made her careful.
Sally-Anne had never before understood the necessity to be absolutely precise in everything she did, and Matey's training, designed to make her a good maid, was beginning to affect her habits in every other part of her life.
She found that she was bringing more and more discipline to her writing for the Clarion , and thinking of this she noticed that it was the Clarion which Dr Neil was reading, or pretending to read, for her new awareness told her that, too.
Finally the coffee was steaming in its pot, as hot as good coffee should be; the cream had been fetched from its place on the cold stone slab in the pantry and poured into its jug, sugar cubes had been decanted into a tiny flowered porcelain bowl which matched the pot and the jug, together with a matching cup and saucer.
Lastly a small plate of Nice biscuits was added, and the whole was arranged on a round tray which McAllister had polished the day before with the rest of the silver, beneath Matey's stern eye.
McAllister placed the tray on the table before Dr Neil and performed the parlourmaid's traditional bob after she had done so.
He put down the paper, said, 'Thank you, McAllister, most commendable,' and then added, 'And now you may fetch another cup and saucer and drink with me.'
'No,' said McAllister, for Sally-Anne was trying hard to be a real maid, not a pretend one.
'That would not be proper.'
Dr Neil banged his fist quite gently on the table, and said in his most severe voice, 'McAllister!
The Master commands.
Jump, girl, jump!'
He had no idea why he was behaving in such a flighty way.
After rescuing McAllister just over a fortnight ago, he had told himself firmly, This must stop! and himself had refused to listen.
After Angela Deverill's cruel treatment of him he had vowed never to have anything to do with women again -- other than to slake certain demanding bodily appetites -- and then even the appetites had disappeared until McAllister had revived them.
He looked covertly across the table at her, at the vital face, the dark eyes, the generous, amused mouth, the blue-black hair loosely knotted on to the crown of her head, so that curls and tendrils clung around the perfect oval of her face.
No, it really would not do.
She was his servant, whatever she had been in her previous life, and she was ten or more years younger than himself, and what a battered old soldier he was, to be hungering after such a fresh young girl.
More, she was passionate and wilful, exactly the sort of creature a man badly damaged by life and love ought to avoid -- only, he could not avoid her.
He and Matey had employed servants before, even quite attractive ones, and he had felt no temptation at all, they had left him cold, but McAllister...
McAllister sat herself down opposite him and poured out two cups of coffee -- he had refused to drink his own until she had joined him.
McAllister could not prevent her eyes from straying to his newspaper.
She had several almost overwhelming desires.
One was to see her own column -- for so J.D. called it -- in print, another was to know whether or not Dr Neil had read it, and still another was to know what he thought of it -- and none of these desires seemed likely to be satisfied.
She took one of the biscuits; they were Nice biscuits in the pretty floral tin today, quite her favourite; she preferred them to Marie.
'You make good coffee, McAllister,' Dr Neil said; he had noticed her eyes stray to his newspaper, now neglected on the table.
'Thank you, sir,' said McAllister primly, 'Americans usually do.
English coffee invariably tastes of dishwater.'
And she added, without thinking, 'Even in the best houses.'
'But this is not a best house, is it?' said Dr Neil with a grin.
'And in the best houses master and servant do not sit drinking coffee, or anything else for that matter, together in the kitchen.
And I suspect that the best houses would not allow the Clarion Cry inside their sacred boundaries.'
'I am a little surprised, Dr Neil,' said McAllister prim again, but managing to sneak another biscuit -- she really did feel hungry and breakfast would not be served until Matey came down at about seven-thirty --'that you have seen fit to buy it.'
'I have my duties,' replied Dr Neil, assuming a pompous voice, 'and one of them is to know what is happening in the world.
For that,The Times is not enough.
I am informed that the Clarion Cry is the thing to read if one wishes to be in the know.'
'Then perhaps I ought to read it,' was McAllister's somewhat impertinent response to this statement.
'Not at all.
It contains matter not fit for a delicate young lady such as yourself.'
'But I am a servant now,' said McAllister incontrovertibly, 'not a delicate young lady.
And, as you frequently remind me, I do not know enough of the wicked world in which I live -- so one might consider the Clarion Cry a useful means by which to educate me.'
'Touche again, you disrespectful minx,' said Dr Neil.
'Why do I ever argue with you, I ask myself, McAllister?
I never seem to win.
Did that remarkable papa of yours instruct you in logic-chopping, as well as in how to play a wicked game of chess?'
'Oh, Mama had a nice line in that, as well as Papa,' said McAllister, remembering for once to use the past tense when speaking of her parents.
She eyed the biscuit plate and the Clarion Cry simultaneously -- surely he would not mind if she took another biscuit?
'You are, I see, determined to be naughty,' announced Dr Neil.
'For not only do you wish to read this disreputable rag, but, according to your wolfish expression, you are yearning for another biscuit.
Go on!
Eat the whole plateful!
Why should I stop you?
And if Matey complains about their rapid disappearance you may tell her that I have developed a passion for them!'
'Oh, thank you,' said McAllister, seizing the plate before he could change his mind.
'All these chores are making me so hungry that it is, as I said before, only fair that you feed me properly in order to keep me going.'
'The problem is not keeping you going, McAllister, but getting you to stop.
Such energy!'
He picked up the paper.
'Shall I read you something decorous while you gorge yourself?
If I can find anything, that is.
Ah, I see that there is a column written by a lady, if we can trust the name above it, which is Vesta.
I doubt me that many ladies write for old J.D. -- a dubious proposition, that.'
He read in silence for a moment, then said, with a laugh, 'I see that you might enjoy this, although I am not sure your mama might not wish you to read it.
It is all about those poor girls who behave as Rose did the other night, and suggests that a decent wage might prevent them from indulging in such conduct.'
He looked mischievously at her over the top of his paper.
'Yes, McAllister, I will continue to further your education.
You may read it.'
And he handed the newspaper to her across the table.
It took McAllister all her will-power not to wrench it from his grasp and to see and devour her work in one go.
Somehow she managed to keep her self-control, opened it at the right page and began reading it slowly, as though she had never seen it before, although she knew every word by heart -- because it was written from the heart.
She had just read the last paragraph when the kitchen door opened and Matey entered, far earlier than she normally did, to stare at the unlikely scene before her: Dr Neil characteristically leaning back in his wooden Windsor armchair, McAllister sitting on her stool, reading the paper, the used coffee-cups and the empty biscuit plate sitting between them.
'Well, well,' she said reprovingly, looking at the unfinished grate and thinking of all the other duties which McAllister should have performed by now.
'You are up early, Dr Neil, and are keeping McAllister from her work.'
'In a way, Matey, in a way,' said Dr Neil, lazily catching at her hand and pressing it as she passed him.
'McAllister's first duty is always to the Master, and so I commanded her to make me a pot of coffee.
You must not reprimand her for doing as she was told.'
'And I suppose that you will inform me that you instructed her to drink a cup with you, as well,' said Matey drily.
'And I can only say that the Master will not like it if his breakfast and lunch do not arrive on time, because he serves his maid coffee, engages her in chit-chat, and encourages her to read disreputable papers when she ought to be blackleading the range, or scrubbing the kitchen floor.'
McAllister leapt to her feet.
'I will start work at once,' she said, casting a wistful eye on the Clarion Cry which she was leaving behind.
Fortunately she had managed to read quickly through her own piece, which looked really professional, but she wondered what J.D. was thundering about on the editorial page.
And then there were the letters.
J.D. had told her on her last visit, when she had handed in the column she had just read, that there had been a large number of letters about Vesta's contribution and he would be publishing some of them in the next issue.
She finished the blackleading -- a truly nasty job which she was glad to be done with early in the day.
Dr Neil had left the kitchen, to finish his toilet, presumably.
Matey was bustling about, doing some of the chores which McAllister had neglected while drinking coffee with him, and on opening the biscuit tin she complained bitterly that Dr Neil would eat them out of house and home if he continued to run through biscuits at his present rate -- a judgement which amused McAllister, but also made her feel vaguely guilty.
Later that day, in the early afternoon, Matey was instructing McAllister in how to knit -- something which had been left out of her previous education.
Matey watched her, dark head bent over her work, hands busy, having mastered the basics of the task with what Matey thought was surprising speed.
She was not sure what she felt about the rapport which was springing up between her treasure and his now not so new maid.
On the one hand he seemed much improved since her advent.
When Matey had ventured a favourable comment about his midnight drinking having stopped, he had simply said, 'You must thank McAllister for that,' and so her influence seemed to be a good one.
On the other hand, however, who and what was she?
The mystery of her origins occupied both Matey and Dr Neil in quite different ways.
Almost as though she knew that Matey was thinking about her, McAllister looked up and said, 'I would never have thought that I would enjoy knitting so much, and the ladies' sewing circle, too.'
'So you will go again next week?'
Matey asked.
McAllister nodded vigorously.
'Oh, yes, I want to help to make things for the big bazaar Mr Sands was talking about, to help those who live here in such misery.'
'Yes,' said Matey encouragingly, 'we hold it every year, and we usually make quite a large sum to give to the various charities connected with the Church.'
'I was thinking...' said McAllister slowly.
'When Miss Purdue showed us those baby clothes -- the little dresses and nighties for the children of the poor fallen girls in the home on Callendar Street -- they seemed so stark and plain.
Would the circle mind if I embroidered some rosebuds on their bodices after the other ladies had made them up?
And perhaps she might agree to allow the circle to make some fine ones in lawn to sell at the bazaar and I could do some white Swiss eyelet embroidery on them.
Mama taught me how to do that when I was quite small.
But I think that the poor girls might like something pretty, too.'
Her face was so earnest that Matey felt a lump in her throat.
She thought that McAllister had never seen misery before, and that her response to it was a warm and loving one -- and practical, too.
Perhaps the fallen girls might behave better if they were not constantly reminded how different they were.
She would support McAllister's idea to prettify the baby clothes; it might make the poor things feel happier and that someone really cared about them.
'No reason why I shouldn't suggest it,' she said briskly.
'But you would have to keep to your promise to do it, you know.
No backsliding.'
'Oh, I always keep my word,' said McAllister, equally brisk.
'Mama and Papa were always most particular about that.'
Matey, like Dr Neil, thought that McAllister might have been spoiled and pampered in her old life, but she had certainly been exposed to many good principles of conduct, even if she was headstrong.
'You can do an errand for me this afternoon, McAllister,' she said.
'One that I think that you would like.
The housekeeper at the rectory said that they have a splendid display of late June roses this year and offered to let me have some.
Would you like to take the trug from the kitchen cupboard and go and collect them for me?'
'Oh, yes,' said McAllister, jumping up; she liked walking and doing things rather than sitting about, even if she did enjoy knitting and plain sewing more than she had ever thought that she would.
'Do I need to change?'
She was wearing her afternoon uniform and Matey thought that she looked charming in it; it was not surprising, after all, that Dr Neil had a soft spot for her.
She wore a black cotton frock with a very full skirt and frilly lace-edged petticoat underneath it, just showing.
Her black stockings were finer than her morning ones, as were her buckled shoes.
Over her frock she wore a spotless white apron decorated with lace; it had a little white top, similarly adorned, and perched on McAllister's lustrous blue-black curls was an elaborate white cap, lace-trimmed, with long white streamers.
'Oh, you will do very well as you are, child,' she replied.
'No-one will object to such a pretty uniform.
Don't be too long, and try not to dawdle on the way there and back.
Remember that Dr Neil is entertaining two of the doctors from the hospital and Mr Sands, the curate, to dinner tonight, and I shall need you back to help me -- there will be a lot to do.'
McAllister nodded.
Hooray!
Freedom, if only for a short time, and she would have a chance to speak to Rose, whom she had not seen since their evening out together up West.
Matey had told her that Rose had not returned home until the small hours, and the housekeeper there, Mrs Parker, a middle-aged widow, had almost turned her away, but had relented, although Rose had lost her next three half-days off as a punishment.
More of a punishment, perhaps, than Matey realised, McAllister thought, seeing that Rose would lose three free suppers, and some spare cash besides.
She found the trug in the outhouse, not the kitchen, and cleaned it before trotting off towards the rectory, which was quite a long walk from Vetch Street, through a rowdy street market where an organ-grinder and his monkey were performing, and a Punch and Judy man stood on the corner, and Sally-Anne -- no longer McAllister now that she was out of the house -- for all of her advanced years stood and watched Mr Punch for some time before she guiltily remembered what she was supposed to be doing.
She trotted off again, through a really insalubrious area known as Bligh's Corner, after some landlord, long dead, and she unconsciously quickened her pace a little so as to arrive in the relatively respectable square where the rectory stood, next door to St Jude's church where Dr Neil's household worshipped every Sunday.
Sally-Anne reached the gates and walked up to the rectory, an eighteenth-century building which had once been beautiful, but like the area around it had gone to seed badly, although Dr Neil had told her that it was elegant, if shabby inside.
She had never visited it before, and looked around her with interest, remembering to go to the kitchen entrance.
She had almost forgotten herself by walking to the big front door, and giggled inwardly at the shock that she would have given Mrs Parker if she had done so.
Rose answered the door, looking rather crushed.
'Ow, it's McAllister from Dr Neil's,' she called into the kitchen, unceremoniously.
'Ow are yer, McAllister?
Got home safely, did yer, that night we spent up West?'
'Eventually,' said Sally-Anne, a little coolly.
'Thought as 'ow you would,' said Rose.
'Ave a good supper, did yer?'
And she gave a grotesque wink.
'Yes,' said Sally-Anne truthfully, if deceitfully.
'Mrs Parker promised Miss Mates some roses, and I've called to collect them.'
'Don't keep the gel on the doorstep, you gaby,' said Mrs Parker.
'Tell her to come in.'
And Rose led Sally-Anne into a large and airy kitchen full of the smells of a good afternoon's baking, its windows giving a splendid view of lawns and flowerbeds, full of blossoming rose-bushes.
The rector, the Reverend Mr Hallam, was a bachelor, and his garden was his only passion -- his ministry quite wilted before it.
Mrs Parker unpinned the large white apron she had worn to bake in.
'Mind you look after the tarts, Bailey, while I collect some roses for the gel here to take to Miss Mates.
What's your name, gel?'
'McAllister,' answered Sally-Anne, remembering to do her servant's bob.
'Well, you come with me, McAllister.
I hope you've got more sense than Bailey here.
You don't look it, though Miss Mates speaks well of you.'
Mrs Parker bounced into a garden shed, came out with a pair of secateurs.
'A dozen and a half red ones should do, I think.
Ena Harkness,' she added.
'Hold your trug out, gel.'
And she cut a dozen and a half splendid blooms and laid them reverently in Sally-Anne's trug.
'Oh, what a lovely garden,' said Sally-Anne, looking around her as the last bloom was placed on her pile.
'Like it, do you?' said Mrs Parker, pleased.
'A Yankee, aren't you?
Do you have gardens in America?'
'Yes,' said Sally-Anne respectfully, thinking of her mother's splendid garden, but she also thought that St Jude's garden was an oasis in the wilderness.
'Sit down before you go,' said Mrs Parker kindly; she thought McAllister a great improvement over Bailey, so soft-spoken and pleasant.
'It's hot today.
I was thinking of giving Bailey and myself a rest and having some lemonade.
I'll bring you a glass when we've finished cooking -- which shouldn't be long if Bailey has done as she was told.
I know Miss Mates won't mind you having a rest before that long walk back.
She's a kind soul.'
Sally-Anne sat down on a wooden bench, the trug at her feet, and drowsed a little in the sun.
She was feeling tired, what with all her hard work and the walk, and writing for J.D. after her long day was over.
Miss Mates had said not to dawdle, but surely she wouldn't mind McAllister having a little rest and a long, cool drink?
Mrs Parker's lemonade would be highly welcome when it arrived...
She dozed off.
Dr Neil arrived home in the middle of the afternoon.
His round had been long and arduous, and he was thinking how pleasant it would be to see and tease McAllister and ask her to serve them all some home-made lemonade.
But the little house was quiet, and the reason was plain.
Matey was seated on her own in the parlour, darning, and there was no sign of McAllister.
'Lost McAllister, have we?' he said, tossing his soft grey hat on to the sideboard.
Not for the first time Matey noted that his first thought was of his maid.
'I sent her to St Jude's Rectory to collect the roses which Mrs Parker promised me,' said Matey, looking at the clock.
'I thought that they'd make the parlour look nice for your dinner party tonight.
She really ought to be back by now.'
'To the rectory?' said Dr Neil, struck.
'You let her go there on her own, through the market and across Bligh's Corner?
It's not really a safe route for a pretty girl like McAllister.'
'Oh, come, Neil,' said Matey, speaking to him as though he were eight again.
'She's getting on for twenty-one years old, and perfectly capable of looking after herself.'
'Oh, no, she's not,' said Dr Neil grimly.
'She's a regular babe in the wood in the East End, however knowledgeable she might be in her own world -- wherever that is.'
He had sunk into his armchair, but jumped up.
'I'll go and collect her, or meet her on the way home.
There's only one route she can take.'
Matey opened her mouth to remonstrate with him, but forbore.
Best to say nothing, perhaps -- and she watched him pick up his hat and dash off to save McAllister from any dragons she might meet between Vetch Street and the rectory.
Sally-Anne was still seated on her bench, the empty lemonade glass beside her.
The rose-filled trug was at her feet.
She had her back to the entrance of the garden and was looking across it at a small orchard whose fruit never found its way to the rector's table, always being pilfered by the small street arabs of the district.
She did not hear Dr Neil come up behind her until he said, 'So there you are, McAllister.
I was fearful that I might find you down another alley.'
She turned towards him.
He had removed his soft hat and the sun was gilding his hair.
'So you came to rescue me,' she said, teasing him for a change.
'But, as you see, I don't need a knight errant this afternoon.'
'No,' he agreed, 'but all the same I shall walk you home.
I need you to help Matey with my dinner party and I shall make sure you arrive in the kitchen in one piece.'
'I shall not,' said Sally-Anne grandly, 'reproach you, Dr Neil, for rescuing me merely to ensure that I shall carry on skivvying for you.
On the contrary, I am wondering how I may reward you...
I know...' she said, and bending down, picked up the most beautiful of the blooms.
'A favour for my knight,' she said, and handed it to him.
Dr Neil bent down to take the bloom from her, and as their hands met his heart gave a great lurch.
The piquant face looking up at him was alive with mirth.
He swallowed and the world steadied again.
'Come,' he said.
'I accept your favour, my lady, although I may not wear it on my sleeve as a true knight should.
And now work awaits us both.
We may not dally here among the roses in Arcadia -- more's the pity.'
'Yes, indeed,' said Sally-Anne, rising.
'I must peel the potatoes for Matey, and you must take evening surgery and be ready for your guests.'
Dr Neil carefully replaced her recent gift to him in the trug, before taking it from her, and watched by Rose, peering interestedly from the kitchen window, they set out together for Vetch Street.
Oddly, Neil's suspicions about McAllister's presence in the East End, dormant since his rescue of her, were revived by his sudden awareness of how much she was coming to mean to him.
It was almost as though he was trying to erect walls around his heart to protect himself from his softer feelings.
'Still enjoying your masquerade, McAllister?' he offered, and although his voice was soft there was a sting in his words.
Sally-Anne -- she would not be McAllister again until she reached Vetch Street -- bit her lip.
She thought that he had stopped his baiting of her, had accepted her, if not exactly for what she claimed to be, but as herself.
'I'm sure I don't know what you mean, sir.'
And her voice was a parody of a servant's showing dissent, a politeness which verged on the impolite.
'You know perfectly well what I mean.'
He gave a little laugh and tightened the hand he had placed on her arm -- Sally-Anne had not offered to take his; that was not a servant's right.
'I live in fear every evening, McAllister, that I shall come home to find that our unlikely treasure has flown the nest, tired of playing with us.'
Sally-Anne wrenched her arm away from his hand, stopped, and turned to face him, regardless of interested spectators.
'I am not playing,' she informed him fiercely, 'and I shall not leave you without due notice, however inconvenient.'
'Three months' notice,'he said, and his voice was kind again -- her high spirits always amused him.
'I shall demand my full pound of flesh, McAllister, be sure of that.
I warn you, the Master will not lightly lose such a hard-working and decorative part of his household.'
This was a declaration of sorts, however lightly made, and one which Dr Neil had not thought to make to any woman in any form, and before Sally-Anne could take in its full meaning he rapidly continued, 'And now let us stop providing entertainment for the passers-by.
You look ready to dance with rage, and although we might collect a few pennies from them in return for our providing such a spectacle I hardly think that we could put it on as a permanent entertainment!'
He tucked her arm in his again, so firmly and gently that without actually fighting him off she had no alternative but to walk on by his side.
'Oh...' began Sally-Anne furiously, and then caught his twinkling eyes, and one raised eyebrow -- He was waiting, she knew, to see how she would react.
And she began to laugh, her temper draining away, and for the first time she realised that he was frequently doing something which no one else had ever succeeded in doing before -- he was defusing her rage.
The famous high temper which no one, not even her formidable papa, knew how to cope with, once she was on fire, was being controlled by a man whose outward-seeming was mild and pleasant, although beneath Dr Neil's bland exterior she was beginning to sense a will as strong and resolute as her own.
What on earth would Mama and Papa say if they knew?
How did he do it, when they couldn't?
Jared Tunstall, who manipulated billions of dollars, made and broke men at will, whose power was legendary, had never managed in a clash of wills with the daughter he adored to overcome her, once she had made her mind up, or decided on a course of action.
And even when, in the end, he had sent her to London, to her uncle Orrin, after one last, disastrous, raging tempest, he had managed her physical transference from the United States to England, but her will had remained unbroken.
She had refused to give an inch.
So why could Dr Neil, with a few words and a sidelong look, lance her indignation so easily, compel her to reassess herself and life?
She looked sidelong at him, walking along beside her, his unspoiled profile presented to her; but the other scarred side did not matter, because it was Dr Neil who mattered, not his face...
Sally-Anne stopped.
Oh!
She had read of Cupid, the little god of love, and of his wicked darts piercing the heart at the moment of understood love, and she had laughed a little at the idea; but one of them had struck home at last, and she knew that on the two previous occasions when she had previously and disastrously thought herself in love it had not been love at all ...
This, then, was love.
So sweet and strong that merely to be beside the loved object, to feel his hand on her arm, was enough.
No!
No!
It wasn't!
She wanted more, much more, than that.
She wanted to please him, to see him smile, and tease her by calling himself the Master, and pretending that she was his slave.
And -- oh, she also wanted -- no, desired -- to stroke his smiling face, to run the tips of her fingers gently along the dreadful scar to show that she loved every part of him, to...
For the first time in many months she contemplated the uncontemplatable, and thought that it might not be so with him, even though her heart began to thud uncomfortably at the very idea.
She had stopped without thinking, stopping him as well.
He turned his lazy smile on her, and asked, 'What is it McAllister?
Not another swoon, I trust?'
'Yes -- no,' she said idiotically.
'It's nothing.
My heart missed a beat...'
And oh, that was yet another lie, for it was not nothing, it was something.
Sally-Anne Tunstall had found her true love at last.
CHAPTER NINE
Of course, Matey was not best pleased that she had taken so long to come home -- or so McAllister thought when the older woman came towards them after Dr Neil had let them both in.
He thought that McAllister had been oddly subdued on the walk home, and had surreptitiously admired her face -- even more beautiful, he thought, when pensive, even though in her early days at Vetch Street it had been her animation which he had admired.
Admired!
What a word!
It was love he was beginning to experience, and what a joke that was.
How could he be falling in love with the beautiful cuckoo in his nest who was exactly the sort of pampered darling a poor doctor ought to avoid?
When she had tired of her games she would take wing and fly away, and even if she stayed, how likely would it be that she could ever settle down as that poor doctor's wife?
He too realised that Matey was not pleased, but it was not their late arrival which was troubling her, but something else, as soon became apparent.
She ignored McAllister, and said sharply to Dr Neil, 'You have a visitor.
I have put her in the parlour; she insisted on waiting until you returned.'
'Oh,' said Dr Neil, staring at Matey.
'Is it Teresa Darrell?'
'Worse than that,' said Matey grimly.
'Lady Macleod.
I told her you'd not be best pleased to see her, but she insisted on staying, as I said before.'
McAllister was fascinated.
Another woman chasing Dr Neil!
Well, she wasn't surprised; he was eminently chaseable.
And was this strange emotion she was experiencing jealousy?
Matey turned her attention to McAllister.
'You've taken long enough to do a simple errand, I must say.
Never mind.'
She took the flower-filled trug from Dr Neil, admired the blooms, adding briskly, 'Come along, McAllister; you might as well make up for lost time,' and all three of them walked into the parlour, Matey and McAllister of necessity, since it was their only indoor access to the kitchen.
A blonde woman fashionably dressed in half-mourning, wearing a violet gown, her hair dressed high under a huge picture hat trimmed with a bird of paradise in full flight, rose to meet them, or rather to meet Dr Neil.
She threw a cursory glance at Matey, inclining her head slightly, made no gesture at all to McAllister, but merely stared inimically at her, offended by such youthful, dewy charm.
'Ah, Neil, there you are,' she announced.
'I thought that you would never come home, but I would have waited until the cock crowed.'
An unfortunate allusion, thought McAllister nastily, who knew her Bible, and that the cock crowed after betrayal, and from what Matey had said this woman specialised in betrayal.
'I have my duties to perform, Lady Macleod,' he said sternly, face and voice frozen alike.
'Oh, not Lady Macleod, surely,' she said sweetly, putting a gloved hand out to him, and then proceeding to remove her gloves as provocatively as she could.
'Angela, always Angela, Neil.'
McAllister could not remember having seen her in her old life in high society, which was fortunate, perhaps.
She was a violet-eyed beauty, her figure full and voluptuous, a regular fly-trap.
McAllister's thoughts were growing nastier and nastier, but she could not endure the sight of the woman who Matey had said had trampled on Dr Neil and rejected him after he had sustained his war wounds and started his new career as a doctor.
She felt Matey take her by the arm and pull her rapidly away.
'Come on, McAllister.
Work to do!'
And her last sight of them was Dr Neil standing cold and upright, the scar on his white face angry and vivid, facing the lovely woman whose whole posture was that of a pleading Niobe -- McAllister wondered where all these classical allusions were suddenly coming from, remembering that Niobe was famous for her tears, and for their power.
Would they have the power to move Dr Neil?
Angela Macleod breathed a sigh of relief at the closing door -- never mind that every word they uttered might be heard in the kitchen.
'You must know why I have come, Neil,' she said, and she held out her arms to him.
'No, Lady Macleod,' he said, his voice still relentless.
'I have no idea why you are here.
And I must inform you that you are not welcome -- not welcome at all!'
'Oh, never say that, Neil.
Remember what we once meant to one another.'
Her eyes brimmed with tears.
'And Hector is dead, Neil, dead these six months.
Think what that means for us.'
'Yes, I know that Hector is dead.
Stair told me so.
It means nothing to me, madam.'
'Oh, Neil,' she wailed, the tears falling at last, great crystal drops.
'You do not know what my life has been these last ten years with him .
As soon as we were married he took me off to Scotland, kept me there in virtual purdah, never let me out of his sight -- his jealousy was manic.
I thought of you every day, of how we could have been together, and now we may start again.'
'Start again!'
He laughed and the sound was frightening.
'Do you remember what you said to me in this very room when I begged you to marry me and you told me that you were going to marry him?
Shall I remind you of what you said?'
'Oh, Neil, do not judge me too harshly.
I was so young and foolish and I did not know what I was doing.'
She had crossed the room to stand before him, and the scent of her expensive perfume was in his nostrils -- and it nauseated him.
'You knew very well what you were doing,' he replied, not giving her an inch of himself, wondering how he ever could have desired her.
Maturity had given her extra beauty, but her predatory nature was written on her face.
'You were marrying a rich old man for his money and rejecting the crippled boy who had been your sweetheart since childhood, and whose engagement ring you wore.
You were not made to marry a poor man and live in an East End hovel, you said.'
'Ah, but you need not do that now, Neil,' she answered him feverishly, seeing him slip from her.
'I am a rich woman; he left everything to me, and so he should have done, after all that I sacrificed for him.
And now we can be together again, and enjoy the life we were always meant to have.'
And she put her arms around his neck and tried to place her mouth on his.
Dr Neil stepped back so swiftly that she almost fell; he seized her by the shoulders and held her from him.
'Do you think to wipe out what you said and did, and ten years' loneliness, so easily?
I worshipped you, Angela, thought that I was coming back to you, and when you left me, you left me with nothing.
My life is here, and I have no intention of sharing it with you, not if you offered me an Empire, let alone a castle in Scotland and the fortune of a man whom I despised and who I warned you would treat you badly.
You are ten years too late with your offer of yourself.
And I have half a mind to take you at your word, give you what you seem to want and turn you out of my house after I have finished with you.'
Her eyes glittered, and, mouth twisting, she flung away from him.
'Oh, you were always a fool, Neil, always -- a fool to think that I would be content to live on nothing with a younger son and a fool to reject me now that I can give you the life which you ought to be living.'
She had begun to pull on the long suede gloves which she had earlier stripped from her beautiful hands.
'But I suppose that you are consoling yourself after the fashion of a fool who lives in the slums.
Does that pert chit you were escorting home satisfy you, Neil?
You used to be more fastidious, used to be above servants --'We must not exploit them,' you told Hector once -- how we laughed over that!'
Neil strode to the door, opened it, and held it open.
'I must ask you to leave, Lady Macleod.
Your insults to myself I can stomach; those you choose to demean an innocent young girl with are quite another thing.
Miss McAllister is in my home under my and Matey's protection.
You forget yourself.'
Angela, Lady Macleod, who had once been Angela Deverill and whom he had thought that he had loved more than life itself, picked up her parasol, an object as lovely and useless as she was.
'And that is your last word, Neil?
Remember, I shall not come to you again.'
'My pleasure, Lady Macleod,' he said curtly.
'I do not wish to see you again.
Pray take your leave; you have outstayed your welcome these past ten years.'
She was by him and was gone, the door closing behind her, this time for good.
He had thought that he had done with her long ago, had assumed that she had lost the power to hurt him.
He did not know, indeed, which hurt him the most -- the memory of the bright young girl she had been, or the knowledge that now, having sold herself to a rich old man, she had come, once that man had died, to try to buy him for herself.
In the kitchen Matey and McAllister had been compelled to hear every brutal word -- neither Dr Neil nor his tormentor had kept their voice down -- bustle about and bang pots as they might.
McAllister wanted to go in and do something, anything, to stop such a harpy from hurting him.
She knew that he would be hurt even though she hoped that he had stopped loving such an undeserving, self-serving shrew, because anyone would be hurt by such a scene.
Once the outer door had banged shut, Matey said, violently for her, 'Good riddance to bad rubbish!
Thank God she's gone,' and the unwonted strength of her language betrayed the depth of her feelings.
'She hurt him dreadfully when he came home from Africa, crippled and his career in the army over, just when he needed her loving support the most, and now she has the gall to chase him again, when I hoped and prayed that he had finally recovered from her brutal treatment of him.'
'She's very beautiful,' was all that McAllister could say to that.
'Handsome is as handsome does,' snapped Matey, who as an ex-children's nurse had a fund of such sayings.
'And by that maxim, McAllister, she's as ugly as sin and high time Master Neil realised it.'
He had reverted to being her youthful charge, as he always did when she was worried about him, and McAllister uttered a fervent inward Amen.
There was silence from the other room.
Usually when he was giving one of his little dinners he came into the kitchen while they were preparing it, joked with them, had his hands smacked smartly by Matey when he lifted saucepan lids and tried to peer into the oven; but today he stayed away.
'Take him some tea,' said Matey, who was rolling pastry for a raspberry tart.
'I'm worried.
He shouldn't be on his own.'
How the British depended on tea -- but it seemed a good idea.
McAllister bustled in her turn, made the tea and carried the silver tray through into the parlour, to find Dr Neil at the sideboard, looking grim, the whisky bottle in his hand.
She thought that he'd not started to drink, but might at any moment.
'Tea's better for you than that stuff,' she said pertly, as pert as the chit Angela Macleod had called her.
She did not know that Neil's anger and disgust with his old love had been fuelled more by her unpleasant reference to McAllister than anything else she had said to him.
But for a moment he looked thunder at her impertinence, then put the bottle down and began to laugh.
'No use supposing that you didn't hear all that passed, McAllister,' he said.
'Neither of us kept our voice down.
Tell me, what should I have done?
Run off with her and the money which she sold herself for?'
McAllister did not quite know how to answer him, and said slowly, 'I suppose that at some time or other we all sell ourselves for something.'
Dr Neil stared at her, took the tea-tray from her hands, put it on a side-table.
'What a wise child you are becoming, McAllister.
Is it living in Vetch Street that is changing you?'
'Or the people who live here,' she said, greatly daring, but she would say anything to keep his mind off the bottle.
'I don't want her, you know,' he said, pouring the tea in an amber stream into his cup.
'But she revived old memories, and sometimes memories hurt more than the acts which created them.'
'I know,' said McAllister, closing her eyes against her own unwanted memories.
'You do?' he said, putting the teapot down.
'Yes, I believe you do, McAllister.
But surely you have no dreadful memories, so young and fresh as you are?'
'They are one thing we all have,' she replied, opening her eyes.
'Our ideas of what is dreadful may differ.'
For a moment Dr Neil was silent, then, 'Come and sit by me, McAllister.'
'Should I, Dr Neil?' she said, but gently.
'Perhaps not, but I command it.
The Master commands it, his slave must obey.'
And a hint of his normal happy, teasing manner to her was in his voice.
McAllister sat down, opposite him, in one of the Maplewood dining chairs.
'You have only brought one teacup with you, McAllister, so we may not share the tea as we shared the coffee earlier today.'
'Matey will expect me back to help her.'
'I suspect Matey sent you in deliberately instead of coming herself,' was his only answer to that.
'You know that I nearly broke my word to you about drinking just now, McAllister.'
'Yes,' she said, and to lighten the situation added, 'And had you done so I would have dropped all the dishes for your dinner party tonight over your unfortunate guests to show that two can break a promise!'
That did the trick, didn't it just!
'Oh, I'm glad that I behaved myself, then,' he said, his face clearing suddenly.
What a fool I am, he thought, still to mourn for a lost past and a girl who existed only in my imagination, the reality being so different, when I have this ardent, hardworking...child...to keep me company, so different from the mercenary beauty who has just left.
I should be running up a flag to celebrate her going, not behaving like a weak boy myself.
And he looked at McAllister sitting primly before him, hands in her lap, the face beneath the lace cap so enchanting that Angela Macleod, nee Deverill, and all her works, flew away forever, and he knew again where his heart now lay.
His face betrayed him to McAllister.
Her own heart thudding dangerously, the atmosphere between them charged, she rose.
'I really must be going to Matey, Dr Neil.
I am neglecting my duties.'
He rose, too, and they stood, face to face.
'No,' he said slowly, put up his hand to brush back a lustrous curl which had strayed from its bonds.
'No, McAllister, don't go.'
And he put up his other hand to clasp her face, gazing ardently into her blue-black eyes, and his intention was suddenly plain.
He began to drop his head to find her parted lips, and McAllister, who was prey to the most extraordinary and conflicting emotions, one part of her welcoming a kiss from Dr Neil, whom she loved, a kiss which she knew would be passionate, the other part hating and fearing being held or touched by any man, including him, said breathlessly, 'No, you shouldn't.
You must not; Matey...'
She pushed at his chest, fear and love fighting for mastery in her.
Something she said, perhaps Matey's name, the lack of privacy for the two of them in the small house, although they had both kept their voices low, stopped him.
He stepped back, smiled wryly, and said, 'I wasn't consoling myself after Angela Macleod, McAllister.
You do believe me.
That...kiss...would have been truly for you, no one else.'
What could she say but, 'I believe you, Dr Neil, and I really must go, or...'
'Matey will think that we are doing what I was about to do.'
He laughed softly.
'Oh, you are a witch, McAllister, a very witch, designed to turn any man's head.
Even she saw that.
Yes, you may go.'
And lightly, now that the danger was over, 'The Master commands you to go.'
'And you promise not to drink Matey's 'nasty stuff'?'
'Yes,' he said, 'and will you keep my guests safe from flying pots and food?'
'Agreed,' responded McAllister gaily, and then in a flash was back in the kitchen, Matey's shrewd eyes on her.
Their interchange had been soft, not loud, and McAllister wondered how much Matey had heard, and whether she approved or not -- and did not care.
Full summer, the heat of late July was on Vetch Street.
The dead flies on the fly-papers which hung in every room bore witness to that.
Even so, one naughty specimen, avoiding the fate of its fellows, buzzed around McAllister's head.
She was engaged in dressing a doll for the bazaar which loomed nearer and nearer.
The doll had been given in all its naked glory by one of the poorer parishioners, a relic of better days, and McAllister had undertaken to make its trousseau.
She had finished the sewing part, and was now knitting its little coat.
When that was finished the doll would be ready for sale, and it would be all her own work.
She had already embroidered a dozen little nighties for the babies of the fallen girls, decorating them with tiny pink rosebuds and pale green leaves.
Since Angela Macleod's visit her rapport with Dr Neil had grown stronger and stronger, although he had never touched her again.
Her duties over, she played chess with him in the evening, or childish card and board games with him and Matey, laughed, talked and read, in some odd way enjoying herself in this hard life more than she had ever done before in her old soft one.
Late at night she kept up her journal and wrote her column for J.D., and on her last visit he had offered her a post on his editorial staff, a full-time one, when she had finished her East End stint, as he called it.
'But not yet,' he had said.
'Your work is still creating interest, and the women of the Fabian Society are talking of following it up with a scientific examination of the social and economic condition of the women in the poorer parts of London, carrying on where Booth in his great survey of poverty left off.'
'No one...' began Sally-Anne passionately, thinking of all that she had seen since arriving in Vetch Street -- the poor creatures in Dr Neil's surgery, and the even poorer ones who could not afford to go there -- and of Dr Neil's own selflessness letting people off their bills, so that only the small income he still received from an aunt's legacy allowed him to keep going at all , when he could have been revelling with Stair.
'No one who has not lived among them knows how hard their lives are, how wretched their condition, could imagine what they suffer.
Anything that I can do to help them, any words I write which may soften their lot, is little enough.'
Like Dr Neil, J.D. thought how much she had changed and was still changing.
The flighty girl she had been was graver, and her compassion was real because the suffering of which she spoke was no longer an abstract thing to her.
This afternoon, however, all that was far from her mind.
She was on her own: Dr Neil had been called out some time ago to attend a woman in childbirth, and Matey had gone to visit an old friend in St John's Wood, and she was the sole mistress of the doctor's house in Vetch Street.
A little earlier the mother of Dr Neil's surgery boy, who carried out the doctor's errands on the bicycle which lived in the outhouse, had come to say that Eddie had broken his arm and would not be fit to carry out his duties.
She had been crying, for the money Eddie earned was precious, and McAllister had not known what to say, only that she was sorry and would inform Dr Neil when he had returned.
McAllister finished the right front of the coatee, as Matey called it, cast off, and looked again at the half-grandfather clock with a sigh.
Half-past three.
Dared she make herself a cup of tea in Matey's absence?
Of course she dared.
She was just lighting the gas on the tiny ring when the knocker on the front door sounded a veritable tattoo.
Irritably -- she really wanted that cup of tea; she was growing as bad as the British -- McAllister turned off the gas ring, blew out the match, and walked to the front door, grumbling to herself, Hold your horses, I'm coming, I'm coming, when another urgent series of knocks sounded.
She opened the door and there was Rose, dirty and dishevelled.
She said urgently, 'Ow, McAllister, Dr Neil says to send his boy wiv his emergency bag and some towelling, and quick about it.'
McAllister stared at her.
'Where is he?' she said practically, and then, 'Oh, shoot, I can't send the boy; he's broken his arm today.'
'Doctor says to be quick,' repeated Rose, 'and he's in our Buildings.
One of them upstairs is avin' a bad time, and 'er only fourteen, too.'
McAllister hopped up and down gently, her reaction the common one of a child to worrying news.
'Oh, dear,' she wailed, 'he must want his bag urgently to send for Eddie,'and then, irrelevantly, 'Why aren't you at work, Rose?'
'Turned off, wasn't I?' said Rose.
'For slippin' aht at night.
The missis caught me, and that was that.'
McAllister made up her mind; if Eddie could not cycle to the Buildings then why should not she?
After all, she knew where he was, and although it was a man's bicycle, not a woman's, she could surely manage to ride it.
She began to pull off her apron.
'I'll go,' she said, and ran to the surgery to collect the bag, and to pick up the roll of towelling and a pair of scissors which Dr Neil kept for emergencies.
She dragged the bicycle from the shed, placed the bag and the towelling, a little insecurely, in the basket at the front and somehow managed to throw her leg, encumbered by her heavy skirts, over the high framework, so different from the shape of the one on which Papa had taught her to ride.
'Ow, McAllister, that don't look safe,' panted Rose as she wheeled herself and the bike into the road, preparatory to setting off for the buildings.
'Never mind,' called McAllister, 'needs must...'and wobbled down Vetch Street, praying that she arrived at her destination in one piece, the balance of the wretched thing being all wrong, especially with the overloaded basket at the front.
Behind her Rose followed her at a wheezing trot, keeping up with her, until she suddenly mastered the antiquated machine, and sped away, feeling like a Swiss mountaineer with his St Bernard dog, setting off on his errand of mercy.
CHAPTER TEN
Dr Neil Cochrane was engaged in one of the worst tasks a doctor had to face: attempting to save the life of a young girl and the child she was trying to bear, after earlier neglect and mismanagement from the moment the birth had begun.
He had cleared the small, fetid room, in which five persons lived, of everyone but the girl's grandmother, who seemed to have some idea of how to help him -- the girl's mother was a defeated, apathetic creature only just over thirty herself.
The grandmother was capable of fetching water from the tap in the courtyard, if nothing else.
Pulling out his watch, he stared at its unkind face.
Would Eddie never come?
Rose should surely have reached Vetch Street by now and Eddie been sent on his way by Matey or McAllister after being summoned from his nearby home.
A face peered round the door at him.
His patient, a girl whose age was uncertain -- he judged her to be about fourteen or fifteen -- was at the moment quiescent, her face as grey as the linen on which she lay, her eyes closed.
Her pains, which had started long before a frightened Rose, acting as an amateur midwife, had thought to send for him, had stopped -- a bad sign.
He impatiently waved the face away; he wanted no useless, interfering spectators, and the impatience was as much for his own continued failure as for the uncaring curiosity of the woman who had looked in -- life was cheap in the Buildings.
The old grandma who was holding the girl's hand looked across at him as he walked to the window to stare down into the street to watch for Eddie.
He really should have had the sense to bring his emergency bag and the roll of towelling with him, but Rose's terrified urgency had sent him on his way with all speed.
He could not have visualised such a desperate situation as the one which he had found.
And then he saw McAllister.
She had mastered Eddie's cycle and came around the corner as though she were taking part in one of the bicycle races which had recently become such a popular spectator sport.
'McAllister!
What the devil...?
Where's Eddie?'
He watched her swing an indelicate leg over the top bar of her man's bike, displaying a neatly rounded calf which would normally have engaged his interest, but in the cruel circumstances failed to register.
She pulled the bag and the towelling from the carrier, hitched up her skirts to show yet more leg -- still unheeded -- and dashed into the Buildings.
He could hear her frantic progress up the uncarpeted stairs and along the top corridor until she threw the door open.
Panting, her hair coming down, her face red with the effort of riding at speed, she thrust the bag at him, saying breathlessly, 'Rose said that you wanted this.
Eddie couldn't come, he's broken his arm, so I came instead.'
She hung on to the towelling and the scissors, and then looked around the dismal room to find a clean surface on which she could put them.
Fascinated, Dr Neil, who had opened the bag and was beginning to pull out strange-looking forceps and a bottle of chloroform, watched her lift up her black skirt to reveal a spotless white petticoat, from which she cut a large square which she placed on the grimy table, having first cleared to one side the used cups and plates covered in half-eaten food.
McAllister, coughing slightly against the stench of the room -- Dr Neil was hardened to it -- laid the towelling and the scissors on the clean square, and said briskly to Dr Neil, 'Is there anything in your bag which you might need urgently?
If so, you can give it to me to keep clean and I can hand it to you when you want it.
Dr Bodkin of New York General told Papa once that surgical instruments should be kept absolutely spotless.
Do we have any boiling water or carbolic soap?'
She knew very little about surgery and nothing about childbirth, but she had a fund of common sense, and the moment she had pushed through the useless, wailing women downstairs, and seen the squalid room in which Dr Neil was working, she had begun to dredge up what little she knew in order to help him.
She hoped that she would not disgrace herself by fainting, or by being unable to help him through fear or disgust of what she might be seeing.
At the moment the girl on the bed looked comatose, but she supposed that would not last.
'What do you think you're doing McAllister?' asked Dr Neil as she walked towards him, just at the moment when the girl began to moan and thrash about, her pains suddenly returning in full force.
'You can't stay here.
It's not a fit place for a young girl to be.'
'And it's not a fit place for a young girl to have a baby,' snapped McAllister, who had just seen that his patient was little more than a child, 'but she's having one, all the same.
Surely even an amateur nurse would be helpful.'
She looked at the grandma, who although willing seemed full of the resigned acceptance of disaster which had ruled the women downstairs.
'I don't want a nurse who might scream or faint when matters become difficult, or would swoon at the sight of blood,' said Dr Neil, a little frantically, trying to care for his patient and send McAllister away at the same time.
'You are really wasting your energy on me,' said McAllister severely, 'when you should be concentrating on your patient.
If I promise not to shriek or have hysterics, will that do?
As for the sight of blood, that has never overset me.'
'Oh, very well,' he said wearily; he might have known that it was useless to argue with McAllister -- her tongue was as long as her will was strong.
'Make sure that the old grandma keeps up a constant supply of hot water to keep the instruments clean.
You may cut the towelling into large squares, and then you can come and be my nurse -- and God help you, McAllister, if you hinder me by starting to have fine lady's vapours.
Things are bad enough here without that!'
McAllister, for once, did as she was told.
Dr Neil's face was nearly as grey as his patient's, and she did not wish to distract him, but to help him, rather.
The towelling prepared, she made a neat pile of it, and then went over to the large bed, on which four or five persons normally slept.
The girl was now writhing and moaning faintly, barely conscious of what was happening to her.
She saw McAllister, though, and caught at her hand.
'That's right,' said Dr Neil encouragingly.
'You hold on to McAllister.
And McAllister, you can keep Effie here steady for me while I take another look at her.
Put that pillow under the small of her back so that I can lift her legs.
Yes, that's right.
Now we can have a look inside.'
McAllister swallowed when he bent between the girl's extended legs, and Effie shrieked aloud, throwing herself about, trying to avoid his gently probing hands.
'Hold her steady,' he commanded.
'She'll not hurt herself then.'
And the internal examination over, he said to McAllister, 'Lay her down.
She'll be quiet for a moment, I believe.'
Which proved true.
The girl had fallen half-conscious into McAllister's arms.
'Now fetch me some towelling, wet it in that bowl of water and hand it to me...'
Trembling and obedient, McAllister did as she was bidden.
She was already feeling faint, as much from the stale heat of the attic room as for any other reason.
Dr Neil wiped his hand on one of the pieces of towelling.
Effie was lying quite still, her eyes shut, her swollen stomach seeming too large for her otherwise immature body.
'What's wrong?' whispered McAllister bluntly.
'I didn't think childbirth was normally as difficult as this.'
Dr Neil discarded the towel, throwing it into a packing-case, checked that his patient was not needing immediate attention, and drew McAllister into the far corner of the room where their conversation could not be overheard.
'I suppose, McAllister, if you're going to be of any real use to me --' He was interrupted by Rose bursting in, a Greek chorus of depressed-looking women following her.
'Ow, McAllister, yer got 'ere safely, then.
Can I 'elp, Dr Neil?'
'Yes, Rose -- and the rest of you, out!
This is not a  raree show.
And you, Rose, if you want to be useful you may keep everyone away, run any errands I may require, and take over from, or help McAllister, when I tell you to.
You can help the old grandma to keep up a supply of hot water.'
He bent over the bed once again.
He had spoken in the voice of the army officer he had once been, expecting instant obedience, showing McAllister a man far removed from the one she had now known for over two months.
'Now,' he said, after inspecting Effie again, and taking McAllister out of earshot, 'listen carefully to me, for I don't want to repeat myself later when things get difficult.
Effie here is in a bad way for several reasons.
She is weak for lack of good food, she has plainly suffered from rickets, her body is not yet sufficiently mature for her to carry a child successfully to term, and on top of all that her baby is coming in quite the wrong fashion.
'A normal baby arrives head first to make childbirth as easy as possible -- and even that would have been difficult for her, given her narrow pelvis -- You do know what and where a pelvis is, I hope, McAllister?'
Yes, McAllister did know all about pelvises, but had the sense not to spark back at him that, of course, she did know to what he was referring.
'But this baby is coming feet first, which is very bad indeed, and makes a successful delivery very difficult.
Had I been sent for much earlier I would have tried to turn the baby -- but now...it's too late.
First she was on her own with Rose and some of the women, then they sent for the midwife, who didn't turn up, and only after that did they think to come for me.
So, we have a long, hard task before us.'
He had said 'we', which heartened McAllister enormously but also meant that she must try to justify his confidence, and not let him down; he must not think her a fine lady only playing at life -- she meant to do her share, and yet try not to hinder him.
'Listen to me carefully,' he continued, his face grave.
'I want you to measure out an ounce of chloroform in the jug you will find in my bag.
With the chloroform we can ease Effie's pains at this stage.
You must only give her an ounce every hour, and you will only start when I tell you, and stop immediately I order you to.
You understand me?'
She nodded mutely, and while she did as she was told he went over to the bed where Effie had begun to writhe and cry out, and her pains, which were erratic, had started again -- that they were so erratic was another bad sign, Dr Neil quietly explained.
'Sit by her,' he ordered.
'Let her take your hand, stroke hers; it will comfort her a little, and when I tell you give her the chloroform -- Good -- that's the way.
Now, when she is fully dilated, which will take some time, I will tell you to stop.
That way, we shall save her a great deal of pain, and she can husband her strength for when she needs to push the baby out.'
The trust which had shown on Effie's face, before the chloroform-induced unconsciousness, was now repeated on McAllister's as she looked up at Dr Neil.
Effie's fear had disappeared a little under his quiet comfort, and it had the same effect on McAllister.
She thought suddenly that she had taken Dr Neil for granted, that she had not fully realised either his hard work or his dedication, and for the first time understood the impulse which had caused him to leave the cushioned life of a younger son of a good family and become an East End doctor instead.
Before, she had understood his choice in a theoretical fashion; now she was seeing what that choice meant, and how hard he tried for his poor patients, and for such little return other than the simple and offered faith of the poor, suffering girl on the bed.
Later McAllister was to remember that afternoon spent in the stifling attic as a kind of nightmare punctuated by Dr Neil's barked commands to her and Rose; she had ceased to exist for him except as a pair of willing hands and legs.
He had ordered Rose to stand by and cut more squares ready for when Effie entered the second stage, and would be conscious and suffering again.
'Wet them, Rose,' he said.
'We can give them to her to suck on when she needs to push.'
Was this what love led to, McAllister thought more than once during the long afternoon: this agony, this terrible suffering?
Had her mother gone through this four times, once to produce herself?
Were all men and women born in such pain?
Reason said, Not always, for otherwise mankind itself might come to an end.
Or were the acts which produced this end so far away from them in time that somehow the temporal distance anaesthetised reality?
Sally-Anne, for thinking this she was Sally-Anne again, not McAllister, Dr Neil's acolyte in childbirth, had never seriously considered such things before, but living among the East End poor she was beginning to ask questions not only about her own, previously pampered life, but about life in general, and the stunted lives of those around her in particular .
One of the defeated-looking women, sullen and grimy, came in once with cups of tea for them all.
Dr Neil drank the foul-tasting stuff gratefully; he had now taken off his tie, and the sweat stood in pearls along his forehead.
When he walked over to hand his empty cup back, his limp was very pronounced and the scar on his cheek red and angry.
The woman took the cups, gave the figure on the bed a dispassionate stare.
'I did say as 'ow she ought to 'ave 'ad Mother 'ampton,' she observed.
'Did you so?' barked Dr Neil.
'Seeing that Mother Hampton was dead-drunk under the table at the Jolly Waterman when Rose Bailey went for her, a fine job she would have made of it.'
He looked at McAllister for the first time since she had begun to help him, and saw her strained white face, the eyes enormous in it.
'Would you like a rest, McAllister?' he enquired.
'You could let Rose take over for a little.'
McAllister knew that if once she stopped her ministrations she would never get up again, and her reply was as firm as she could make it.
'Thank you, but no.
I shall go on as long as you do, Dr Neil.'
She would show him that there was no chance of her giving way to a fine lady's vapours!
Nevertheless Dr Neil encouraged Rose to help; he said that he really needed two pairs of hands now that Effie was fully dilated.
He ordered McAllister to stop giving her chloroform and the two girls worked busily side by side.
Once conscious again, Effie began to scream, pain taking her over, and only McAllister's soothing hands, and Dr Neil's voice, urging her to push with each pain, stopped her from becoming a mindless howling thing on the bed, unable to help herself.
Rose supplied McAllister with the damp towelling squares, and she alternately wiped Effie's forehead, and gave her the towelling to bite on when the pains came strong and fast.
'Now,' said Dr Neil fiercely, for the baby had suddenly decided that, whatever position it was in, it was going to be born, 'give her the chloroform again, McAllister; the baby is ready to be delivered.'
He ordered Rose to take one of Effie's legs and lift it on to her shoulders, and McAllister the other, to make it easier for him to help the insistent baby on its way.
'And when you need to give her more chloroform,' he said, 'hand the leg you are holding to Rose -- you may take them both on your shoulders if you will, Rose; it's important that she doesn't regain consciousness during the actual birth.'
He bent down, preparing to lever the baby into the world with his special forceps without either crippling it for life, and killing it, or the mother, or both of them.
McAllister really did think that she was going to faint this time, not through shock or disgust, but sheer physical exhaustion, except that the will which her family deplored kept her on her feet.
And when Dr Neil commanded her to give Effie chloroform again while he finally manoeuvred the little body out the relief she felt was almost overwhelming, enhanced by the sudden strange joy she felt at the sight of the tiny purple baby boy.
Effie, spared the final agonising pangs of the actual moments of birth, recovered so slowly that McAllister feared that she had overdone the chloroform.
She looked even more ghastly after the birth than before it, frightening McAllister.
Dr Neil's attention was, for the moment, on the baby, and it was left to McAllister to wipe the mother's livid, sweating face, only the perspiration on Effie's face was cold, not hot.
Dr Neil had the baby upside-down and was smacking him to make him cry -- having been born, he no longer seemed to want to survive in the harsh world outside his mother's warm body.
He had damaged the baby's shoulder during birth, but not so much, he thought, that it would be a cripple.
The baby finally gave a feeble and pitiful squawk, quite unlike the enraged roars which McAllister had been told that her sturdy brothers had made on entering their privileged world -- and which she had undoubtedly made herself -- but at least he was breathing.
Dr Neil, cradling him in his arms for a moment before handing him to McAllister to care for, wondered not for the first time on such occasions whether he had done the baby a favour by enabling him to live.
With no acknowledged father, a child mother -- if the mother survived, which he was beginning to doubt -- and an attic in a tenement as its home, his future seemed uncertain, to say the least.
But he was not God, to decide such matters, only a competent and caring doctor, and his duty to humanity and to his Hippocratic oath -- to save life, not destroy it -- left him no choice.
He laid the naked red scrap briefly in Effie's arms, and said abruptly to McAllister, 'Let Rose do that while you take the baby from me and find something to wrap it in.
We'll leave bathing it for the moment; I must take care of Effie and the afterbirth.'
The afterbirth!
And hold the baby!
She had never held a baby before that was not a fat, well-dressed, sweet-smelling thing, but not one of them had ever given her the strange sense of fulfilment which the pitiful scrap of humanity in her arms, wrapped in the square which she had earlier scissored from her petticoat, did.
As she cuddled him, seeing his tiny face, his perfect little hands, a great surge of emotion swept through her.
Never mind the narrow life that he had been born into; his tiny presence seemed to make all the suffering and pain his birth had caused worthwhile.
Except that while she was loving him, gazing wonderingly at his blind eyes, Dr Neil said suddenly and hoarsely, 'Quick, McAllister; put the baby in the drawer over there and bring me as much towelling as you can.'
For earlier, the grandma, who had left them some time ago, had pulled out a drawer from a deal tallboy, and put a noxious blanket in it, ready for the baby's birth.
Once again she leapt to it, put the baby down regretfully, and carried the towelling over to the bed, to see a dreadful sight.
As was often common after breech births, the afterbirth had not come away as it should have done, and Effie had been so torn during the birth, Dr Neil told her later, that she had started to bleed, and then the bleeding had turned into a violent haemorrhaging, the passage of the afterbirth completing the damage already done to Effie's poor little body.
Effie, Dr Neil, Rose, McAllister and the bed were covered in blood.
Dr Neil made vain attempts to stifle the bleeding, his own face now as grey as the bedlinen had been before it turned red.
'Oh, Gawd, she's goin',' shrieked Rose.
'No!' said Dr Neil violently.
'No, I won't have this.
I thought we'd won this time.'
And McAllister remembered his silent distress when he had come home after losing a woman in childbirth not long after she had arrived in Vetch Street.
But the battle was in vain, and McAllister, like Dr Neil and Rose, had to watch while all their efforts to save Effie were useless, and slowly her life leached away with her blood, leaving her white and drained on the bed, only the shouts of her now motherless baby breaking the silence of the room.
Two great tears ran down McAllister's face.
She neither sobbed nor wept aloud as Dr Neil sighed, closed Effie's eyes, pulled up the sheet, and rose to walk to the window, to look out of it, at nothing, McAllister suspected.
She could only imagine what he must be suffering.
Rose had begun to sob gently, not at all like her usual wild self, and someone must tell the poor mother downstairs that she had a grandson, but not a daughter.
CHAPTER ELEVEN
Dressing the doll, after her experience with Effie, took on a different meaning for McAllister.
Since the baby's birth and its mother's death her whole view of life seemed to have changed abruptly, as though someone had shaken a giant kaleidoscope and she saw the whole world and herself in a different light.
She was not even sure any more who she was -- Sally-Anne Tunstall, the rich American heiress, or the humble servant McAllister, whom Dr Neil, Matey and the East End knew -- except that, like Sally-Anne Tunstall, McAllister had never been humble.
She was, perhaps, an amalgam of the two, and her masquerade, begun, she now knew, almost as the game Dr Neil had called it, had, in turning into something more serious, changed her as well.
Dr Neil, coming into the parlour where McAllister was permitted to knit and sew, caught his breath at the sight of her with the doll.
She had finished knitting and sewing all its little garments, had arranged its fine hair, and was now busily engaged in slipping them on.
He had caught a glimpse of her tender expression once before when he had handed her Effie's baby to hold, and later when she had bathed it before they had left the Buildings.
Since then, since the whole-hearted and selfless manner in which she had helped him, and had comforted him after Effie's death -- he remembered her saying gently to him when he had railed against Fate and his own incompetence, 'Don't, Dr Neil, don't.
I know that you did your best, and you said yourself that the poor thing had little hope of surviving the birth'-- his attitude to her had changed.
The raillery with which he had spoken to her in her early days at Vetch Street had changed its nature.
From being a little bitter it had become increasingly affectionate, as Matey had shrewdly noted.
He had adopted his slighting manner, he knew, to protect himself from the attraction which she had possessed for him from the first moment that he had seen her.
He could not allow himself to fall in love with a girl so obviously a part of the world of wealth and consequence which he had abandoned.
It was foolish.
Like Angela she would not want a poor doctor, and sooner or later would tire of her play-acting and retire back to that world -- he wondered why she had left it, and what she thought that she was doing here in the East End, so far from her family and friends.
At first he had thought that her ebullient nature, the evidence of her having been loved and indulged, would mean that her response to life in and around Vetch Street would inevitably be shallow.
But, as day after day she tirelessly pursued her harsh and relentless duties, never complaining, other than by the occasional wry joke about following such a demanding routine, and then, in this latest episode, showing such selfless devotion and practical compassion, his judgement of her had changed.
She seemed to be maturing before his eyes.
Her sturdy common sense, the downright attitude to life which never ceased to surprise him, the constant loving references to a papa and a mama who sounded remarkably practical themselves, even if they had spoiled their beautiful daughter, informed him that she came from a background very unlike any that Dr Neil had ever encountered.
Perhaps it was because she was American, but all this supposition failed before the one fact that mattered to him: he had come to love her passionately, wanted to fulfil that love, and he did not know what to do about it!
He remembered with affection how she had refused to allow him to send for a cab to take her back to Vetch Street.
'No,' she had said, with her usual fiery spirit still dominant, even after the long and tiring hours which she had worked with him.
'You always walk home, I know, after your work is over, and I shall do the same.'
She had laughed, looked at him with some of her usual mischief on her face, erasing for a moment the memory of what had so recently passed.
'You seem to forget that I am your skivvy, Dr Neil, not a fine lady to be cosseted.'
Her temperament had not changed, that was plain, but her moral nature had, as his had changed after he had discovered the wretched and hopeless poverty of the homes from which the soldiers who had served him had come.
After that he could never feel the same again about his own fortunate life.
He was sure that McAllister was undergoing the same experience.
She held the doll up, her work with it finished.
'There,' she said gaily.
'Salute the Master, Belinda, although I warn you that he will not be your master for long.
You will find a new one at the bazaar, I hope.'
He put out a hand to take the doll from her, to admire her exquisite handiwork, and when, in the doing, her hand touched his, a shiver of something which she had never thought to feel again passed over her.
That Dr Neil was similarly affected was plain to her.
And her pleasure was mixed with fear -- but less so than she had expected.
'A most convenient baby, this,' he remarked lightly, 'seeing that she never cries, or does anything to provoke her parents -- whatever they do to her.'
He was trying to exorcise the wave of desire which had swept over him when McAllister's blue-black eyes were fixed so earnestly on him.
He was so conscious of her that everything about her was vividly present to him.
He almost felt, as well as saw, her curls, which, as usual, had escaped from the bonds which confined them, clustering on her hairline, falling on to the beautiful oval face, the lips, so rosy and tempting against the pure complexion of her face.
When she laughed he saw her teeth, not only shining tributes to American dentistry, but also to her remarkable constitution.
Professionally, he had seldom seen anyone in such a perfect state of health -- that alone was sufficient to mark her out in Vetch Street.
His passion, his arousal which the mere sight of her had provoked, was so strong that he wondered that she could not feel it too; it seemed to hang, a living thing, in the air between them.
Dr Neil tried to calm himself by a grave examination of the doll, as though it were one of his patients, holding the tiny wrist to take the pulse, only to see the laughter on her face, and for that to provoke him to further inward excesses.
He wanted to undo the shining blue-black hair, to kiss the parted lips, to unfasten the high collar of her cheap black dress to find the exciting treasures beneath, to reveal the splendid breasts which strained against the cloth, to stroke them, to...
Desire had become a physical pain.
The months of living in close proximity to her were taking their toll, combined as they were with his previous ascetic life.
He handed the doll back to her, avoiding the touch of her work-reddened hands, tried to say something stupendously dull, wondered wildly whether running upstairs and pouring cold water over himself might stop this unseemly conduct.
'I see that you have become a valued member of Mr Sands' ladies' sewing circle.'
'Oh,' said McAllister demurely, 'he may organise us and look in occasionally, but he does not exactly take part, you know.'
'I could hardly imagine him sewing a fine seam,' murmured Dr Neil, 'despite his ladylike proclivities.'
'That is very unkind of you,' said McAllister severely, but her severity was a mock.
'But I do know what you mean.
We had a clergyman like that at home, and Papa naughtily called him Miss Nancy.
Not to his face, you understand, and he did make up for it by giving lavishly to the church...'
'Before or after he lost his fortune?' asked Dr Neil wickedly.
'Who, Papa?' said McAllister, who had forgotten the whoppers she had told about her papa's bankruptcy, but, remembering them now, recovered herself rapidly.
'Oh, before, of course.
It was such generosity which helped to bring it about,' she added helpfully.
'No doubt,' murmured Dr Neil, who was finding that this interchange, far from dowsing lust, was fuelling it, so that he had a terrible desire to fall on his knees before his skivvy, crying, 'Be mine, McAllister, be mine, immediately,'like a hero, or perhaps a villain, in a stage melodrama.
Instead, he said prosaically, 'I thought that you were determined to avoid all such bread-and-butter occupations, McAllister.
What brought about this change of heart?'
McAllister, who had put the doll down, and was now fetching out her work basket to embroider pansies on some fine lawn dresses made for the bazaar by the aforesaid ladies, said, 'I didn't mean to become involved, you know, but Matey has been so kind to me -- when not slave-driving me, you understand -- that when she asked me to accompany her I had not the heart to refuse, and strangely, after I began to work for the bazaar, I found that it was most rewarding.'
The look that McAllister gave him when she said this was so killing that Dr Neil decided to remove temptation by removing himself.
He made for the surgery, saying, lying in his teeth, 'I have some work to do, McAllister.
You may bring me a pot of tea at four o'clock.
Please do not ply me with biscuits; I have no desire to set you a bad example by pigging myself.'
McAllister cuddled her doll when he had gone.
Ever since she had held the baby she had found herself treating the doll in the same way.
And I never liked dolls when I was a little girl, she thought with wonder.
How strange!
Mama was always a little put out by my preference for boy's play and toys, and that grew worse after she married Papa and had three boys for me to play with.
She remembered Rob, the eldest boy, named after her mama's father who had died in Arizona, saying to her one day, after she had been caught riding bareback at her uncle Orrin's ranch in Wyoming, 'What a pity you weren't a boy, Sally-Anne; we could have had such fun together.'
She had been just fifteen at the time, and had agreed with him heartily.
She had not at all looked forward to being a debutante and wearing fancy women's clothes, and not being allowed to climb trees, and accompany Papa, who always, said her mama despairingly, let her do exactly as she pleased.
And now what pleased her was loving the doll, and sewing and embroidering baby clothes, and looking after Dr Neil -- particularly looking after Dr Neil -- and what would Mama say to that?
Or Papa.
For Papa, too, had wanted her to be a proper young lady after she had reached eighteen, and, for the first time, remorse struck her, for what had happened once that magic date had been reached, and for what had happened after that, culminating in...
She clutched the doll to her heart, and told herself to think of poor dead Effie whose fate had been so much worse than her own, and that should teach her to count her blessings as her dear aunt Nella had always said.
And how strange it was that, after all these months of resolutely not thinking about her old life in the States, she should begin to remember it with such pleasure!
J.D. had told Sally-Anne Tunstall that if it was inconvenient to visit him she could always post her column to him; he had seen her growing more and more responsible as the weeks went by, and he was no longer so worried about her safety.
The previous week she had posted it on Tuesday, having written it the day after Effie's death in a white heat of sorrow and indignation.
She had called it 'Death in the Buildings', and had based it not only on the tragedy she had witnessed, but had also written of the temptation for young girls to make money by selling themselves than by working long hours for poor pay, and had followed that by writing of women's disabilities in a world where care in childbirth was minimal, and how only the kindness of humane doctors made it possible for them to have any skilled treatment at all.
She was always careful never to write anything which might betray where she lived and the identity of those whom she wrote about, while stressing that the column was based upon her experiences and verifiable fact.
Not having to visit Fleet Street meant that her afternoon and evening off could be genuinely relaxing, and she usually spent the time knitting, sewing or reading, with only the occasional foray up West.
Matey had gone out, not to return until late at night, leaving Dr Neil's tea in the larder on a plate carefully positioned under a fine gauze hood to protect it from the ubiquitous flies; he had said not to leave him supper, for he might visit the local evening market where prepared food of all kinds was on sale, from oysters to whelks; he was partial to shellfish, he said.
McAllister thought that when it was cooler, in the early evening, she might walk towards St Paul's to see the sights there, and was content to sit, her feet on a tapestry-covered tuffet, finishing off her embroidery for the bazaar; she was astonished to find how much she was looking forward to it.
Dr Neil also had his evening off from his practice on Wednesday, but after lunch he had vanished into the surgery, on the pretext that he had work to do there, but in reality to escape McAllister's siren call, which had unconsciously grown stronger over the previous week.
He thought that he had heard the house door close: McAllister must have left for a stroll, or perhaps even a visit up West, and it would be safe for him to leave the surgery where he had been reading Mr H. G. Wells's scientific romance The Time Machine , and return to the comfort of his armchair.
He entered the parlour, only to discover that he was not alone, after all.
McAllister had gone out, but only to speak to Rose, who had managed to obtain a new post as a maid-of-all-work, and, her half-day off also being Wednesday, had come to ask McAllister to go up West with her again.
McAllister had apologised to Rose for refusing her invitation -- she had sewing to do for the bazaar, she said -- when the front door had banged shut behind her and she had been compelled to run round to the back and come in through the kitchen.
Breathless, she entered the parlour, to see Dr Neil arrive, his novel in his hand.
They had not been alone together since their conversation on the day on which she had finished dressing the doll, both of them avoiding disturbing contact with the other, for their own good reasons.
Hearts thumping, they confronted one another.
'I thought you'd gone out, McAllister,' said Dr Neil inanely.
'Oh, is that why you thought it safe to come back to the parlour after your arduous work in the surgery?' said McAllister dangerously, casting a sarcastic eye at the half-read novel in his hand.
'Now, now, McAllister,' said Dr Neil unwisely.
'It's not for the servant to reprimand the Master.'
'It is, if the Master is telling whoppers,' replied McAllister grimly, her heart bumping with a combination of fear and excitement, as she continued to be provocative.
And why was that?
Was it annoyance that he had been avoiding her, as she had been avoiding him?
'I was not telling 'whoppers',' said Dr Neil with dignity.
'What an unpleasant word.
American, I suppose?'
'Oh, you suppose, do you?' sparked McAllister.
'And what is wrong with American words?
Most of them were good British words first, and you haven't had the sense to keep the best ones, but have left it to us to do that for you.'
'And when do you start singing 'Yankee Doodle', McAllister?' said Dr Neil, provocative in his turn, but oh, she looked remarkable when she was all fired up, as he supposed she called it, and despite his resolution to have as little to do with her as possible he could not resist continuing to tease her.
'Better than singing boastful tunes like 'Rule Britannia' as though you Brits weren't beaten hollow by us poor rebels,'was her riposte to that.
'Only because the French were obliging enough to do a lot of your fighting for you,' said Dr Neil, who as an old soldier had a grasp of military history.
'Oh, pooh to that,' said McAllister, wondering how this ridiculous conversation had started, and how it would end.
'Trust a Britisher to blame someone else for their own lack of success.'
'Wellington and Marlborough to you, miss,' said Dr Neil, his grin growing wider.
'And what are they, Dr Neil?
Railway stations or public schools?
Let me remind you that we Yankees have won all our wars, which is more than you can say.'
'Only because you Americans insist on fighting among yourselves,' said Dr Neil, and began to whistle 'Dixie', 'which means you can never lose...so there!'
This came out so comically that they both began to laugh together, Neil putting his book down, and McAllister whooping into the apron which she had thrown over her head at his last sally, as though she were truly the skivvy she pretended to be.
She dropped the apron to show him her laughing, scarlet face, tears of amusement in her eyes.
They were by now close together, and the devil, who had got into both of them, whispered in both their ears; McAllister had meant to ask, Now, how did we arrive at that? and Dr Neil, who had meant to say something provocative to her, decided to do something provocative instead.
He took her laughing face by the chin, tipped it up, bent and kissed it.
He decided that he liked what he had done -- and did it again.
McAllister, still beset by that strange mixture of desire and fear which gripped her whenever he began to make love to her at even the lightest level, found herself shivering, but allowed him to continue, and when the next kiss found her lips she responded to him, timidly, it was true, but still a response.
Desire, long reined in, had Neil Cochrane in its grip.
The longed-for mistress of his heart was in his arms at last, soft and apparently willing, and if her lovemaking was as fiery as the rest of her, what delights might he not taste?
His arms around her, he began so gently that although McAllister was already feeling stifled, and the fear of men which had beset her for so long had begun to tighten its grip on her, she not only allowed him to kiss and fondle her face and neck, but let him undo her hair, so that it tumbled about her shoulders, as magnificent in its abandon as he had imagined it in the long nights when he had been unable to sleep.
She tried to tell herself not to be afraid, for was not this Dr Neil, whose gentleness and compassion she had often witnessed, even if his tongue was sharp?
McAllister knew that she must truly love him, for that could be the only reason why she could allow him, a man, to touch her at all.
And oh, she wanted to touch him, but a strange timidity, so unlike her, was controlling all her actions.
Dr Neil was not in any way surprised at McAllister's reaction to his lovemaking; he expected such modesty from a well brought up young girl, and her arms around his neck, her timid responses, fluttering though they were, told him that she felt for him what he felt for her, and further inflamed his own passion, while warning him to go gently.
But going gently was difficult, to say the least, for his active mind was giving way to the essential basic desire for a man to possess the woman he loved and who loved him, and his kisses and caresses grew more and more urgent.
For McAllister it was torture.
The war between desire and fear was consuming her utterly.
She willed him to continue, but instinctively needed him to stop.
Oh, this could not be happening -- that she loved him so, and yet could not bear him to caress her!
Her trembling gave him the impression of her own passion being fuelled, which in its turn led him on to further, and deeper caresses.
And then, his right hand, rising to undo the buttons of her high-collared black dress, his other arm unconsciously straining her to him, closer and closer so that she could feel his arousal brought on a memory so dreadful to McAllister, a memory which she had fought against for months -- and fear suddenly won the battle.
She was stifling, she was being crushed, overwhelmed, and the hard man's body against her was not Neil's but that of the man who had...who...had...and panic ensued, all-enveloping panic, uncontrollable.
McAllister began to scream, to fight him off, shrieking, 'No, no!
You are not to!
Don't!
You are hurting me!
No-o-o!
Stop, please stop!'
Dr Neil hardly knew what was happening.
One moment he had McAllister in his arms, soft and willing, making relatively innocent love to her, and while her response had been gentle, rather than passionate -- which was to be expected -- there had been nothing to show that she had not welcomed it.
And suddenly he had a  maenad in his arms, fighting him, screaming, striking at him, the tears running down her face which fear had made unrecognisable, pushing him away, and when he let her go, for very decency's sake, because he saw no way to calm her unless he did, she sank on to the sofa, still sobbing and crying, her face hidden in a cushion, her whole body heaving and shaking.
'McAllister, don't be frightened.
Dear God, McAllister, I shan't hurt you.
What is it?
What's wrong?'
For, after all, he had done nothing to bring on such an astonishing reaction, for behaving as though he were trying to rape her...
Rape her...!
He was suddenly a doctor before all else, remembering other women he had seen, and tried to examine, victims of a brutal assault, who had been unable to bear a man near them, let alone a man attempting to make love to them, and a score of odd things about her behaviour, hardly noticed at the time, fell into place.
'McAllister,' he said hoarsely, falling on his knees beside her, trying not to touch her, for that might bring on her violent struggles again.
'Tell me what's wrong; I promise not to hurt you,' he repeated.
'It's Dr Neil, McAllister; you know that you can trust me -- I...love you, McAllister, truly love you and want to help you.'
It was finally out, the declaration which he had never thought to make to any woman, after Angela, but the huddled figure on the sofa was not placated.
'No, you can't love me.'
Her voice, muffled by the cushion in which she had hidden her face, was so despairing that it almost broke his heart.
'You can't love me.
I want to love you back...and I can't, I can't, I'm so afraid...and so ashamed.'
She looked up at him, her face so blotched and swollen that despite himself he put out a hand to comfort her, only for her to let out a frightened cry, trying frantically to burrow into the sofa to avoid any contact with him.
All the months which Sally-Anne Tunstall had spent trying to erase the memory of what had happened to her, her refusal to remember any part of it, were as nothing.
To be once again in the arms of a man, even one whom she had welcomed, not one who was attacking her, and whom she hated, had undone the self-control which she had so rigorously exercised.
Immediately after it had happened just the sight of any man, even kind Uncle Orrin, had been enough to make her feel faint, and nauseous.
She had conquered that by an effort of the will which had once been almost her enemy, so strong was it, but had, in enabling her to live calmly among so many enemies in male form, become her friend.
She had almost thought that she had recovered completely, until, once in Neil's arms, the memory of the rape inflicted on her had struck her down.
She had thought, wrongly, that she could allow the love which she had begun to feel for Dr Neil to express itself, only for her deepest instincts to take over when in beginning to make love to her he had unwittingly reminded her so strongly of what...he had done to her.
Even now she could hardly bear to remember him, what he had done to her...his very name.
Dr Neil was speaking again.
'What was it, McAllister?' he said, his voice hardly recognisable, love mixed with anger at the wretch who had violated her.
'Were you attacked...?'
He hesitated, barely able to say it; the word was a violation in itself.
'Raped?'
She could not speak; her throat was in spasm.
She nodded, swallowed, began to whisper yes, could not, would not say it, said instead, 'Oh, I am so ashamed.
I am soiled...and I...when you touch me...
I cannot...oh, I cannot...
I wish I were dead.
I cannot make myself...and who would want such a...thing...?'
And the sobbing began again, and to see her so defeated -- remembering how gallant she had always been, even in the alley when she had attacked her assailant with her shoe -- overwhelmed him.
Realisation struck him.
She had cared nothing for the man who was attacking her then; her despair now was because she could not bear the caresses of the man whom she loved, and she knew loved her.
Oh, how doubly difficult it was for them both.
It had been tricky enough for him to help the women to whom he had no commitment even though he felt the deepest pity for them, but to help McAllister, whom he loved so dearly -- how was he to do that?
'Oh, McAllister,' he said softly.
'You are not soiled.
And the man who attacked you should feel shame, not you.
You are hurt, I know, and unable to believe in any man's decency, but try to believe that whatever happened to you I still love you; you have not changed for me, it makes no difference, no difference whatsoever -- except that I want to kill the man who hurt you.'
No difference?
How could he believe that?
How could she believe what he was saying?
Of course it made a difference.
She could no longer bring to the man she loved her untouched innocence, and, worse than that, she could not bear his lovemaking because it reminded her so bitterly of what had been done to her.
He would not want a spoiled thing!
And now she was frightened of him -- Dr Neil, who was so kind and good, and had already been so badly damaged by life...
She tried to control herself, but the will which had driven her for so long was broken, whether forever, or temporarily, she did not know.
Dr Neil rose.
Useless at the moment to try to touch her, to convince her that what he felt for her was love and a pity so enormous that it was almost unmanning him.
Like McAllister he was discovering that it was easy to bear the grief and pain of others when you were not emotionally committed to them.
He had felt distress and pity for the sufferings he had seen about him, but they had not struck home to his heart as seeing McAllister brought so low had done.
He must let her cry herself out.
He suspected that perhaps McAllister, always so gallant in facing life, as he had now seen on several occasions, had tried to suppress, to crush down the awful memory of what had been done to her, had refused to give way to grief, to shed healing tears, until in his arms something had reminded her so strongly of what had passed and broken the barriers her will had erected.
And once broken, once the grief and tears had begun, it would take some little time for her to recover, although now that she had faced what had happened she might do so more quickly and easily than she would have done had she continued suppressing her memories, and refusing healthy grieving.
'Let me fetch you some water,' he said gently to the distressed figure on the sofa after a little time had passed.
She had stopped crying and lay there rigid, her face averted from him, her whole body rejecting him and the cruel world.
He left her to go to the kitchen, began to run the tap, then heard her rise, to dash from the parlour -- not, he hoped, to leave the house; but, going after her, he saw her mounting the stairs to her room, to be away from him, simply because he was, for the moment at least, no longer Dr Neil, but only another member of the sex which had betrayed her.
He drank the water himself, the puzzle of McAllister half-resolved, for there were still some pieces missing to explain the mystery of her presence in Vetch Street.
CHAPTER TWELVE
Sally-Anne Tunstall, beloved daughter of Senator Jared Tunstall, arguably the richest man in the USA, and his dear wife Mary, niece of Orrin Tunstall, the American ambassador, society beauty, heiress, spoiled child of fortune, who had once thought that the world was her ball to play with, sat on her bed in an East End attic, dressed in her skivvy's clothing, grieving because she could not consummate her love for a poor doctor who had renounced the world over which she had once reigned.
For the first time in her life she faced the truth of her past, clear-eyed as her formidable father's daughter ought to be, devoid of pride and love of self, and inspected what had brought her to this pass.
One thing, and one thing only, she knew beyond a doubt.
Unless she could face that past, she could not bury it.
She had evaded it long enough...
'So, here I am, Uncle Orrin,' she had said gaily, 'your naughty niece, exiled because she was silly enough to fall in love with a poor man who was only interested in her money.
How many of the men whom I meet in London society will be after that same money, do you think?
Only, they will have something to offer for that money, will they not?
Titles and lands, and a position in the world.
They will buy me for that.
Terry Rourke sold me for enough dollars to buy himself and his true girlfriend a small business.'
Her bitter disillusionment was plain in her voice, her uncle thought.
They were in his study -- his office, he called it, American-style.
Now in his early sixties, he was still a big, handsome man, his shrewd common sense a byword.
He was not as clever as his half-brother, Sally-Anne's father, but he shared the same intensely practical outlook on life.
'Now, my dear,' he said, 'no need to take on so; you are young.
You have the sense, I am sure, to avoid those who wish to exploit you...'
'Especially after my recent unfortunate experience,' she remembered saying lightly, trying to disguise her profound hurt.
'Tell me, Uncle Orrin, shall I always be regarded as my father's heiress, never seen for what I truly am?'
He answered her in his usual calm manner, which was easy for him, thought Sally-Anne, for was he not one of the lords of the world and a man to boot?
What did he know of the pains and penalties of being a female?
But she responded to the kind note in his voice when he said, 'Now, my dear, you are a clever girl and a brave one.
Life is not easy for any of us, and despite what people think can be hard for those of us who have the luck to be well endowed with this world's goods.
Use the common sense which I know you are blessed with, and little can go wrong.'
Well, she had shrugged her shoulders metaphorically at that, and oh, how she wished that she had heeded him, instead of going on what she now saw clearly had been her wilful way.
Who had she been trying to impress?
Her mother and father, or the young man who had made her his easy prey?
Terry Rourke had been her father's chauffeur, a handsome, black-haired Irishman, with a lively, impudent face, quite unlike Dr Neil's.
She thought that he had put her off handsome, impudent men for life.
How it had happened she did not know, but back there in New York he had begun to pursue her, and she had found ever-increasing opportunities to talk to him; he was so gay and lively, quite unlike the rather stiff young men whom she met in New York and Newport society.
He had taken her out one day, her and Mama, and when he had handed her out of the big Daimler, her papa's pride, he had slipped a note into her hand, inviting her to meet him when his duties were over, and go out with him -- perhaps for a ride on the Brooklyn Ferry.
Sally-Anne loved a dare, and Terry Rourke's appeal had been frank and animal -- he had excited her, and it had been simple for her to lie to her mother, to set out to go to a girlfriend's home on the following afternoon and meet Terry instead.
'I never thought that you'd have the guts to come,' he had said.
He was well dressed in imitation of those whom he served, was clever and determined to make his way in life: he would not be Jared Tunstall's chauffeur forever, that was for sure.
He had already decided not to seduce Sally-Anne -- that would be too dangerous -- but to manoeuvre her and her father into a position where he could be bought off.
His plan was simple.
He would persuade her that he loved her, ask her to promise to run away with him if her father refused his consent to their marriage, and, when he finally asked her father for her hand, threaten him with the prospect of his daughter's elopement -- he and Jared Tunstall both knew how self-willed his daughter was, and that any threat from her would not be idle.
On the other hand, for a reasonable sum, he, Terry Rourke, would consent to be bought off.
He was shrewd enough to know that to be too greedy with Jared Tunstall might lose him everything.
Sally-Anne had known nothing of this -- only that standing with him on the ferry that day was sheer enchantment.
'Freedom,' she had said, shouting into the wind, glad to be rid of chaperons, lady's maids, and, yes, fond mamas and papas.
Together they had done all the things that she had wanted to do and never been allowed to: gone to Coney Island, licked lollipops on the front there, enjoyed herself in funfairs, and ridden on the Big Wheel.
He had always been the gentleman with her, and if, at times, he had regretted what he would ultimately do to her, because he had begun to care for her more than his plan demanded, he could not let that deter him.
Jared Tunstall would, he knew, ruin an employee who stole his daughter away from him.
But, 'I love you, Sally-Anne,' he had said one day, giving her the lightest of butterfly kisses.
They had wandered into Central Park, and Sally-Anne had pretended that they were simply another happy pair of young lovers.
'Are you prepared to risk everything for us?'
She had looked deep into his black eyes; for six weeks she had enjoyed, in the brief times they had spent together, the happiest and most carefree days of her life, surreptitiously stolen, with the help of a willing friend, from her guarded life, the life of a rich heiress.
She had lifted his hand and kissed it.
'You know I am, Terry.'
'And, if I asked your father for your hand, you would support me?
Say that if he refused to allow us to marry you would even run away with me?
You know that he will be angry with me, of course, but he will be angry with you a little .'
Sally-Anne had lifted her head proudly.
'Oh, Terry, what does that matter?
If he does not approve, why, I am ready to leave with you.
So long as I love you, I wouldn't mind being poor, and I don't want his money.'
Terry Rourke had thought cynically that Sally-Anne Tunstall had no idea what being poor meant, and he wanted Jared Tunstall's money, or some of it, at least.
'That's the spirit, my darling,' he had said, 'so long as you are prepared for storms...'
Her trust in him had been absolute.
But it had all been a trick, a dreadful trick.
He had gone to her father, and what happened she never quite knew.
Afterwards Jared Tunstall had said to her mama, 'It was a question of a little hustler meeting a big one, and if it weren't that he has hurt Sally-Anne so badly I could almost admire the swine for his gall.
He knew that I would never let him marry her, and, besides, he has a girlfriend of his own, so when I offered him money to go away and leave her alone he stood me off as much as he dared, and collected a big enough stake to marry his girl and start a small business, which, he kindly informed me when he left, he intends to turn into a large one, one day.'
When her papa told Sally-Anne of Terry Rourke's betrayal -- she never saw him again -- she had not believed him, at first.
Finally, 'But he said that he loved me,' she had wailed.
'He loved my dollars more,' Jared Tunstall had replied, hard for once on the daughter he had always indulged.
'And you lied to me and to your mama and poor aunt Nella again and again in order to be with him.
I thought that you had more sense, Sally-Anne.
Couldn't you see how flash he was?'
'Oh, you're so cynical,' she had flung at him, head up; she was never down for long.
'And I shall never forgive you for tempting him away from me.
He loved me and I loved him.'
Jared Tunstall had not meant to tell his daughter of Rourke's sweetheart, but, seeing her obduracy, he had done so.
'I don't believe you,' she said, white to the lips.
'Believe me,' said her papa.
'Time you grew up, Sally-Anne.
Your mother is right -- I've spoiled you.'
And that had been that.
True love had never existed, except in her imagination.
But she would not be put down; she would not.
She had hated the pity she thought that she saw in her mama and papa's eyes, and when her papa had suggested that she might like to spend a year in London with her Uncle Orrin, be presented at court, meet his old friend the Prince of Wales, now King Edward VII, again, she had agreed with alacrity -- she, who had hated being parted from her mama and papa.
She would show them.
She would, she hoped, find her true love there.
There must be many in London society so rich that her inheritance was not a temptation to them; it could not be made up only of rogues and fortune-hunters.
The American Princess, for that was her nickname, was an instant and stunning success.
She was rich, beautiful, clever and amusing, and her wit became renowned; she was featured in the picture papers, in the scandal sheets, and everyone of consequence was reported at one time or another to be going to marry her.
And if there was something wild and desperate about her London career which disturbed her uncle and aunt a little there were few to criticise.
Everyone from the King and Queen down seemed to love her.
When she appeared at a ball people stood on chairs to see her.
It was Lillie Langtry all over again, the old ones said wisely.
She was introduced to the novelist Mr Henry James, and took tea with him at his home, Lamb House, in Rye, and he thought that she was all the beautiful American girls he had written about rolled into one!
And then she had met Havvie Blaine, or, more properly, Havelock Torquil Roderick Blaine, Marquess of Blaine, heir to the Twelfth Duke of Innescourt, the most handsome man in society, the most run after by every girl from the age of sixteen upwards, and most of the married women, too, said the cynics.
Who else but the American Princess was good enough for Havvie?
From the moment that they were introduced it was plain that they were meant for one another.
Sally-Anne could not believe her luck!
Love, the oldest title in England, after the Crown's, of course, and all for her.
She lived in a dream of love from the moment that he had bowed over her hand when she had been introduced to him by Daisy Warwick at Warwick Castle, and murmured, 'Not an American Princess, surely, but a true one,' and later that evening, dancing with her for the third time, he had said into her ear, 'And can a princess stoop to a mere marquess?'
'An American Princess can stoop to anything,' she had replied gaily, 'because we don't need to stoop.
We are all equal, you know, in the land of opportunity.'
Well, they had said she was witty, and Havvie frequently got the benefit of the wit -- as well as the beauty -- and, of course, the future benefit, if all went well, of her money.
For what no one knew was how strapped for cash the Blaines were.
If necessary, Havvie Blaine would have married a one-eyed dwarf to restore their fortunes, but luck was with him -- he had Sally-Anne Tunstall in his sights, and the hope that she and her papa might not enquire too carefully into the Innescourt means, and if they did, why, transforming the daughter of a vulgar American robber-baron into Great Britain's premier duchess was surely worth a dollar or two.
Sally-Anne had no idea that Havvie Blaine was Terry Rourke all over again, and even her uncle Orrin was deceived by Havvie's name, his charm, and the feeling which he seemed to possess for his niece.
So that when, one evening, Sally-Anne, radiant in pink and silver, a fortune in pearls around her neck, the mere sight of which made Havvie salivate internally, was gently led by him into a conservatory -- at the Keppels' this time -- and proposed to, there was only one answer which she could give him, and that, of course, was yes.
'Yes,' she breathed, 'oh, Havvie, yes,'and if, when he kissed her, with the perfect decorum with which a well brought up young peer should treat a single girl -- even one who had promised to be his wife -- she did not feel quite the surge of passion which she had expected, she put that down to her inexperience, and his tentative handling of her, which would change with time, she knew.
Oh, they made the handsomest of couples, Havvie and his American Princess -- the latest Yankee heiress, unkind gossip said, to buy herself a title and a place in English society.
Her manners, as well as her looks, were rather better than most, it was agreed -- which was fortunate for the Blaines.
It was to be the wedding of the year.
Papa and Mama, informed, were coming over, the lawyers had gathered to draw up the marriage settlement, everything seemed set fair.
Except, even before disaster struck, Sally-Anne had begun to awaken from her dream of bliss.
After all, before he had proposed, she had not seen so very much of Havvie, and she had been so flattered that it had not struck her how banal his conversation was, and how limited.
And if she had heard nothing of the gossip about his private life before she accepted him, certain ladies he had discarded, both married and single, took care that she overheard quite a lot now.
Even then, all might have been well.
Wearing Havvie's ring, a diamond heirloom given to all Blaine brides, laughing, talking and playing with him as the London season began, Sally-Anne persuaded herself that she was as happy as a girl could be, Terry Rourke's betrayal wiped out and forgotten.
She would see Mama and Papa again soon, they were coming over well before the wedding, and what bliss to greet them as the future Marchioness of Blaine, beautiful blond Havvie by her side, eager to meet them -- or so he publicly said -- his private comments were somewhat different -- more to the effect that he could swallow Sally-Anne and her dollars, but her parvenu papa was quite another thing!
So when, on that fateful afternoon, enchanting in pale amethyst, unpinning her huge straw picture hat, crowned with a bouquet of early spring flowers, she was told that she had a visitor waiting to see her --'A Mrs Greville.
Most insistent, Miss Tunstall,' the English butler said.
'She was adamant that she would wait for you, however long you were in returning'-- she thought nothing of it.
She and Havvie had been to a flower show.
'And you are the fairest flower of all,' he had said, and presented her with a spray of pink carnations.
Mrs Greville sat quite alone in her aunt's drawing-room.
Her aunt and uncle were out at some official function.
She rose as Sally-Anne entered.
She was tall, in her early thirties, and had once, it was plain, been a great beauty; but time, and grief, Sally-Anne thought perceptively as she moved to greet her, had worn her down.
'You do not know me, Miss Tunstall,' she began abruptly.
'And I must thank you for consenting to see me.
What I have to say, though, may not, I fear, please you.'
This bewildered Sally-Anne, but for once her intuition was acute.
Dulled since meeting Terry Rourke, it had inconveniently come to life again, and was telling her things about life -- and Havvie -- which she did not wish to know.
'But you feel that it is necessary to speak to me,' she said gently, wondering why this woman, whom she did not know, wished to say unpleasant things to her -- but of what?
'I must,' said Mrs Greville.
'It is my duty, to myself, and to my children.
You know that I am Mrs Honoria Greville?'
'So the butler said.'
'You have not heard of me from any other person?'
'No,' said Sally-Anne, still bewildered.
'No?
Then I must inform you.
I have been Havelock Blaine's mistress since he was turned twenty-one, nine years ago.
I was married then, but I had left my husband because of his cruelty.
Havvie and I are exactly of an age.
We could not marry, but he has kept me in comfort ever since we became lovers and he fathered my two children.
He...promised me marriage, should my husband die.
Six months ago my husband broke his neck on the hunting field.
Havvie did not offer me marriage, after all; he offered it to you, instead.
I ask you to break that pledge you made with him so that he may marry me, the mother of his children.'
Sally-Anne, suddenly pale, stared at the woman before her.
She had known that, beyond a doubt, Havvie must have formed liaisons, had mistresses -- few young men of his rank and class arrived at the age of thirty innocent.
What she could not have thought of was this sort of permanent liaison, almost a marriage.
'Oh, I am so sorry,' she said.
'But...'
'But what?' exclaimed Honoria Greville proudly.
'I have every prior claim; even you must see that.'
Sally-Anne did not know what to say, what to think.
That Havvie could so lightly do this...
'But he asked me to marry him,' she said feebly.
'He loves me now.'
And, saying it, thought, No!
And can I want him, a man who can do this?
'Loves you?'
Mrs Greville began to laugh.
'Oh, no.
He does not love you , believe me.
He loves your dollars.
You see, he does not intend to cast me off, not at all.
We are, he says, to continue as we are, only your money will make my life more comfortable, he says.
But it is not enough, Miss Tunstall.
I have, in effect, been his wife for nine years, and now I wish to be his wife in truth.
But no, he does not love you.'
'I don't believe you,' said Sally-Anne, white to the lips now.
Whether she wanted Havvie or not after this revelation was one thing; to learn that he wanted her only for her money was quite another thing.
'Believe me,' said Mrs Greville, and there was almost pity in her voice.
She had opened her handbag, rummaged in it, produced a letter.
'Here, read this, see what it says.'
And she handed the letter to Sally-Anne.
'No,' said Sally-Anne, trying to hand it back.
'No?' said Mrs Greville, her lip curling.
'Are you afraid to read what the man who loves you says about you?'
To call a Tunstall afraid was almost, as Sally-Anne already knew, enough to raise the devil.
She opened the letter, drew out a blotted sheet of paper -- Havvie, she was well aware, was not the most fluent letter-writer in the world.
But what she read was fluent enough.
'My dearest Honoria,' it began, and then went on to enquire lovingly after her and the children.
But it was the second paragraph which did Mrs Greville's work for her.
I am really impatient with you, my dearest.
You know perfectly well that you, being penniless, and myself, being little better, cannot marry.
And you know, too, that you need have no fear of losing me when I marry.
How could she compare with you?
Like all Americans she is both greedy and gullible -- greedy for a title and gullible enough to think that I could love such a raw provincial chit as she is.
Were it not for Papa's dollars I would not look at her twice!
No, she may give me her money and my legitimate heir, and that is all.
My heart is always yours, even though the laws of God and man rule that we may never marry.
Sally-Anne thought that she was going to faint.
Havvie Blaine, for all his name and lineage, was no better than Terry Rourke.
As with Terry she meant nothing to him, except as a means of extorting money from her papa.
Sally-Anne Tunstall was a thing to be handled, used, of value only as an avenue to her papa's wealth.
'Why are you doing this?' she asked hoarsely.
'When I refuse him --' and Honoria Greville noticed the when -- no mention of an if '-- he will not marry you.'
'I think he might,' said Mrs Greville.
'No,' retorted Sally-Anne, 'not Havvie Blaine.
He will marry for money, and if not me then another.
He wishes to have it all.
The money -- and you.'
'Well, at least he will not have you,' said Honoria Greville, reading the girl before her aright.
Sally-Anne was not at all what she expected, and she was almost sorry for her.
'I think that you had better go,' Sally-Anne said -- and she remembered with pride that she had not given an inch, had shown no emotion as her life had crumbled around her.
For, whatever else, she would not marry Havvie Blaine now, and she would lose no time in informing him of that fact.
Let some other heiress finance his illicit family.
Sally-Anne Tunstall would rather drop dead at his feet than accept him after this.
But oh, she thought, tearlessly watching Mrs Greville leave, how could I have been so deceived -- and twice at that...?
CHAPTER THIRTEEN
Greedy and gullible!
Greedy and gullible!
The words rang again in Sally-Anne's head as they had done that afternoon in her aunt's pretty drawing-room once Mrs Greville had gone.
For what reason the woman had done it Sally-Anne did not know.
She could not really think that Havvie would marry her, once Sally-Anne jilted him -- which she must, for in the face of that letter could she believe a single word he had said, or written, to her?
And she could not believe that he would marry Honoria Greville, might even throw her off if he knew what she had done, which was to lose him his heiress.
She knew what she must do.
She was her father's daughter, after all.
Pride -- and honour -- demanded one thing of her.
Cost her what it might, she would break off her engagement, and if it meant social ruin, so be it.
She could not build her life on another's pain, or on Havvie's lies.
He had looked into her eyes too often, told her that he loved her too many times for her ever to believe him again.
He would marry another heiress, she knew that quite well, almost certainly one who would take him for his title, regardless of his character and conduct.
Maybelle Foy, the daughter of one of her papa's rivals, had, she knew, almost caught him before she, Sally-Anne, had arrived on the scene.
Well, she could have him and welcome to him.
Once in her suite of rooms she sat down at a little Louis Quinze escritoire, its pale grey panels painted with carnations and pinks, and wrote a short letter to her faithless lover, asking him to call on her urgently at the embassy at eleven the next morning.
She could not face him before then; nor would society approve of a private evening visit.
Then, pleading a headache and fatigue, she cancelled her outing for the evening, and went to bed.
But not to sleep.
Eleven o'clock the next morning saw her immaculately dressed in a pale grey and pale pink creation, a great froth of fine lace at her throat and at her wrists.
She did not know it, but pallor had conferred an almost grave beauty on her, so that Havvie Blaine, arriving on the minute, thought, not for the first time, that he might be selling himself for dollars, but that he was getting more than dollars in return.
He did not love her, but she would make a remarkable duchess, and the fact that she had inherited more than her share of her papa's famous shrewdness was a bonus which would assist the Blaines and make their future secure.
She received him alone, a privilege granted to her because they were so soon to be married.
'My love,' he said, 'what was so urgent that you had to summon me so soon after we parted yesterday afternoon?
And I was sorry to miss you at the Keppels's thrash; I hope that you are recovered?
You do look a little pale.'
His manners to her had always been impeccable, which made what he was shortly to do all the more shocking.
'Oh, Havvie,' Sally-Anne murmured; this was going to be even harder than she had thought -- and she had said nothing to her uncle Orrin -- she could not, not until her engagement was irrevocably over.
'I have asked you to call,' she said, 'to tell you that, after all, I cannot marry you.
It would not be fitting.
You do not love me, and I, I fear, was in love with love and not with you.
Far better that we should part now, before all the arrangements go too far, and we are bound by them, and not by our desires.'
Havvie stared at her, his handsome face as ashen as hers had been when she had read his letter.
'I cannot believe that I am hearing you aright, my darling.
What has happened?
Yesterday afternoon we parted in such love and friendship -- and now -- now -- you say that you do not wish to marry me!
Only consider what you are saying.
I am sure that you are only feeling a young girl's modest misgivings at the approach of marriage.
Let me reassure you...'
Like the Player Queen in Hamlet , thought Sally-Anne dully, he protested too much.
He flung down his soft hat and cane, crossed to where she stood, took her hand.
She rejected it, and said sorrowfully, 'Oh, Havvie, I wish that I could believe you.
But I know only too well that you do not love me, that you are marrying me for Papa's dollars, and that I cannot bear.
I thought that you loved me , and not my money.'
'But I do love you,' he said, impetuously for him; he had always been measured with her, one of his charms for a girl who tended towards impetuosity.
'How can you not believe me?
Have I not given you sufficient proofs?
I have always held back, my darling, from being too ardent, fearing to frighten you.'
'You are not telling me the truth.'
She was measured, too.
'You know perfectly well that it is not me whom you love, but Mrs Greville, and the two children who are your responsibility, and, knowing that your heart lies with them, and not with me, I cannot marry you.'
She had thought to do it in a civilised fashion, not confront him with her knowledge of what he had written about her, but he was persistent; he could see his prize and her dollars slipping away.
Many heiresses could give him dollars, few could offer him the appearance and manner of a lady -- it was that which had attracted him after the cruder charms of such as Maybelle Foy.
'How can you say such a thing?' he exclaimed.
'And who has been filling you with canards about Honoria Greville?
I own that I had a liaison with her, but you must know that all such young men as myself have an...experienced life before they marry.
But I have given her up, and in any case she has meant nothing to me for years.'
Sally-Anne closed her eyes, thought of his letter, dated that very week.
He was determined to lie to her to the very end.
But it would not serve him.
'How can you say that, Havvie?
You have no intention of giving her up, and, in honesty, I cannot urge you to do so.
You have a duty to her and to your children.
I cannot overlook that -- nor what you truly think of me -- which is very different from the soft and loving words you have used to me so often.'
'Nor what I truly think of you,' he echoed; he was no fool.
'And what is that, Sally-Anne, and how do you know that?'
He was across the room in a stride, caught her cruelly by the shoulders.
'What do I think of you, and how do you know?'
Sally-Anne was suddenly frightened.
'Let me go!' she exclaimed.
'You are hurting me, Havvie.
Suffice it that I know that your true opinion of me is scornful in the extreme.'
And before she could stop herself she came out with it.
'Greedy and gullible, I believe.
You cannot want such a person for a wife, and I certainly do not want a husband who thinks that of me -- even before marriage.'
He did not let her go, and said hoarsely, 'Honoria Greville.
I might have guessed,' and then he uttered a string of oaths, so oddly at variance with his usual smooth and civilised manner and appearance that Sally-Anne shivered and tried again to pull away from him, but he held her more tightly than ever.
'You cannot believe the word of a jealous woman, my darling,' he said.
'Reconsider, I beg of you.
Together, we can build a happy life, I am sure of it.'
'It was not Mrs Greville's word,' said Sally-Anne, goaded unwisely into further indiscretion, still trying to break free, 'it was your handwriting -- the same that told me that I was the light of your life and that you loved me dearly.'
'You are determined, then?'
'I must be, I am.
Please let me go, Havvie.
You are hurting me.'
'Irrevocably no?
You are determined to humiliate me?'
'I must; I have no choice,' said Sally-Anne.
'I cannot marry a man who secretly despises me.
I could have done that back in the States.'
'Oh, you might do anything, back in the States,' he jeered.
'You are all savages there.'
He let go of her, and stepped back.
'Say it, then, Sally-Anne Tunstall; let me know the worst.
Remember, I can make you a duchess one day.
Surely that is worth paying a little for!'
'You want me to pay too much.'
She was pulling off the beautiful antique ring.
'There, my Lord Marquess.
There is your ring.
Do you believe me now?'
'Oh, I believe you,' he said, his handsome face transformed through rage and spite into something quite different.
'I believe you.
You speak of payment, Miss Tunstall, and I shall certainly take mine now.
You owe me something for the humiliation which you are about to put on me.'
He leaned forward again, and caught her to him, savagely, all his perfect manners gone.
'Payment, indeed,' he repeated.
'And if I may not have you permanently, then a little temporary use would satisfy me, I think.'
Sally-Anne was helpless before him.
She was not a small girl, was reasonably athletic for a woman, but there was little she could do to fight off a determined man, particularly when she was neither mentally nor physically prepared for such a brutal assault.
He bore her down to the carpet, his face hard against hers, one hand pulling up her skirts, and when she still struggled struck her a blow which left her half stunned on the floor, for him to do as he would with her.
She could not believe what was happening to her.
Frustration and lust tore at him, and so he tore at her.
She tried to scream, but he put his hand across her mouth, and when, pulling herself free for a moment, she panted at him, 'Stop, Havvie, stop; you will regret this later,' he replied,
'Not I; I shall remember this with pleasure, you American whore.'
And there, on the carpet, the worst thing of all happened to Sally-Anne Tunstall, and there was nothing she could do about it, nothing.
She tried to distance herself from it all, as Papa had once said that the Indians did, but it was useless; he hurt her too much, and, what was worse, took pleasure in her pain...
And then it was over.
He stood up, face scarlet, looking at her and speaking to her as though she were the whore he had called her.
'I've had better tarts,' he said, and, walking to the door as she struggled into a sitting position, trying to restore her ruined self, for at some time she must leave the room and face her shattered world, he flung her his final insult.
'Try explaining that on your wedding night, Sally-Anne Tunstall, and tell your husband who it was who was there before him.'
He was gone.
And even Terry Rourke had not shamed and handled her so.
She wanted to vomit.
She tried to rise to walk, but her legs would not let her.
He had been careful not to tear her dress, but everything else seemed to be ruined irrevocably, not the least the inward psyche of Sally-Anne herself, which seemed to have been raped more than her body, if that were possible.
Somehow she got herself out of the room, and up the stairs before anyone appeared to speak to her, tore off her clothes, and then hid the betraying underwear in a Gladstone bag which she later threw off Waterloo Bridge after she had left the embassy, pretending to go with Laura Parslow on her European tour, but actually having hired herself out to J. D. O'Connor, and gone to the East End.
But that had all been in the future that morning.
That, and the discovery that she could not bear a man near her in the early days after the rape; so she had feigned an illness, explained as the consequence of her breaking her engagement with Havvie, which, of course, had caused an immense furore inside and outside of society, and Mama and Papa had put off their visit to England, and sent her loving letters, for she had written that she had discovered that she did not really love Havvie at all , had merely been beglamoured by his appearance, name and title.
Which explanation, she was sure, had deceived neither Mama nor Papa, for both of them had written saying that if she needed them they would come at once after all.
No, she had not wanted that, not Papa's shrewd eyes on her; she had shivered at the very thought.
Later, perhaps, when she had recovered a little.
What had really caused her, in the end, to escape from her life as an American Princess was overhearing her English maid talking with one of the senior parlourmaids.
She had thought that the woman had liked her, but the comment she had caught had been to the effect that spoiled children like Miss T. got what they deserved!
It was probably the Marquess who had jilted her, and not the other way round.
'Useless,' she had jeered, 'like all spoiled rich girls.
Never done a hand's turn in her life, and wouldn't know how!'
Sally-Anne sometimes thought that her career as a housemaid had been sparked off by that remark as much as by anything else -- that and discovering how hard life was in the East End, and her determination to write about it from the inside, rather than as a privileged outsider looking in.
And now Havvie's cruelty had wrecked her true love for her, for she was unable to bring to Dr Neil either an unspoiled body or a spirit which was prepared to accept his lovemaking.
The bitter sobbing which had consumed her immediately after her coming upstairs had stopped.
Instead slow, silent tears ran down her face, and she lay down on the bed, trying to forget, not to remember.
The truth shall set you free, but oh, the truth of her past had not done that -- it had simply shown her what a selfish, unthinking thing she had been that first Terry Rourke, and then Havvie Blaine could have exploited her so.
Self-knowledge had come too late.
And with that thought her tears stopped.
For it had come, and perhaps, but only perhaps, she could build on that.
There was, almost in response to this, a tapping at the door, the knob turned, and he spoke.
'Let me in, McAllister.
Please let me in.
I need to talk to you.'
On hearing him, desolation struck harder.
Sally-Anne sat up, the sodden sheet in her hand.
Her handkerchief had been a useless rag long ago.
She said nothing.
He spoke again.
'I shan't go away, you know.
Let me in, McAllister.
I need you.'
That moved her.
If he had said, 'You need me ,' then that would have been the end for him.
'You can't need me,' she croaked.
'Oh, but I do.'
Silence, and then, 'We need one another, and I shall sit here until you let me in.
Think how scandalous that would be, McAllister, when they find me starved to death outside your door on the top landing.
Think how dear Matey will grieve!'
This poor attempt at humour produced a broken sob, but also the reaction he wanted.
Sally-Anne had a flash of memory of him just before their abortive lovemaking, remembered how dear she had felt him, so dear that she no longer saw his scarred face, but only Dr Neil, whom she loved.
She rose, ran across the room, unlocked the door, and then swiftly retreated to the bed again -- not to lie on it, but to sit on it, huddled against the end, one shoulder against the iron railings of its head, hands primly in her lap, feet side by side, her whole posture defensive.
The door opened, and Dr Neil came in.
CHAPTER FOURTEEN
Dr Neil saw her on the bed, face swollen and unrecognisable, all her bright lustre gone.
He swallowed, and said gently, 'McAllister, may I sit beside you?
I promise not to touch you.'
Sally-Anne shuddered, but nodded.
She had recovered the sheet, and held it as a baby might, with both hands.
One thumb stole into her mouth, for comfort.
Dr Neil swallowed again as he sat down on the bed, not too near, not too far away.
He knew the symptoms she was displaying.
They were classic, and whatever had been done to McAllister had been so severe and unwanted that unless she was cured soon she would be stricken for life, all her bright spirit running to waste.
'Who hurt you, McAllister?' he enquired.
'It would help you if you could bring yourself to talk about it.'
Tell Dr Neil?
Oh, no!
She could hardly think about it herself.
How could she tell him what Havvie had done, and said?
She shook her head at him.
Humpty-Dumpty was quite broken, soiled in the dust, and not Dr Neil, nor anyone else, could ever put her together again.
'Try to tell me, McAllister.
I'm a doctor.
You were...raped...were you not?'
'Yes,' she whispered, avoiding his eyes.
It was the first time that she had ever told anyone of what had happened.
'And not too long ago.
Not long before you came here?'
The thumb had crept out of her mouth.
She looked at it, and thought contemptuously, What a baby I am, a cry-baby.
She put it in her lap, and, her voice a thread, said, 'Yes, oh, yes,' and then, 'I did nothing to provoke it, nothing...
We...were...not...loving at the time.
Not at all.
Quite the opposite.'
She stopped, and Neil waited patiently, not attempting to touch her, thank goodness.
She stole a look at his face.
He did not appear to be judging her, either.
'I was surprised.
I suppose I trusted him, and then...oh, I can't tell you, but...he hurt me so, and I am unclean, so unclean.
Who would want me now?
The only consolation I have is that he did not make me pregnant.
At least I was spared that.'
She knew that she had pushed her dreadful experience away, had refused to be broken, had tried to pretend that it had never happened.
She had held her head high, kept up her indomitable front.
She had been fiery Sally-Anne still, and then fiery McAllister, and a few moments' loving from a man she loved had undone her, had brought her not joy, but a trauma of the remembered pain and anguish which Havvie had inflicted on her.
No good man, and Dr Neil was a good man, could want her after that.
Dr Neil knew that she was speaking the truth about what had happened to her.
'Your employer?' he said gently, not probing, remembering what she had said on the first day.
He was still not sure that he believed that story, but in her misery was prepared to go along with her.
Sally-Anne nodded.
She could not tell him about Havvie; neither could she speak the lie to him, not to Dr Neil, but she could not tell him the truth, for that would mean telling him who she was, and she could not tell him that, not here, not now; it would spoil everything between them if he knew that she was the spoiled and pampered American Princess.
'McAllister,' he said, and oh, his voice was so kind.
He could not want to be kind, could not love the tarnished and ruined thing Sally-Anne Tunstall was, but she looked at him all the same.
'I love you, McAllister,' he said.
'Believe me when I say that to you.
And nothing you have told me makes any difference to that.
Believe me when I say that I would never hurt you.
Hold my hand, McAllister.
Hold my hand.'
And he put his hand out to her, warm and loving on the counterpane between them.
That wretched thumb was in her mouth again; it left her mouth and she watched the hand it belonged to inch its slow way across the counterpane towards his, almost as though it had nothing to do with her.
As it stole towards him, Dr Neil made no move, did not move himself or it, when she put her own hand on his, but let it lie lax, so that she felt no threat from him.
And so they sat, side by side, for some time, the afternoon sun gentle on them, through the attic window.
Once or twice, Sally-Anne shuddered, and then, as he made no move, she looked at him, and gave a watery smile.
'There, that's not so bad, is it?' he said, giving her the smile which always transformed his ruined face.
Sally-Anne nodded.
He might call her McAllister, but since she had fled from him she had become Sally-Anne again.
Could she trust him?
Dared she trust him, or any other man, ever again?
Her own strong and unthinking rejection of him had shocked her as much as it had shocked him.
Emboldened, Dr Neil inched towards her.
He pulled his hand gently from hers, and, still gentle, put his arm around her to hold her by the shoulder.
He felt her grow rigid at his touch, made it so light that she hardly knew that it was there.
'Trust me, McAllister,' he repeated.
Oh, she did want to trust him, but...but...
But she did not pull away, and when he moved nearer still, holding her as though she was infinitely fragile, and turned her so that she was fully in his arms, her head on his shoulder, he did it so slowly that Sally-Anne felt not fear, but reassurance.
How long they sat unmoving Sally-Anne never knew.
All she knew was that she was turning into McAllister again, Dr Neil's pert and lively maid, and that this was the haven which she had unconsciously been seeking in the months since Havvie Blaine had assaulted her.
She could feel the strong beat of his heart, the warmth of his body, smell the masculine scent of him, and still she was not frightened, for this was Dr Neil, was it not, and why should she be frightened of him?
'You see how easy it is, McAllister,' he whispered suddenly, and he kissed the top of her head, 'when it is love that lies between a man and a woman, and not hate and fear.'
The kiss had been such a butterfly of a thing that she did not reject it, and when he bent and kissed her cheek she did not reject that either.
She even thought that she might like to sleep like this, held in one strong arm.
But Neil had other ideas.
If he was to save her from a lifetime's retreat from life and love he had to move her along the right road as quickly as he dared and as quickly as she would let him.
Some gentle lovemaking on this first occasion to prove to her that she was not repulsive to him because of what had happened to her, and that all men were not brutes, was in order.
The rest could wait for another day.
He kissed her again, on the forehead this time, and whispered in her ear, 'Kiss me back, McAllister.
On the cheek, if you like.
The Master promises not to bite.'
This even brought a ghostly chuckle; Dr Neil's being the Master again was enough to bring McAllister back, she thought, if only to provoke him to further jokes.
Still she could not bring herself to touch him, until, turning his head, she saw the scar, and without thinking leaned forward to kiss that, a butterfly kiss like his, to show him that she loved him -- and his scar.
Time stood still in the attic room while they exchanged the most innocent of kisses and embraces, and McAllister faced her future, the bruises slowly disappearing from her wounded spirit under the healing power of true love.
Dr Neil -- she could not think of him as Neil -- was careful to hold her in such a way that she felt no restraint, although his own self-control was slowly beginning to slip.
Holding her in his arms, warm and loving, feeling her fear drain away, was beginning -- shamefully, he thought -- to rouse him.
Oh, this is wrong, he reminded himself fiercely, when all she needs is mild affection and the ability to trust, and if I go too far it will only be to betray her again -- or so she will think.
But oh, I love her, and if truth be told I want to prove that love in action.
He dared not pull away, even though he did not want her to feel his changing body, but at this point one of her gentle caresses touched the corner of his mouth and before he could stop himself his mouth was on hers, and restrain himself as he might it was a lover's kiss, not a friend's.
McAllister did not reject it, but answered it.
Something was beginning to happen to her; as love began to drive out fear, the tension in her body relaxed, and the feelings which she had experienced when Dr Neil had first begun to make love to her, earlier that afternoon, returned again.
His restraint, exercised for so long, paradoxically excited her, because she felt no fear of him.
So that when his mouth closed on hers she responded passionately, and a great wave of love and affection for him, and for his patience, engulfed her, and carried her away.
She raised her left hand to stroke his scarred cheek and on impulse kissed it again, saying, 'Oh, my love, my poor love; you were hurt even more badly than I was.'
On Dr Neil this had an effect even more erotic than any more intimate caressing of his body might have produced.
He possessed just enough self-control to continue his gentle stroking of her, but both his arms now held her, and almost without her willing it McAllister's crept around his neck, and unconsciously they sank, from their sitting position, on to the bed, to lie with McAllister's head on the pillow, Dr Neil's by her, he careful to lie still, to do nothing which might suggest any kind of bodily union.
They lay half entwined like this for some time, occasionally exchanging kisses.
By now Dr Neil was in an agony of desire, the loved one in his arms, so near and yet so far -- But not so far as he thought, for McAllister suddenly wanted more from him than he was giving.
She felt herself dissolving and the nearness of him, his gentleness after Havvie's remembered brutality, was having its effect on her.
She gave a little cry and turned fully towards him, inside his arms, pulled his head towards her, and began to rain kisses on him.
To return them he, too, moved, and he was now above her, and for a second she felt fear again, only for him to murmur into her neck, 'Don't be frightened, McAllister, I won't hurt you,' and he began to kiss her throat, exposed above the white blouse she was wearing, while his right hand stroked her breast through its cloth.
At what point McAllister's fears finally disappeared beneath his loving patience, and they began to make love in earnest, neither of them could have said.
He said huskily, 'Let me, McAllister,' and began to unbutton her blouse, 'I want to stroke you, McAllister, and not your clothes,'and she made no effort to stop him, and when he bent his head to kiss the breasts he had fondled with his hands the cry which she gave was one of pleasure, not fear, for now it was Dr Neil loving her so carefully that the flood of pleasure was almost on her from that alone.
No, he would not hurt her, and now she wanted to be him, to be part of him.
Her hands moved to unbutton his shirt, and, that done, she not only wanted to stroke the golden down on his body, but to see that body, to see him .
And, quickly now, the irrevocable moment was approaching, when his passion, which she could feel hard against her, would demand fulfilment, and in doing so might destroy the fragile palace of love which they were building together.
Only, both their minds, their tortuous, questing minds, the busy intellects which glorified them, but tormented them, had been consumed, if only momentarily, in the fires of passion.
McAllister no longer knew herself, nor Dr Neil either; if he demanded fulfilment, so did she; the desire to be one, not two, had overcome them both.
And as proof in the healing power of accepted love her hands undid him, as he undid her, until, naked together, he lifted himself to enter her, and if for McAllister there was a moment of trapped fear it disappeared when she was truly his, and they were, at last, one thing, moving together in harmony in an experience quite unlike the shock and terror which she had felt with Havvie, and had always feared would happen to her if she ever made love again.
Dr Neil had not meant to consummate his love for McAllister, but having begun to heal and reassure her he had started something on its way which was beyond his control, or McAllister's either.
They had lived in proximity for weeks -- nay, months -- they had shared in the coming of life and the going of death, had suffered and exulted, and finally they offered their shared love as a sacrament, he crying out her name, and she his, in the blessed joy of fulfilment.
Time started again as they lay panting, their hearts beating as one.
'Oh, forgive me, my dear love,' he said at last.
'I had not meant that, only to reassure you.
Oh, I am as bad as the man who ravished you.'
She put her hand on his mouth.
'Oh, no.
Never say that.
I wanted it too, so much, so much.
The other was violation.
It was not like loving you...not like that at all...'
She could say no more, only after a moment when he did not answer, 'Please don't reproach yourself, my darling, when you have given me back to myself again.'
And, after that, she still in his arms, they slept a little, two tortured spirits, who in finding each other had found love and peace together.
McAllister awoke to find him above her again, put her arms around his neck to draw him down to her, but he shook his head.
'Too late, I fear.
Matey will be back soon, and -- oh, what would she say to me for seducing you as I did?
We must return to the world, I fear, but first --' And he kissed her cheek, so gently that her tears flowed again, only they were tears of joy, not pain.
'First?' she whispered.
'Will you marry me, McAllister?
I must call you that.
It is McAllister who is dear to me, not Sally-Anne.
Can you bring yourself to marry a poor doctor, be his wife, and share his burdens?
It is little enough I have to offer you -- apart from my love, that is.'
Now was the time to tell him who and what she was, but she could not spoil this magic moment; it was McAllister he loved, not Sally-Anne Tunstall, and she would not ruin her perfect afternoon.
Later, later, she would tell him -- of the embassy, Papa and J.D. -- but for the moment it was enough for her to be the wounded girl whom he had healed.
'You have offered me more than anyone has ever done before,' she said.
'For it is McAllister the poor servant whom you love, and that is sweeter to me than any fortune you can offer me, Dr Neil.
Of course I will marry you.
I want to share in your work, not live in idleness.'
Dr Neil was pulling on his clothes, hiding his scarred body, the scars which she had embraced in their mutual passion.
'You?
Idle?' he said, thinking of the tireless girl she had been.
'I mocked you one night, my darling, for working so hard, remember.
I thought that it would soon pass, and it did -- for you to work harder still.
Yes, you may share my life, and live a little more easily, I hope.'
He kissed her again, almost shyly, as though they had not so recently been as intimate and close as a man and woman could be.
'Come, we must dress.
There will be other times, my love.'
CHAPTER FIFTEEN
'Not with the Parslows?' said the American ambassador incredulously.
'You are telling me, Warrender, that my niece is not travelling with the Parslows?
Why, I know for a fact that she left here to go with them, and half a dozen trunks went with her.'
'Nevertheless,' said Jordan Warrender, one of his senior counsellors who had recently been travelling in Europe, ostensibly on holiday, actually on a tricky diplomatic mission, but who had looked up the Parslows on finding them in Venice, 'she is most definitely not with them.
Cried off at the very last moment, Mrs Parslow said.
Fortunately, and by pure chance, I had said nothing that could lead her to assume that we thought that she was travelling with them -- as arranged.'
Orrin Tunstall reverted to being a Westerner, as he often did in moments of crisis.
His smooth speech was saved for English society and the Court of St James.
'Godammit, feller, where in Hades is she?
She's been sending me postcards -- how's she managed that, hey?'
Jordan Warrender sighed.
He had had a few hopes for Sally-Anne Tunstall himself, had been sorry to see her go.
He had read her banal cards with more interest than they deserved.
'Oh, I bearded Miss Laura, on the quiet,' he hastened to add.
'I know you want no scandal.
She confessed that Sally-Anne had written them on old ones Miss Laura had collected on earlier trips, and Miss Laura had posted them for her, to deceive us.
No,' he forestalled his superior, 'she has no idea where Miss Tunstall is.
Talked vaguely about work -- she didn't know what -- that Miss Tunstall had hinted she was going to do.'
'Work!' grunted Orrin, unkind for once about the niece he loved.
'That, Warrender, will be the day.
She's done nothing but play for the last two years.
I thought better of her once.'
Warrender proved his mastery of diplomacy by looking enigmatic as the ambassador continued to fulminate.
'Good God, feller, she could be anywhere, and a regular Babe in the Wood at that.
Caught by two swindlers already; what's she doin', eh?
Tryin' to find a third?
Yore duty, feller, is to find her, and discreetly.
If not it'll be for me to tell her lovin' papa that his baby daughter is lost, and what's the bettin' he'll be over here on the fastest boat runnin', to cut my throat for me if anything's happened to her?'
A fortnight after Orrin Tunstall's unwelcome discovery of his niece's absence, Sally-Anne sat writing her journal at midnight.
She had just finished an article about pawnshops, the meat of which had been supplied by Rose, who had come round again and invited Sally-Anne to tea at the Buildings.
'To see the baby,' had been the bait, and so she had had tea with the Baileys and the baby, whom Mrs Bailey had adopted -- a piece of altruism which staggered Sally-Anne, who thought that Mrs Bailey had enough to do on too little with her own brood.
'Couldn't let it go to the Home,' Rose had said briefly.
'Like to die there --' As Dr Neil had agreed, after the funeral of the little boy's mother, attended by Sally-Anne, Matey and Dr Neil.
'What's one more mouf to feed?'
Common sense said otherwise, but the unheeding kindness of the very poor took little account of that, Sally-Anne had found.
Now she was writing in her journal that she would tell Dr Neil who and what she was on the coming Saturday, after the bazaar was over.
'For,' she put down, 'I want to go there as humble McAllister who keeps the house clean, dressed the doll and made and embroidered the baby clothes, not as the American Princess doing the gracious, ladling our Papa's money, not one penny of which I have earned, smiling and patronising those not so fortunate as myself, and then leaving them to go home to the embassy -- to forget them!'
When Matey had returned on that fateful afternoon, Dr Neil had told her that he had proposed to McAllister, and that she had accepted him.
Earlier he had tried to hint to McAllister that she ought, perhaps, to inform her family, but she had been deliberately obtuse.
She wanted so much to remain with him on the terms which had always existed between them, did not want the wealth and consequence of her real life to come between them before it needed to.
Matey was not surprised by their news, was pleased by it, in fact.
McAllister's influence on her treasure had been so good, he was quite transformed.
'And I am still your maid,' McAllister had said gaily to him, but, of course, she was now less and less of a maid and more a member of the family, working side by side with Matey in cheerful equality, living and playing with them in the evening, and when Dr Neil spoke of the wedding day again she said that she must write to her uncle at least , before anything could be arranged, and let him think that she had done so.
The bazaar drew nearer and nearer -- and what a strange thing, McAllister thought, to be such a central part of her life, and she and Matey worked together now, in that, as in everything else, and Matey's liking for her grew daily, for the eyes McAllister turned on him were always so full of love, and he was easy and happy with her, as he had not been since he was a boy before Africa and Angela Deverill's betrayal.
'Such a fine day,' said McAllister happily, serving breakfast to the three of them.
She sat and ate it with them now, and oh, it was torture that in deference to propriety and Matey they had not made love again as they had done that golden afternoon, always to be remembered.
But their eyes and looks made love for them, and spread happiness around them, and their talk was always of the future they were to spend together.
'The Master commends his slave for the quality of the breakfast as well as the weather,' said Dr Neil, catching at her hand and pressing it as she walked by.
'You may thank me for the breakfast,' McAllister said, sitting down to attack her own meal with vigour, 'but the weather -- now that, alas, is beyond me.
If it were not, you may be sure that I would have arranged a better climate for the Brits than the Lord has seen fit to give them.'
'Witty as well as useful,' said her lover appreciatively.
'I can see, Matey, that breakfast conversation is always going to be lively when Mrs Dr Cochrane is at table, if it resembles McAllister's, that is!'
McAllister looked at him from under the long dark eyelashes which had won his heart from the very first moment when he had seen them, on his sofa, adorning the unconscious girl he had carried in from the street.
'And you are both ready for the bazaar, I see,' he said, for he had already watched them carry through the packing-cases filled with their work, ready to be transported on Mr Hanson the grocer's small cart, due to call at Vetch Street at ten o'clock, for them to start setting out the stall.
'Sands tells me,' he observed, 'that some grand rich people, relations of his, have promised to come along to buy -- to ease their consciences, one supposes, they having so much, and the poor devils whom the bazaar is to benefit having so little.'
'All grist to our mill,' said McAllister cheerfully, never thinking that Mr Sands's grand relations might prove to be Nemesis, arriving in group form, to smoke out Miss Sally-Anne Tunstall, American heiress and the ambassador's niece, the Marquess of Blaine's one-time intended, AKA -- also known as -- in the language of American police reports -- Miss Sally-Anne McAllister, skivvy and Dr Neil's fiancee.
On the contrary, all seemed set fair, and when they had loaded the packing-cases on to the cart Matey and McAllister set off for the church hall, to lay out the stall, to have everything ready for the afternoon's visitors, Dr Neil having promised to come along to help them.
'Try to keep me away,' he had whispered, 'although it is exquisite torture for me these days, my dear McAllister, to have you so near, and be unable to stroke every bit of your delectable body.
I only wish that your uncle would hurry up with his consent; I doubt whether my resolution to behave myself can hold out much longer.'
'You are being very naughty,' McAllister whispered back, pretending to be severe, 'and the only thing that I can say in your favour is that I am in nearly as bad a way as you are, and know exactly how you feel -- but be patient, it cannot be long now,'for she knew that once the bazaar was over, and she was at last able to tell him the truth, they could go ahead with their marriage plans, the only delaying thing being the time it would take Mama and Papa to cross the Atlantic.
And suddenly she knew how much she longed to see them, to introduce him to them, and to tell them how sorry she was for her wild behaviour over Terry, and for her conceited folly over Havvie.
Except, of course, she could never -- ever -- tell them what that folly had led to -- even if , looking back, she could now honestly say that she had done nothing to provoke it, that she could have expected anger from Havvie at her changing her mind, but never that he would do as he did.
And even now, after Dr Neil's healing influence, daily exerted, for although he could not make direct love to her he was able to let her know in a thousand different ways how much he loved her, and how precious she was to him, she still felt the odd shudder of shame and fear.
But it was a happy McAllister who arranged the stall with Matey, joked with Mr Sands, and then in the afternoon after lunch walked with Dr Neil and Matey to the church hall.
At first, the afternoon was as much a pleasure as McAllister thought that it might be.
Wealthy ladies, doing a little charitable slumming, as well as a few shopkeepers' wives, bought the pretty baby clothes, and the doll she had dressed went for a fabulous price, to be given to some little girl more fortunate than those for whom the money was being raised.
Dr Neil, with many a joke, helped them, made change, and once went out to return with lemonade and glasses for all the ladies who were so busy manning the stalls, Mr Sands helping him -- the rector being a vague benevolent figure who came in and blinked at them all, said, 'I am sure, ladies, that you will all gain treasure in heaven for your noble work here today,' and went out again.
Such a modest occasion to make a girl like Sally-Anne Tunstall, used to kings and courts, happy, but it was, she discovered, true happiness to hold her future husband's hand, drink the lemonade he had brought her, and watch her work being praised and, more to the purpose, sold.
Afterwards, she was to wonder what might have happened if Mr Sands's grand party had not turned up.
As the afternoon wore on he began to look unhappy, muttered to Dr Neil that he might have supposed that his cousin James would forget to come -- he lived a busy life, after all.
He had promised, Mr Sands said, to bring some friends with him, but by four o'clock there had been no sign of them, until, suddenly, there was an immense bustle outside, and Mr Sands, rushing out, came back to announce triumphantly that Cousin James had arrived --'Better late than never' being his unoriginal version of the truth, and the Hon.
James Sands and his friends processed in to salve their consciences by buying bookmarks, needlecases, embroidered handkerchiefs and even some of the exquisitely embroidered baby clothes on offer.
McAllister, happily unaware of who made up the party, watched these inhabitants of the world in which she had lived since she was eight years old stare and chatter as they made their way through the doorway, Mr Sands bowing and scraping at them as befitted a poor relation to whom they were doing a favour, the rest of the bazaar's patrons staring at these strange beings, male and female, as though they were visitors from another planet, perhaps one described by Mr H. G. Wells.
Dr Neil muttered in her ear, 'I never thought poor Sands such a flunkey; even to persuade them to buy our knick-knacks this is surely excessive.'
McAllister could not but agree with him, and had to stifle a grin as he led them around, the women staring, and the men, manifestly bored, dragged along to accompany wives and girlfriends who would, that night, describe their visit to the East End in terms that would do justice to a journey up the Amazon.
And then her smile froze on her face, and she saw Dr Neil's change.
His face changed because his brother Stair was one of the party, escorting the young heiress to a beer barony in Milwaukee, which was as much as such a penniless man as himself could aspire to, especially as his title was a mere baronetcy.
But walking along beside him, sucking his cane, and looking ineffably bored as he escorted his new marchioness, nee Miss Maybelle Foy, the meat-packer's heiress from Chicago, was none other than Havvie Blaine, of unblessed memory.
The mere sight of him was enough to make McAllister's toes curl.
She retreated round the stall; she had been at the front, helping a customer to choose something from what remained unsold.
Stair saw his brother and came forward to greet him, Havvie still at his side, his wife picking up and putting things down as though they were contaminated -- she had not married Havvie in order to slum it, as she later said.
'Neil!
What are you doin' here?
Hardly your line, hey?
Sellin' baby clothes, old thing, and Matey --' he turned his bright eyes on his old governess and nurse --'still with him, eh?
Gets the cane out to you, does she, Neil?
Haw, haw!'
His feeble wit amused his party, if no one else.
'This is my parish, Stair,' said Dr Neil, 'and, seeing that I live here, such occasions as this deserve my support.
All for a good cause, and I hope that you and your friends have come here prepared to spend.'
He moved round to the front of the stall, pulling McAllister with him.
For some unaccountable reason she had bent down to inspect the floor, hoping, though Dr Neil did not know this, to escape the attentions of Havvie and his friends, all of whom had known her well in her old life.
Dr Neil would have none of it.
'Come on, McAllister,' he said good-humouredly, 'do your bounden duty.
The Master commands you,' he then murmured, for her ears alone.
There was nothing for it.
McAllister, clad in her shabby bottle-green dress enlivened at the throat by a bit of cheap lace, her serviceable shoes on her feet, her hair simply tied up in a knot on top of her head, her hands red raw from hard and constant manual work, resignedly straightened up, to meet the gaze of Havvie and his friends.
Recognition came slowly, but, as she had feared, it came, and when it did so Havvie's cruel eyes ran up and down her body.
And there was no doubt that Stair Cochrane had recognised her as well.
He looked from her to his brother, his mouth open, about to speak.
The American Princess had disappeared from view months ago, and here she was, in an East End church hall, wearing shabby clothes, in the company of his brother and his old nurse.
Havvie forestalled him.
His expression wicked, he continued the stare which seemed to strip poor McAllister of her clothing, a stare so cruel that Dr Neil immediately saw its import and began to bridle at the sight, putting out a protective hand to his beloved, to feel her tremble beneath it.
A second or two later, Maybelle, her attention drawn from what was on the stall to what stood beside it, also recognised Miss Sally-Anne Tunstall.
'Well, well,' drawled Havvie Blaine, at last, 'what have we here?
If it isn't the oh, so charming Sally-Anne -- missing from our midst to turn up in the East End -- and in fancy dress, too!'
His derisory gaze swept over her, and then over Neil, who, at the jeering note in Havvie's voice, and the addressing of McAllister by her correct Christian name, had suddenly stiffened.
'And the good Dr Neil Cochrane with her, the saintly refugee from the best society, who prefers the company of the dregs to that of his equals -- well, well, well!'
Dr Neil was so shocked -- both by Blaine's manner and the cavalier fashion in which he spoke to them both, that his normal composure left him, and he could think of no more to say than, 'You know Miss McAllister?' and his voice was incredulous, so incredulous that Havvie's eyebrows shot up in gleeful, unholy delight at the prospect of yet another opportunity to humiliate the girl who had jilted him.
'Know her?'
His voice carried every shade of innuendo in it.
'Of course I know her -- as what man does not?
Very well indeed.
I think you would say I know you very well, Sally-Anne, wouldn't you?'
And the smile he gave to the frozen victim of his indecent amusement was pure poison.
'Miss McAllister, is it now, Sally-Anne?
A strange setting for the Tunstall heiress, all this --' And he swept his hand around the shabby room, noting the protective arm which Dr Neil had placed around Sally-Anne's shoulders.
By his side Stair Cochrane's grin and Maybelle Blaine's inimical stare were equally hateful at the sight of the suddenly stricken expression on Dr Neil's face.
He took his hand from Sally-Anne's arm, gazed first at Havvie, and then at McAllister, and said, in a hollow voice, 'The Tunstall heiress?'
'Didn't you know?'
Havvie's grin grew broader.
'Allow me to present, since you seem to be in the dark about your companion, Cochrane, Miss Sally-Anne Tunstall, the American Princess, niece to the American Ambassador, heiress to Senator Jared Tunstall, of whom, even you, Cochrane, shut away as you are, have surely heard.
The American politician and robber-baron, the biggest and richest rogue in the States, and she so free with her charms before marriage that marriage to her was neither necessary -- nor desirable.'
Afterwards Neil Cochrane thought that he should have struck Havvie down then, but shock at learning who McAllister really was had him in its grip, and he let the shameful accusation go by.
He heard Maybelle Blaine titter, saw McAllister's -- no, Sally-Anne Tunstall's -- stricken face.
All of the courage, the new-found confidence which Dr Neil had given her, with his love, drained away from Sally-Anne at the sight of Dr Neil's face when Havvie told him who she was.
She put up her hands in a gesture reminiscent of the one which she had made in the attic, when she had been still fearful of him and of all men, but the gesture was as much for Havvie as for him.
'Please, please, Havvie,' she said hoarsely.
Both men looked at her, and the expression on Dr Neil's face struck her to the heart.
Oh, why had she not told him who she was straight away?
Why had she waited -- to be confronted with this?
'Oh, I like that,' said her former fiance softly.
'Oh, yes, I do, indeed.
What do you want of me, Miss Sally-Anne Tunstall?
More of what you had before?
The virtuous doctor not satisfying you?
Tut, tut...
Go home, Sally-Anne, and I'll think about it, so I will -- if you ask me nicely again, that is...'
Sally-Anne put up her hands again, could barely look at Dr Neil.
Why did he not speak, stop Havvie?
But Neil's world lay in ruins about him, as he thought of all the lies which McAllister had told him, and listening to Havvie began to wonder where the truth lay.
All of his suspicions about women since Angela Deverill's treatment of him came back with a rush, and he could not tell whether McAllister was registering shame or guilt.
He swung on her, said, 'He's telling the truth?'
'Yes.'
Sally-Anne thought that she was about to faint.
'That I'm Sally-Anne Tunstall, yes.
But not the rest, Neil, not the rest.
He's lying about that.
He's the one who...hurt me.'
Havvie's wicked eye was on Dr Neil again.
It was as though everyone else in the little hall -- Mr Sands, the curious spectators, Havvie's party -- did not exist -- only the three of them left in the world.
Oh, was not this delightful, to hurt again the girl who had rejected him, the girl who when all was said and done would have been a greater prize than the vulgar woman he had married for her money?
Sally-Anne deserved punishment for forcing him to do that, if for nothing else.
And by her expression the wound he was dealing the pair of them now was a mortal one.
'So innocent,' he jeered.
'But can you believe her, Cochrane?
Did she put on her pretty act for you?
Show you her little deceits?
The bashful, wounded virginity, such coy shrinking to inflame the passions even more -- as I know to my cost -- and then, when she's finished with you, rejection.
Using her money and her charms to trample on you, to humiliate an ex-lover.'
Neil was ready to knock Blaine down, had come out of the trance which had held him silent under Havvie's insults to the girl he loved -- but he saw Sally-Anne's face, and read there only guilt, not the shame which gripped her.
'No, Neil,' she said, swaying, 'he's lying.
How can you let him say such things about me, before so many?
Are all men liars and cowards?'
'Do but look at him, sweetheart.
He knows, doesn't he, that I -- and possibly others -- was there before him?'
Havvie said.
Neil broke on that, and on Sally-Anne's cry.
She had put her hands before her face, and by now the whole room was watching them, although most could not hear what was being said.
He pivoted, and despite his lameness, his sedentary life, he was still a powerful man, and he knocked Havvie Blaine flying, for him to be caught and steadied by Stair, who fended his brother off with one hand.
The women standing on the edge of Havvie's party shrieked and Maybelle Blaine cried viciously to the distraught Sally-Anne, 'Haven't you done enough damage to poor Havvie, Sally-Anne Tunstall, without egging on another of your lovers to attack him?'
Another of your lovers!
Sally-Anne's paralysis cracked.
She advanced on Maybelle, her hands clenched into fists, as she had advanced on her little brothers, sometimes seriously, sometimes in play.
'You lie, Maybelle.
You know you lie.
I'll teach you to tell the truth.
If no one will fight for me, I shall fight for myself.'
'No,' said Neil; both Dr Neil and McAllister, his love, had disappeared.
He caught Sally-Anne's wrist in his hand with such strength that she cried aloud.
Ignoring Havvie Blaine, Stair, the astonished Mr Sands who could hardly believe what he had been hearing, and all the other spectators, he dragged her across the room and through a side-door by the small stage, into a long corridor.
Still holding her, he locked the door behind them, then released her wrist to take her by the shoulders and pin her against the wall.
For some reason she had snatched up her bag as he had dragged her by the stall; she remembered seeing Matey's grieving face, Matey's hand on his arm, Matey saying, 'No, Neil, no.
Think what you are doing,' and Neil ignoring her.
His face was ashen and he was shaking.
'You lied to me,' he said, his voice thick with passion.
'Not once, but many times.
I thought that you were telling lies, but, by God, not such lies as you told.
The poor girl with a dead father and mother, ruined in the Depression.
Such a touching story, Miss Sally-Anne Tunstall --' and he spoke her name as though he were striking her with it '-- and all a lie.
Lies, all of it.
And the rape -- was that a lie too, to catch the silly doctor?
Why did you do it?
Did it amuse you?
If you lied so badly about yourself, how can I know that you did not lie about your innocence?
God forgive me, Sally-Anne Tunstall, I have spent my life avoiding such flytraps as you plainly are.
The American Princess -- even the poor doctor has heard all about her , and then you came into my home...
Did it amuse you to seduce the poor doctor after the noble Marquess?
Do you like playing games with men?
Does it feed your vanity?
Is that why you refused to name a day for our wedding?'
And he shook her at the end of each question.
Sally-Anne's world had collapsed about her.
Half an hour ago she had been in a heaven of love with the enraged man before her.
And why did he believe Havvie, and not her?
Oh, they were all the same, men, even the best of them.
Her hope of happiness was gone, all gone, and if she had to bear a little of the blame for not having told him immediately who she was, then surely that was not a thing to hang her with.
Could he not trust her a little, just a little ?
Was Havvie's word worth so much more than hers?
Up went her head.
If she was to be betrayed yet again, then she would go with banners flying, trumpets braying, and no tears to speed her on her way.
'Do you -- did you -- love me so little, Neil, that you believe him without question, and me not at all?'
'Oh, my sweet cheat, but he knew you, didn't he?
Knew your little deceits.
He knew that you weren't a virgin when you came to me.'
'He knew because he raped me,' she answered him, head still high.
She remembered Havvie's last words when he had left her that afternoon, and knew that what he had said was true.
In taking her virginity he had taken away her power to convince any man of her innocence, even the man before her who had said that he loved her.
'I told you the truth about Havvie.
I lied to you about who I was because --'
He interrupted her, savage again.
'The truth, Miss Tunstall?
Do you know what the truth is?'
What could she say?
She had lied to him because she wished to evade the unpleasant truth of her life as a rich heiress, to try to make a new life free of old ties and old mistakes, where she would be loved as McAllister, who had nothing.
And how right she had been to lie!
For look what had happened, once he knew who she was...
She flung her head back.
She was her father's daughter again, the daughter of the predatory pirate who controlled Wall Street, who was a powerful man in government as well as in the world of society, was King Edward's friend, who kowtowed to nobody.
'The truth, Dr Cochrane?
Do you want it all?
You shall have it.
First of all my name is neither Tunstall nor McAllister.
You fell in love with an illegitimate girl who carries her father's name by grace and favour, not by right -- never mind that I was not meant to be illegitimate; that is what I am.
And the first man who pursued me did so to gouge a small fortune out of my father, to marry his true sweetheart and set himself up in business.
And the second man who pursued me did so to gain a large fortune, to spend on his long-term mistress and their children, and to laugh at the silly girl and her father who gave it to him.
And when the silly girl found out and rejected him he raped her for her pains.
'And the third man who...said he loved me...' and her voice faltered at the words '...loved me so dearly that on hearing the slanders of my assailant he believed every word that he said, and none of mine.
I loved you, Neil Cochrane, truly loved you, that was no lie, and I shall make sure I never love another man.
And all the lies that I told you were to try to escape the lie that was my life.
'And what did you love?
Tell me that.
For sure, not Sally-Anne McAllister, but some doll of your imagination you took to bed to comfort yourself.'
He turned his ravaged face on her, and the ravage was grief, and she had never loved him more, doubt her though he might.
'How can I believe you, when all that I thought I knew of you was false...?'
'Try believing your heart, Neil.
Or is that too much to ask?
Let go of me,' she said fiercely, as his hands tightened on her shoulders.
'Unless you wish to do to me what Havvie did.'
He dropped his hands instantly, as though she had struck him.
'Was it a game for you, Sally-Anne?
To trap Stair's poor brother and add another to your list of conquests?'
And he almost groaned at the sight of her.
She had never looked more beautiful, standing there in her poor clothes, defying the world -- and him.
'And if I had told you the truth, Neil, that day when Jem Higgins knocked me senseless, what would you have done then?
Sent me on my way?'
'Better that than you should practise all your sweet deceits on me.'
And his voice shook.
'My sweet deceits.'
Sally-Anne closed her eyes, opened them to see him lift his hands to cover his face, his head bent, shoulders bowed, his world in ruins about him.
'Oh, I see, I see,' she said scornfully.
'You are determined to wallow in self-pity again, determined to think me a whore so that you can return to the bottle, and blame all women, and not yourself, that you are frightened to face life.
So be it.
I shall be a whore, and leave you with a whore's kiss.'
And she leaned forward, pulled his hands away from his face, placed her mouth on his in a kiss of such passion that they drowned in it, and then, with a sudden twist, she was away from him, out of the arms he had placed about her, running down the corridor towards the outer door.
She turned and said, voice shaking, 'Goodbye, Neil, goodbye.
Think what you will of me.
Think that, for me, you were but a summer's play, but summer is nearly over...
Goodbye.'
And she was out of the door -- and out of his life.
For a moment the paralysis which had struck Neil from the moment that Havvie had begun his assault on her reputation was on him again.
Sanity returning, he ran after her, to find that she had taken the key out of the lock -- and locked the door from the outside, against him.
Desperate to find her, to say -- what?
Anything to get her back again.
He ran through the hall, the mockery of Havvie Blaine's party following him as he dashed out of the main doorway -- no sign of her there -- and then back down the alley at the side -- and still no Sally-Anne.
Distraught at what he had thrown away -- and, whether she was guilty or innocent, he only knew that he wanted her back -- he ran towards Vetch Street like a man possessed, head bare, careless of curious glances and jeering comment at his headlong progress, reaching Vetch Street -- to find that she was not there.
Her sewing basket sat on the kitchen table where she had left it, after lunch.
And then he was up the stairs to her room, to find there her few poor possessions -- but still no Sally-Anne.
And he had no idea of where to go, and how to find her -- nor of what he would say to her when he did find her, either at the embassy or anywhere else she might have gone.
He only knew that he had lost her, and she might be the liar, cheat and whore he thought she was, or the injured innocent she claimed to be -- he did not care which, for whatever she was she had taken his life and his hopes with her.
CHAPTER SIXTEEN
'And still no sign of her?
No clue to where she might have gone?'
Jared Tunstall was distraught; his much adored daughter had disappeared, but as was his usual way no one could have guessed at his inner feelings.
Only his wife Mary, always calm, and still, at the age of forty, a great beauty, could guess at the real distress which tore at him.
'Nothing,' said his brother Orrin.
'Nothing.
We have been most discreet, you realise.
I know that you would want no scandal attached to her name.'
He hesitated; best not to tell Jared of what Havvie Blaine had been saying of her; he was fearful that if his brother knew he would do Havvie a mortal injury -- Best perhaps to warn Mary Tunstall privately.
Jared and Mary Tunstall had arrived at last in London, summoned by Orrin when it became apparent that Sally-Anne could not be traced.
'Dear God,' said her father.
'What can she be thinking of?
For all we know she could have been made away with.'
He rose, began to pace the room, stared at the butler who entered, and said to Orrin, 'Beg pardon, sir, for interrupting, but the housekeeper says that a young woman, claiming to be Miss Sally-Anne Tunstall, has arrived at the servants' entrance, and what is she to do?'
'Do?' said Orrin energetically.
'Send her in immediately, Baines.
Have you seen her?
You know Miss Sally-Anne.'
'Begging your pardon again, sir.
Mrs Wren says that the young person does not look as an ambassador's niece should...'
'For God's sake,' said Jared Tunstall, beginning to feel a little hope at this news that his daughter might, after all, be safe.
'Send the girl in; we shall soon know if she's Sally-Anne.
Surely the housekeeper can recognise her?'
'Mrs Wren is new, sir, since Miss Tunstall left to go on holiday,' explained the butler.
'But I will send the young woman in, if that is your wish.'
'English servants,' snorted Jared, to have his wife put her hand on his arm, while Orrin rose, saying,
'I will see her, and if it's Sally-Anne, which I suspect it might be, and not someone masquerading as her for some God-forsaken reason, I'll send her to you.
You won't want me around.'
'My brother,' said Jared, when Orrin had gone, 'is the truest gentleman I know; beats all these so-called aristocrats here hollow.'
He resumed his tigerish pacing of the room, to stop only when there was a timid knock at the door, to call impatiently, 'Come in, come in.'
Clutching her battered handbag, graceless in her servant's clothes, Sally-Anne Tunstall advanced into the ambassador's private drawing-room to confront her mother and father, who stared at the strange sight she presented.
'What are you doing here?' was all that she could find to say.
'I thought that you weren't coming over until the fall.'
'Good God, Sally-Anne!' exploded her father.
'Your uncle Orrin tells us that you have been missing for months, not travelling with the Parslows as we thought, and when you surface again you have the gall to question us .'
Seeing them after so long, after all that had happened to her, it was as though they were strangers.
She had not realised that they were so young.
After all, Papa was only just forty-one, and Mama was younger than that, and they were both so handsome that it was a wonder -- or perhaps not -- that they were so faithful to one another, so much so that it was almost a joke, Papa having been a great ladies' man in his youth, she had discovered.
The sight of them made her feel old -- old and soiled.
'I decided not to go with the Parslows,' she announced, as though she were saying the most reasonable thing in the world, but her lips were numb with fright, Papa was such a hard man to defy.
'I changed my mind at the last minute.'
'And that's it?' said her papa quietly; he was always at his most dangerous when he was quiet.
'You turned down Havvie Blaine at the last minute, too.
You then disappear, frighten us all to death, and suddenly, without warning, reappear, wearing fancy dress --' this was an unfortunate echo of Havvie, to be sure '-- and all you can say for yourself is that you changed your mind!'
'Yes,' she said, again, as though it were the most reasonable thing in the world, but she saw that her mother looked worried, and that doubtless Papa was behaving as he did because he had been worried about her, his darling, whom he had sent away from him, only to lose her in a foreign country -- for that was what Britain was.
Jared Tunstall, enraged by his daughter's cool refusal to be intimidated in any way, advanced on her, his face stern as she had never seen it before, although his business rivals would have recognised it.
He was obviously even more angry with her than he had been during the contretemps with Terry.
'By God, Sally-Anne, your mother said that I spoiled you rotten, and I think that she's right.
Answer me!
Where have you been?
What have you been doing?
Why did you not have the common decency to inform your uncle Orrin of where you were going, even if you no longer wish to oblige your mother and me?'
Perhaps what disturbed him most about his daughter was her uncharacteristic calm.
'I have been learning about life,' she replied.
'Life outside the bastions of privilege, rank and wealth.
As I believe that you both did, when you were my age -- if you can remember so far back -- but I knew you would prevent me from doing so if I told you the truth.'
This piece of insolence shot out without her willing it.
'I found it very illuminating,' she continued, and then, as her papa, his face now black with rage, advanced on her, his hand upraised, her mama's suddenly fearful hand on his arm, as she cried to him,
'Be careful, Jared,' Sally-Anne answered him almost indifferently -- for after what Dr Neil had said to her nothing seemed to matter any more.
'Must we wrangle so soon after we meet again, Papa?
I'm very tired.
I should like to have a good bath -- I haven't had one for weeks -- and something to eat, before you begin to cut me up as though I were one of your business opponents laid out for you to hand them the coup de grace .'
She thought that he was about to choke, or to strike her down, and that would be the end, the very end; but did she not deserve it, and more, for the way in which she was speaking to him?
But for the life of her she could not behave as she ought when she stood among the rubble to which her life had been reduced.
Something in her expression, the hint of a mortally wounded animal inviting further hurt, touched the almost feminine intuition which was one of Jared Tunstall's greatest assets in the cruel game of life.
He mastered his rage, dropped his hand, said, his voice as indifferent as hers, but the feeling beneath it deep indeed, 'I should have thrashed sense into you long ago, Daughter, as I would have done had you been a boy.
You're too big now, more's the pity -- and, in any case, I don't think it would answer at the moment.'
For he had suddenly realised that she was at the end of her tether -- for what reason he did not know, and neither did he care, only that he must go warily with her, lest he damage her beyond repair.
'The pity is, I wasn't a boy,' she told him, her voice still indifferent, her face calm, quite unlike any Sally-Anne her parents had ever known.
'But there it is.
You must make the best of me, Papa.
I will tell you what I have been doing later.'
If he was adamant, so was she.
They stared at one another, and as her mama had so often said, and thought again as she watched them, they were so alike in every way except their sex that they might have been two sides of the same coin.
'At least tell me that you have come to no harm in your adventuring,' he said, and, wonderfully, his voice was pleading.
Her face broke into a mischievous grin that was pure Sally-Anne.
'Not since I left the embassy, no.
On the contrary, I have been most carefully guarded and looked after.'
Her mother spoke at last.
She had left the two of them to duel, not even acting as referee, waiting for them to tire, perhaps even to appeal to her.
'Leave off, Jared, do,' she said briskly.
'The child looks done in; let her have the bath and the food she asks for.
She's not one of your junior clerks to be bullied.'
She advanced on her daughter, caught sight of the red and ruined hands holding the shabby bag, drew in her breath, and said, 'What on earth, Sally-Anne, have you been doing to, or with, your hands?'
'Working,' she said briefly, and turned to her papa, who was, for once, at a loss, both his women against him, Sally-Anne in an unlikely alliance with her mother, who had put a protective arm around her daughter's shoulders, which her daughter did not throw off impatiently in her usual brusque manner.
Something of his momentary, and unusual, lack of confidence where she was concerned penetrated her dulled senses; she said penitently, 'I will explain later, Papa, I promise.
At the moment I could not boil water successfully.'
He had to be content with that.
He watched them walk away.
She was right.
With that spirit and her inventive brain, she should have been a boy, and together they would have run the world.
Bright though his three sons were, they had nothing on Sally-Anne, the daughter conceived in the purest love, out of wedlock, to bring him joy and sorrow.
She had her bath, which seemed to help, washing away her sorrows as well as the accumulated scum of living in a house where, for a female, baths were difficult -- she remembered Dr Neil standing in the outhouse in the yard, pouring water over himself, the water running beneath the short door -- all that she could see of him being his head and his bare feet.
Shamefully, she had had a vision of his body naked, even before they had gone to bed together, her first intimation that she was beginning to feel more for him than she ought, for the vision had not frightened but intrigued her.
And later she had found that his body, scarred as it was, was better than the one which she had imagined for him, as a living and breathing man was better than a marble statue or painting, and her vision had lacked life.
How could she think of such a thing?
She should try to forget him, but the memory of their afternoon together had the power to make her cheeks rosy, and so they stayed while she put on the plainest gown which she could find in the wardrobe which she had left behind, with her personal maid, when she had pretended to go with the Parslows.
Her dress sense, as well as her character, seemed to have changed while she lived at Vetch Street.
Finally, when she was fastening the long row of buttons on a classic dark blue creation, her mother came in, her lovely calm as untouched as ever, as though missing daughters suddenly returning were quite a normal thing.
Usually this calm annoyed Sally-Anne; today it soothed her.
'Come here, my child,' said her mama gently, seeing her struggle with the tiny miniature pearls.
'Let me fasten that.
Your poor hands are making you clumsy.
They look as though you have been blackleading grates!'
Only for her to step back, the calm a little shattered, as Sally-Anne blushed to her hairline.
'You haven't,' she said faintly.
'Yes,' said her daughter.
'And scrubbed the floor, swept the house, and laid the fires, too.
As well as carried in the coal, made the beds, run errands, carried out the slops.
Everything.
I'll tell you and Papa about it over tea, if I may.'
'There is something which I must say to you before we join Papa,' announced her mama.
'It is about Havvie Blaine.
You were right to break with him if you decided that you had made a mistake in accepting him, but oh, my dear, your uncle Orrin tells me that he dare not inform your father of the dreadful things Havvie is hinting about you for fear of what he might do to Havvie.
Your uncle thinks that your papa might even kill him.'
'What dreadful things?' said Sally-Anne, who knew what Havvie might be saying, for had he not said it to her, before his court?
'I can't tell you,' said her mama.
'Say it,' said Sally-Anne steadfastly.
'That you are promiscuous, and that he jilted you, not you him, because he found you out.'
Sally-Anne closed her eyes, and said, 'Papa could kill Havvie for all I care about him.
Except that the Brits here are not so straightforward as we are, and might hang Papa for giving Havvie what he so richly deserves.
I agree that we must not tell him, but he is sure to find out, you know.
And I am more worried about what he thinks of me than of what he might do to Havvie.'
And despite herself her lip quivered, and her face began to crumple.
Mary Tunstall looked at this greatly changed daughter whom she felt she hardly knew, but was sure that she was going to admire, as well as love.
'You are a good girl, Sally-Anne,' she said, 'and in a way that is your biggest cross.
You have had to learn how wicked people can be.'
This was perceptive and she was rewarded by a brief display of her daughter's old impulsiveness, Sally-Anne kissing her mother fervently on the cheek before they joined Papa in the drawing-room.
'So, Sally-Anne,' he said, the three of them on their own, before the tea-tray, Uncle Orrin and Aunt Nella having tactfully found that they had other engagements.
'What have you been doing?'
Thankfully he seemed to have recovered himself, she noted.
The worry over her disappearance which had caused his relief that she was safe to express itself in what she acknowledged was quite justified anger had abated.
'I know that I owe you both an explanation and an apology,' she said slowly.
'I am so sorry that I caused you such grief by acting as I did, by going off without telling anyone.
When I left here, I never thought...'
She hesitated, said shyly, 'That is something I have learned, I suppose -- that in the past I never stopped to think before I spoke or acted...though, goodness knows, you both told me to, often and often.'
Both her parents were surprised at this admission, she saw; and, seeing, she was ashamed of her past wilfulness.
'And,' she went on, 'I was compelled to end my engagement to Havvie when I discovered that he did not love me at all...merely wanted Papa's dollars, and, worse, despised us, all of us, for being rich and vulgar Yankees.
He intended to take Papa's money and use it to keep his long-term mistress and their children; he laughed at me behind my back...while he pretended to my face that he admired and loved me dearly.
I shudder to think what kind of life I should have had with him.
You could not wish me to marry such a...cur, for that is what he is, despite his title.'
She saw her papa's face change at this news, and went on rapidly, 'He took it very badly...' which was, she thought wryly, the understatement of this, or any year '...and...
I felt so dreadful.
First Terry Rourke...and then Havvie Blaine...
I thought that there must be something wrong with me, and I had to do something to prove that I was not just a thing which other people used, of value only because you, Papa, are so enormously rich.
You do understand?' she said, a little fearfully.
Thankfully they both nodded and her papa said gently, 'Yes, but, Daughter, why did you need to disappear?'
'To escape it all,' she said, her fiery pride suddenly back again.
'My reputation, your wealth, everything.
I needed to be nobody, to find out how ordinary people lived.
And then J. D. O'Connor, who edits the Clarion Cry , had asked me to write for him -- almost as a joke at first-- about women's lives.
He suggested that I write about poor women, and the only way in which I could honestly do that was to go and live among them, and I knew that you and Uncle Orrin would never agree to that.
'So I got Laura Parslow to help me, and went to live in the East End, and the only work I could find, being such a fine lady and quite untrained, was as a maid-of-all-work, a skivvy, which I was by day, and wrote my column at night, when my work was over.'
'A skivvy?' said her papa, almost disbelieving, and she saw that Mama felt the same.
'On your knees, doing the dirty work, Sally-Anne?'
'Really, truly, a skivvy.
Sixteen hours a day.'
She held her hands out for them to see.
'And you need not worry about whether I was safe or not.
I was in a doctor's house.
Dr Neil -- he's Stair Cochrane's younger brother -- he works for the poor in an East End practice --' Her voice almost broke when she said his name, and her mama knew immediately that Sally-Anne was not telling them the whole truth about her East End adventures.
'And his old nurse and governess is his housekeeper, a kind dragon who kept me in order.
And I learned so many things,' she said, 'I am not sorry I did it, not at all, even if it was desperately hard work.'
She swallowed.
'I learned discipline, you see.
And J. D. O'Connor liked my articles -- he said they created a great deal of excitement and interest; I wrote under the name of Vesta.
And Dr Neil and Matey -- Miss Mates -- taught me so much about...life...so much that is useful...and I learned how hard it is for most people to live even halfway decent lives...and now I am home again.'
She fell silent.
'And that's it?' said her papa, as he had done earlier, but in a different and kinder tone of voice altogether.
He rose from his chair, came over to her and took her hand in his.
Like his wife, he had heard her voice change when she had spoken Dr Neil's name, but he said nothing of that.
He bent from his great height to take her right hand into his own large one, turned it over to look at the scars and callouses hard manual labour had inflicted on its once pink and white delicacy.
'Were you a good parlourmaid, Sally-Anne?' he enquired gently.
'Miss Mates said that I was the best and most hard-working girl she had ever had,' she answered proudly.
'But it was very tiring at first, there was so much to do.'
'So, you were a good, hard-working skivvy; and the work that you did for the Clarion -- was that worthwhile, too?'
'Oh, yes.
J.D., Mr O'Connor, has invited me to join his permanent staff of reporters now that I think that my work in the East End is...over.
I want to accept, Papa, I really do.
I don't want to be a society lady any more, but I'm not quite sure whether I want journalism to be my life's work.
I have thought about being a doctor, after seeing Dr Neil -- he didn't start to study until he was about my age; but he's a man...'
She fell silent.
She wanted to talk about him, but she could not.
She could not forget his betraying, rejecting face.
Her papa was speaking again, his voice grave, his expression kind.
'I think I have been very wrong in treating you as I have done, Daughter, in the way I brought you up.
You are too like me to be idle, to be a useless fine lady.
I should have known better.
You should have had a tough education.
You need occupation to direct your energies.
But I was so proud of my beautiful daughter, and I wanted to make up to you for the years we lost before I met your mama again; I wanted you to have the best of everything, and I thought that that was what I was giving you.
I was wrong.'
'And you gave me the best!'
For the first time she spoke with all her old unbridled fire and passion.
'You thought that you were doing the right thing.
It is I who am wrong, not you.'
Only to meet his shaking head.
Oh, how she loved him, her strong and handsome father, not the ogre she had thought and called him over Terry -- and to hear him admit that he was wrong!
Sally-Anne saw her mama's loving, watchful eye on them both, pulled her hand from her father's, ran to her mother to drop on her knees before her.
'You will not mind, Mama, if I take only a small allowance from you and Papa?'
For she thought that to reject his money altogether might hurt him.
'And not live in high society, but earn my own living and try to do something worthwhile with my life?'
'And stay in England?' asked her mama, not wishing to lose a Sally-Anne who, for whatever reason, seemed to have changed so greatly since she had last seen her, as a spoiled, petulant and wilful child, thinking only of herself.
'For a little, perhaps; I don't know,' said Sally-Anne.
'To find out what I want to do.'
She resolutely pushed away the memory of Dr Neil and the life which she had been going to share with him.
She stood up and faced them.
Pain and love were mixed in equal proportions on her face.
'Please say that you will agree, will at least let me try to earn my own living, find out if I am strong enough to do so -- not go back to being useless, merely a symbol of Papa's wealth, not really his or your daughter.'
Of course, they agreed, and if afterwards both of them also agreed that there was more to Sally-Anne's story than she had cared to tell them, about both Havvie and the mysterious Dr Neil -- they had noticed that his name was constantly on her lips -- they did not tell her what they had guessed.
She wished to live her own life, and they must respect that wish.
Before she left them, she said one more thing.
'I have asked Uncle Orrin to tell everyone who calls for me that I am not 'at home'.
I have no wish to resume my old life.'
Jared Tunstall thought that something more than hard work had brought his daughter fulfilment and a new maturity -- and, with them both, a deeper beauty.
He put his arms about her, held her to him.
'When Orrin told us that you were missing,' he said, 'I knew how much we both love you.
And now that you have come back I want to tell you that.
And of course you shall live the life which you have decided on -- as Mama and I have lived ours -- even if it means that we lose you.'
She had lost them, and found them again, and if she had had to love -- and lose -- Dr Neil in order to do so, then so be it.
In life, she was discovering, gain and loss frequently went hand in hand.
Heads turned that night when Senator Jared Tunstall and his wife entered the ballroom at Devonshire House.
At forty-one, six feet two and a half inches tall, with the face of a handsome pirate and the body of an athlete in tip-top condition, Jared was worth two of anyone's looks, and his serenely beautiful blonde wife Mary was no less remarkable.
It was easy to see that he was the Princess's father: her likeness to him was astonishing, as more than one person commented.
Maybelle Foy, now Lady Blaine, tittered to her husband as they walked across the floor, after mounting the stairs.
'Your one-time intended's folks,' she said -- she did not see him wince at the Americanism; she did not yet know him as well as Sally-Anne did.
He laughed, however, and said something derogatory about the Tunstalls to his little court of parasites.
Later, in the supper-room with the tall Senator in earshot, he could not resist doing it again, with Maybelle egging him on, loudly drawling his opinion of Sally-Anne's reputation -- or lack of it.
'Like her father's, more than a little dubious,' he announced.
Jared heard this, as he was meant to.
He put down the plate he was carrying, spun round on his heel and surveyed Havvie leisurely.
His clothes and his voice were perfect.
A great mimic, he had developed for English society an accent which outclassed the Brits around him.
'So you,' he said, his mien so dangerous that Mary beside him, fearful, put a restraining hand on his arm, 'must be the Marquess of Blaine, seeing that you are escorting the former Miss Maybelle Foy, of Chicago.
Do not expect me to congratulate you, madam, I am too busy congratulating myself that I am not the father-in-law of a man who can speak so demeaningly of any young girl in public.
May I, instead, commiserate with you, Lady Blaine?
You have made a bad bargain in marrying Havvie.
It's becoming a family habit, I fear.
Your father has brought off some very bad bargains lately.'
'You are insulting, sir,' began Havvie, scarlet.
'Oh, I take my cue from you, sir,' said Jared cheerfully.
'Be thankful we are not in the States.
Were we so I should have been happy to kill you for speaking so of my daughter.
As we are in England your punishment must take another form.'
'Oh, we are civilised here,' sneered Havvie, feeling safe in the crowd, although Tunstall's size and presence were intimidating, to say the least.
'Then I shall be happy to take a civilised and modern revenge,' said Jared gravely.
'Lady Blaine, I have spent the last week debating whether or not to ruin your father Hiram, and his syndicate, and provide the information to send them, and him, to prison for stockmarket fraud.
Thinking that dog must not eat dog, even if my meal be only that of a mongrel, I had almost decided to spare him.
Between you both you have changed my mind for me.
I shall immediately take all steps to destroy your father's fortune so that you, my noble Lord Marquess, will have married your Yankee bride for nothing.
All she will have brought you is her own ill-bred person.
I wish you both joy of your poverty.
There will be no American dollars for you to spend on mistresses and in night-houses where children are on offer, my lord.'
He had had Havvie Blaine investigated when he had heard of Sally-Anne's engagement to him, and her rejection of him had saved him, and her, a deal of pain and trouble.
He paused.
He had spoken in so low a tone that only the few near to them had heard him.
Maybelle Blaine flushed scarlet, and said shrilly, 'You forget yourself, Jared Tunstall.
This is not Washington DC.'
'As well I know,' he said, 'or I might be tempted to thrash that cur beside you until he is senseless for what I hear that he has been saying about London of me and mine.
I give you ten years, sir,' he said to Havvie, 'but you are in bad condition, I see, so sweeten your language in future, or it will be the worse for you.'
He turned to his wife, and said, 'Mary, my love, you will excuse me if I ask you to leave immediately.
I have cables to send, and work to do.
It is day in New York, and there is still time to ruin a man before dark.'
Maybelle, face ashen, was transfixed as she watched the Tunstalls leave.
Of them all she was the most aware that Jared Tunstall's threats were never idle, and that her new husband had ruined her father, and themselves, with his loose morals, and looser tongue.
'Go after him, Havvie,' she almost shrieked.
'He means what he says.
He will ruin us all.
Apologise -- anything -- to stop him.'
Jared turned again, his face stone.
'Too late, madam; if you had defended my daughter when he slandered her...but now, if you both walked on your knees from here to Chicago, I would not change my mind.'
'Jared,' said his wife in the motor returning them to the embassy, 'was that necessary, or wise?'
'Both,' her husband said grimly.
'No one will ever dare to slander her again, and besides, I have the feeling -- and you know that I am seldom wrong about such things -- that that effete swine did our daughter a great hurt.
I know her well, and how brave she is, and by God, for whatever he may have done he shall pay, and pay again.
I don't think he'll badmouth her in future, but he will have learned his lesson too late.'
CHAPTER SEVENTEEN
She had gone.
He had lost her.
The house was empty and all that Dr Neil Cochrane had left was Matey's grieving, reproachful face.
'I'm surprised at you, Neil,' she said -- he was not even eight years old to her now, more like five or four.
'Yes, I know that she deceived us as to who she is, but how could you believe a word Lord Blaine said about her, and she such a good hardworking girl?
I remember him well.
He was always a horrid little boy for all his pretty face, and now he's a horrid man.
And you're not to start drinking that nasty whisky again.
It's not like it was with Angela Deverill -- if you've lost poor McAllister it's your own fault.'
And it was cold tea and coffee, burned toast and chops and sniffs all the day long.
He was no longer her treasure, but a naughty child, too old to be smacked.
She had seen salvation for him in McAllister, that was for sure, and if Neil was more doubtful, about that and about her, with Havvie's hateful words still ringing in his head, and his memory of her tortured face growing the more painful whenever he recalled it, the passing of time only accentuated the agony of his loss.
A night without sleep found him red-eyed and haggard.
Midway through the next morning a footman and a supercilious lady's maid arrived to collect her few belongings, so that the last traces of her life with them had gone.
On the day on which he had lost her he had sat for hours on the bed in the attic where she had lived during her time with him, and where they had shared their afternoon of love, and now, on the evening of the second day, he sat there again.
The evening sun fell low upon him.
The empty room mocked him, emphasised the lost vitality of her presence.
Everywhere he went in the tiny house he saw her: sitting at the table, sewing, on her knees, blackleading the grate, laughing at him as she handed him his cup of tea or coffee, and holding up the doll she had dressed for him to admire.
Guilty or innocent, it did not matter; he wanted her back, and he knew at last that, whatever he had felt for Angela Deverill, it was not this.
Earlier, avoiding Matey's compressed lips, he had picked up the whisky bottle and reached for a glass -- anything, so long as he achieved oblivion, surcease from pain -- and then he had heard her voice again, in the corridor at the church hall, mockingly telling him to retreat to it.
He had stared at the bottle in his hand, walked outside, smashed it against the outhouse wall, watched the liquor run down, dark against the brickwork, and returned to the house, almost reeled past Matey as though he had drunk all of it, and mounted the stairs to her room to remind himself of what he had lost, hurt though it might.
In the morning he would go to the embassy, fall on his knees before her, beg her to return, to forgive him -- anything -- because he loved her, and love, if it were true, not only conquered all, but accepted all --'Love is not love, Which alters when it alteration finds,' Shakespeare had said in his greatest sonnet, and he had discovered that for himself -- but too late, too late.
He rose, paced about the little room, stared at the oak bureau, remembered it in his old room, in the nursery; it had come here with Matey -- he remembered her showing him the secret drawer it held, quite capacious, and unthinkingly worked the mechanism which opened it --
To find it, not empty, but holding a leather-bound book, and, beneath it, a small pile of newspaper cuttings.
Neil picked up the book.
It had 'Sally-Anne's Journal' stamped in gold lettering on the cover.
The clippings, he saw, were all from the Clarion Cry and were the articles on East End life, written by Vesta.
A thousand questions clamoured in his head, and the journal held the answers to them all.
The clippings were another, more puzzling matter.
He stared at the book.
The time for gentlemanliness had passed; in a moment he could learn the truth of her, and if to do so involved not behaving like a gentleman, then so be it.
He wanted her back, regardless, but some part of him wanted desperately to know the real truth.
He flipped open the book whose pages were filled with her beautiful, careful script, which he had seen many times on the shopping lists which she had made up under Matey's instructions.
On the cover, facing the first page, was an inscription: 'For Sally-Anne, to keep her English adventures in, from her loving papa.'
The writing was bold and strong and black.
Feeling more than ever like a cur, Neil turned the pages -- but it was all of her that was left to him -- and, he told himself firmly, he would read just enough to discover the truth about her...and why she had hoarded the cuttings.
Later, much later, he put the book down, and the cuttings which he had picked up at the end of his reading.
Halfway through the journal he had jumped to his feet -- he was ready to rush out of the room, to find Havvie Blaine, to kill him for what he had done to her, and for lying about it afterwards.
Common sense prevailed, tears pricked behind his eyes, he sat down again.
He could not help her by attacking Havvie and creating an even worse scandal; that would be self-indulgent, merely seeking to exorcise his own guilt for not believing her.
He learned the truth about the cuttings, closing his eyes when he thought of her sitting in the attic, her long day done, painstakingly writing for O'Connor the articles whose brilliance and feeling told him of the intellect which lay behind her beautiful face and emphasised again what he had thrown away.
But the worst of all was to read what she had finally written on the night before the bazaar, the night before he had added himself to the list of those who had betrayed her -- It was the worst hurt of his life.
I am so happy, as I never expected to be, knowing that he wishes to marry me, has not rejected me as soiled goods as so many would have done.
He is so different from the others, from Terry and Havvie; he loves me truly for myself, which is the best thing of all, not because I am Papa's rich daughter, the King's friend.
We are all going to the bazaar tomorrow.
I have so looked forward to it, making all the things with Matey, and thinking of what we shall do with the money we make -- which is stupid really, when I know that I could so easily give them so much more -- But that would be nothing, for what I have done with Matey has been done by me, and not by Papa, for that is what giving them his money would mean.
And when we get home again I shall tell him the truth -- about who I am, and about Havvie, not the lies about being a governess, and Mama and Papa being dead.
And I shall tell him about the journalism, too, for I want to go on writing, and I want nothing false and wrong to lie between us.
Oh, my darling Dr Neil, you will never know quite what you have done for me.
Not only have you given me your love, freely, but you have made me able to face life again, without fear, be able to face Mama and Papa when I next see them, enabled me to love again, with all my heart.
This is the true, right end of my English journal.
Neil Cochrane, holding the book in his hands, could not even weep over it; he was beyond tears, his hand on the page on which his dearest love, whom he had sent away from him with scorn and mockery, had told of her love for him, and the hope and joy which he had given her.
All turned to ashes because he had not trusted her, or the love she had given him.
He never knew how long he sat there.
Only, very late, Matey knocked on the door, and said, timidly for her, 'Neil?
Neil?
Are you all right?
I am worried about you.
Neil, answer me.'
He put the book on the bed.
In the morning he would take it and himself to the embassy, to beg her to forgive him.
He hardly dared hope that she would change her mind and marry him.
In the meantime there was Matey to reassure -- Matey who had always given him her love and support, whom he had taken for granted, as though her love and loyalty were simply his due.
He was no better than Stair, or Havvie, in his lack of concern for the feelings of those about him, he thought.
Charity began at home, and, though there were larger claims on him in the world outside, what was the worth of his work there if he ignored and hurt those nearest to him?
'Don't worry, Matey,' he said to her, leaving the room of many memories, putting his arms about her, seeing with new eyes how old she had grown, and that he was all she had, the last of the many children for whom she had cared in a long life of selfless service.
'I did not mean to be selfish.
And you were right to reproach me over McAllister.
I treated her abominably, but, God willing, I shall try to mend matters in the morning.'
'Tomorrow's always another day,' she said, trotting out one of her many sayings, and he neither smiled nor mocked at her, but went downstairs to eat the supper which she had prepared for him, because that was the way she showed him her love and concern, and the least which he could do was respond to it.
Jared Tunstall and his wife, splendidly dressed to spend the morning shopping -- Jared to see his English tailor, Mary to examine the latest Paris models at a small shop in Bond Street where she was well-known to the staff -- walked into the embassy hall to hear the sounds of altercation.
The butler was arguing  with a tall fair man, threatening to send for the footmen to turn him away physically if he persisted in demanding to see the person who was refusing access to everyone, and particularly to anyone answering to the name of Cochrane.
Sally-Anne's name floated across to them as they prepared to leave.
Jared's eyebrows rose.
He dropped Mary's arm, and walked over to where Baines confronted the visitor.
He thought he knew who the tall young man might be.
Two days ago, after his daughter had told him her story, he had immediately contacted an enquiry agent whom he knew to be trustworthy and told him to find out all that he could about Dr Neil Cochrane, Sir Alastair Cochrane's younger brother.
The information had arrived by the first post that morning, and Jared Tunstall already knew more about Neil Cochrane than Neil Cochrane did about him.
He had read of a distinguished army career, of a decoration won on the field of battle for bravery, his invaliding out of the army, the fight against injury, and his selfless life as an East End doctor.
And here was the man himself, scarred face, heavy limp, and impressive bearing, demanding to see Sally-Anne, but he gave away nothing of this, merely said, 'What is it, Baines?' and then to the intruder, 'And who the devil are you, sir?
And what do you want of my daughter?'
Neil Cochrane knew immediately who it was who was speaking to him.
This tall and handsome man, not so many years older than himself -- he must have been a mere boy when McAllister was conceived -- could only be by his manner, and his resemblance to her, McAllister's formidable father, so often referred to.
And the beautiful blonde woman with him, although so unlike her, could only be her mother.
'You must be Senator Tunstall,' he said, 'and I am Neil Cochrane, Stair Cochrane's younger brother, lately your daughter's employer.
I desperately need to see her.'
So, this was Dr Neil, so often referred to, thought Jared Tunstall, in his turn.
His daughter's latest love -- and why she had so precipitately left him, he was about to find out.
'Miss Tunstall particularly said that no one was to be admitted to see her,' said Baines plaintively.
'Then you may admit him to see me,' said Jared Tunstall curtly, immediately informing Neil where McAllister got her logic-chopping ability from.
'Mary, my dear, you will excuse us for a moment.
This way, sir.'
And he indicated the library door to Neil.
Oh, thought Neil, following him, you could see at once what a man of power he was, for all his relative youth, and why Sally-Anne was what she was, her father's daughter in every way, so alike in will, intellect and temperament, he dared swear.
The suppressed energy radiating from the man before him reminded him again of McAllister, who possessed all those traits which women were not supposed to share -- and how difficult that had made life for her.
'Well, sir?'
Tunstall was saying to him.
'Why do you need to see my daughter so -- desperately, I believe you said?'
'I don't know what she has told you, Senator Tunstall, but until forty-eight hours ago she was my housemaid; I thought that she was penniless, and that we were to marry.
She talked of an uncle whose permission she might need.
I never thought that her uncle was the American ambassador!
I love her dearly and, God forgive me, when unexpectedly Havvie Blaine came into our lives and lied about her I believed him, and not her.
I debated whether to call on him to kill him for his slanders before I came here to try to see her, and ask her to have me back -- if you think I would make her a suitable husband, that is.
God knows, I have little enough to give her in a worldly sense.
As for killing Havvie, I did not think that she would thank me for the scandal,' he finished simply.
'Oh, you have no need to kill Havvie,' said Jared Tunstall drily, pulling his watch out of his pocket to look at it.
'I should estimate that about three hours ago I ruined his father-in-law and started him on his way to prison, thus effectively ruining my Lord Marquess, too.
You say that you wish to marry her.
She has said nothing of this to me, and as her father I really ought to know, I think, what she, and you, propose to do, even if in the modern fashion you do not choose to ask me for my blessing.'
This last came out in true McAllister style, satiric and sure, another way in which father and daughter were alike.
Had this self-contained man been as impulsive as McAllister when he had been her age?
'She said that you and her mama were dead -- you died of grief after bankruptcy.'
Neil saw Jared's mouth twitch.
'Yes, I know now that it was a whopper, as she says, and I also know now why she said it.
I wish to see her to renew my offer of marriage, although...' and his voice trembled '...
I shall understand why she refuses me, if she does.'
'You may be sure,' said Jared Tunstall, 'that if ever I do go bankrupt I should certainly die of grief, or of mortification at such a turn.
She was right about that.
Perhaps you would like my permission to address her.
You are a man of good family, I believe, but of small means, as you said.
I take it that you would like a handsome settlement on marriage?'
'God forbid,' said Neil fervently.
'Begging your pardon, Senator, but unearned wealth does not seem to have made McAllister -- I mean Sally-Anne, you understand -- happy.'
'No,' said Jared.
'But I think that you might indulge me by accepting a small one -- if she accepts you, that is.
After all, I am her papa.
You think that she would manage as the wife of a struggling doctor, be able to cope with the work and the responsibility?'
'Cope with the work?' said Neil, a little astonished.
'You should have seen her as our maid!
I have never known anyone work like it -- and I discovered today that she had been writing articles for the Clarion Cry at midnight on top of everything else she did in the house.
Being my wife will be a great deal easier than that.'
'So one hopes,' said Jared, amused at these revelations confirming what his daughter had told him two days ago.
'Well, you have my permission to approach her.
Anyone who can inspire her to such industry deserves to win her, and you look as though you will be able to cope with her.
She really needs a strong man by her side; I hate to think what marrying a weak one would do to her -- and to him.
She is a most determined child, you know.'
'Yes,' said Neil, almost humbly.
'But then, I'm determined, too.'
'So one hopes,' said Jared again.
'Stay here a moment and I will try to persuade her to see you -- For what it's worth, you have my blessing, but the decision is hers, of course.'
Her father had gone.
He had not put any pressure on her to see Neil.
Merely said, 'I understand that he is a worthy young man, quite unlike Rourke, or Havvie Blaine, and I think from what he says, and the manner in which he says it, that he cares deeply for you.
But it is your decision, Daughter.'
Well, that was fair enough, and, twisting her fine lace handkerchief in her hands, she had said steadily, 'Yes, I'll see him, but I can't tell you what my answer will be.
Two days ago, I never wanted to see him again --' And her lip had quivered as she spoke.
'And now you're not sure,' her father had said.
'Yes, Papa.
Quite sure that I love him.
Not sure what to do about it.'
Sally-Anne now opened the door to the library, and there he was, his face pale, so that the scar blazed on it -- but what of that?
He was Dr Neil, and she wanted to throw herself at him, and would have done so three months ago, but he, and Vetch Street, and Effie and Rose, and everything, had changed her.
As she walked into the room she saw Neil's eyes hard on her, registering her beautiful gown, her carefully dressed hair, and even though the gown was plain she looked like a queen of society in it, she knew, and she wished that she were wearing McAllister's shabby bottle-green shirtwaister again.
'Dr Neil,' she said, her voice quite steady, which surprised her.
'McAllister -- no, Miss Tunstall,' he said.
He was carrying something, a parcel, and she wondered what it was.
'Oh, McAllister,' she murmured.
'I shall always be McAllister to you, I think.'
Neil did not know what to say.
He had rehearsed a thousand fine phrases, but now before this grave beauty none of them seemed to answer.
He held out the parcel to her.
'I have brought you your journal, and the cuttings,' he said.
'You left them in Matey's secret drawer.
Did she tell you about it?'
Which, he thought, was inane enough, in all conscience, to say to your true love with whom you wished to be reconciled.
Sally-Anne, McAllister, Miss Tunstall, the American Princess -- which was she?-- put out her hand, then coloured and drew it back, her face turning pale.
'I found the drawer by accident.
I forgot that I had left them there,' she said, and then, 'Oh, Dr Neil, you didn't...?'
'Yes' he said, 'I was not a gentleman, McAllister, my darling -- for you are my darling, and come what may, if the whole world separated us, you will still be that.
I had to read it, a little of it, and if listeners never hear any good of themselves, then the little which I read shamed me forever.
Even before it I wanted you back with me, was ashamed of what I said, and afterwards...
And if, because of that, you never want to see me again -- because, after all, I am no better than Havvie; I simply raped you in a different way --' And, having lapsed into complete incoherence, he fell silent, not knowing what to say, except to add humbly, 'I think that your father approves of my suit, but that is no matter.
Oh, dear McAllister, I have stopped making sense.
I have been out of my mind since you locked the door on me on Saturday, and I richly deserved it...'
He was so wild and different from the usually cool and sensible Dr Neil that Sally-Anne stared at him, saying nothing, because unshed tears were choking her, and he took this as a sign of rejection, and began to speak feverishly again.
'I really cannot expect you to accept me after the way in which I have behaved, and then was wicked enough to read your private book, and the cuttings are there, I knew that you would want them back, and I expect that you will wish to make a career in journalism, and why accept a poor doctor, no need to do that, you can always live on your father's allowance and what a remarkable man he is, so like you, or earn your living by your pen...'
How on earth was she to stop him -- her dear Dr Neil who was so distracted by love that all his sense appeared to have left him?
'Yes,' she said simply, breaking in on his torrent of words.
'Yes, McAllister, what do you mean by yes!
That you can earn your living by writing?
Or that you will return to being an American Princess, marry into the Royal family, perhaps?'
He was almost shaking with the intensity of what he was saying.
She crossed the room to him, put her hands on his lips, and said, 'Hush, my dear Dr Neil, hush.
The servant wishes to say something to the Master.
Yes, I will accept you, and it has nothing to do with Papa, or common sense, or anything else.
For the last two days I have known what I lost when I left you and, while you should not have believed Havvie, I can understand why you did.
It would be madness for me to refuse you through hurt pride when all I want to do is be with you, as I have just found out.'
And now she could not stop talking, and he had put the book down, and they were shaking in each other 's arms, and he was kissing her, and she was kissing him, and really this would not do -- they were in the embassy library, and anyone might come in, and, 'Dr Neil, we cannot make love here; it would not be proper,' she said breathlessly.
'Even if you do intend to marry me -- you do intend that, don't you?'
'And you hush, too, McAllister -- yes, you will always be McAllister to me,' he said.
'And yes, of course I wish to marry you, and do you think, my dearest love, that, as Matey would say, we have both been struck by the asylum mop, we are carrying on so wildly, and shall I ever be sane again?
It is relief, I think, and we must marry soon, I cannot bear to wait, and you won't mind living in a little house in Vetch Street when you are used to this, will you?'
And he waved at the magnificent room, and her whole luxurious life which she knew beyond a doubt that she was willing to give up forever.
'You and the little house are all I want, my darling, and a small allowance from Papa, so as not to hurt him and Mama, but not enough to corrupt us -- perhaps to educate the children.'
'The children,' he said fondly, holding her at arm's length.
'What a girl you are, McAllister.
Have you planned how many we shall have, and what schools they will go to?
And, truly, no regrets?'
'None,' she said.
'None.
I have been plotting all morning how to get you back again.
I thought of going to Vetch Street while you were on your rounds, waiting for you, and begging you to take me back, as a skivvy, nothing more, just to be near you, however much you despised me, and hang all worthless pride, and then you came.
And why are you not on your rounds, Dr Neil?
Did you want me so much that you gave them up?'
They were beyond sense, and later, much later, when he met her mama and papa, who had not gone shopping after all, but were sitting patiently in Uncle Orrin's drawing-room, and later still when Uncle Orrin and Aunt Nella arrived, and they all had luncheon together, the lovers' happiness was so patent that Jared Tunstall thought that his wilful daughter had found her true love at last.
And no, there were never any regrets; the Princess had found her palace in Vetch Street.
Chapter 1
Introduction
The last two decades have witnessed vast improvements in the capabilities of computers.
They have become smaller, faster and able to store huge amounts of data.
Despite these technological advances there have been few developments in the method of communicating with the computer, which remains keyboard-based.
Alternatives are via spoken or written mode of communication.
Neither of these are possible, in any general sense, at the present time.
The development of reliable text recognition procedures would serve two important functions.
Firstly, it would make it easier for users who are unfamiliar with keyboards to communicate with the computer by using their normal handwriting.
Secondly, existing paper documents could be scanned into the computer enabling them to be further processed without the necessity of reproducing the original.
Existing systems for performing text recognition are susceptible to errors.
Improvements in these systems could be achieved by incorporating linguistic knowledge into the systems to assist in the recognition of the text.
1.1.
Text Recognition
The difficulty of the problem facing text recognisers is related to the format of the text.
For the purpose of this project text is categorised in terms of how it is produced (i.e. printed by a machine or handwritten by a person), and how it is input to the computer for recognition (i.e. is it recognised dynamically as a person writes or already existing text which is scanned in).
Printed text may be produced by a computer or typewriter.
The text may either be in a single font or a mixture of fonts (e.g. Courier, Roman, Helvetica, Bold, Italic etc.) and different font sizes.
With the increasing popularity of desktop publishing systems (DTP) the use of a variety of font styles is becoming increasingly common.
Handwritten text, either cursive or hand-printed, is much more difficult to recognise than printed text.
Hand-printed text is less natural for adults to produce but is simpler for a computer to recognise.
Cursive script is the mode in which a person normally writes text but is more difficult still to recognise automatically.
The method by which the text is to be input to the computer is also important.
There are two general methods for inputting text into a computer.
With static (off-line) input the text already exists on paper and is digitised by a scanner.
The scanner converts the text to a pixel representation of the text.
Recognition software then attempts to determine the text that was written.
Dynamic input of the text involves the user generating the text at run-time using either an electronic bitpad or electronic paper.
With dynamic input the order in which the pixels are formed is known and provides accurate stroke-position and time-sequence information.
This simplifies the recognition problem since overlapping points may be simply discriminated between and the problem becomes one of pattern recognition.
Static input lacks temporal information -- there is no indication about the pen's direction or speed, nor are the locations of the pen-up and pen-down movements known.
Processing static input is also more difficult because the scanning procedure often introduces additional noise to the input.
The application of dynamic input is more restricted than static recognition since it demands that the user have available a suitable input device at the time of writing and is obviously only applicable to handwritten text.
Static input allows for a greater variety of input formats and makes no requirements about the availability of hardware at the time of writing.
1.2.
Motivation for Text Recognition
There are two principal reasons for doing text recognition.
The first is to provide a more natural method for communicating with computers.
The second involves conversion of already existing paper documents into a format that the computer can process.
1.2.1.
Natural Communication with Computers
Communication with computers and computer controlled machinery is normally achieved using a QWERTY keyboard.
The QWERTY keyboard has its origins in the days of early mechanical typewriters.
The keys were organised in such a way to prevent the typist from typing too quickly and thereby prevent the mechanically-operated rods of the typewriter from sticking.
(An alternative story suggests that the top line of the keyboard was arranged to enable typewriter salesman to find the word 'typewriter' easily.)
The arrangement of the keys is now so firmly established that attempts to alter it have met with failure.
Millions of typists throughout the world have trained on this keyboard layout and are (understandably) resistant to change.
Hence the QWERTY keyboard remains.
Unfortunately the problem with the QWERTY keyboard is not simply the order of the keys -- good typists are able to achieve very high speeds.
The major problem with the QWERTY keyboard is that it needs to be learnt.
Non-typists have difficulty finding the desired keys and may take a long time to type even a short word.
Longer communications with the computer are extremely difficult for novice users and almost certain to produce errors.
An additional problem is that while concentrating on the keyboard when using an interactive program the user may miss important happenings on the screen.
Attempts have been made to improve upon keyboard communication with the computer.
Most noticeably the mouse has become a prominent computer accessory.
Quite complicated functions may be performed with mice via the use of menus and icons.
However a mouse is not a replacement for a keyboard, simply a supplement.
In order to input text it is still necessary to use the keyboard.
As computers become increasingly available to non-typists more natural means of communicating with them are sought.
1.2.1.1.
Speech or Writing
There are two natural methods of communication available -- speech and handwriting.
A commonly used argument in favour of speech recognition is that it is the most natural communication medium.
Children learn to speak and understand speech first and without the formal instruction needed when learning to read and write.
As a result speech is often regarded as being easier to recognise than text.
Although speech is the more natural communication method for humans, literate people feel no sense of priority between speech and writing -- indeed, many people also find it easier to organise their thoughts using pen and paper.
It should be remembered the situations in which speech is used.
Speech is normally used in face to face interaction whereas writing is used across barriers of space and time.
Unless the speech is being used as part of an interactive dialogue system then it is likely that writing is more natural for the situation.
Other advantages proposed for speech are that the hands are free to perform other tasks and that communication may be carried out over a standard telephone line without the requirement for additional hardware.
There are arguments in favour of handwriting recognition.
Writing allows private communication with the computer that is not possible with speech recognition.
A frequently cited example of this situation is a doctor's surgery where the doctor may wish to take notes during a patient's examination.
There may be details about the patient's condition which the doctor wishes to record, but not discuss in the presence of the patient.
Writing is also more flexible than speech.
Drawings and sketches may be produced (a picture paints a thousand words) with a pen, but not easily by voice.
Many facets of spoken language are absent from written language.
Different degrees of planning are associated with speech and writing.
Spoken language is spontaneous which results in hesitations (errm, um) and false starts, and a grammatical structure can be cut-off in mid-production and replaced by another (Hindle, 1983a).
The transient nature of speech does not permit editing of the speech signal.
False starts do occur in written language but they are far less common.
Writing also provides a permanent representation, hence the writing may be manually edited, thereby simplifying the identification of false starts.
Written language also tends to be more grammatically acceptable than spoken language.
Non-standard grammatical forms such as'we was' do not tend to appear in written form (except in literature when trying to convey speech).
Written language also tends to be more conservative with regards to the introduction of new words.
Neologisms come and go very quickly in spoken language but tend to be less frequent in writing.
For an extended communication with a computer the ability to edit the input is an important consideration.
It is often inconvenient (or impossible) to complete a communication with a computer in a single session.
The ability to modify and alter what has gone before is therefore a necessary part of any system.
This point is conceded by Jelinek, a leading speech recognition researcher, who admits 'To get a really good interface, it may prove necessary to use writing tablets and achieve automatic recognition of hand-produced diacritical marks and possibly of handwriting as well.'
(Jelinek, 1985b).
Speech is heavily supported by gestures and other cues from context.
Much of the meaning is implicit and therefore absent from the speech signal.
For handwriting, the writer and reader are apart and so there is much less support from context in the signal.
1.2.1.1.1.
Problems for speech and writing recognition.
There are three main problems involved in the recognition of speech and handwriting.
These are variability, segmentation and operational difficulties.
The principal problem for both speech and handwriting recognition is the variability of the input.
One person's speech differs from another's due to both physical (physiology of vocal organs, age, sex and state of health) and sociological (accent) reasons.
Even words spoken by the same person are never identical (Vaissiere, 1985).
A speaker makes heavy use of prosodic features to convey meaning.
There are three parameters by which a speaker is able to modify the meaning of the utterance -- pitch, volume and tempo.
The same sentence may be expressed as either a question or a statement simply by varying it's pitch (e.g. 'They're here!' and 'They're here?').
There are a wide range of variations in pitch available to express nuances and degrees of emphasis.
The volume of the utterance may be loud to express anger or whispered for secrecy.
Furthermore individual phonemes may also be stressed.
The tempo of speech differs depending on the urgency (rapid) or deliberation (slow) of the speech.
The combination of these features emphasises a large problem for speech recognition -- the same word may be spoken in many different ways and never match exactly what the system was trained with.
Writing is also very individual.
Writing is a motor skill which ensures that the same words written by different people are distinct -- allowing signatures to be used as verification of identity.
There are a number of dimensions in which writing can vary including size, slope and connectiveness.
Even a single writer is seldom able to reproduce exactly the same writing.
However, many of the prosodic problems found with speech are absent from writing.
The orthographic representation of a question or statement is achieved by the use of punctuation.
Written language is generally structured.
For example, it tends to be divided into paragraphs to simplify the structure of an argument.
The presence of such features can provide assistance in recognition of the text.
Segmentation is another problem that affects both media.
The problem arises for speech because speech sounds are not discrete.
Rather they overlap and as one sound is produced the vocal apparatus is preparing to speak the next one.
As a result segmentation of the signal into the constituent units of speech is difficult and requires knowledge of the language.
Likewise, in cursive handwriting it is often difficult to determine where one letter stops and another starts.
The segmentation problem for speech may be illustrated by listening to a foreign language about which one has no knowledge and trying to locate the word-breaks.
This is not simple.
Normal speech is so rapidly and informally articulated that there are seldom pauses between the words (this may be verified from the waveform of the speech signal).
The situation is different for native speakers of the language who automatically perceive the speech as being chopped up into discrete units.
Words spoken in connected speech often sound different than when spoken in isolation.
Connected speech can lead to elision (omission of sound, e.g. 'go away' is often pronounced as 'go way') and assimilation (adjacent sounds are modified to sound more like each other, e.g. 'ten bikes'is pronounced as /tem baiks/).
Therefore training a connected speech recognition system with isolated words may not be satisfactory for a connected word recognition system (McInnes & Jack, 1988).
Handwriting recognition is also plagued by problems of segmentation.
The detection of word boundaries in written text is relatively simple (due to the custom of leaving spaces between them despite the absence of pauses in speech) however the segmentation of a word into its constituent characters is difficult.
This is especially true for cursive handwriting where a number of alternative segmentations may be possible for a word.
For example the letter d is visually similar to the letters cl.
A further problem is that handwritten letters are contextually dependent on the surrounding letters (e.g. the letter e is often written differently in the words gent and rent).
Despite these problems the detection of possible segmentations is less problematical than for speech.
Both speech and handwriting recognisers suffer from operational difficulties.
For speech recognition the principal operational difficulty to be faced is the interference to the acoustic signal from background noise.
It is necessary to filter out the sound that does not correspond to the speech from the input signal -- in a noisy environment this is problematical.
A further problem can be the microphone performance -- different microphones can produce different acoustic signals which need to be standardised.
Until recently a problem with handwriting recognition was that the tablet on which the writing took place and the screen on which it was displayed were physically separated.
However, recent hardware developments have resulted in electronic paper -- a tablet-LCD device which unifies the computer monitor and graphics tablet.
A pen (or stylus) is able to draw on the LCD-screen and be shown as electronic ink.
Perhaps the most telling argument for the use of handwriting instead of speech as a natural communication medium is the current state of the commercially available systems based on speech and writing.
Despite considerable research and investment a large discrepancy exists between human and computer recognition of speech.
Currently available speech recognition systems impose a selection of constraints on the input to simplify the speech recognition task.
Typically these constraints may include (however these constraints do not apply to all speech recognition systems):
Limited vocabulary
The vocabulary size for a system is often very small (of the order of hundreds of words).
Speaker dependence
For a new speaker to use a system the speaker must train the system to her voice.
Training procedures may be both long and tedious involving repetition of sets of words.
Disconnected speech
Brief pauses are required in the speech to discriminate between different words.
The resulting speech is unnatural both to produce and to comprehend.
In contrast pen-based systems have received much less attention than voice-based systems but a number of commercial products are currently emerging on the market (e.g. Penpoint, Paragraph and PenWindows).
In general, the handwriting recognition systems supplied with these systems are based on printed handwriting.
The commercial success of even such limited systems would suggest that now, more than ever, handwriting presents a medium for natural communication with computers that is convenient, flexible and possible.
1.2.2.
Static Recognition of Documents
Although communication between computers is becoming increasingly common, paper (and FAX) remains the principal medium of communication in the business world.
Unfortunately text on paper cannot be manipulated by computer.
The need to enter printed text or data into computers without the effort of re-keying it is the driving force behind the development of optical character recognition (OCR) technology.
An added bonus is that much faster input processing rates are possible with OCR than manual entry of the text.
Using this technology text is entered into the computer via a scanner and software then classifies the characters from the pixel representation.
Typically OCR systems are used in offices to create electronic copies of letters received from other offices, and for the inclusion of spreadsheets and tables in printed reports.
They are also of use in transferring text between incompatible word processing systems or in the restoration of an electronic file that has been accidentally erased.
1.3.
Linguistic Processing for Text Recognition
A number of factors can cause interpretation errors by OCR systems.
They may arise due to poor quality of print, letters being printed together, a variety of fonts being used, the document being submitted to the scanner at an angle, an unknown font being used, etc.
There is also the problem that some letters are visually very similar.
For instance the letters h and n, or c and e are easily confused.
There may also be problems in deciding on the segmentation of a word into letters e.g. the letter combination cl can easily confuse with the letter d.
Rosen (1991) classifies the errors that can arise and these are shown in table 1.1.
Of these errors the most common is substitution.
Handwriting recognition is much more difficult than OCR due to the variety of handwriting styles between writers.
Handwriting recognition systems experience similar errors to those of OCR systems but to a greater degree.
Recognition of cursive handwriting is especially prone to errors due to the difficulty of determining the correct segmentation of a word.
The general output from a character recogniser is a lattice of the possible characters with indications about how close the match was between the character and character templates stored in a database (see $1.4.2).
Selection of the correct character combinations from the character lattice requires some higher-level knowledge.
1.3.1.
The Human Expert.
A primary motivation for using higher-level knowledge to improve the performance of a text recognition system comes from the human ability to recognise text.
Humans are able to identify and discriminate a word from tens of thousands of other possible words.
They are also able to do this very rapidly (approximately 240 words per minute) and convert the written symbols on a page into a mental concept.
This capability far exceeds the performance of any computational system.
The examination of the techniques that humans employ when reading and try to incorporate them into a computational system is therefore worthwhile.
1.3.1.1.
The use of higher level information.
A wealth of evidence exists to prove that reading involves much more than simply recognising the individual characters that form a word.
Early experiments (Cattell, 1885) found that letters are easier to read when they form a word than when they do not (the word superiority effect).
This finding came from testing the ability of readers to recognise strings of letters when only briefly exposed.
If the string of letters were totally random and did not form words then the subjects could accurately report a maximum of three or four individual letters.
However, if the strings of letters formed words then they were able to accurately report a number of short words containing more than three or four letters given the same exposure.
Hence the time taken to recognise a string such as FONHGTAEW is much longer than that taken to recognise the arrangement FOG HAT NEW.
More sophisticated tests (Reicher, 1969) confirmed Cattel's deduction that familiar written words are perceived as units, not as strings of letters.
Shannon (1951) demonstrated that there is over 50% redundancy in the English language by investigating the ability of readers to predict what letters were missing from words.
He found that a reader's familiarity with the language enabled his subjects to fill in missing letters.
The experiment consisted of presenting an unseen passage of text to a person and asking her to guess one letter at a time.
If the guess was correct the subject was told so and moved on to the next letter.
If the guess was wrong then she was told so and asked to guess again.
Below is some of the text that Shannon used in the test with spaces between words being shown as underscores.
If more than one attempt was required to guess the correct letter then the character is emboldened and the number of attempts shown as a subscript.
The results obtained are impressive.
Of the 102 symbols, 79 were correctly guessed first time and only 8 symbols required more than 5 guesses.
Eye movement studies provide further evidence of the redundancy present in English.
The eye does not sweep smoothly from left to right when reading but instead makes a series of fixations with rapid movements (saccades) between fixations.
Analysis of the eye fixations that people make when reading provides some insight into what visual information is being processed.
From such studies (Just and Carpenter, 1987) it has been determined that only about 65% of the words in a passage are fixated -- substantial proof of the importance of higher level processing.
It has been shown that readers differentiate words depending on whether they are content words (e.g. adjectives, adverbs, nouns and verbs) or function words (e.g. conjunctions, articles and prepositions).
Readers fixate a high proportion (over 80%) of the content words in a text compared to only about 40% of the function words.
This differentiation between function and content word is further substantiated by letter cancellation studies which showed that letters in function words are more often missed than letters in content words (Smith and Groat, 1979).
Eye movement studies have also been used to show that humans employ syntactic information when reading.
Carpenter and Just (1983) recorded the fixations made by readers when reading a syntactically ambiguous sentence (' The pickpocket stood before the black-robed judge entered the courtroom to convene the jury.')
Most subjects initially interpreted 'before the black-robed judge' to be a prepositional phrase meaning 'in front of the judge'and were surprised by the word 'entered'and consequently spent longer time on it.
A control sentence which substituted the unambiguous conjunction 'while' for 'before'provoked no pause at the word 'entered'.
This provides evidence that readers try to determine the syntactic role of 'entered' while fixating it.
In summary, humans employ much more than just the visual stimuli when reading.
Many processes operate beyond the simple pattern matching of the individual characters.
In order to understand language humans must use their knowledge of word structure, world knowledge, syntactic structure, semantic information, idiomatic expressions and word frequency when reading.
Simply improving the performance of the pattern recognition module will not produce a recognition performance comparable to that of a human.
For an automatic text recognition system to succeed it should exploit as much of the higher level information as is computationally possible.
There are a number of different knowledge constraints that can be applied to improve the performance of the pattern recognition information.
1.4.
System Overview
For a text recognition system to succeed it is necessary to combine a number of knowledge sources (e.g. syntactic, semantic, lexical and pragmatic).
The work reported in the present thesis was carried out in the context of the development of a complete recognition system (Boes et al, 1989).
This system incorporates several higher level knowledge sources and a general overview of the system is provided in figure 1.1.
1.4.1.
Character Recognition
The first level of recognition involves character recognition of the input.
A number of processes are involved in this level (Wright, 1989).
For dynamic input (i.e. handwritten on a graphics tablet or electronic paper as recognition occurs) the text consists of a sequence of x-y co-ordinates which are vectorised using Freeman encoding.
Fig 1.2 gives some indication of the Freeman encoding process.
The character a is written in Fig 1.2(i).
The way that this is input to the recogniser is as a sequence of x-y points with an indication of the order in which the points were written (Fig 1.2ii).
Adjacent points are then converted to vectors resolved into the directions shown in Fig 1.2iii.
By knowing the vector directions the character can now be converted to a numerical sequence.
Hence the letter a shown would be the numerical sequence 3456712601.
The permutations available from using an unrestricted number of vectors are too vast to provide a database matching scheme.
Therefore a reduction stage is used to reduce the vector sequence to a maximum of five vectors.
The reduction process determines which five vectors are the most significant of a vector sequence based on the lengths of the vectors and how different the vector's direction is from its predecessor.
A typical result of this reduction process is shown in fig 1.2iv.
Hence the letter can now be classified as the numerical sequence 45716.
The Freeman coding can be further enhanced by supplementing it with an indication of what proportion of the whole character each vector represents.
A database is created that contains a selection of such vector sequences and the letter that each sequence represents.
At recognition time the input is reduced in the same manner as the training database.
The vector sequence obtained is matched against what is stored in the database.
The selection of letters that are similar to what was written are retrieved with an indication of how close a match was obtained.
A similar recognition process is carried out for static input.
The difference in this case is that, given the character points in fig 1.2ii there is no indication about which sequence the points were created in.
Hence a pre-processing level is required which determines the probable sequence (Wright, 1989).
This process is not completely reliable, introducing further difficulty into the recognition procedure.
Other systems use various other techniques, along the same lines (Tappert et al, 1990).
1.4.2.
Word Recognition
The rationale behind using word structure to aid recognition is that people normally write words rather than random character combinations.
Hence character combinations that form words are considered to be more probable than character combinations which are not words.
Also recognition experiments (e.g. the word superiority effect) show that humans recognise on a word-level basis.
Within the speech recognition community word-level restrictions are also utilised.
Checking of word structure may be achieved by looking up each of the possible character combinations in a list of words.
More than one word may be created from the character lattice.
For example the letters d and cl are physically similar.
Hence it is easy to envisage the words 'dogs' and 'clogs'being confused.
The character recognition phase may propose a selection of possible characters for the input.
This results from uncertainty about the segmentation of the input string (i.e. where does one character stop and the next one start in a cursive word).
Typically the characters are presented as a lattice of possible characters.
Table 1.2 shows the contents of such a lattice that was obtained for the word its (the graphical representation is shown in Fig 1.3).
Each segment of the input word has a set of possible character combinations with a probability score to indicate how close the character matched the character stored in the database.
There is also a list of the next segments to be tried in the lattice.
The word recognition module must traverse the lattice and determine which of the possible character sequences form valid words (i.e. words in the lexicon).
For the example shown there are three possible words to be found in the lattice: its, us and ox.
The syntactic and semantic information about each of these words is then made available to the relevant processors.
A numerical index (the root index) is also returned which is an indication of the morphological root form of the word.
Other systems use sub-word information to verify whether character combinations are legal (e.g. Higgins and Whitrow (1984)).
Rather than store the complete words in a lexicon, such systems store the probabilities of letter transitions.
Very few systems attempt to use any information beyond the level of word recognition.
1.4.3.
Compound Recognition
The identification of common phrases and word combinations may assist in selection of the correct word.
For example, if the words prior to dogs/clogs had been raining cats and then it would be fairly safe to say that the intended word was dogs.
An additional necessity for performing compound recognition is that many of the compounds will have different syntactic and semantic properties than the individual word combinations.
Without this level of recognition incorrect assumptions will be made about the words.
Compound recognition determines whether consecutive words are able to form idiomatic or compound lexical entities (i.e. words containing spaces).
Essentially it involves reducing the individual words to their root index.
The combinations of words that form compounds thereby become number sequences which are stored in a compound tree.
This will be dealt with in detail in chapter 4.
1.4.4.
Syntactic structure.
Humans find it much simpler to process grammatically well-formed language than ill-formed input (Miller & Isard, 1963).
One can imagine the use of a recognition grammar which states what language is well-formed and prefers such input to ill-formed alternatives.
Hence such processing would prefer combinations of the form the cat sat on the mat to cat mat on sat the the.
The application of such information to assisting in the recognition of text forms the basis of this thesis.
1.4.5.
Semantic Processing.
The aim of semantic processing is to demote word combinations that are not meaningful.
What constitutes meaningful is the basis for a philosophical argument.
There have been a number of semantic theories proposed.
One factor that unites all of these theories is that no computational system exists which is able to cope with unrestricted English.
The manner in which semantic information is applied in the current system is to utilise dictionary definitions and collocational information from corpora (Rose, 1990).
Collocational information is obtained by taking a corpus and finding how often pairs of words are found together.
A collocation 'definition' is then stored for a word.
The possible words in the input are given semantic scores based on how well they combine with possible neighbouring words in the input.
The way in which the definitions are used is as follows.
Two words are awarded a high semantic score if the definition of one word contains the other.
A weaker indication is that their definitions contain a number of similar words.
For further details see (Rose & Evett, 1992).
1.5.
Sources of Knowledge
Having determined the kind of information it would be advantageous to utilise in a handwriting recognition system the next problem which must be addressed is where such information may be found.
There are three main sources of machine readable information: machine readable dictionaries, corpora and existing parsers.
Machine Readable Dictionaries (MRDs)
To find out how a word behaves the obvious place to look is a dictionary.
Dictionaries are the results of a lexicographers analysis of a language.
The information contained includes the syntactic and morphological behaviour, phonetic description and a definition of the words contained.
A number of types of dictionaries exist, some of which are in a computer-readable form.
Each dictionary differs in the information it contains and the style in which the information is presented.
For example the information within the Oxford English Dictionary gives a description of almost all of the words in the English language and their etymological history.
A learner's dictionary will define far fewer words and concentrate on examples of usage for those words.
Corpora.
Corpora provide a representative sample of language.
Corpora may be raw (text only), tagged (each word is assigned a syntactic category) or analysed (parsed)(see $4.5.2).
Many of the syntactic phenomena that occur in English are rare.
In order to capture as much information as possible corpora must be large.
Corpora are becoming increasingly popular within linguistics to evaluate existing natural language systems, investigate the occurrence of linguistic features and the production of probabilistic models of language.
Existing parsers
Much of the work within linguistics has involved the development of grammars to specify the linguistic regularities of language.
Application of grammars requires knowledge of the grammatical behaviour of the words of the language (stored in a lexicon) and a parsing algorithm to specify how to apply the knowledge of the grammar and the syntactic behaviour of the words to the input.
Rather than duplicate the work carried out in the creation of these parsers, the use of existing parsers should prove to be beneficial.
1.6.
Summary
Alternate ways of communicating with computers are required which do not require a keyboard.
This chapter has considered the two most natural methods of recognition used by humans -- speech and writing.
Despite a large amount of research into automatic speech recognition the results have been unimpressive.
Although recognition of handwriting has received much less attention the task is simpler and promising results have been obtained.
Commercial systems are currently available that can recognise handprinted characters.
An ulterior motive for performing text recognition is to convert existing printed material into a computer format that permits further processing.
Available OCR systems are able to perform this conversion very rapidly, thereby saving much effort and avoiding the necessity of re-typing the text.
There are difficulties associated with automatic text recognition however.
With printed material the font may be unknown to the system or the copy may be of poor printed quality.
The recognition of handwriting is even more difficult due to the number of different ways in which letters are written and the visual similarity of many characters.
Furthermore, with cursive handwriting, words are difficult to segment into the constituent letters.
Humans experience few of these problems when reading.
The principal reason for this is the high degree of reliance a human places on linguistic information.
Humans do not read character by character.
Instead recognition is based on the context in which a word is found.
Incorporation of some of the linguistic information that humans employ is necessary to improve text recognition systems.
The current project is imposing lexical, syntactic and semantic constraints on the input to enhance the recognition rate of text processing systems.
This thesis examines the use of syntactic information for assisting in the text recognition process.
Chapter 2
Literature Review
This chapter reviews research into handwriting recognition and natural language applications that involve syntactic knowledge of some description.
This chapter will provide
A description of the use of linguistic information in handwriting recognition systems
A brief history of the use of syntactic information in NLP systems.
A description of the main theoretical paradigms that have been developed and consideration of how they are suited to the text recognition task.
2.1.
Handwriting Recognition
In comparison to speech, automatic handwriting recognition has received little attention.
Early attempts at the automatic recognition of handwriting date from about 1960 (e.g. (Earnest, 1962),(Eden & Halle, 1961),(Eden, 1964),(Frishkopf & Harmon, 1961) etc.).
A hiatus followed until the development of Sayre's static system (Sayre, 1973) re-kindled interest in the problem and numerous systems followed (e.g. (Tappert, 1984),(Brown & Granapathy, 1980),(Higgins & Whitrow, 1984)).
The approaches taken by these systems to recognise cursive script fall into two categories -- either whole word matching or segmentation followed by letter or stroke recognition.
With whole-word recognition the features of a complete word are extracted and matched against a stored database to find the closest match (e.g. (Brown & Granapathy, 1980)).
This approach has the advantage that it is fast since it requires no segmentation of the word.
However this system is restricted to recognising only a small number of words since it must be explicitly trained for every word in the vocabulary.
The alternative approach uses segmentation methods to recognise the word.
The complete word is taken in and segmented into individual characters or strokes (e.g. (Higgins & Whitrow, 1984),(Wright, 1989)).
Segments are then matched against a database of character shapes to produce a lattice of the likely characters.
Word level restrictions may be imposed on the lattice by only allowing character sequences that form words (thereby allowing semantic and other higher-level to be applied).
With this approach the vocabulary can be infinite and the system requires significantly less training.
2.1.1.
Linguistic information in text recognition systems
Despite evidence that reading encompasses much more than simple character recognition little of this language information has been exploited in text recognition systems.
The language information most often incorporated is that of word structure.
Knowledge of word structure is often applied to the results from the character recognition module.
The manner in which the word-structure knowledge is applied is dependent on the output produced by the character recogniser.
For text input some shape recognisers produce a single character candidate for each possible character, whereas others produce a string of alternative character candidates with an associated weight for each alternative for each possible character.
If only a single character is suggested then word knowledge is used to detect and correct errors in the hypothesis.
Examples of this approach are provided by Srihari and Bozinovic (1982) and Ullmann (1977) who developed a probabilistic model of the substitution, deletion and insertion errors that were likely to occur which modified the output from the shape recognition module.
When a shape recogniser is used that produces alternative candidates the application of word structure involves determining which of the possible character combinations form words.
Two general techniques have been used to determine whether a string is a legal combination of characters.
The first is lexicon-based and involves storing a list of the allowed words.
Character combinations are permitted dependent on their membership in this list (e.g. (Wells, 1992),(Srihari, Hull & Choudhari, 1983)).
The alternative approach uses n-gram techniques.
The probability of bigrams and trigrams (letter pairs and triplets) is determined for the language.
This information is then used to rank the possible character combinations that are found (e.g. (Riseman & Hanson, 1974),(Whitrow & Higgins, 1987)).
Both systems have advantages and disadvantages.
Use of n-grams requires little storage or processing but is less efficient at discriminating between acceptable and unacceptable letter strings.
Lexical lookup is more demanding of storage and processing but is better at rejecting unacceptable letter strings.
A major disadvantage of a lexicon-based system is that if a word is absent from the lexicon then it will be rejected by the system.
To reduce the likelihood of this a large lexicon must be stored.
A recent review of research on handwriting research (Tappert, Suen & Wakahara, 1990) showed no practical application of syntactic information for improving handwriting recognition.
More recently, however Crowner and Hull (1991) have carried out work on this application.
The methods used are similar to the statistical techniques that have been arrived at independently in this project.
2.2.
Computational Linguistics
There are diverse motivations for developing systems for processing natural language.
For computational linguists the over-riding concern is their practical use.
In general, computational linguists are more willing to accept computer systems that are able to process a useful range of language input.
Incomplete systems are accepted as a fact of life.
Features of language that interest theoretical (academic) linguists include linguistic competence (the way in which people decide whether a sentence is grammatical or not), language universals (grammatical principles that apply to all natural languages) and in finding the simplest, computationally most restricted theory that can account for natural language.
It is characteristic of theoretical linguists that they select example sentences that computational linguists would categorise as pathological.
Another important motivation is that of the psycholinguists who develop computer parsing systems as test-beds for hypotheses about human linguistic processing.
These systems are mainly interested in comprehension and the over-riding aims of such systems do not include wide-coverage or computationally efficient methods.
The different linguistic fields have influenced each other but there have also been marked differences between the fields.
Wilks (1983) argues that linguistic theory has had little direct influence on parsing techniques.
Specifically he points out that parsing is performance motivated and directional (i.e. from surface to structure)-- the very opposite of the aims of competence theory.
It is also the case that the elegant theories created by theoretical linguists need not necessarily evolve into computationally effective techniques.
The most notable case of this is the application of transformational grammar to automatic parsing systems which proved theoretically very appealing, but computationally unsuccessful (c.f. $2.2.3.1.2).
2.2.1.
Syntactic Processing
Syntax is the study of structural patterns in word order and structure.
It is the first level at which human language diverges in principle from the sign systems of other animals.
It is also the area in which most work has occurred in computer systems for language processing.
This is not surprising.
Lexical processing may be adequately dealt with either by the use of wordlists or simple morphological processing.
Higher levels of language cannot be started on until the syntactic processing is complete, since the relationships that they are concerned with are denoted by structural properties of the input.
A much fuller review of many of the syntactic theories and applications is provided by Winograd (1983), which contains detailed descriptions of many syntactic theories and applications.
Details of the theories will be given briefly with attention to how successful they have proved within natural language systems.
2.2.2.
Practical Applications Of Linguistic Processing.
There have been a number of practical (large coverage) computational applications which have utilised linguistic information.
The most common of these applications are machine translation, information retrieval and human-computer interfaces.
2.2.2.1.
Machine Translation
Machine translation (MT) was one of the first computational applications of linguistic information (Weaver, 1949).
This research followed on from the success of code breaking techniques in World War 2 and the supposition that translation was simply a more complex coding of words.
Despite considerable funding the results of these 'first generation' MT systems were far from encouraging.
Frequently the results were so error-prone that it took more effort to correct the translation than it actually did to manually translate the text.
There were a number of reasons for the lack of success.
One must first consider the computers available at the time, which were roughly equivalent in power to an average modern programmable calculator.
Also the translation process amounted to little more than looking up words in bi-lingual dictionaries.
Obviously such methods are far from sufficient (an apocryphal example of the problems that arose involves the translation of the sentence 'The spirit is willing but the flesh is weak' into Russian and then back-translated to English as 'The vodka is strong but the meat is rotten').
Some syntactic knowledge was utilised in these early systems but it tended to be ad hoc and pre-dated current work on formal linguistic theory.
The 1966 ALPAC committee report on machine translation systems concluded that human translation was superior in terms of speed, accuracy and cost, and recommended that funding be curtailed.
Following on from this the amount of research carried out into MT was seriously reduced, although there has been some renewed interest (e.g. SYSTRAN (Toma, 1977) and EUROTRA (Raw et al, 1988)).
These second generation systems employ much more linguistic information, particularly semantic, than their predecessors.
A recent (and controversial) development within MT is the possible use of purely statistical information to achieve translation.
Initial work by Brown at IBM (Brown et al, 1989) has produced results of similar quality (48% correct translations) to those achieved by commercial MT systems.
Many of the near-miss translations in the IBM study result from the complete lack of linguistic information.
Brown suggests that large improvements may be achieved by incorporating simple morphological and/or syntactic analysis based on probabilistic methods.
It is doubtful that MT will ever achieve the performance once believed possible.
However there are currently available a number of MT systems which perform limited translation (of controlled syntax or vocabulary) or translation-related tasks (machine-aided translation).
2.2.2.2.
Information Retrieval
Another application to receive early attention aimed to produce the 'library of the future'.
Two applications have been developed within this field -- text retrieval and question-answering systems.
Text retrieval seeks to interpret what the user says and retrieve the appropriate document from the database.
Most text retrieval systems have been based on statistics and probabilities.
In question-answering systems the computer generates answers to the user's query based on stored information.
In order to achieve these aims it is necessary to store a knowledge representation of the contents of the documents.
Question answering systems have been a central project within natural language understanding.
Numerous projects are based on question-answer dialogues and the development of natural language front-ends for databases.
2.2.2.3.
Human-Machine Interfaces
Natural language is the most convenient method for communicating with interactive systems.
The development of a natural language interface to a database has proved to be more tractable than other applications.
The type of language used for such applications is typically very simple (Diaper, 1988).
Also, the interactive nature of such systems permits uncertainty about the input's meaning to be resolved by asking the user to rephrase the question.
Another aspect of human-machine interfaces is the conversion of text to speech.
This application is of benefit for queries to a computer that take place over a telephone and also for providing reading aids for the blind.
To produce natural sounding speech it is necessary to utilise linguistic knowledge to produce the fundamental frequency and duration of the produced signal.
2.2.2.4.
Speech Processing
Speech processing has been investigated at many levels -- from the shape of speech waveforms to the meaning of whole sentences.
The most significant research effort into the application of linguistic information to speech processing arose from the ARPA initiative (Klatt, 1977) which resulted in a number of systems.
Each of the systems produced was based on diverse, co-operating knowledge sources to handle the uncertainty in the signal and processing.
The systems differed in the types of knowledge, interactions of knowledge, and control of the search.
The most famous of these systems were HEARSAY-II (Lesser et al, 1977) and HARPY (Lowerre, 1976).
Perhaps the most sustained research into speech recognition has been carried out by the IBM speech group which started work on the production of a large vocabulary continuous speech recognition system in 1972.
The group pioneered the use of Hidden Markov Model techniques and were able to demonstrate performances far better than HARPY (Bahl et al, 1983).
Among the systems that resulted from this work is TANGORA, a 20,000 word isolated-word dictation machine (Averbuch, 1987), and a 5,000 word continuous-speech recogniser (Bahl et al, 1989).
These systems are based on HMMs and trigram language models (c.f. $2.2.4.1).
Other recent successful speech systems include BYBLOS (Chow et al, 1987) and SPHINX (Lee, 1989).
BYBLOS used context-dependent phonetic HMMs to produce accurate speaker-dependent continuous-speech recognition of a 1,000 word vocabulary.
SPHINX used multiple code-books, generalised triphones and function-word HMMs to obtain high accuracy speaker-independent performance of the same task.
2.2.3.
Generative Linguistics
The dominant linguistic school over the last thirty years has been generative linguistics which has its foundation in Chomsky's revolutionary Syntactic Structures (Chomsky, 1957).
Generative grammar rejects the empirical nature of structural linguistics and instead uses linguistic intuitions of native speakers.
A structural approach is rejected on the basis that it is unable to capture the creativity of language.
Central to this approach is the distinction between linguistic performance and competence.
Linguistic competence is a person's knowledge of the rules of the language.
Linguistic performance is how a person uses those rules in real situations.
Chomsky believed that the linguists should be studying the competence of language rather than the performance which was studied by the structuralists.
His reasoning was that language is infinite and the study of performance is limited to a very small fraction of the possible sentences of the language.
Within the generative paradigm the following definitions hold: A language is a set of sentences, where each sentence is a string of one or more symbols (words) from the vocabulary of the language.
A grammar is a finite, formal specification of this set.
The grammar is able to take many forms.
If the language is finite then it is possible to simply list the valid sentences.
If the language is infinite then the language may be specified by producing a recogniser program.
The function of the recogniser program is to decide whether a given combination of words forms a sentence of the language.
A parser is an extension of a recogniser which assigns grammatical structure to the input.
The construction of a generative grammar represents an attempt to formulate a system of rules for the formation of the sentences of a language.
The aim is to produce a grammar that generates all of those sentences -- and only those sentences -- that a native speaker judges to be well-formed.
In practice however the complexity of language is such that no generative grammar has come anywhere near providing this degree of coverage of the language.
The rewrite rule is an effective method of representing the rules of a generative grammar.
An example of a simple rewrite rule is:
which can be interpreted as: A noun phrase, NP, consists of a determiner, det, followed by a noun, N.
The Chomsky Hierarchy
Although a complete generative grammar of English does not exist, one approach linguists have taken is to formulate different types of grammar and determine whether any particular grammar is in principle adequate.
Chomsky (1957) formulated three types of grammar -- finite state grammar, phrase structure grammar and transformational grammar.
He was able to prove that finite state grammars are in principle incapable of generating certain sentences
Since the appearance of Chomsky's Syntactic Structures linguists have been deeply concerned with the form or type of grammar that, in principle, could adequately describe natural language.
The Chomsky hierarchy of languages (table 2.1) identifies four classes of grammar and is a crude dimension of the power of that grammar.
The least powerful grammar is type 3, the most powerful type 0.
A language is said to be recursively enumerable if a program could be written that would print out each sentence of the language.
A recursive language is one for which a program could be written that would either specify whether any given sentence belonged to the language or not.
Not every recursively enumerable language is recursive.
Only if the language is finite is it possible to state with certainty that any given sentence belongs to the language.
Context-sensitive grammars are phrase structure grammars of the form:
where the length of x is less than or equal to the length of y (i.e. there must not be more symbols on the LHS of the rule than the RHS).
e.g. 
This restriction ensures that the grammar is recursive.
Although they are more powerful grammars than context-free grammars, context-sensitive grammars have been found to be less well suited for stating grammatical constraints than indexed languages.
In a context-free language, every grammar rule is of the form:
where A is a non-terminal form (i.e. not a word) and x is a sequence of zero or more terminal and non-terminal symbols.
Below is a small context-free phrase structure grammar (CF-PSG).
The key features of a context free grammar are the use of a set of finite grammatical categories, and a finite set of rules for specifying how the LHS of the rule may be created from the sequence of RHS elements of the rule.
The term context-free indicates that there is no restriction on the context in which the rule occurs.
Context-free grammars are widely used for specifying both formal and natural languages.
Between context-free and context-sensitive grammars come the indexed grammars.
Indexed grammars are context-free grammars that are augmented by the ability to test for feature constraints.
This ability removes the restriction on context-free grammars that only a finite set of grammatical categories are allowed.
Indexed grammars have been usefully employed in ATNs (c.f. $2.2.3.1.3).
Finite state grammars (also called regular grammars) are the weakest grammar in the Chomsky hierarchy.
Finite state grammars are based on the view that sentences are generated via a series of choices made 'from left to right'.
In other words once the first element has been selected every subsequent choice is determined by the element immediately preceding it.
Other than this we do not need to know anything about the portion of the sentence we have already generated.
Finite state grammars are conveniently expressed by a state diagram.
Fig 2.1 shows a very simple finite state diagram.
In the diagram, grammatical categories are shown by letters (d for determiner, a for adjective, n for noun).
This represents the language d(a)n where the brackets indicate that the category may be found zero or more times.
Hence the diagram represents a determiner followed by any number of adjectives followed by a noun.
Finite state grammars are unable to represent anything other than a very simple sequence.
An obvious difficulty is the specification that an element may be found one or more times.
The major failing with finite state grammars is their inability to deal with any dependencies that exist between non-adjacent words.
2.2.3.1.
Grammatical Framework
Perhaps the most significant decision to be made when developing an NLP system is the selection of the underlying theoretical framework.
There are a number of grammatical architectures that have been proposed and implemented during the last 30 years.
The following briefly reviews these systems.
2.2.3.1.1.
Finite State Grammars
Although finite state grammars are weak they have been used in computational systems.
One application that benefited from their use is given by Rabiner and Levinson (1985) for speech recognition of flight reservation information.
This grammar took advantage of the limited vocabulary (127 words) and constructions permitted in the domain.
Part of the grammar used is shown in fig 2.2.
The full grammar had 144 states, 450 transitions and 21 terminal states.
The language specified by this grammar had in excess of 6*109 sentences which were all syntactically and semantically well formed.
Significant improvements were found in the recognition performance of the recognition system.
Unfortunately, development of the system to cope with more general language is not feasible because the development of this grammar is inherently based on the sublanguage task.
2.2.3.1.2.
Transformational Grammar
One of the most dominant syntactic theories involves the use of transformational grammar (Chomsky, 1957)(Chomsky, 1965).
The underlying concept is that pairs of sentences that have constituents in common can be related to each other by a linguistic transformation.
For example the sentences 'The dog chased the cat' and 'The cat was chased by the dog'are related by the passive transform.
Other transforms include the cleft transform which results in 'It was the dog that chased the cat' and the question transformation giving 'Did the dog chase the cat?'.
Transformational grammars consist of a context-free (base) grammar and a set of transformational rules that map syntax trees onto new (derived) syntax trees.
The base component produces a deep structure tree.
The transformational component is a set of tree-rewrite rules which specify how to produce surface structure trees from the deep structure tree.
Transformational grammars are inherently used for sentence generation not analysis.
Grishman (1986) specifies a number of problems in reversing this procedure to use transformational grammars for the a recognition procedure:
1
Assigning to a given sentence a set of parse trees including all of the surface trees which would be assigned by the transformational grammar.
2
Given a tree not in the base, determining which sequence of transformations may have been applied to generate this tree.
3
Having decided on a transformation whose result may be the present tree, undoing this transformation.
Methods have been used to overcome these problems to some degree.
The first problem could be dealt with by the development of a context-free covering grammar -- a set of all of the surface trees that can be generated by the transformational grammar for all of the words in the input sentence.
Many false trees will be generated from such a grammar, especially given the syntactic ambiguity of many English words.
A set of reverse transformations could be developed to deal with problems 2 and 3.
However, it is still very difficult to search for the sequence of transformations which produced any given tree.
Matthews (1962) suggested the use of synthesis (the generation of all possible sentences and checking to see whether they match the input) to apply transformational grammars.
Obviously this is computationally infeasible (and was never implemented).
More practical attempts at using transformational grammars for analysis incorporated an 'augmented grammar'(e.g. MITRE (Zwicky, 1965) and TQA (Petrick, 1981)).
The augmented grammar contains the rules of the original grammar plus rules which characterise the structures that the transformations can add.
The input is then parsed into a tentative surface structure on which the transformations can be performed to determine the deep structure.
Given syntactic ambiguity and the recognition systems ambiguity of words this can be computationally prohibitive (e.g. MITRE takes 36 minutes to parse an 11 word sentence).
2.2.3.1.3.
Augmented Transition Networks (ATNs).
Following on from the impracticability of transformational grammar the introduction of ATN grammars (Woods, 1970) provided a practical application of linguistic theory.
For a full description of the ATN formalism see Bates (1978).
ATNs represent the grammar as a set of networks displaying the possible orderings of constituents in a grammar and the various options that the parser will have at any stage in the processing.
A successful ATN parse is obtained by traversing a path of arcs from the initial node to a success node using up all words in the sentence.
If this is achieved then a sentence is produced otherwise failure is announced.
Through the use of registers ATNs are able to enforce further restrictions on the input such as number agreement.
The resulting structure from an ATN is dependent on the grammar writer and can take many forms.
One method is to take the information stored in the registers at the end of the parse and build the resulting structure from this.
Fig 2.3 shows a portion of the ATN grammar used in the BBN speech understanding system (Woods et al, 1976).
This network defines acceptable noun phrases as consisting of the categories determiner, optional adjective string, noun and optional prepositional phrases.
ATNs have been widely used in natural language processing systems over the last two decades (e.g. LUNAR (Woods, 1973), PROGRAMMAR (Winograd, 1972), SOPHIE (Burton & Brown, 1979)).
The speech understanding system, SPEECHLIS (Bates, 1978), contained one of the largest ATN grammars (the grammar contained 448 states, 881 arcs, and 2280 actions).
This system was written for the purpose of understanding travel-expense reports.
The grammar took advantage of the limited domain by employing semantic categories and contained many domain specific word combinations.
Much of the use of ATNs stems from their clarity when used in small systems.
As a formalism ATNs give extensive possibilities for optimisation and tight control over the precise details of the parsing strategy.
The principal drawback with ATNs is the complexity that arises when a system becomes large.
The ATNs are bound up with inter-relations to other transitions making it difficult to determine the effect and purpose of any given transition.
As a system gets larger the logic becomes more obscure, modification more risky and debugging increasingly problematic.
Due to the increased complexity arising from multiple word candidates in the input and the requirement that the system provide a wide coverage of English the use of an ATN-based grammar was considered unsuitable for the current application.
2.2.3.1.4.
Marcus Parsers
Alternately referred to as Wait and See Parsers (WASPs)(Marcus, 1980) these parsers reject the backtracking control strategy used by ATNs to cope with local ambiguity.
Marcus's parser, PARSIFAL, was initially designed to mimic human linguistic ability.
The parser was shown to be incapable of processing garden-path sentences (i.e. sentences which cause incorrect syntactic premises to be made e.g. 'the cotton clothing is made of grows in Mississippi').
That humans also fail on these sentences was regarded as support that it reflected how humans process ambiguity although arguments exist against this inference (Briscoe, 1987).
Marcus parsers behave in a strictly deterministic manner; no backtracking is permitted and all of the structures created by the parsers are permanent.
Two data structures are used: an active node stack and a three-cell constituent buffer.
The active node stack is a pushdown stack of incomplete constituents.
The constituent buffer contains constituents which are complete but whose higher level grammatical function is uncertain.
This buffer is used to 'look ahead' at the data which will follow before making a decision about the parse to assign to the input.
An advantage of this type of parser is that it performs recognition in linear time.
The disadvantages of a Marcus parser for the current application include:
1
Extending the system to deal with multiple candidate words could be problematic.
2
The rules needed for this type of parser are difficult to develop for a large grammar.
3
Totally hypothesis driven i.e. It is unresponsive to the actual data.
4
Availability of a large-scale parser based on this approach.
2.2.3.1.5.
Unification Grammars
In recent years unification grammars have come to prominence in the computational linguistics community.
The key idea in the unification-based approaches is that informational structures (known as feature structures) are used to encode a variety of linguistic information (lexical, syntactic and semantic) in a uniform manner.
Manipulation of these structures is possible by means of a small number of well defined operations, the most important operation being unification.
The feature structures consist of features and associated values which may be either atomic or complex (i.e. feature structures themselves).
The operation of unification creates new feature structures and together with some string combining operations pairs the feature structures with strings (Joshi, 1987).
The appeal of unification grammars is that they are able to combine both linguistically theoretical and computationally tractable methods for processing language.
A number of grammars are unification based including Functional Unification Grammar (FUG)(Kay, 1985); Generalized Phrase Structure Grammar (GPSG)(Gazdar et al, 1985); Lexical Functional Grammar (LFG)(Kaplan & Bresnan, 1982) and PATR-II (Shieber, 1984).
Shieber (1986) lists the properties that unification-based grammatical formalisms share:
Surface-based: i.e. directly related to the word order in a sentence.
Informational: The strings have information associated with them from some informational domain
Inductive: The associations of strings and informational elements are defined recursively.
New pairings are derived by merging substrings based on prescribed string combining operations.
The associated information from the strings is merged according to prescribed information combining operations.
Declarative: The associations between strings and the informational domain are declared in terms of what associations are permitted, not how they are to be calculated.
Complex-feature-based: Associations between features and values are taken from a well-defined structured set.
A prototypical unification grammar (e.g. the Alvey Natural Language Tools $3.3) consists of a context-free skeleton which is supplemented by a set of feature-value specifications in the grammar and the associated lexicon.
The feature specifications may involve variables and may be recursive (that is the values may be interpreted as referring to a whole category).
Unification is the controlling factor that determines whether categories are able to combine to satisfy a grammar rule.
Consider the grammar rule:
This states that X may be formed from Y and Z combining.
X receives Y's value for feature foo, and Y receives the value that Z has for feature baz.
This percolation of values is governed by the unification.
The concept of unification is analogous to that found in set theory.
The unification of two categories is the smallest category that extends both of them, if such a category exists, otherwise the unification is undefined.
In this way the problem of feature agreement and percolation of feature-values between nodes in a phrase structure tree is dealt with.
Unification grammars are very strongly based on lexical lookup and employ a lexicon containing very detailed information about the possible uses for the words.
The use of complex lexical structures makes it possible to use simpler grammars.
Unification grammars take advantage of lexical orientation to cover:
i.
Agreement of subject and verb.
ii.
Subcategorization for noun phrases and verb phrases of all types.
iii.
Auxiliaries.
Unification-grammar formalisms have a number of advantages for natural language processing.
Among these advantages are their ability to represent partial information in an elegant way; the inherent potential for structure sharing; the declarative description of information flow, and a mathematically clean and computationally tractable type system with inheritance (Bouma et al, 1988).
2.2.3.2.1.
Representation of lexical information and lexical coverage.
The representation of syntactic information in the lexicon is inextricably linked with the grammar being used.
Unification-based grammatical formalisms tend to employ very detailed information within the lexicon.
The use of such lexical representations allows knowledge to be embedded within lexical representations, rather than rules about those representations.
Existing natural language systems take either the demo or the book approach (Miller, 1985).
Demo systems are generally developed by linguists to explore and explain linguistic phenomena which can later be increased in their coverage of the language.
Lexicons for such systems tend to be small and hand-crafted, reflecting the proprietary nature of lexicons and the difficulties in obtaining the required grammatical information.
Book systems tend to derive their lexical information from existing machine readable dictionaries (MRDs).
The information derived requires modification to create a computationally useful formulation of the information.
For a practical system it is essential to have a reasonable (though not necessarily complete) coverage of the words of the language.
The actual number of words that are necessary is open to debate.
Allen (1980) determined that 12000 morphemes are capable of covering 95% of English text, Jelinek (1985b) investigated the use of personal databases acquired by maintaining a dynamically varying vocabulary of the previous N words used.
5000 words covered 95.5% of text and could be derived from 56000 words of text, 15000 words covered 99.0% and required 640,000 words of text to determine this many individual words.
Of the established parsing systems, very few have substantial lexicons.
Exceptions to this are the Linguistic String Project (Sager, 1981) which had a lexical coverage of about 10,000 words (although many of these are specialist medical terms); the CLAWS system (Leech et al, 1983)(Atwell, 1983)(Atwell et al 1984) used to tag the LOB corpus had a lexicon of approximately 7,200 words plus a 670 SUFFIXLIST which added word-tags to most other words; IBM's CRITIQUE (formerly EPISTLE) system (Heidorn et al , 1982) with a lexicon of approximately 100,000 words and the Alvey Natural Language Tools (ANLT)(Carroll & Grover, 1989), an early version of which contained a lexicon of nearly 7,000 morphemes (probably about 20,000 words).
Recent versions having been further augmented.
2.2.4.
Corpus Linguistics
Bloomfield's publication of Language (Bloomfield, 1933) represented the first major statement to combine both the theory and practice of linguistic analysis.
Bloomfield's approach to linguistics was based on observation of the language.
Any data that could not be observed or directly measured were rejected.
Hence this school of thought held that the tool for the linguist to work with is a collected corpus of naturally occurring language.
This approach represents a behaviourist view of the language.
Structuralist linguists retaliate to the generative linguist's claim for competence over performance with the argument 'An automatic language-processing system which works adequately for competent language but fails on performance is a futile system, because all there is to be processed is performance.'
(Sampson, 1987).
Such issues as whether a system is able to provide an adequate model of how humans process the language are of limited interest to this school of research.
Natural language corpora have been available in computerised form for over thirty years.
In the 1960's computers were rare and very expensive and the only text acquisition method for corpus compilers was manual input.
This placed restrictions on the size of the corpus that could be produced.
With the computer now established as an indispensable accessory in the workplace the volume of text available in a machine-readable format has become massive.
With this easy-availability of text the size of corpora becoming available are orders of magnitude larger than those originally envisioned.
The COBUILD corpus (Sinclair, 1987a) has about 20 million words and the American TEI corpus (Walker, 1989) and British National Corpus (BNC)(Leech, forthcoming) are expected to be in excess of 100 million words.
A by-product of this is that corpus-based linguistics, for a long time ignored in favour of logic-based linguistics, is becoming more accepted and attracting a lot more attention.
Corpus-based systems can be very simple.
A model of the language is created by training it on some text that is already available.
When new text is presented to the model this can be processed according to how closely it matches what has gone before.
The statistical approach reduces the problem facing the syntax analyser from a logic-based task to one of pattern recognition.
2.2.4.1.
Hidden Markov Models
A HMM is a collection of states that are connected by transitions.
Each transition has two probabilities: a transition probability which specifies the likelihood of the transition between the states and an output probability density function (pdf) which defines the probability of emitting a symbol from a finite set given that a specific transition has been taken.
One important application for Markov models has been in automatic speech recognition (ASR) in which a speech signal is modelled as a probabilistic function of a (hidden) Markov chain.
Further developments within the speech paradigm have extended the use of Markov models from the acoustic level to the linguistic (Jelinek et al, 1983)(Brown et al , 1983).
Jelinek took individual words as the states in his Markov model for speech recognition.
There were 5,000 different states in his model, corresponding to the number of words in the lexicon.
For a corpus of 2 million words the frequency of adjacent occurrence of groups of 2 and 3 words was determined and stored in a transition matrix.
With 5,000 different states the size of the bigram and trigram matrices are enormous (2.5e07 and 1.25e11 elements respectively).
It is inconceivable that any corpus should provide occurrences of all of these possibilities.
Since Jelinek's corpus was about 2 million words it was impossible.
To compensate for the absence of transitions Jelinek used backing off formulae to account for missing transitions.
Hence if a trigram was not present in the corpus then a probability was generated for it based on the bigrams, and if the bigram was absent the probability was based on the unigram.
Normalising factors were employed to ensure that the probabilities always added up to 1.
A probability was assigned to any word string that could be formed from the spoken input.
The sentence with the highest score is then selected.
A failing of systems such as Jelinek's is that they are heavily dependent on the size of lexicon used.
A large lexicon system demands a huge corpus to obtain a satisfactory model and a large amount of storage for the resultant transition matrices.
An alternative technique is to use syntactic categories rather than individual words as the states in the model.
Such an approach has distinct advantages.
By using syntactic categories there is no limit on the size of the lexicon used.
Since there are substantially fewer syntactic categories than words in the language, the number of states in the model is reduced.
As a result a good model of the system may be obtained from a much smaller corpus of data.
Deroualt and Merialdo (1984) used such a system to perform stenotypic transcription of unconstrained French with a vocabulary of 250,000 words.
Stenotopy is a means of recording shorthand by machine and is used mainly for conference recording.
The phonetic information is manually input using a special keyboard having 21 keys.
There are two problems associated with converting the stenographed record to typescript.
The first is that there is no end of word mark in stenotography, hence all of the phonemes run together.
The second problem is the large number of homonyms (sound-alike words) in French.
The use of syntactic information led to a 30% reduction in the number of words that were incorrectly transcribed.
Another important use of statistical grammatical information has been for the grammatical tagging of natural language corpora.
Grammatically tagged corpora gives valuable information to researchers investigating the syntactic nature of language.
In English many of the words may belong to more than one grammatical category and it is the function of a tagger to select the correct tags for the words.
The first large computer corpus to be tagged was the Brown corpus of American English.
This corpus was tagged using TAGGIT (Greene & Rubin, 1971), a rule-based lexical disambiguation program.
The variety of language found in the corpus was such that TAGGIT met with only limited success.
Of the text that was encountered about 78% was successfully disambiguated automatically, the remaining 22% was manually disambiguated.
This was still a sizeable amount of text to process manually.
When faced with the same problem the compilers of the LOB corpus opted for a statistical approach.
The resultant CLAWS program (Leech et al 1983)(Atwell, 1983)(Atwell et al 1984)(Garside, 1987) proved to be very simple and successful, achieving 96-97% correct tagging of all words in the corpus (90% correct of ambiguous words).
Although many words in English are syntactically ambiguous, much of this ambiguity arises from rare usage.
If one simply uses a dictionary to determine the grammatical category of a word then there is little indication of which are the more likely syntactic categories.
The CLAWS team incorporated information about the likelihood of a word belonging a grammatical category into their system in a very gross way -- some of the available tags were labelled as being unlikely.
Church (1988) improved the stochastic approach to corpus tagging by including statistics based on 'lexical probabilities'.
The lexical probability that Church refers to is the likelihood that a word will behave in a particular syntactic manner.
Church's approach to the problem derived information about the lexical probability of words from existing tagged corpora and dictionaries.
The dictionary provides comprehensive coverage of the categories for a word, the corpus supplements this with information about the actual usage of the word.
There have also been attempts to produce parse trees for input using probabilistic techniques.
Garside and Leech (1987) developed a system that parsed the tagged output from the CLAWS system.
There are three stages in producing the parse.
1
A set of possible parse continuations are assigned to each word in the text from a lexicon which lists the parse continuations from each pair of tags.
2
A search is made for a number of special tag patterns and parse fragments and the parse structure suitably modified.
3
Each possible parse is completed, a score associated with it and the highest scoring parse selected.
This task is more difficult than the tagging problem and the results have been, understandably, less impressive (approximately 50% success rate for producing an acceptable parse).
Garside believes that the system can be significantly improved to produce a robust, economic parsing scheme able to operate accurately over unconstrained English text.
Church augmented his tagging program to locate noun phrases.
The training data was obtained by (semi-automatic) parsing of a 40,000 word section of the Brown corpus.
The probability of each part of speech starting and ending a noun phrase was then determined from this data.
At run time, given a tagged sequence of words the parser uses the training data to assess how likely a combination of adjacent tags were to form a noun phrase.
Impressive results were obtained, with only 5 out of 243 noun phrase brackets being omitted.
Development of this method to cope with other types of phrases would require a larger training set that included these phrase types.
This in turn would require a corpus larger than the Brown to provide sufficient examples and considerable effort to obtain the parses.
The APRIL parser (Haigh et al, 1988)(Sampson et al , 1989a) was based on statistical techniques.
The parser was trained by taking a database of manually parsed sentences and extracting statistics that refer to the likelihood of a non-terminal parse being obtained from a set of constituents (e.g. what are the possible set of constituents that can form an adjectival phrase and how likely are they to do so).
Rather than attempt to determine every possible parse and associate a goodness score to each parse, APRIL employs simulated annealing techniques.
This approach requires that a possible parse is randomly chosen and assigned a score.
Local changes are then progressively made to the parse tree and the resultant parses evaluated.
The system made no attempt to determine deep structure parses, limiting itself to finding surface parse trees.
Although initial results with APRIL were promising the system encountered some development problems.
Within the machine translation community there has been the introduction of purely statistical techniques to translate between French and English (Brown et al, 1989).
The corpus used for this was the three million sentence bilingual proceedings of the Canadian Hansard (parliamentary proceedings).
The probability that one word in a sentence in one language corresponds to 0, 1 or 2 words in the translation is calculated.
A glossary of word equivalences is obtained which lists a set of possible translations for each word with a corresponding probability (e.g. the translates as le with a probability of 0.610 and la with a probability of 0.178, etc.).
When translating, the probabilities are combined in various ways and the highest scoring combination selected.
An algorithm is then applied to put the words in the correct order.
Despite contravening almost all the rules that have been thought necessary for providing machine translation the system manages to achieve correct translation in 48% of cases.
The developers also believed that many of the near-miss cases could also be solved by using some simple morphological and/or syntactic analysis.
Statistical methods have been used to automatically deduce the linguistic categories of words from raw data (Atwell, 1987)(Atwell & Drakos, 1987)(Atwell & Elliot, 1987)(Finch & Carter, 1992).
The method used is based on collecting and clustering bigram statistics using a rank correlation metric.
A large (33 million word) corpus was taken and sets of ten nearest neighbours of words derived for the thousand most common words in the corpus.
Words were then categorised based on the similarities of these groups.
Although the technique is very simple it is sufficient to differentiate between verbs, adverbs, adjectives, prepositions, determiners, and between singular and plural nouns.
It is also able to discover some semantic groupings such as numbers, compass directions and animate objects.
What is impressive is that these distinctions can be made without any prior knowledge of the syntactic and semantic categories of the words involved.
2.2.5.
Hybrid Systems
The boundaries between rule-based and probabilistic approaches to language processing are gradually being eroded.
A number of hybrid systems that incorporate the methods from both camps are beginning to appear.
One hybrid approach attaches probabilities to the rules in the generative grammar (Wright, 1992)(Pocock & Atwell, 1993).
The grammatical analysis is performed using a generative approach but each derivation is assigned a probability based on the probabilities of the rules used in the analysis.
The highest scoring final derivation is then selected as the parse.
The probability score can also improve the efficiency of the parsing algorithm by pruning out low-probability alternatives.
A similar approach is taken with the grammar used by the METAL machine translation system.
The grammar used by METAL is rule-based but the rules each have an associated weight (Caeyers & Adriaens, 1990).
To improve the efficiency of the analysis each rule in the grammar is ordered based on its frequency of use.
This ordering ensures that standard phenomena are covered first, exceptions and non-grammatical structures later on.
Lower level rules are only used if the standard grammar is unable to cope with the input.
A hybrid system was used in the 'best first' parser developed by Paxton and Robinson (1973) for the SRI speech understanding system.
This system combined the probability scores of the words that were produced by the recogniser and a score dependent on how grammatically well-formed the sentence was deigned to be.
A number of minor grammatical violations were allowed for the input, with each violation the score representing the 'grammatical quality' of the phrase was reduced.
Deroualt and Merialdo (1986) combined their Markov model with a probabilistic grammar for French.
This enabled them to handle global constraints on the sentence.
The grammar consisted of 200 context-free rules with each rule having an associated probability.
The grammar did not offer total coverage of the language (it successfully parsed 65% of the corpus it was tested on) but was intended as a supplement to the results from the Markov analysis of the stenotography.
If a successful parse was obtained then its probability was calculated by determining the product of the probability of each rule used to achieve the derivation.
(Unfortunately no details are provided about how the probabilities of the rules were determined.)
For word sequences having no complete parse then the probabilities of any partial phrases that had been generated were used.
Their tests evaluated the rule-based grammar and Markov model separately and then in combination.
The best results were found from the combination of the two processes, the worst when just the rule-based method was used.
2.2.6.
Combining syntax and semantics
Although the task for the current system is recognition rather than comprehension, the use of semantic information is still important since it can provide a further level of constraint for selection of the correct word from the candidate words.
As with syntax there are two ways of approaching semantic processing.
The artificial intelligence community aim at producing computer programs that can 'understand' the input.
To achieve such a goal it is necessary to perform detailed and comprehensive semantic analysis of the input (e.g. (Schank, 1972)).
Such systems are capable of dealing with only very restricted language domains.
For certain tasks (e.g. recognition or stylistic processing) it suffices to use much simpler techniques.
The most common method used is concordance analysis.
More recently use has been made of machine readable dictionaries and thesauri to obtain semantically-based groupings of words (e.g. (Alshawi, 1987),(Amsler, 1981)).
Humans perform both syntactic and semantic processing when reading.
The actual manner in which these processes occur and/or co-operate is the subject of much debate within the psycho-linguistic community.
A central issue is whether syntactic and semantic information contribute independently or interact in the comprehension process.
The first approach is to perform syntactic analysis first then have a second pass convert the syntactic tree to a semantic representation.
This autonomous approach concurs with the language model Forster (1979) proposed for human understanding.
This model allowed only semantic analysis, output from the syntactic analysis and feedback to the lexical and syntactic analysis was forbidden.
A second approach combines syntactic and semantic processing.
Woods et al(1976) suggest the use of ATNs for semantic analysis.
In such systems the arcs and nodes of an ATN are labelled with both syntactic and semantic categories.
This method more closely agrees with the language model suggested by Marslen-Wilson (1975).
This model incorporates all different categories of information which interact in an ongoing manner to constrain the processing of a sentence.
From a computational perspective there are advantages for both approaches.
The autonomous method permits the program to be written as separate modules, which both simplifies the program and allows parallel processing.
The combined approach requires a complex control structure to integrate the individual knowledge sources.
For a generative grammar, the concurrent application of semantic processing can reject unlikely partial structures, thereby reducing the amount of processing the parser has to perform.
However an efficient, practical approach to semantics that is able to cover general language is not yet in existence.
Until such a system has been produced the question of how such systems should integrate with syntactic processing remains very much within the area of theoretical linguistics.
2.3.
Summary
To date developers of handwriting recognisers have tended to concentrate on improving their recognition systems by concentrating on the pattern recognition level of the process.
The use of linguistic information in these systems has been extremely limited -- at best these systems perform lexical checking to determine whether the input forms a valid word.
Unfortunately handwriting is often too poorly formed for such techniques to work reliably.
A number of computational applications do use linguistic information however.
Most importantly the use of syntactic information has a long history in computational linguistics.
Its use has been recognised to be of value in applications such as machine translation and speech recognition.
Syntax is also important in other fields such as theoretical linguistics and psycholinguistics.
Because so many different people are working in the field of syntax there are a large number of different approaches taken.
Severe contention exists about whether a rule-based or statistical approach should be used to apply syntactic constraints.
There are strong arguments for both cases and the arguments between these two approaches are unlikely to be resolved in the near future.
Some workers have recently attempted to combine the statistical and the rule-based approaches in hybrid systems with promising results.
This issue is considered in detail in the discussion.
Chapter 3
Investigations with a Generative Grammar
3.1.
Introduction
The character recognition level of a text recognition system proposes a number of possible character alternatives for any input.
The lexical processor then determines which of these are able to form words.
It is then necessary to make a choice between the words.
Information about syntax can be used to select words which form grammatically allowable strings; or to reject strings which are grammatically unacceptable.
One approach to this problem is to develop a parser that selects the words that can combine to form valid sentences.
Generative linguistics has been the dominant approach to syntactic processing of language since the late 1950's.
There have been a large number of syntactic theories proposed and an even larger number of parsing systems developed.
The majority of systems that have been developed have been for restricted domains and small grammars.
The construction of a large grammar represents a major problem in it's own right which has only been attempted by well-funded projects.
Without such funding the best way to process general language is to acquire rather than develop a parser (Patten, 1992).
Existing parsers encapsulate the linguistic intuitions of their developers.
Hence it was decided to select an available parser and modify it to cope with the recognition task.
A large number of parsers exist.
This chapter reviews the considerations that must be taken into account when selecting a parser for a specific task.
The parser selected is the Alvey Natural Language Tools (ANLT).
The ANLT system is described and the necessary modifications are made to it to perform the recognition task.
Results obtained in testing its ability to deal with the recognition problem.
3.2.
Selection of a Parser
There are a number of selection criteria that should be taken into consideration when selecting a parser.
The main considerations are:
The underlying grammatical formalism
How large a fragment of the language the system can deal with
The parsing algorithm
Practical considerations
3.2.1.
Grammatical Framework
Perhaps the most important selection to make is the type of grammar on which the parser is to be based.
One must decide whether to use a unification-based approach, an ATN system, a Marcus parser or a transformational grammar etc.
The most important grammatical formalisms were discussed in the previous chapter.
Of the available types the unification-based grammars appear to be best suited to dealing with a large coverage of English.
There are a number of reasons for this.
Firstly they enable large systems to be produced yet still retain clarity within the grammar.
They also have the well-defined approach to semantic processing and have led to theoretically clean and fairly efficient computational representations.
3.2.2.
Language Coverage
Parsers frequently fail to parse a sentence, either because the words in the sentence are not stored in the lexicon or because the grammar available to the program is unable to handle a particular grammatical construction.
It is axiomatic that 'all grammars leak'(Sapir, 1921).
Such is the variety of expression that people use that no definitive grammar of English exists.
Large authoritative grammars have been written (e.g. Quirk, Greenbaum, Leech & Svartvik, 1985) but not incorporated into computational systems.
Even this grammar is not able to deal with all sentences, requiring constant improvement and amendments to be made.
Since the possibility of full grammatical coverage is not feasible many systems have been created that are targeted to a specific application domain.
Such systems are able to take advantage of the limited constructions that are found for a specific task and the reduced vocabulary size.
By accepting such a limitation in application domain it is possible to achieve much more detailed analysis of the text.
The emphasis in such systems has been to concentrate on the representation, organisation and use of linguistic knowledge as encapsulated and expressed by linguistic rules and procedures.
The lexicons have been regarded as being of only secondary importance or their production considered to be trivial.
The prototypical systems tend to deal with only a small number of words, with the lexicons being created in an ad hoc manner.
Furthermore the structures of the lexical entries have tended to be proprietary to the system making it impossible to exchange lexicons between systems.
Early examples of this approach include SHRDLU (Winograd, 1972) and LUNAR (Woods, 1973).
SHRDLU was able to understand natural language commands on how to move blocks.
LUNAR was a natural language front end for a data base containing moon rock sample analyses.
Both of these systems were able to understand relatively unconstrained language for these very narrow domains.
Both were also found to be non-portable and non-extensible to other applications.
Hence, when selecting a parser it is vital to determine how wide a variety of grammar the parser is able to process and how easily it can be extended to a larger domain.
Since 'real world' applications require the lexicon to contain entries for tens of thousands of words it is also necessary to consider how large the available lexicon is, the format of the entries and how much effort would be required to create new entries.
3.2.3.
Parsing Algorithm
A grammar is simply a declarative statement of what forms a valid sentence.
To determine the derivation of a sentence it is necessary to use a parsing algorithm.
The parsing algorithm specifies the manner in which lexical information is to be retrieved, which grammar rules are to be applied to the input and how the information obtained is to be stored.
A variety of parsing algorithms are available and incorrect selection can result in much greater processing requirements.
Top down and bottom-up parsing
The first consideration to be made with the parser is the manner in which the grammar rules are applied.
Early parsing algorithms worked in either a bottom-up or a top-down manner.
Bottom-up (data-driven) parsers work from rules whose left-hand side matches the grammatical categories of the input.
That is, given a sentence to parse the parser first determines the grammatical categories of the words in the sentence.
The parser then checks these categories against the right hand sides of the ruleset.
When the right hand side of a rule is found the symbols involved are replaced by the symbol on the left hand side of the rule.
This process is repeated until the sentence symbol, S, is reached or the parse fails because no other rewrites are possible.
Top-down (concept-driven) parsers work in the opposite way.
These parsers build the parse tree by starting with a rule whose right-hand side is a sentence, S. The rules are progressively expanded down to the rules having the lexical categories that match the input.
There are problems with both bottom-up and top-down parsers.
Bottom-up parsers are very susceptible to problems arising from lexical ambiguity.
The major problem with top-down parsing is the time wasted in expanding rules that cannot possibly be satisfied by the input.
Depth-first and Breadth-first search.
Parsing may be regarded as a search for the syntactic representation of the input.
There are many alternative paths available to follow and many of these paths will turn out to be dead ends.
It is necessary to give the parser instructions on how to search for the correct parse.
One possible solution is to use a depth-first approach.
This involves going as far down a single path as possible.
Should the path fail then the system backtracks to the previous decision point and takes a different path.
This method has been used in such systems as ROBOT (Harris, 1977).
The efficiency of this system was improved by ordering the arcs of the ATN grammar.
This led the parser to explore the more likely solutions first.
Another option is to use a breadth-first approach.
Instead of trying out just one hypothesis at a time, a breadth-first approach retains a number of possible hypotheses.
Each hypothesis is in turn extended by a single step.
As the parse progresses down the input the incorrect hypotheses should fail; thereby reducing the search space to be explored.
For the purposes of the current project a bottom-up approach is an obvious one to use.
Since more than one successful path through the grammar may be necessary (more than one syntactic combination may be possible from the candidate words) a parallel process would also appear to be a logical choice.
Some method is required to organise the partial structures produced so that a more efficient parse is obtained.
A chart provides such a structure.
The simplest form of charts (Kay, 1986) stores all of the hypotheses that have been tried out on the input.
Use of a chart prevents repetition of work that has already been carried out by blocking any attempt to reapply the same rule in the same place.
Practical Considerations
There are a number of practical considerations that must be taken into account in the selection of a parser.
Firstly there is the question of efficiency which may depend on the parsing algorithm, the machine used and the computer language that the system is written in.
Certain systems are written to exploit specific features of a particular machine.
Requirements may exist about the amount of memory that is required to run a system.
It is also likely that some tailoring of the system will be necessary for a different application.
The grammar and lexicon may be incomplete for the task at hand and may need to be supplemented.
It is important to determine how easily such additions can be made.
More substantial modifications may be necessary to the system for a particular task.
Practical natural language systems are by necessity large and complex and modification of such systems can lead to unexpected results.
If the system has been written in a modular way (i.e. the individual components of a system are intelligently partitioned, allowing each module to see all and only the input it requires) then the modification is greatly simplified.
Early systems were not written in this way.
For example the early machine translation systems tended to contain the grammar intrinsically within the program.
Modification of such systems required a great deal of effort.
3.3.
The Alvey Natural Language Tools
Of the available parsers the one which best satisfied the above criteria was the Alvey Natural Language Tools (ANLT) system.
The ANLT is a powerful piece of software designed for use in natural language processing applications.
A large lexicon is provided which is compatible with a large grammar based on the Generalized Phrase Structure Grammar (GPSG) formalism (over a 1000 rules are in the object grammar).
GPSG is computationally economical and has been developed to a point where a large number of interesting syntactic phenomena may be dealt with in an integrated formal framework.
The parsing algorithm provided is based on chart-parsing techniques.
One of the aims of the ANLT is to provide a general purpose system which may be easily adapted for new tasks.
With this in mind a Grammar Development Environment is provided as an aid to the development of a natural language grammar.
It is useful for bringing together the separate parts of the ANLT and the grammar and lexicon can be developed and altered within this environment.
A description of the major components of the system is necessary in order to determine their applicability to a recognition task and what modifications are necessary.
3.3.1.
Morphological Analysis and Dictionary System
The morphological analysis and dictionary system of the ANLT (Ritchie et al, 1987a) have been developed to provide a large coverage, morphological lexicon.
Morphology is the mechanism by which morphemes combine to form words.
The use of a morphologically-based lexicon can lead to a large reduction in the storage requirements for the lexical information.
The use of morphological methods ensures that it is necessary only to store the lexical information for the root forms of words and a small set of affixes (-er, -est, -ing, pre- etc.).
Some method is then used to derive the syntactic representation for the derivative form from combining the root form and the affix.
Given the detailed lexical information used in the ANLT a morphological processor leads to a large reduction in storage space required.
(Furthermore with the LISP system available to the current project it was not possible to store a very large lexicon -- necessitating morphological processing to reduce storage.
This will be dealt with later in this chapter.)
The ANLT lexicon aims to provide consistency (i.e. syntactic information is only explicitly provided for words that cannot be derived from extant morphemes within the lexicon).
The lexical entries are of a form compatible with a GPSG grammar (Appendix 1).
This compatibility requires that the syntax information is defined in terms of feature-value pairs.
e.g. ((N +)(V +)(BAR 0)(INFL -)(PLU -))
indicating that the morpheme is a singular noun which cannot be further inflected.
Accessing the syntactic information for a word in a morphologically based lexicon requires the following steps:
1
Splitting the word into its constituent morphemes.
Given a string it is necessary to determine what morphemes the string consists of (e.g. the string boys comprises the morphemes boy and +s).
This process is trivial for a native speaker of a language, however for a computer system the solution is more difficult.
The two major problems involved are related to the spelling changes that can occur when morphemes combine and the multiple segmentations that are possible for many words.
Spelling changes may occur at the morpheme boundaries resulting in slight changes to the surface string produced by the morpheme combinations.
Table 3.1 shows some examples of the regular morphological effects that occur in English.
The morphemes may simply be concatenated (1), letters may be added between morphemes (2 and 3), deleted (4) or changed (5).
It is interesting to note the similarity between the morphemes in (1) and (2) but the differences in the words produced.
Spelling rules exist to specify when particular spelling changes can operate.
The ANLT system represents these in a set of Koskenniemi rules (Koskenniemi, 1983).
It should be noted that there are a number of morphologically irregular words in English which will require explicit statement in the lexicon (e.g. children instead of childs, caught instead of catched).
Determining the constituent morphemes is further complicated because there may be more than one possible segmentation for a word.
Consider the word preached.
Humans have no problem in determining that this is the past tense of the verb preach.
Fig 3.1 shows some possible segmentations for the word preached that a computer could posit.
From the diagram, the possible segmentations are pre-ache-ed (where ache may be nominal or verbal) or preach-ed.
2
Accessing the syntactic information for the constituent morphemes.
The syntactic information for the morphemes (e.g. boy and +s) must then be accessed.
Large lexicons are generally unable to reside in main-memory and the required information must be retrieved from disk.
3
Combination of the syntax information for the morphemes to generate the syntactic information for the complete word.
Having obtained the syntactic information for the morphemic constituents of the word, this information is combined to determine the syntactic information for the complete word.
The combination of the morphemes requires a 'word grammar' to determine the legality of the combinations and the resultant representation.
A word-grammar is used to make decisions of the form:
The word grammar used in the ANLT is based on Gazdar's GPSG model of syntactic features and is similar in structure to the sentence grammar.
Instead of assessing the feasibility of words combining to form a sentence and providing a syntactic representation for the result, a word grammar determines the feasibility of morphemes combining to create a word.
The actual grammar rules and certain of the features employed are different (e.g. sentential grammars have no feature to indicate whether a word can be further inflected) but the underlying principles and mechanisms are similar.
The syntactic information for boy and +s below could be combined by a word grammar to produce the syntactic information for boys: e.g. 
3.3.1.1.
The Lexicon
The lexicon provided with the second version of the ANLT contains over 62,000 entries.
Not all of these entries are unique however.
There are approximately 33,000 different morphemes.
The term 'morpheme' is not used in the strict linguistic sense.
In this case any word, or even phrase is classed as a morpheme.
This allows nominal compounds, such as ice cream, to be treated as morphemes, and allocated the corresponding syntactic information.
Approximately 4,500 of the entries are compounds.
The lexical entries of the ANLT are presented as 5-tuple entries:
Although there is obvious redundancy within the lexical entries (the phonological form and semantics field are simply repetitions of the morpheme) this reflects the aim of the ANLT project to produce a general-purpose, flexible system.
Words possessing more than one meaning or usage must be allocated multiple entries within the lexicon.
3.3.1.2.
Problems
The morphological system provided with the ANLT is one of the most complex morphological systems ever created.
This complexity permits the compaction of the data that is stored and is ideal for research into the morphology of a language.
The authors of the system have stressed (Ritchie et al, 1987b) that the system is prototypical and our experience with using the system for a practical application has encountered some problems:
The major problems with such a morphological system are
1
The computational effort required to determine the syntactic information is excessive (see below).
2
Although morpheme combinations are syntactically valid they may be semantically invalid.
For example the system allows the existence of the words such as overbelieve and interbelieve
3
Unexpected syntactic categories may be returned for a string.
The morpheme liver is not explicitly stored in the first lexicon distributed with the ANLT.
The system returns the syntactic information for the morpheme combinations live +er which although possible, represent infrequent usage of the word liver.
Any rule-based system of morphology is susceptible to these errors.
The major problems for the recognition application are the multitude of possible syntactic representations postulated by the look-up and the computation required.
Recent investigations into lexical ambiguity (Church, 1988) have shown that although words are capable of existing in many syntactic categories, in reality the problem is not so widespread and words actually possess unique 'best' parts of speech.
For example, the word ball may be either a noun, a transitive verb or an intransitive verb but for 99% of cases may be treated simply as a noun.
It would be beneficial to know the 'best' grammatical category for any particular word when attempting to determine the probable parse to assign to a sequence of words.
This is made more difficult by the use of a morpheme-based lexicon because many parts of speech may be derived from each morpheme.
Many aspects of the system are computationally intensive.
The morphological system can require a long time to obtain the syntactic information for a word.
The user documentation quotes that on a SUN 3/160 machine running Kyoto Common Lisp (KCL), the time required to look up a large word (excluding garbage collection) is 11.8 seconds of CPU.
Our experience has shown that on average obtaining the syntactic information requires about 20 seconds of run time per word (including garbage collection).
For morphologically complex words this can be much worse, for example the word unbelievable on the same system (including garbage collection) required over 200 seconds of run time.
Further problems were encountered with the lexicon supplied with the second version.
The lexicon was much more complete, being derived from the LDOCE dictionary (for details see Carroll and Grover (1989)).
The lexicon supplied with the first ANLT contained 6,800 morphemes and suffered from omissions.
The larger lexicon supplied with the second version remedies this problem but is too large to load into the KCL LISP environment.
An alternative morphological system had to be devised in order to test out the system (see below).
Another problem arises from the nature of the text recognition system.
Before accessing the information for a word, it is necessary to first know what word to obtain information for.
In a recognition system this is non-trivial, requiring a 'word-lookup' stage to determine which valid character combinations exist within a character lattice.
(Wells et al, 1990) Two options are available -- one can select the character combinations based on transitional probabilities between characters or one can specify a vocabulary list and check the character combinations within that list.
Although the former method requires little storage it has been found that the benefits from a fixed wordlist far outweigh a statistical approach when determining possible words.
Prior to the use of the ANLT morphological system in the current system it is necessary to first check the existence of a word.
This means that every word that the system can process has to be known beforehand, otherwise it will not be accepted.
One of the benefits of the ANLT's morphological system is that it does not need to explicitly store all of the words it knows.
The ANLT is able to generate the syntactic information for any word whose morphemic roots are in the lexicon.
Hence the number of words 'known' to the ANLT is unbounded.
The imposition of the current system's wordlist on the ANLT system nullifies this advantage.
3.3.1.3.
Alternative Morphological Systems
The current system (Keenan and Evett, 1989) requires a significantly simpler morphological system which is able to take the character lattice output from the recognition stage and return the words that may be formed from the lattice and the grammatical information for those words.
Some work has been carried out towards making this possible.
As has been previously stated one of the major problems with morphology is determination of the word boundaries in a string.
Such approaches, though simple, are error prone (e.g. mother - moth + er etc.).
The ANLT system uses a set of rules to segment the string into possible morphemes and then determines the validity of the segmentations using a word grammar.
In the past simple morphological systems have used simple affix-stripping routines to determine the morphemes present in a word.
A typical algorithm for accessing a dictionary using an affix stripping routine is as follows (Holmes, 1988),(Ramsay & Barrett, 1987).
Strip off the longest set of letters from the end of the word which matches a (usually small) set of standard suffixes.
Then look up the root word in the lexicon.
If it exists then the root word has been found, otherwise look for the next longest suffix present.
Adjustments of roots (e.g., re-inserting final e) is performed.
Irregular words and those not dealt with by the set of suffices are marked and have their own entries.
These decomposed words are then used to access syntax.
A small set of rules are used to derive the syntactic information for the word.
At least two problems exist for using affix stripping as a morphological technique.
The first problem is that it is easy to determine the incorrect stem of a word (e.g. mother - moth + er).
The other problem is the amount of lexical checking that is required.
Each time an affix is removed the remaining stem must be checked to determine whether it is a valid word.
This requires constant checking which is trivial for a 'demo' system but for a system with a realistic vocabulary it can be inefficient.
A new method has been developed for the current project which attempts to provide a much simplified morphological system with little increase in the storage requirements and remove the errors permitted by the affix-stripping methods.
The fundamental idea is to determine the identity of a word's morphological root during system development.
This root is then stored and may be accessed in the lexicon at run-time rather than determined on the fly.
The grammatical information for the lemmata is then accessed.
This information is the ANLT lexical entry, or entries, for the word.
A sub-lexicon is produced for the words in the text under analysis.
This is combined with the lexicon of bound morphemes supplied with the ANLT.
There are a number of advantages from taking this approach.
Firstly the complete lexicon supplied with the ANLT is too large to load into the KCL.
Only by including this process is it possible to access every word in the lexicon.
This method also prevents selection of the incorrect morphemic roots of the word (preached is found as a derivative of preach rather than ache) since the selection of the correct root form can be made with certainty at development time with the aid of dictionaries.
The imposition of a wordlist on the system forbids the selection of semantically invalid morpheme combinations (e.g. underbelieve).
There is still the development time task of finding the correct root for a lexicon of approximately 70,000 words.
The first step towards achieving this task is to use the information present in the machine readable dictionaries.
A large number of the words are simple inflections which were solved using the information in the Text710 version (Mitton, 1986) of the OALDCE.
One form of this dictionary contains just the root-forms of words and a set of codes which indicates the manner in which a word may inflect.
For example there are codes for nouns which form their plural like the word pony (i.e. replace the final y with ies).
The inflections of each word are generated and the root linked to the inflection.
Only a limited set of inflections are used in the Text710.
The Collins English Dictionary (CED) lists other derivatives that can be formed from a headword.
This accounted for many derivations that are not simply inflections (e.g. happy, happiness, happily).
The use of dictionary information greatly reduced the task but it did not complete it.
A number of words contained affixes that the dictionary ignored (e.g. un- and dis-).
A program was written that searched for such affixed words and prompted a human with the word and the probable root form which could be selected or rejected.
Although a tedious task it ensured that the correct root forms was found for words.
3.3.2.
The Grammar
The ANLT provides a unification-based formalism, derived from GPSG, that covers the major syntactic constructs of English.
GPSG (see Appendix 1) is a grammatical framework which extends the expressive power of context-free grammars enabling complex systems of regularities and restrictions to be stated easily.
The GPSG framework employs meta-grammatical rules to generate new rules from existing rules.
This allows rules with a wide variety of realisations to be specified with only a small number of explicit statements.
There are a number of problems with the system however.
The main ones are the number of parses that are produced and the limited extent of the grammatical coverage.
Many of the words in the lexicon have a high degree of syntactic ambiguity.
Much of this ambiguity is as a result of the feature-based grammar.
(The most ambiguous word in the lexicon is put which has 638 different lexical entries.
One cause for this ambiguity is the specification of the particles that put can occur with to form phrasal verbs.
Each different particle that put is able to combine with must have a separate entry.)
As a result of the syntactic ambiguity and the number of grammar rules that can be satisfied, multiple parses are possible.
In the case of ANLT however the number of parses suggested can be excessive (e.g. 'I enter orders until the markets close' produced 4,848 different parses)
A problem with any rule-based grammar is that complete coverage of a natural language is not possible.
The search for a complete grammar of English has been likened to the search for the Holy Grail (Jensen & Heidorn, 1982).
One of the principal authors of the ANLT grammar system has estimated that the system is capable of covering about 25% of English (Briscoe, 1988).
Grammatical omissions from the ANLT include punctuation, titles, dates, addresses, digits, parentheticals, ellipsis, anaphor binding, conditionals and comparatives (Grover et al, 1989).
3.3.3.
The Parsing Algorithm
Having developed a lexicon and the set of grammar rules it is necessary to provide a mechanism by which these rules are applied.
The parsing algorithm used is unique to the ANLT (Phillips, 1986).
The parsing algorithm is based on chart-parsing techniques and works breadth-first across the input.
The parser works its way along the text string word by word.
The basic mechanism is to build an edge from left to right.
When an edge is completed then any partial edge immediately preceding it in the chart is examined to see if the newly completed edge is required for completion.
If this is the case then the edge is added to the preceding edge.
If the preceding edge is in turn completed then any partial edge preceding it is tested to see if it needs the newly completed edge etc.
A diagram best describes the situation:
A complete edge (complete1) is found in (a).
This is required by partial2 for completion.
Partial2 is extended to include complete1 and in turn becomes a complete edge itself, complete2 in (b).
Partial1 requires complete2 for completion, it extends to include complete2 and becomes complete3 in (c).
For example consider the ruleset:
and the input:
The parser works its way along the text string word by word and modifies the list of possible (partial) parses by incorporating each word as it is encountered.
At position 1 a determiner, det, is found which is on the right hand side of rule (r2).
There is not sufficient information available at this point in the parse to complete all of this rule.
Hence a partial NP edge has been found between positions 0 and 1 which requires a noun, N, to complete it.
At position 2, N is found, this completes rule (r2).
Hence a complete NP edge is created between positions (0 -- 2).
This in turn provides a partial edge for rule (r1) between positions 0 and 2 which requires a verb, V, to complete it etc.
Experiment 1
The first step in the evaluation of the ANLT processing system was to select a few sentences from the Longman/Lancaster corpus and samples of recognised text (the HP business text in Appendix 5 and estate agents details).
Each test sentence used is therefore a valid member of the English language (although it may not necessarily conform to prescriptive syntactic rules).
Some general rules of thumb were used in the selection of test sentences.
Firstly there should be no punctuation (beyond sentence terminators) due to the acknowledged failure of the ANLT to process such phenomena.
Simpler sentence forms were preferred and no sentence containing more than 18 words was used.
The set of sentences selected are shown in table 3.2.
Each of these sentences was submitted to the ANLT (with the alternative morphology described in $3.2.1.3) to determine whether the sentence parsed, how many parses were obtained and how long the processing took.
Processing was carried out on a SPARC 4/75 with 48 MBytes of memory.
With a number of these sentences the ANLT was unable to process the complete sentence.
In such cases simpler versions of the sentence were considered.
Table 3.3 contains the results for these sentences.
'Top category' is a flag used by the ANLT.
If this flag is OFF, any category which covers all the input is accepted as a valid parse; if the flag is ON then only parses whose root category is an extension of [T +]are accepted.
As can be seen from the table a significant reduction in the number of parses is achieved when the flag is ON.
When 0 is given as the number of parses, this indicates that the parser exited cleanly with no parses being found, a dash (-) indicates that the system crashed before a parse was obtained.
Hence sentence 10 does not have a number of parses since it exhausted the storage of the LISP system, having already allocated 25 MBytes.
A similar problem was also found with sentence 13, whereas sentence 12 crashed while printing the bracketed parses.
The quoted CPU times do not reflect the true extent of the time actually required to obtain parses.
For longer sentences there is a vast amount of garbage collection required.
It is not uncommon to wait for 5 hours for the result to be obtained.
On a SUN 3/160 the effect was even more devastating.
Sentences could require in excess of 9 hours for a result (or crash) during which time it was known for the system to refuse logins to other users due to insufficient swap space.
Although swap space may be increased it is a drastic solution simply to obtain a parse for a single sentence.
It is understood that the ANLT is by no means complete (no rule-based grammars are) and therefore no complaint can reasonably be made when a sentence such as(3) fails to produce a parse.
However the length of actual time required before the parse fails is unacceptable.
The problem is further compounded for the recognition task.
Unlike the test sentences above, recognition input can contain many alternative words for each one written.
Hence there is a significant increase in the processing due to the increased numbers of words to be assigned grammatical information.
A basic modification was incorporated into the system to allow for the processing of the multiple candidate words present in the input lattice.
The basic algorithm provided with the ANLT is able to deal with the lexical ambiguity (the ability of words to belong to more than one syntactic category).
A short-term modification was made which forces into this framework not just the lexical ambiguity of single words but also the other candidate words for any given location in the input.
Due to the computational intensity of the system, this was only successful for the simplest of sentences with very few alternative words.
3.4.
Discussion
The following criticisms are based solely on the applicability of the ANLT to the process of script recognition with no attempts to extend the grammar for any particular domain.
The ANLT is much more than just a parser, it is a development tool for the production of testing of grammars and morphological processors.
The generality of the ANLT system will obviously be ill-suited to certain aspects of any specific problem.
From the investigations carried out it appears that the use of the ANLT system for a recognition task is computationally infeasible.
These findings are similar to those of Briscoe (personal communication) who attempted to modify the ANLT for a speech recognition task.
The decision he reached was that existing computer hardware was not sufficiently powerful to cope with the problem.
From the description provided of the system it is clear that more testing of the system to determine it's full capabilities would be desirable.
However the computational demands of the ANLT system severely restricted the amount of testing that could be carried out.
Many of the 'real' sentences that were processed took so long and used up so much of the computational processing of the SUN system that full testing was not feasible.
There are a number of linguistic phenomena that the ANLT system does not attempt to deal with.
The most striking example is the ignorance of punctuation.
In fact very few generative systems make any attempt at processing punctuation.
(Exceptions to this include Allen (1987) who used a very basic implementation of a rule-based grammar for text-to-speech processing.)
Punctuation performs a valuable role in written language.
It enables passages of language to be read in a coherent manner and gives an indication of the rhythm and colour of the speech.
The principal use of punctuation is to separate units of the grammar (e.g. sentences, clauses, phrases, words) from each other.
Without the capability to deal with punctuation, rule-based systems are unable to deal with many naturally occurring linguistic constructions.
Syntactic ambiguity is a major problem with the system.
The extent of this problem may be appreciated by referring table 3.2 which shows the number of entries in the ANLT lexicon for some highly ambiguous words.
There are ways to reduce the problems posed by syntactic ambiguity.
Automatic tagging programs are extremely efficient and accurate (e.g. CLAWS (Garside, 1987), PARTS (Church, 1988)).
The PARTS tagger is currently being used to reduce the syntactic ambiguity problem for the FIDDITCH parser (Hindle, 1983b).
This parser is currently being used to provide skeletal parses of the ACL/DCI corpus (Marcus, 1990).
Rather than the parser consider each of the possible grammatical categories of the input words, the tagger pre-processes the text and labels each words with its most likely part of speech.
Hence the parser has only to decide on the syntactic structure that can be made from combining these different parts of speech.
A similar approach would be of some benefit to the ANLT system.
If the tagger indicated that a word was a noun then the non-noun lexical entries could be given reduced priority by the ANLT.
Unfortunately a tagger can only provide gross syntactic categories; the ANLT requires very detailed grammatical information.
The principal reason for the vast number of parses that may be suggested for relatively simple sentence constructions appears to be the subcategorization of verbs.
The highest level of verbal information that can reasonably be expected from an automatic tagger is the transitivity of the verb.
Therefore only a limited reduction of ambiguity will be obtained.
The matter is further complicated for the recognition task.
A tagger is designed to work with certain text.
That is, the actual words in the sentence are known, and the function of the program is to determine the grammatical category of each word.
For the recognition task there are a number of alternate words suggested.
The accuracy of the tag assignment will be seriously reduced by such input.
The combination of syntactic ambiguity problem of the ANLT system with the word ambiguity from the recognition system adds a further dimension to the difficulty of performing the parsing task.
The number of potential parses increases exponentially, often beyond the capability of the hardware to process the input.
The resolution of syntactic ambiguity is one of the tasks to be performed by semantic analysis.
If no semantic or pragmatic restrictions are applied during the parsing of the sentence then many parses are suggested that are semantically or pragmatically infeasible.
The imposition of semantic or pragmatic restrictions could occur after the parse is completed to remove improbable combinations.
This approach is inefficient since the computational effort to produce the parses has already been expended.
It can also mean that there will be more work for the semantic analyser to perform since there are so many more combinations to check.
A better way would be to apply semantic restrictions to partial parses.
Any partial parse that is rejected will be removed -- thereby removing any parse that would have contained this component.
Unfortunately there is no semantic theory currently available that is able to work in such a way for a broad coverage of English.
3.5.
Summary
The issues that need to be taken into consideration when selecting a parser have been discussed.
The ANLT was selected as the best available rule-based parser to use.
There are a number of reasons for this.
It has a has a large coverage (for a rule-based parser) grammar, based on GPSG.
The lexicon is very large compared to many other systems.
The parser works in bottom-up, breadth-first manner and uses a chart for efficiency.
In addition the system is provided with a grammar development environment that is designed to support a linguist develop and edit grammar rules and for building a lexicon.
The system is written in a modular manner allowing modules to be supplemented and replaced.
Investigations with the system have proved disappointing.
The system is computationally intensive.
The complete lexicon cannot be loaded into memory.
Also the fragility of the grammar -- there are a number of features that are not covered by the grammar and it often takes a long time to reject the sentence.
Sentences that can be analysed often produce a huge number of possible analyses.
The morphology of the system was simplified in a manner suited to the recognition task.
Only through the use of such a modification was it possible to have access to the complete lexicon provided with the second version of the ANLT.
This did not however make sufficient difference for the system to be feasible.
Overall the computational complexity of the system rules it out at the present time for application to the recognition task.
Even if a computer was available which could cope with the load, the system has serious difficulties not least of which is coverage.
The large number of parses produced are unnecessary and unhelpful in the recognition application.
While little data is presented here it is clear from the few examples given that the ANLT has problems with 'real' language -- the number of possible parses produced and the failure to parse -- which would not be solved by additional computational power.
Chapter 4
A Probabilistic Syntactic Processor
4.1.
Introduction
An alternative methodology to rule-based parsing employs probabilistic techniques.
The use of statistical techniques is not new to computational linguistics.
As long ago as the 1930's Zipf was investigating the statistical nature of language (Zipf, 1936).
However the use of statistical techniques in linguistics fell from general use apart from a few 'marginal' applications.
One application that has continued to make use of the statistical properties of language is stylistic analysis.
Stylistic analysis aims to 'fingerprint' authors and verify the authorship of disputed works (e.g. (Ellegard, 1962),(Morton, 1965)).
There are a number of measures that can be used in such analyses including the lengths of words and sentences used, the frequency of use of individual words and syntactic analysis.
The use of such techniques for language processing has traditionally been frowned upon by the linguistic community.
In recent years the use of statistics for language processing has regained support.
There are a number of reasons for this.
The difficulties encountered by workers taking the generative approach has led to the search for alternative techniques.
At the same time computers have become much more powerful and common with much more information available in machine readable form.
Whereas a few years ago a mainframe computer would have been essential to perform corpus processing, a desktop computer now suffices.
Nor is storage of large amounts of data a problem anymore, CD-ROM technology offers the ability to store many million words on one small disk.
Hence statistical processing of language is once more an important technique used by computational linguists.
This chapter describes the methods employed in creating a statistical syntactic processor to assist in text recognition.
It is a long chapter.
A small amount of the theory of statistical models is provided followed by an overview of the methods used to create a statistical processor.
A preliminary investigation is shown that uses a small scale statistical analyser to process text that had caused great problems for the rule-based system.
This system is extremely basic but the results are convincing enough to lead to the development of a large scale recogniser.
The remainder of this chapter deals with the production and testing of a statistical analyser capable of dealing with general language.
The production of such a system requires the development of a large lexicon and extraction of language information from large samples of the language.
Such information can be found in machine readable dictionaries and corpora.
The information present in these sources and its development into a useful format is described.
A number of investigations are carried out with the scaled-up analyser to determine what factors can lead to improvement in performance.
Implementation details are provided, indicating the system operation and communication with the other recognition modules.
4.2.
Language Modelling
Language may be modelled in terms of its statistical behaviour.
For a recognition application the aim of the model is to participate in the choice of candidates recognised (Jelinek, Mercer & Bahl, 1983).
Since the sequence of words in a sentence is subject to grammatical and semantic constraints it is possible to build a probabilistic model for this purpose.
Probabilistic models have commonly been used to investigate the semantic associations between words.
Rigorously, the probability for a word to be produced should depend conditionally upon the whole of the previous word sequence.
Let W, denote a string of n words.
where
W = w1, w2,..., wn
The probability of this word sequence, P(W), is the product of conditional probabilities:
That is, the probability of obtaining a word string, W, is given by the probability of finding the first word, w1, multiplied by the probability of finding the second word, w2, given w1 etc., multiplied by the probability of finding the last word, wn, given that all of the previous words were found.
Therefore, in general the probability of finding a specific word, wi, is given by the entire past history of the text.
In reality it is impossible to estimate the probabilities P (wi|w1,..., wi-1) for all but very small values of i, and only for quite a small vocabulary.
The reason for this is that for a vocabulary of size L, there exist Li-1 different histories.
This would require a huge corpus, would be extremely difficult to estimate and would require too much storage.
One approximation that is often used is to simplify the model by considering only N-grams.
With this approximation only the N-1 previous states are considered.
An example of this is the trigram model used in the TANGORA speech recogniser (Jelinek, 1986) which assumed that histories are equivalent if they end in the same two words.
Thus 
If one considers that a practical vocabulary size will contain thousands of words then even with this assumptions the number of estimates required can become astronomical (e.g. if L=5000, and i=3, it is necessary to estimate values for 1.25*1011 states).
Hence it is clear that N-gram models are computationally practical for only small values of N. (Jelinek's system used a vocabulary of 1000 words and i=3.)
It is also often the case that there is insufficient data available for a reliable determination of all of the parameters of a Markov model.
Many of the word trigrams that are possible are not present in the corpus.
Jelinek trained his system on a corpus of 1,500,000 words and tested it on a test corpus of 300,000.
Of the trigrams found in the test corpus, 23% had not been found in the training corpus.
Such sparse coverage necessitates the estimation of values for the missing transitions.
For a given corpus size, if one uses coarser classification then more reliable but less precise predictions are obtained.
By using syntactic categories rather than words as the states of the model we are able to improve the accuracy of the model's ability to predict.
Hence we investigate transitions not between individual words but grammatical categories.
The use of grammatical categories rather than individual words means that the number of different states is much smaller.
(Although there is no universally accepted set of grammatical categories, the size of tagsets used in corpora tend to be less than 200.
In contrast, the number of words in the language is unbounded.)
This provides a number of advantages.
Firstly the number of transitions which can occur is significantly reduced -- with L=109 (the number of different tags used in the current project) and i=3 there are 1,295,029 possible transitions (compared to 1.25*1011 for a vocabulary of 5000 words).
The probabilities of these transitions may be more reliably obtained from available corpora.
Also the size of the model is independent of the number of words in the dictionary.
Hence the vocabulary of the system is effectively unlimited, unlike models working at the word level where the size of the model increases very rapidly with the size of the vocabulary.
The investigation of syntactic relations in this way is much less common than semantic relations.
There are two main applications for which this approach has been used:
1
Automatic corpus tagging.
Examples of this are the CLAWS system (Garside, 1987) for tagging the LOB corpus and the PARTS tagger (Church, 1988) which is being used for tagging the ACL/DCI corpus.
2
Syntactic error detection (Atwell, 83)(Atwell, 87)(Atwell, 88)(Atwell & Elliott, 83).
4.3.
Overview of Methods.
The underlying approach taken is based upon n-gram statistics.
An n-gram is an ordered sequence of n symbols (letters, words, grammatical categories).
The frequency of occurrence of each n-gram in a continuous stream of data constitutes the n-gram statistics of the data set.
Hence the 1-gram statistics of a data set are simply the frequency of each symbol within the data set.
The 2-gram (bigram) statistics are the observed frequencies of each pair of symbols etc.
In this case the symbols used are the grammatical categories of the words in the corpus.
There are two distinct processing phases involved -- a training phase and a recognition phase.
During the training phase a corpus of text is taken and investigated to determine the frequencies of transitions between grammatical categories (e.g. how often is a determiner followed by a noun).
The frequencies obtained are then normalised in some manner and stored in a transition matrix.
The recognition phase is the run-time application of the information derived during development.
At run-time, given a lattice of possible words, the syntactic processor ranks the possible words according to their ability to combine with neighbouring words.
Two items of data are required to make this decision.
Firstly one must determine the grammatical categories of the words in the lattice.
This information is obtained from the lexicon.
Secondly the probability of each of the possible transitions between the categories is retrieved from the transition matrix.
Each words in the lattice is then given a syntactic rating based on the probabilities of the transitions it participates in.
4.2.1.
Training Phase.
During the training phase the probabilities of bigram and trigram transitions between grammatical tags are determined.
A sequence of grammatical tags in the corpus is taken and split into pairs and triples.
For example, the sequence of numbers 1 2 3 4 5 would split into the bigrams: 1 2, 2 3, 3 4, 4 5 and the trigrams 1 2 3, 2 3 4, 3 4 5.
To determine the probabilities of the bigrams and trigrams a count is made of how many times each bigram and trigram occurs in the corpus.
(Within a unix environment this may be easily achieved using the sort and uniq commands.)
These frequencies are then normalised to correct for any bias towards tags that occur very frequently (e.g. singular nouns) over rare tags (e.g. pre-qualifiers).
The probabilities of the transitions are then stored in transition matrices for use at recognition time.
4.2.2.
Recognition Phase.
The run-time application of syntactic information uses the transition matrices and the lexicon to rank the words in the lattice.
The grammatical tags for the words in the lattice are retrieved from the lexicon.
The probability of each of the possible tag combinations is then found from the transition matrices and the words assigned the relevant grammatical scores.
The algorithm used is best described by considering an example.
The sentence this parrot is dead was written and recognised by the handwriting recogniser.
The lower levels of recognition were unable to uniquely determine the words this, parrot or dead however, and the actual input to the syntactic processor is shown in fig 4.1.
The grammatical categories (tags) for the words in this lattice are shown in table 4.1.
Grammar tags will be described later.
(For computational ease the grammar tag is represented as an integer, for clarity a description of the tags is included in the table.)
Of the words in the lattice dead, perfect and tail are syntactically ambiguous.
In fig 4.2 the grammar tags of the words are shown as superscripts (tag 7 indicates a sentence boundary).
Different syntactic classes of words are treated as distinct nodes in the lattice.
The introduction of grammatical information increases the number of different paths through the lattice from 24 (4*2*1*3) to 72 (6*3*1*4).
The application of bigram information to the lattice is shown in fig 4.3.
Rather than investigate each complete path across the lattice, a window of width 2 is moved across the lattice.
(The main reasons for taking this approach is that for dynamic handwriting recognition it is necessary to select the correct information shortly after the word was written -- it may not be possible to wait for the user to finish a sentence or clause.)
Table 4.2 shows the possible transitions to be retrieved from the bigram matrix for the windows of fig 4.3.
Each word in the window participates in a number of transitions.
For each window position the scores of the transitions are determined from the matrix.
This score is a normalised version of the transition frequency.
For each word in the window the score of the best transition it participates in is added to its syntactic score.
The window is then moved along to the next position.
Once the syntax window has passed beyond a character position, the score for each word is scaled, relative to the highest scoring word in the position i.e. 
The methods used for trigrams are analogous to those in the bigram case except that the transitions are of length 3.
4.4.
Preliminary Investigation
An initial investigation was carried out using text which the rule-based systems had failed to deal with adequately (estate agent's literature).
The lexicon used consisted of the Oxford Advanced Learner's Dictionary of Current English (OALDCE).
Use of the OALDCE affords a lexicon of almost 70,000 words.
Compared with the usual size of lexicons found in natural language systems this is a sizeable lexicon.
An advantage of using the OALDCE is that it is targeted at non-native learners of English.
Hence the dictionary aims not at completeness but rather at covering the more usual words of the language and avoids rare usages of these words.
(More recent editions of this dictionary ensure that this is the case by drawing on sizeable corpus resources.)
For a syntactic analyser it is necessary that the words be given some identifier to show the grammatical category of the word.
Unlike the detailed grammatical information used in the unification grammars, statistical techniques tend to employ a set of grammatical tags.
A tag is simply a marker of the grammatical category (e.g. NN may indicate that a word is a common noun, PN that it is a pronoun).
A tagset is the selection of grammatical wordclasses that are used by a system.
The tagset used for the initial testing of the method consisted of a subset of the grammatical tags provided in the TEXT710 version of the OALDCE (Appendix 3)(supplemented by some extra tags to indicate punctuation and sentence boundaries).
There are 60 grammatical categories specified within this lexicon indicating such properties as transitive verb, plural noun, proper noun etc.
The grammar tags have been converted into numerical codes (1...60) for computational simplicity.
For the initial investigation no tagged corpus was available, instead the text used for training consisted of a selection of 'raw'(i.e. just plain text) estate agent's details.
Hence it was necessary to first determine the grammatical categories of the words in the corpus.
There were two options available at this stage -- either to manually tag the corpus or to try out some automatic determination of the tags.
It was decided to first try out the simplest possible computational technique.
First the corpus was taken and analysed to determine where the sentence1 boundaries occurred.
This is not completely trivial.
Sentence boundaries in real text are often indicated by textual layout (e.g. a blank line may be sufficient to indicate that a statement is finished).
Neither is it valid to assume that a full stop indicates sentence termination.
Full stops are also present in real numbers, abbreviations and after numerals to indicate a point in the argument.
Once the text had been divided into sentences the words and punctuation were replaced by the corresponding grammatical code(s) from the OALDCE.
For the preliminary investigation no attempt was made to ensure that the correct code was assigned.
This meant that the method over-generated i.e. the correct tag assignment was present as well as incorrect assignments.
4.4.1.
Results of Preliminary Investigation
The processor was tried out on a set of estate agents texts that the generative grammar had difficulties in processing.
These texts are examples of real-life language.
The results for the syntactic processing of a typical sentence (the cliff top with its access to sandy bathing beaches is also within easy walking distance) are shown below.
The score assigned to the word by the syntax analyser is shown beneath the word.
The correct words are given a shaded background.
4.4.2.
Discussion
Similar results to those shown in the example were achieved for the other estate agents sentences tested using this simple analyser.
That is, for approximately 85% of words for which alternatives were available, the correct word was selected by the analyser.
These results compared very favourably with those obtained using the rule-based system.
With the ANLT system these sentences had achieved very poor results whether as a result of the grammatical construction being absent from the grammar or through computational failure of the system.
In contrast this very basic technique is robust and computationally simple.
The processing time was approximately 3 seconds per sentence using the same computer (a Sparc 4/75 with 48 MBytes of memory), and LISP system (KCL) as the ANLT investigation.
For the statistical processor the processing time is linearly proportional to the sentence length, as opposed to rule-based method which was exponentially related.
The computation required for each sentence is extremely simple and does not require the complicated intermediate storage of possible parses to be kept in memory.
Of course there are a number of modifications that need to be made to this analyser before it is able to cope well with general text.
The most obvious improvements to be made are the use of a large tagged corpus and development of a compatible, large vocabulary lexicon.
The translation to a more portable programming language is also important for compatibility with the other modules of the recognition system.
The rest of this chapter will deal with such modifications and improvements to the probabilistic analyser.
4.5.
Machine Readable Corpora of English Texts
A corpus is a body of text or speech that provides a representative sample of a language.
The importance of corpora to workers in linguistics is that they provide real instances of the language (as opposed to interesting phrases that a linguist may conceive) and may be used to train and test natural language systems.
For the current project a corpus is used to train the higher level modules (i.e. lexical, syntactic and semantic) with information about the behaviour of text.
A number of corpora exist for text (e.g. the Brown Corpus (Kucera & Francis, 1967), the Lancaster-Oslo-Bergen (LOB) Corpus of British English (Johansson, 1980, COBUILD Corpus (Sinclair, 1987a)) and a small number of speech corpora (e.g. London-Lund (Svartvik & Quirk, 1980)).
4.5.1.
Uses of Corpora
Rule-based linguists have traditionally made little use of corpora.
Generative linguists argue that corpora are finite and degenerate and are therefore unable to deal with many of the phenomena present in language.
Such are the difficulties that have been found when trying to scale up toy grammars that computational linguists are drifting away from the creation of demo-sized systems and aiming at larger systems.
Rather than become side-tracked by interesting but relatively insignificant phenomena, some generative linguists are now using corpora to identify where a grammarian's efforts can best be concentrated.
Corpora are also useful to generative grammarians for providing an objective and reliable evaluation of a grammar's coverage and performance.
The use of corpora for automatic acquisition of lexical information has been advocated (e.g. (Brill et al, 1991),(Finch & Carter, 1992),(Atwell, 1987)).
Before this becomes a realistic proposition some developments are necessary.
The main reason for this is that extraction of many types of information from corpora usually pre-supposes the capability to automatically analyse the raw text in various ways.
This will itself involve the development of substantial lexicons.
Statistical analysis of language requires investigation of corpora to derive probabilistic models of the language.
Within the current project use is currently being made of corpora to derive information relating to the frequency of grammatical transitions, word frequency, concordances of words and probable grammatical categories of words.
The use of corpora is becoming increasingly important in the production of dictionaries.
A revolutionary approach to dictionary creation was taken with the COBUILD dictionary (Sinclair, 1987b).
In this dictionary each definition was derived from a 7.3 million word corpus.
There are other ways in which the COBUILD dictionary is unique -- explanations of words are given in sentence form and supported with real examples from the corpus.
While other dictionary publishers are not taking the extreme stance taken by COBUILD, corpora are being recognised as increasingly useful tools for creating up-to-date dictionaries that accurately reflect the language.
As proof of the importance that dictionary publishers are placing on corpus use, three rival British dictionary publishers (Oxford University Press, Longmans and Chambers) are currently collaborating in the production of the British National Corpus.
4.5.2.
Classification of Text Corpora
Corpora may be classified according to how much analysis the corpus creators have performed on the text.
Three formats of corpora exist (as defined by Leech (1990)):
Raw Corpora.
These consist solely of the 'pure' text, no attempt having been made to annotate the text.
Such corpora are very common since any body of text falls into this category.
Tagged Corpora
Tagged corpora contain not only the raw text, but also the grammatical category of each of the words.
Although less common than raw corpora, tagged corpora are becoming more common with the development of reliable (>95% correct) automatic taggers (e.g. CLAWS (Garside, 1987), PARTS (Church, 1988)).
Examples of this type of corpus are the LOB and the Brown corpus.
Analysed Corpora.
Analysed corpora contain not only the tags of each of the words but also a syntactic analysis of each extract.
The rarity of such corpora reflects the effort required in the analysis.
Analysed corpora are produced either by manual parsing or automatic parsing with manual correction.
Although tools have been developed allowing trained users to perform analysis quickly the process is demanding.
4.5.3.
Corpus size and Composition
When building a corpus of text the general maxim is that bigger is better.
There is a huge imbalance in the frequency of words in the English language.
A large proportion of a corpus is made up of only a few words (about 25% of a corpus is made up of just 15 different words).
Even for a very long text, about half of the words found will only occur once.
In order to study the behaviour of a word it is necessary to have a large number of occurrences available which necessitates that modern corpora contain many millions of words (Sinclair, 1991).
The size of a corpus is not the only important factor, it is also vital that the corpus be 'balanced'.
That is, the corpus should be representative of the language according to a number of variables (e.g. time-span, geographical and social range, authorial background, publishing history, discourse model and type (Sweeney, 1992)).
In order to achieve this balance care must be taken in the selection of sources from which the corpus is created and how much of each source is used in the corpus.
Corpus compilers base the composition of a corpus on some model.
In order for a corpus to be representative it is necessary to define the whole of the text that the corpus is to be a sample of.
For a finite domain (such as the Bible or the works of Shakespeare) this is simple.
For unconstrained language this definition is much more difficult.
A number of approaches have been used to achieve this definition.
Yang (1985) assumed that a library was a microcosm of the written language; creators of the LOB corpus relied on the comprehensiveness of established bibliographic sources (Hofland & Johansson, 1982).
Clear (1990) suggests that perhaps the size of a corpus is more significant than its composition although the two parameters are inter-dependent.
Testing of this hypothesis is difficult since probabilistic language models tend to be tested on data from the same domain as the training corpus, hence domain specific effects are unlikely to show up.
4.5.4.
Availability of Corpora
Corpus availability is currently limited due to legal constraints although some have been cleared for academic research.
The most widely used text corpora are the Brown and the LOB corpora, both of which have been tagged.
Both of these corpora are about a million words in size and are derived from a variety of domains.
It is now realised that a million words is insufficient to produce an adequate model of a language since many of the phenomena of the language are so rare that they will be absent from such a corpus.
Recent corpora are significantly larger, the Birmingham Collection is reputed to be about 100 million words (although sadly this corpus is very difficult to obtain) and the Longman/Lancaster corpus of 30 million words.
Currently being created are and the Text Encoding Initiative in America (Walker, 1989)(Marcus, 1990a) is producing a corpus of 100 million words.
A similar initiative is also underway in Britain to create the British National Corpus which will also contain 100 million words, due for completion in 1994.
With such large corpora it is believed that improved language models may be created.
Analysed corpora are rare and only small examples of this genre are currently found.
These include subsets of the LOB corpus (Lancaster-Leeds Treebank, 45000 words) and the Brown corpus (Gothenburg corpus, 128000 words) although some others exist (e.g. Nijmegen Corpus 130,000 words and the Polytechnic of Wales Corpus of 100,000 words of children's speech).
4.5.5.
The LOB Corpus
The LOB corpus has been the principal corpus used throughout the current project.
The LOB corpus is made up of 500 British English (printed) text samples from the 1960's.
Each sample contains approximately 2,000 words and the complete corpus contains about a million words.
Each word is tagged with its grammatical category.
The texts were selected by the corpus compilers to provide a wide representation of text types to allow research on a broad range of aspects of the language.
To allow for testing of the analysers in the current project approximately 5% of the corpus has been removed.
The composition of the test corpus and the reduced LOB corpus is shown in table 4.3.
The grammatical tagset used in the LOB corpus (Johansson, 1987) is shown in Appendix 2.
There are 132 members of this tagset (minor variations of this tagset occur however, for details see Sampson (1987b)).
The simple wordtags either stand for classes of grammatically similar words (e.g. NN for singular common noun) or for closed class words that have a special function (e.g. TO indicating when the word to is infinitival).
The tagset will be discussed in more detail later.
4.6.
Machine Readable Dictionaries (MRDs)
The natural place for a person to look for information about an unknown word is in a dictionary.
A dictionary contains lexical entries for a large number of words.
Each entry has been hand-crafted by a skilled specialist in lexical analysis.
As the production of dictionaries becomes computer-based so dictionaries are becoming increasingly available in machine readable formats.
Over the last few years work has been carried out on using machine readable dictionaries as a source of linguistic knowledge (Amsler, 1981),(Boguraev & Briscoe, 1987)-- thereby enabling the production of a large vocabulary lexicon for considerably less effort than was previously required.
Given the (relatively) small size of the LOB corpus, supplementation with information from MRDs provides advantages for the current system:
Vocabulary Information.
The LOB corpus contains information for less than 40,000 types, available MRDs contain information for a larger number of types.
Hence the use of an MRD provides a larger vocabulary for the system.
Syntactic Information.
Many types occur just once in the LOB corpus.
More instances that this are required to be certain that the possible parts of speech of the word are found.
The supplementation of the grammar information from the corpus with that from the dictionary alleviates this problem since a dictionary aims to provide a list of all of the possible parts of speech (although a dictionary does not indicate the likelihood of a word behaving in that manner).
The Longmans Dictionary of Current English (LDOCE) and the Oxford Advanced Learners Dictionary (OALDCE) are particularly rich sources of syntactic information.
Both of these dictionaries use structured grammar codes to indicate grammatical patterns in which the words may participate.
More general dictionaries tend to define grammatical information less formally than these learners dictionaries.
Semantic Information.
The dictionary definition of a word is one form of semantic information.
(other forms include collocational information, encyclopaedic knowledge, semantic networks).
There are problems associated with extracting semantic information from dictionaries.
Most importantly no widely recognised classification of semantics exists.
Even if such a system did exist it is unlikely that it would appeal to lay users of a dictionary.
For a human reader a discursive natural language definition is a more sensible format.
If one investigates the format commonly used in dictionary definitions it is clear that there are some conventions that are followed.
A restricted style of language is used in dictionary definitions with a number of phrases being extremely common (e.g. of or pertaining to, also called, compare).
Researchers within computational lexicography have attempted to exploit this style to extract semantic information.
One method involves parsing the dictionary definitions to identify the genus term in the definition and produce semantic relations (e.g. sets, hierarchies and networks)(Ahlswede & Evens, 1988).
Beyond this implicit semantic information the machine-readable version of LDOCE also explicitly contains semantic tags which have been used (e.g. (Slator, 1989)).
Morphological Information.
Regular inflected forms of words are not given their own specific dictionary definitions.
This duplication of definitions would serve little purpose.
Dictionary authors expect users to perform regular morphological processing themselves.
Within the definition of a root form of a word is contained information about the inflected forms of the word.
Hence a reader looking up the definition of a word such as dogs would find the relevant information stored under the word dog.
For words that inflect irregularly this is not the case.
The irregular inflection will be presented as a headword with a reference to the appropriate root form.
Hence the definition of wept will contain the brief description that it is the past tense of weep thereby informing the reader where to seek further information.
Morphological processing is not quite so trivial for a computer as it is for a human.
Within the Text710 version of the OALDCE are inflexional codes which provide a means of producing the standard inflections that a human would be able to generate automatically.
The dictionary definitions contain the information about the irregular derivations.
For example under the headword child is the information that the plural is children rather than childs.
Such information has been employed by the current system to produce a computationally simple means of morphological processing and automatic lemmatisation.
Compound Information.
Compounds can pose problems for automatic text processing systems since many compounds of words are idiomatic and have developed meanings and grammatical properties different from the combinations of the individual words.
Dictionary compilers recognise this problem and provide separate definitions for such compounds.
Hence from using a dictionary one could determine that hot dog bears little relation to canines.
There are many such compounds in dictionaries (e.g. the Collins English Dictionary contains about 30,000 compounds).
The use of compounds is discussed later.
4.6.1.
Availability of MRDs
Until very recently MRD's have been available only as a typesetting tape of a printed dictionary (Meijs, 1992).
Problems often exist in persuading dictionary owners to part with these typesetting tapes for understandable reasons (the production of a pirate copy from a typesetting tape is a trivial matter and can have serious repercussions for dictionary publishers).
Over the last decade the major dictionaries that have become available (in a limited manner) to researchers are:
Collins English Dictionary (CED)
Longmans Dictionary of Contemporary English (LDOCE)
Merriam-Webster 7th edition (W7)
Oxford Advanced Learners Dictionary of Current English (OALDCE).
Text710 (a limited version of OALDCE, elsewhere referred to as CUVOALD and CUV2)
Dictionaries are produced with specific markets in mind and their contents vary accordingly.
Learner's dictionaries (e.g. LDOCE and OALDCE) are designed for people for whom English is their second language.
Fewer assumptions are made in such dictionaries about the linguistic competence of the user.
Hence the definitions aim to be very clear and precise, perhaps produced using a restricted defining vocabulary.
The grammatical and morphological information in learner's dictionaries also tend to be very explicit.
At the other extreme is the Oxford English Dictionary (OED).
The definitions contained in this dictionary are voluminous descriptions of the etymology of the words, the dictionary being designed for people with a deep interest in the English language.
Between these two extremes are the general purpose dictionaries such as the CED.
A major difference between the dictionaries is the number of words defined -- learner's dictionaries tend to contain about 30,000 headwords, general purpose dictionaries (e.g. CED) have about 100,000 definitions, Websters Third New International has 450,000 and the Oxford English Dictionary has about 500,000 definitions.
4.6.2.
Extracting Information from Collins English Dictionary
The use of MRD's to provide information for computational linguistic applications is a very attractive proposition.
The potential benefits from using MRD's include the availability of lexicons with syntactic, semantic, morphological, phonological and idiomatic information for a very large number of words.
Unfortunately there are some difficulties involved.
The main problem is that a dictionary is designed to be used by a human being, not a computer program.
The entry in a lexicon required by a NLP researcher may require full and formal treatment of morphology, syntax and semantics.
Such a dictionary is unlikely to appeal to a member of the general public wishing only to check on a word when solving a crossword or playing Scrabble.
However, such users are a substantial market for dictionary publishers, therefore the information in dictionaries relies on the linguistic skills and background knowledge of the human user, rather than being designed for machine consumption.
(Gross (1984) argued that the information in MRDs is too unreliable and unsystematic for use in NLP applications).
There is also a difference between machine readable and machine usable.
A sample of a machine readable typesetting tape (from the CED) is shown below, indicating the definition for the word abandon.
Dictionaries make heavy use of typography to convey information to the reader.
Typography within a dictionary indicates the function carried out by particular parts of the text.
For example the headword of a definition is generally in bold font of a particular size.
To achieve such changes in typeface typesetting codes are interspersed with the actual text.
Information contained in the dictionary (lexical, syntactic, morphological, semantic and compound) was necessary for the higher level modules of the current project.
This information could only be obtained by conversion of the CED typesetting tape to a usable representation of the information.
The transition between a typeset machine-readable form and a lexical database requires considerable effort (Amsler, 1989),(Alshawi, Boguraev & Carter, 1989).
Removal of the typesetting information from a dictionary to leave just the bare ASCII text loses useful information that the typography makes visually apparent.
The aim is to replace the typographical codes with computer-usable codes that make explicit the function of the text.
Unfortunately there is no simple translation between typesetting codes and role played by the text.
Dictionary publishers have limited fonts available and the same font may indicate different roles in different parts of the definitions (human readers are easily able to use the definitional context to determine the function of a particular font change).
Separate typesetting codes are used to indicate the typesize and the font of the text.
It is the combinations of such codes that determine the visual presentation.
The processing required to convert the typesetting format to a usable format was essentially iterative and needed to be performed with care.
Many of the problems involved in the processing derived from the size of the typesetting tapes.
The original CED typesetting tape occupied about 27 MBytes.
It was impossible to completely check the global effects of each alteration to the file.
Another source of problems was the presence of some errors within the text.
Although such errors were rare, processing had to be defensive against allowing an error (such as unbalanced parentheses) to affect a large piece of the text.
An early stage of refinement delimited each individual definition.
All processing was carried out on one definition at a time.
This ensured that errors were localised to the definition in which they occurred.
Unbalanced brackets were automatically detected by counting the brackets within each definition and inserting extra brackets in the most likely location to correct the error.
Dictionaries contain many non-ASCII characters such as phonetic and mathematical symbols (e.g. h, p, q, S etc.).
Typesetting methods produce such characters by combining typesetting codes (e.g. q is produced by PSGg, e by $e).
The first step in processing the typesetting tape was to find the mapping between the typesetting codes and the special characters.
This involved a painstaking search of the paper dictionary to find examples of each of the special characters.
Having found the location in the dictionary the computer file was then searched to find the corresponding location.
An initial estimate was made of the typesetting combination responsible for the character.
To check that the correct combination had been selected the typesetting file was searched for more examples of the combination and the effect on the paper dictionary checked.
After verification of the effect of the typesetting code combinations they were substituted for more computer-recognisable codes to simplify further processing.
Once the special character combinations had been replaced the file was iteratively refined.
Such cleaning up involved determining the structure of the dictionary definitions and the effects of font changes at particular locations within the definition.
This processing recovered such information from the text as the domain of use of a particular word, sense numbering, phonetic information etc..
The font information contains only some of the information about the roles of the definition's contents.
Further information was extracted by taking advantage of the use of dictionary-specific phrases (elsewhere referred to as defining formulae (Markowitz, Ahlswede & Evens, 1986)) to indicate the roles of their subjects.
For example'esp.
in the phrase...' will indicate a common phrasal usage of a word; 'foll. by...'indicates words which commonly follow a word; 'Comp.:'provides a word with which the reader should compare a word with.
These phrases are often embellished in the text with modifiers and slight variations (often, usually, also, mainly) which give some extra information to readers of the dictionary.
The presence of such optional words further complicated the matching of the word patterns into usable codes.
Making explicit this information serves to highlight the semantic relevance of words within the definitions that may be used by the semantic processor to construct semantic relationships between words.
The grammatical information present in the CED is very basic and of little use for developing the lexicon for a parser.
Such information was derived from the computer usable form of the Oxford Advanced Learners Dictionary.
The use of this information will be described later.
Not all the information present within dictionary definitions is of use for handwriting recognition.
Some information has necessarily been removed from the dictionary to cut down on storage requirements.
Examples of this are the phonetic information and the etymological descriptions of words.
Modification of the typesetting tape of the CED has produced a version in which the following information is encoded:
The sample definition of abandon shown earlier is now represented as:
Within this format codes are used to show the particular role of the text.
In the example PSH indicates headword; PSx indicates sense numbering; PSW indicates derivative forms of the headwords; PSG the major grammatical category; PSg a grammatical sub-category.
As can be seen from the example this is much easier to use than the initial typesetting format.
Further modification is possible.
Most importantly, since this work on the dictionary was completed Standard Generalized Markup Language (SGML)(Bryan, 1988) has been accepted as a standard for encoding document information.
The hierarchical structure of the SGML system is substantially different from the system adopted for the processing of the CED.
Since this work has been carried out other dictionaries have also become available.
For example the OALDCE is now available in SGML format, also the restrictions on the availability of the LDOCE have been relaxed.
4.7.
Developing the Syntactic Lexical Information
The essential action of the probabilistic syntactic analyser is to find out how often each set of grammatical categories are found together in the corpus.
When candidate words are presented to the analyser by the recogniser the grammatical categories of these words are looked up.
The neighbouring grammatical categories are examined to determine the possible grammatical transitions.
The likelihood of each transition is retrieved from the transition matrix.
The words are then assigned a rating based on the probability of the grammatical transitions that their tags participate in.
The main information that the probabilistic syntax analyser needs from the lexicon is the grammatical tag of each of the candidate words.
This grammatical tag must be compatible with those that the corpus was tagged with.
The syntactic information stored in the lexicon represents a combination of information from the Text710 version of the OALDCE dictionary and the LOB corpus.
Both the Text710 and the corpus contain information about words and their grammatical tags.
Information is wanted from both sources since the Text710 has a larger vocabulary (approximately 68,200 words) than the corpus (approximately 38,000 purely lexical strings).
The dictionary also provides a more complete set of grammatical tags than the corpus.
Many of the words in the LOB occur just once, an inadequate basis for determining grammatical behaviour of a word.
The use of the corpus is vital for determining the probability of grammatical transitions.
Also, when words are syntactically ambiguous (i.e. they can have more than one grammatical tag) the corpus can provide information about how often each of the possible tags is used for that word.
A definitive set of grammatical tags does not exist.
Existing tagsets tend to agree about the major word class categorisations but there are often differences in the sub-categorisations used.
The tagset used by TAGGIT (Greene & Rubin, 1971) for tagging the Brown corpus contained 86 different grammatical tags.
The CLAWS tagset used for tagging the LOB corpus is very similar to that of TAGGIT but is somewhat larger with 134 tags to achieve greater delicacy of analysis while preserving compatibility with the Brown corpus (Johansson et al, 1986).
The differences between the LOB and Brown tagsets are due to minor alterations made by subcategorising some classes and redefining the boundaries of others.
The grammatical tagset used by the Text710 (appendix 3) is considerably smaller than the LOB tagset2.
As the difference in the number of tags in the tagsets indicates, the Text710 and the LOB tagsets are not directly compatible.
To use the information from both sources a translation process is required.
The result of this translation is a single lexicon having a set of tags derived from combining the Text710 and LOB tagsets.
The tags in the corpus must also be translated so that they conform to the same tagset as this new lexicon.
The first step of this translation process is the amalgamation of the LOB and the Text710 tagsets into a single common tagset.
The aim of this merger is to retain as much information as possible between the two tagsets.
There are problems associated with combining the two tagsets.
Both tagsets contain information that is absent from the other.
The Text710 tagset contains 27 distinct tags for verbs indicating transitivity, tense and details about anomalous verbs.
The LOB tagset uses just 5 tags for general (i.e. open class) verbs.
The only way in which one could incorporate such fine-grained information from the Text710 would be to completely re-tag the LOB corpus.
This was not feasible for the current project, hence these tags were replaced by more general tags.
However many of the tags used exclusively in the LOB corpus can be retained.
The LOB tagset contains a number of tags for closed class words (e.g. personal pronouns, conjugations of to be and to have).
These can easily be incorporated into the Text710 tagset.
A number of tags are missing from the Text710 tagset because they are for word types that are not found in dictionaries.
For example, genitives are not stored because they can be easily identified, nor is punctuation classed as a dictionary part of speech.
The combined tagset that results from amalgamating the LOB and Text710 tagsets contains 109 tags (Appendix 4).
The tags in the LOB corpus and Text710 dictionary were then substituted for those of the combined tagset.
This was achieved using a simple NAWK program (Appendix 7).
To translate the corpus information a file was created containing two columns -- the first field on each line contains the original LOB tag, the second field contains the tag that it is to be translated to.
The initial step for the program is to load in this file and store the translation of the original tag with its replacement using a NAWK associative array.
The corpus is then processed a line at a time.
The LOB tag is extracted from the line (they are located in columns 13-17), it's replacement found from the associative array and the line printed out using the replacement rather than the original LOB tag.
A similar process is used for the OALDCE dictionary.
Full checking that the translation process is valid is not possible.
There will obviously be some differences between the tag assignments made by lexicographers and corpus compilers.
A simple indication that the mapping between tagsets had been successful was made by investigating the words that were common to the corpus and the Text710.
These were checked to determine that the part of speech assigned in the corpus was one of the possible tags for the word in the Text710 lexicon.
Of the 920,297 tokens in the LOB having entries in the Text710, the assigned tag was contained in the lexicon 889,629 (96%) times.
This represents an acceptable level of compatibility.
4.7.1.
Syntactic Ambiguity
Ambiguity in written English may arise from different words having the same spelling (homonyms) or the same word having a number of meanings.
Humans seldom notice ambiguity in language unless it is being exploited in word play (such as puns).
However for computer processing of natural language ambiguity causes a large problem.
In fact some researchers consider syntactic ambiguity to be the largest problem facing NLP (Gazdar & Mellish, 1989).
A large number of words in the lexicon are syntactically ambiguous.
Using the simple tagset of the current system the degree of syntactic ambiguity found is shown in table 4.4.
Much of this ambiguity arises through relatively rare usages of the words.
A contributing factor to this is that a dictionary aims to provide for every usage of a word with little indication of how common that usage is.
Church (1988) asserts that the most frequent category is almost always sufficient.
Consequently, the grammatical tags for each word in the lexicon are given a grammatical frequency factor (GFF) to distinguish the importance of the different tags.
The GFF derives from the equation:
That is, the GFF represents the frequency of a tag for a word relative to its most frequent tag.
Since not every word in the lexicon is present in the corpus and those present need not appear with their rarer forms, the corpus is supplemented by the lexicon for determining the GFF (i.e. each word in the lexicon is assumed to occur once in the corpus with each of the tags in the lexicon).
This GFF value can then be used when processing the input to reduce the significance of rare syntactic forms of words.
4.8.
Word Index System
The morphological processor described in chapter 3 is used for a number of the post-processing components (compound recognition and semantic processing) of the current system.
The principal action of the morphological processor is to associate a word with its morphological root (e.g. run, runs, running, ran will all be associated with run).
The method used to achieve this is to pre-process the information into the system.
At development time the root forms of each word in the lexicon has been determined utilising information from MRDs and manual checking of the unresolved words.
At run time, when a word candidate is proposed the compound recogniser (described below) and the semantic processor receive the root form of the word candidate.
Hence if running were a candidate then these two modules of the system would receive as their input run.
Rather than explicitly store the root forms as strings an indexing method has been developed.
It is more demanding for a computer to process and store strings than numbers.
Neither the semantic nor compound analysers in the current system are concerned with the actual string representation of the root form of the input word.
By storing an integral index (the ROOT-INDEX) which uniquely identifies each morphological root it is possible to reduce the storage requirements and processing is simplified.
This ROOT-INDEX can then be stored in the lexicon.
Therefore within the lexicon run, runs, running, ran each have the same ROOT-INDEX stored (19590) to indicate that they are all derived from the same morphological root.
4.9.
Processing Compounds
It is a common misconception that a word is simply a contiguous sequence of letters.
This is not always the case.
Phrases can often behave as though they were single words.
The meanings of certain types of phrases have come to mean more than simply the combination of words from which they are composed (sometimes they bear no relation to their constituents).
Furthermore they behave like words in that they are always used together in appropriate contexts and they appear to be represented as words in the mental lexicon (Wilson, 1984).
Such combinations are generally difficult for natural language systems to process due to the changes in the syntactic and semantic properties of the constituent words (and the acquisition of information about the compounds).
For the probabilistic syntax system compounds are troublesome since the window may not contain all of the compound.
Even if the complete compound exists within the syntax window it is likely that the incorrect parts of speech will be assigned to the constituent words resulting in an erratic selection of words by the analyser.
For the complete recognition system however the existence of idioms (and compounds in general) may be exploited to further reduce the uncertainty as to the correct input as the following example shows.
The word recognition alternatives for the words tennis courts was:
The only similarity between the words in these lists is that the pattern recognition system perceived them as being orthographically similar.
The presence of words that form a compound in consecutive positions other than by intention is improbable, hence from knowing the existence of a compound tennis courts one would give priority to these individual words over their alternative candidates.
4.9.1.
Acquisition of compounds
Entries for a large number of compounds (about 30,000 in the CED, 11,000 in the OALDCE) can be obtained as headwords and subentries (e.g. 'in one's right mind', 'by rights' occur under the headword 'right'; 'right about', 'right angle', 'right away'occur as headwords).
In fact, dictionaries of compounds, idioms and common phrases also exist, although none are available in machine readable form.
Within corpora, compounds are tagged with a single tag for the whole compound allowing a set of compounds to be easily extracted.
The LOB corpus employs ditto tags to indicate words whose syntactic roles differ from the role of the same words in other contexts.
Hence as to is marked as as/IN to/IN'; so as to is tagged as so/TO as/TO' to/TO'.
It is also possible to take advantage of compounds, phrases and idioms which are specific to certain domains and/or contexts.
For example in Estate Agents documents the phrases 'within easy walking distance' and 'sought after'are very common, and for use in such a context, these would be included in or added to the compound lexicon.
In a closed domain a list of relevant idioms is easily attainable (Zernik, 1989).
Some care must be taken not to store every compound that one finds.
Many compounds present in dictionaries are extremely rare (e.g. whirligig beetle and ruby spinel) and their storage would serve little useful purpose for normal text processing.
To avoid such unnecessary storage, compounds are only stored if the frequency of the least frequent word in the compound is above a threshold.
Although not foolproof (rare compounds of high frequency words will still be permitted) this has removed many of the rare compounds.
There are a number of compounds (particularly personal nouns) that will not be found in a dictionary which will be of use to an NLP system.
4.9.2.
Representation of compounds.
The number of compounds is large and an efficient method must be used to store such information.
It is necessary to organise the compounds in such a way that they are treated in a similar manner to individual words.
That is, when a compound (e.g. ice cream) is recognised an index relating to the location on disk of the lexical information relating to this compound should be returned.
One way to do this would be to treat a space as a legal character and build the whole of the set of compounds into the word recognition tree with an index at the end.
The space required to do this is prohibitive however.
An alternative is to build a second tree consisting of the words in the compounds, i.e. a 'compound tree'.
This tree consists only of words that partake in compounds and at the leaf nodes are the indices relating to the disk address of the lexical information.
The ROOT-INDEX derived from the morphological processing provides a simple and efficient method of storing such information.
Rather than build the compound tree from the actual strings that form the compound, the tree is built from the root-indices.
Each word, when it is verified by the word recognition stage, returns an index that points to its morphological root.
By taking the set of compounds which are to be used and replacing each word within the compound by its corresponding root-index the storage requirement is drastically reduced.
The following example should clarify the process:
The CED contains the following compounds of the word 'accommodation' for storage in the compound tree: accommodation address; accommodation bill; accommodation ladder; accommodation platform; accommodation rig
The space required to store each of these as strings is prohibitive however.
Each of the words within the compounds has already been assigned a root-index as in table 4.5.
By replacing the strings with the root-indices the information that needs to be stored is reduced.
The actual information which is stored for the compounds above is shown in table 4.6.
Hence, rather than actually store the string accommodation ladder in a tree, it is necessary only to store the integers 2049 12467.
There are a number of advantages to be gained from using an index-based compound tree.
Firstly the storage requirement is significantly reduced since integers may be stored in a much smaller space than strings.
Numbers are also better suited to computer operations with the result that the process of searching the tree is made computationally simple.
A further advantage of using the root-indices to build the tree is that inflectional compounds need only be specified once.
For example, the root-indices for the words court and courts are the same.
Hence the indices to be stored for the compounds tennis court and tennis courts will also be the same.
There are also some disadvantages.
Firstly it is not possible to store the syntactic tag for the whole of the compound within the compound tree.
This is because the compound could relate to a set of different word combinations (e.g. 'tennis court' is a singular noun but 'tennis courts'is a plural noun).
The result of this is that the syntactic analyser will not be able to select the correct part of speech for the compound.
However, if a sequence of words is flagged as a compound then the control program can ignore the syntax results which will be unpredictable.
An additional problem is that certain combinations of words may form a compound whereas different morphological forms of the words may not.
For example'running water' is a valid compound but 'ran water'is not.
In practice the probability of such words occurring in adjacent positions is so low that the problem is negligible.
4.10.
Lexicon Storage
For a large lexicon it is necessary to use a storage technique for the lexicon that allows both rapid search for information and compact storage of the information.
The lexicon in the current project is stored in two structures: a word lookup tree and a compound tree.
Both structures are represented in memory as tries.
4.10.1.
Structure of the Word Lookup Tree
Various types of information about the words are stored at the end of word nodes in the trie.
This includes the syntactic information and the root index for the word.
The structure of the trie is shown diagrammatically in fig 4.4.
Each node comprises information about which letter is allowed as well as pointers to child and sister nodes.
Of most importance to the higher level processes is the information stored with the end of word flag.
If the trail through the trie has successfully found a word then the linguistic information for the word is found there.
This information comprises the root index information, the grammar tags and their respective GFFs as well as word frequency information for the word.
A system of flags is also stored here to indicate whether the word must begin with a capital to form a valid word, if the word is a common mis-spelling and a flag detailing whether the word can begin a compound stored in the compound tree.
For more information see Wells (1992).
The sample lexicon shown is the representation for the words cat, catch, cot, cots, dog, doggy and dogged (from (Evett et al, 1991)).
The index of each member of the compound is represented at a node in a trie (Wells, 1990).
The node has three pointers.
One points to the next member of the compound.
One points to an alternative next member of the compound.
The third either points to the index for the whole compound or is a dead end.
The existence of an index here indicates a possible end of compound, and the index gives the address of that compound in the lexicon, where a full entry for that compound will be found.
Fig 4.5 shows the structure of the compound tree for the compounds 'rain cats and dogs', 'tennis ball' and 'tennis court'.
Within the diagram the root index of the word is indicated by preceding the word with PS.
Hence 'PSrain' refers to the integral root index for the word 'rain'.
Access to the compound tree is achieved through the word recognition tree.
Each word in the word recognition tree is flagged to indicate whether or not it is a member of a compound.
The compound flag can be one of two alternatives: start of compound/phrase; or part of compound/phrase.
Given an allowable candidate word which is known to be able to start a compound or commonly used phrase, if the next part of the compound occurs in the list of alternatives for the next word position (and so on until the end of the compound), then it is likely that the alternatives that make the compound are the correct choices.
It is therefore worth checking the position(s) adjacent to that in which a start of compound occurs for its next element, until the end of the compound (or phrase) is reached, or until the next element is not found.
When words are looked up in the word look-up tree, if the flag for start of compound is set, the compound tree is checked.
As adjacent word positions are also looked up in the word look-up tree, they are checked for whether or not they are the appropriate words to complete the compound.
If they are, these words are given priority over the other alternatives available for each word position.
It is not possible for a lexicon to provide full coverage of language.
Among the reasons for this are the continued creation of neologisms, the introduction of foreign words and the vast array of proper nouns that exist.
One of the most important considerations to be taken into account when creating a lexicon is how full a coverage of the language it is able to provide.
There have been some investigations into determining how fully lexicons derived from machine readable dictionaries are able to cover unseen text.
There is some discrepancy to be found between the available results however.
Walker and Amsler (1986) compared the words found in Webster's Seventh New Collegiate Dictionary (W7)(G. & C. Merriam, 1963) and an 8 million word sample of the New York Times News Service (NYTNS).
Between the two sources there were 119,630 different types.
Of these only 23% (27,837) were common to both sources, with 36% (42,695) existing exclusively in the W7 dictionary and 41% existing exclusively in the NYTNS sample.
That such a large proportion of the types in the news wire sample were not found in the W7 dictionary is a worrying statistic.
Analysis of these omissions from the W7 showed that about a quarter were simply inflexions of words that already existed in the dictionary.
Such absences can be simply remedied by improving the morphological processing.
Another quarter of the omissions arose from previously unseen proper nouns.
This class of omissions are much more difficult to deal with computationally.
Sampson (1989) investigated the coverage provided by the OALDCE for a 50,000 word sample of the LOB corpus.
The results of this were more encouraging than Walker and Amsler's.
Of the 45,622 tokens in the corpus sample that would ideally be found in the dictionary (ignoring non-alphabetic combinations and enclitics), 95% (43,490) were found in the dictionary as they stood.
This figure increased to over 97% (44,448) after some minor modifications such as removing hyphens and changing word-initial capitals to lower-case.
A similar investigation was carried out with the lexicon used in the current system to determine it's ability to cope with unseen text.
In order to test the coverage of the lexicon, the test part of the LOB corpus was retrieved and each word in the text checked for its existence in the lexicon.
This lexicon contains over 70,000 words and is a combination of the Text710 dictionary and words from the (training) LOB corpus.
The test corpus is a representative 46,003 token sample of the LOB (the composition of which was detailed in table 4.3) containing 7,853 different types.
Of these 1746 types (1121 tokens) were not stored explicitly in the lexicon (ignoring numbers and formulae).
Further processing to remove hyphenation and possessives reduced the number of absences to 1186 types (669 tokens).
Hence more than 97% of the tokens and approximately 91% of the types in the test corpus are covered by the lexicon.
The largest proportion of absences are proper nouns and foreign words which account for 78% of the missing tokens (75% of the types).
Of the remaining absences only 42 types occurred more than once in the 46,000 words.
From this it is deduced that the lexicon provides adequate but not complete coverage.
The poor coverage shown by the Walker investigation must be largely as a result of the high proportion of proper nouns found in newswire text.
The improvements to be attained by extending the lexicon further also appear to be minimal.
There are a number of reasons for this.
Firstly it is not guaranteed that the required information will be in the larger lexicon.
When one considers that the main reason that words are absent is that they are proper nouns, which will also be unavailable from the larger lexicon, it is clear that omissions are unavoidable.
Furthermore with a text recognition system, as the lexicon gets larger the problems increase.
The situation is shown in table 4.9.
A 521 word text was written from a randomly selected business text, not associated with the LOB corpus.
The text was processed by the handwriting recognition system with lexicons of different sizes (based on lemmatised word frequency values).
The number of word candidates suggested by the recogniser is dependent on the size of the lexicon.
As the lexicon size increases the number of acceptable words that the character combinations can produce rises.
Furthermore a number of the alternative words suggested when using the large lexicon, would be unknown to many native speakers (e.g. betel, bey, littoral, lour, rentiers, ret).
By reducing the size of the lexicon the number of words that the author wrote that are missing from the lexicon increases.
There is obviously a trade off to be obtained by varying the number of words that are stored in the lexicon.
An additional problem that arises from using large lexicons is that, should the pattern recogniser fail to detect the correct word, then incorrect words are more likely to be suggested.
For the business text above there were 111 words in which the correct word was not recognised by the pattern recogniser.
With the 70,498 word lexicon alternative (incorrect) words were suggested for 89 of these cases.
With the 16,825 word lexicon the situation is slightly improved with incorrect alternatives being suggested for 78 of the unrecognised cases.
4.12.
Investigations with the Probabilistic Syntax Analyser
The syntax analyser developed in this thesis works in the following way:
1
The lexical lookup module determines the possible words that can be formed from the characters and the grammar tags for these words.
This information is received by the syntax analyser.
2
The possible grammar tag transitions in the lattice are determined and probability values for these transitions are obtained from the transition matrix.
3
A grammatical score is assigned to each grammar tag based on the probability of the transitions that it occurs in.
4
Each word is then given a syntactic rating based on the score assigned to the word's grammar tag(s).
At development time it is necessary to determine a probability for each of the transitions and store this in the transition matrix.
The aim is to obtain a model that is simple to apply, and has an acceptable error rate.
(It is not possible, regardless of how complex a model is used, to analyse all of the text perfectly).
As with the Jelinek model ($4.2), it is necessary to impose some restriction on the length of the n-gram transitions that are used.
The use of longer transitions requires a larger training corpus, increased storage for the transition matrices and is computationally more demanding.
However longer transitions provide stronger constraints on the tag combinations.
A number of investigations have been carried out to explore ways to best use the probabilistic analyser:
Investigation 1
Examines the ability of the LOB corpus to provide a model based on transitions of lengths 2, 3 and 4.
Investigation 2
Compares the use of three different sets of formulae to determine which provides the best model to employ for bigram and trigram models.
Investigation 3
The incorporation of GFF values to reduce the effect of rare syntactic forms of words and the recognition likelihood of the words (or lexical probability factor, LPF) into the model is investigated.
Investigation 4
A modification is made to the usual method of applying probabilistic information.
One method of applying probabilistic information is to assign a value to the final element in a transition based on its predecessors.
The modification is a summation of the probabilities assigned to a word in each window position in which the word exists.
Hence rather than simply investigate the ability of a tag to follow other tags, it also incorporates the ability of the tag to precede it's successors.
Investigation 5
The effect of different quality of input to the syntax processor is investigated.
It is rarely the case that the correct word is always amongst the alternatives suggested by the lower recognition levels.
This investigation examines the ability of the syntax processor to function with lattices of different quality.
Investigation 1:
Ability of the LOB corpus to cover transitions
The first consideration is whether the LOB corpus is able to provide enough data for a particular transition length.
The sufficiency of the million word LOB corpus for providing different orders of model for the combined tagset is shown in table 4.8.
This contrasts the number of possible transitions of each order with the actual number found in the million word LOB corpus.
An added indication is the percentage of the transitions that are found in the corpus only once.
This table can only give a general feel of the capacity of the corpus to cover the transitions likely to arise in unseen text.
There are many transitions that will never occur, this is especially the case with longer transitions.
However, so few of the possible quadgrams occur in the LOB (and of those found, 70% only occur once) that it would appear that the LOB corpus is not a large enough sample of text to provide a reliable quadgram model.
For the trigram case it is still only a small percentage of the possible transitions that do occur, and 40% of those transitions only occur once.
For the bigram case more than half of the possible transitions are found with only 18% occurring just once.
A further indication of the ability of the corpus to provide a reliable model is gained from observing how the number of transitions found varies with increasing corpus size.
The variation of number of different transitions found as a function of corpus size is shown for bigrams (fig 4.6) and trigrams (fig 4.7).
From examining the shapes of these graphs it would appear that increasing the size of the corpus would have only a small effect on the percentage of bigram transitions found.
With the trigrams however the graph appears to still be ascending quite rapidly, suggesting that a larger corpus will discover many more trigram transitions.
Although incomplete coverage for the trigram transitions can lead to problems when dealing with unseen text there are possible solutions which will allow the use of trigram information.
The problem with using incomplete data is that should a transition occur that was absent from the training corpus then the model would assert that such a transition was impossible, assign zero probability and produce errors.
There are methods available for dealing with this problem.
The first solution is to increase the size of the corpus (assuming that a larger corpus were available).
This should obviously lead to an improvement in the situation but regardless of the size of the corpus there will always be some transitions that are not found.
Furthermore as the matrix becomes less sparse the problems associated with storage increase.
Also if another corpus is available, if the tagset is different then it will be necessary to repeat all of the tagging detailed in the previous section.
Hence although a larger corpus will alleviate the problem of insufficient data it will not cure it and can lead to other problems.
The second solution is to use 'backing off' formulae.
This is the approach that has been taken.
The essential concept is that if a trigram does not exist then an approximate (non-zero) value is calculated based on the lower order transitions.
A number of backing off formulae are possible which aim to provide the best approximations to the probability of the transitions.
The aim for the current project is to employ the simplest method that provides an acceptable error rate.
Many of the backing off formulae that are available have been derived for semantic applications where the matrices are extremely sparse.
Sparsity and the necessity to back off is less of a problem for a model based on part of speech transitions.
Hence the more complicated backing off formulae are not necessary.
The CLAWS system also adopts 'backing off' formulae but seen from a different perspective -- a bigram model is used except for special-case tag-triples which empirical results showed would be wrongly tagged.
For these a trigram probability is used instead.
Atwell (1983) calls this an 'augmented first order Markov model' rather than an 'incomplete second order Markov model'.
Investigation 2
Recognition performance with bigrams and trigrams
Experiments were carried out to determine the best formula to use for scoring the possible transitions in a text of 372 words derived from the Malhotra corpus (Malhotra, 1975), a corpus traditionally considered to be syntactically difficult to process.
Recognition confusions for the text in the corpus were generated and the lattices produced containing 1246 words.
Two sets of experiments were performed on this input.
The first set employed bigram transitions, the second trigrams.
The best formula was determined by checking how many times the correct word was selected by the analyser using three different formulae (Table 4.9).
The weight, wt, given in Formula 1.1 is simply a count of the frequency of a tag sequence in the corpus, T1-T2; 1.2 is a modification of Bayes formula and 1.3 is suggested by Marshall (1983).
(For computational reasons scaling factors were used for the Marshall and Bayes formulae to avoid underflow).
From these results it can be seen that the Marshall formula produces the best formula to use for bigrams and the least successful is the simple frequency count of tag transitions.
The effect of the extra tag frequency value, N(T2), in the denominator of Equation 1.3 has the effect of weakening the tendency of high frequency individual tags.
A similar investigation was carried out using the trigram versions of these equations (Table 4.10).
From these values it can be seen that, as with the bigram results, the Marshall equation again produces the best results.
Furthermore, increasing the transition length to deal with trigrams gives better results than the equivalent bigram equation.
Although the increase in performance is relatively small, the difficulty of the recognition task warrants the extra cost -- an increase in matrix storage and a slight increase in processing.
Investigation 3
Lexical Probability and Syntactic Ambiguity
The next step in the processing is to investigate the possibility of reducing the influence of low probability words in the lattice and syntactic ambiguity.
Two factors are introduced -- a lexical probability factor (LPF) to reflect the recogniser's confidence in the candidate words and a grammatical frequency factor (GFF) to demote rare usage of words.
The Lexical Probability Factor, LPF
The character recogniser for the handwriting analyser (Wright, 1989) produces a ranked list of (up to 6) character candidates for each possible node in the character lattice.
The character that best matches the character database is assigned rank 1, the next best rank 2 etc..
This is best explained by referring again to the character lattice from chapter 1 (fig 4.8) where the ranks are given beneath the character candidates.
A confidence value for each candidate word is derived from the character ranks as follows.
The average rank of the characters that makes up the candidate word is determined.
Hence the average rank of its is 1 ((1+1+1) /3), us is 1.5 ((2+1) /2) and ox is 4.5 ((4+5) /2).
The best possible average is 1 (i.e. all of the characters were top ranked), the worst is 6 (i.e. all of the characters were ranked 6).
These values are then scaled to produce a recognition score for the word in the range 0...1, such that the best ranked word receives a score of 1, and an average rank of 6 would receive a zero confidence score (such occurrences are extremely rare).
This value is referred to as the lexical probability factor (LPF).
It is necessary to employ the LPF since in some cases low confidence words occur in high frequency transitions.
The LPF is used to scale the transitions according to the likelihood of the constituent words.
However, the grammatical score assigned to a word should be independent of it's own LPF (or it would merely serve to reinforce the pattern recogniser's decision) but it should be influenced by the LPFs of words in neighbouring positions.
Also when the assignments from the different modules are combined words having with low LPF will be demoted anyway.
For all but a few types the LOB corpus alone is too small a source from which to reliably derive information about how likely a word is to belong to a particular grammatical category (many words in the LOB occur just once).
Supplementing the information in the corpus with the information in the OALDCE provides a solution for the less frequent words.
That is, the OALDCE is taken as being an addition to the corpus, each word and each of its tags is considered to have occurred once in the corpus.
The GFF is derived from the equation:
That is, the GFF represents the frequency of a tag for a word relative to its most frequent tag.
An alternative equation is:
The disadvantage with this equation for the recognition task is that it would automatically demote any word that is syntactically ambiguous.
The aim of the GFF is simply to demote the rarer occurrences of words.
Given the small size of the LOB corpus and the reliance on the OALDCE to supplement this information the first equation is safer.
The Marshall formulae,(1.3) and (2.3), are modified by taking into account the relative grammatical frequencies of the words involved in each specific transition.
The aim of these is to bias the results in favour of the dominant grammatical categories.
The situation is subtly different for the GFF than the LPF.
The score being assigned by the syntax analyser is the syntactic belief that a word belongs in the input.
Hence if a word is unlikely to behave as a particular tag then it is valid for the syntax analyser to decrease its likelihood in every transition.
The effect of using both the LPF and GFF was also investigated using the equations:
Results
There is no single measure of how well a model has performed.
The number of words correctly selected is a strong indication, but there are some other results that should also be considered.
Due to the relatively small number of tags and the frequency with which some of these tags are found (e.g. over 20% of the words in the lexicon are usually found behaving as common nouns) there is often more than one word given the top rank in any position.
Referring back to Fig 4.2, the syntax analyser is unable to discriminate between the words tail and trio (both having tag 11 in the same position) unless the GFF of these words is different.
Hence, although the syntax analyser may be correctly detecting a high proportion of the words, the actual number of words that are top ranked by the analyser is also an important consideration.
With each equation the average rank given to correct words is better than that given to incorrect words.
The major improvement to be found by including the GFF is the reduction in the number of words selected as the best.
Inclusion of the GFF and LPF into the bigram experiments appears to be less stable than for the trigram experiments -- the number of top ranked correct words tends to drop for the bigram cases but remains almost constant for the trigram cases.
The most beneficial case appears to be when the trigrams are used in combination with both factors.
The major advantage gained by inclusion of these factors is the demotion of the incorrect words, rather than the promotion of correct words.
Of course the LPF information is proprietary to the pattern recognition system used.
For a poor recognition system, or simply one in which the text is unsuited to the recognition algorithm employed or character database used.
In this experiment, with this quality of text and recogniser, improvements were found from incorporating this information.
Investigation 4
A Modified Recognition Algorithm
One way in which probabilistic models are employed is to progressively classify the input based on predecessors and move along, using this information as the predecessor for the next element in the input.
For example if one is evaluating a trigram x-y-z, x and y are decided upon, it is the nature of z which is to be classified.
That is, one is classifying z, given x and y.
With something as uncertain as output from a text recogniser, it is also valid to consider that the trigram value stored for x-y-z, also contains information about the likelihood of x given y and z as successors, and the likelihood of y to have x as a predecessor and z as a successor.
The modification to the algorithm assigns values to each element within the transition window, not just to the final element of the transition.
Hence for a trigram window, the final score assigned to each word will be the summation of three recognition scores: the score assigned when the word was in the first position of the window, the score assigned when the word was in the middle of the window and the score when it was the final element of the window.
The effect of this method was tested using a handwritten text and an OCR recognised printed text.
The handwritten text is a 521 word business text (appendix 5) recognised by the ESPRIT-295 handwriting recogniser (Wright, 1989).
The OCR text (appendix 6) is an 804 word text supplied by Hewlett Packard using a developmental OCR recogniser.
In each text the correct word is always present as one of the alternatives.
The transition matrices used were trigram matrices, based on the Marshall formula (equation 3.3) and backing off to bigram information (equation 1.3) when no trigram information was available.
The GFF factor was used in both cases.
The number of correct words assigned each rank by the syntax analyser is shown in table 4.13.
The cumulative number of words assigned the correct ranks for these texts are shown in diagrammatic form in fig 4.9a and 4.9b.
From these results it is clear that the use of the 'full window' produces better results than simply assigning a score to the final element of the transition.
The extra constraints that are placed on the input by considering the tag in the context of each of the three positions in the trigram window (i.e. start, middle and end of the trigram) improve the selection of the correct word from the lattice.
Investigation 5
Effect of Recognition Performance on the Syntax Analyser
Such is the difficulty of the text recognition task that often the word that a writer intended does not appear in the list of candidate words that the syntax analyser receives.
The detection and correction of errors is an extremely difficult task in such circumstances, beyond the purview of this thesis.
The aim of this investigation is to determine how well the syntax analyser behaves when the lattice is successively degraded.
In order to isolate the recognition performance from the other variables affecting the text recognition the same written text was taken and 'corrupted' to produce lattices of different quality.
The quality, Q, of each lattice is denoted by the percentage of intended words that are found in that lattice.
Hence the best lattice is referred to as Q100 (the correct word was present among each of the candidates) and the worst lattice used is Q50 (only half of the words that were written are found in the lattice).
Eight lattices of the same text have been created of varying quality.
The corruption of the lattice was achieved as follows.
Each word in the lattice has a recognition score.
The target word that the user intended to write at each position is also known and stored in a file.
From knowing the actual word that should have been written at a particular position in the input, a simple NAWK program was written which loaded the target word for each position into an associative array, then went through the Q100 lattice and printed out the recognition score assigned to each target word.
From this a distribution of the word scores assigned to the target words was derived.
A series of threshold recognition scores were then determined from this distribution which would provide the desired lattice qualities.
(i.e. From looking at the distribution it is possible to determine that setting a threshold score of 88 (say) would remove 30% of the target words from the lattice giving a lattice quality of Q70)
A second NAWK program was written that was run a number of times with different threshold values.
This program also loaded in target word for each position into an associative array to allow it to determine which word in a position is the target.
If the target word in the position had a recognition score lower than the threshold then it was omitted.
Words other than the target word in the position were left unaltered, regardless of their recognition score to retain the same degree of branching for the different lattices.
If the target word was the only word selected at a particular position and was below the threshold score then it was replaced by '???'-- the recogniser's indication that it is unable to recognise any word.
The handwritten 520 word business text was used as input.
A series of degradations were performed on the lattice produced such that the worst quality lattice, Q50, retained just 260 of the correct words as alternatives.
Apart from the successive deletion of correct words from the input, the lattices are identical.
A simple method for comparing the performance of the analyser, given the different lattices, is to compare the ranks given to the (correct) words in the worst quality lattice.
That is, the rank assigned to the 260 target words that are common to each lattice is used as an indication of the performance.
Table 4.14 shows the ranks of the 260 target words that are common to each of the lattices.
The results are shown graphically in Fig 4.10.
As one would expect the best results are obtained when all of the correct words were present in the lattice, with quite a large decrease in performance being found when the lattice quality drops to 89%.
The performance of the analyser remains consistent until the quality is reduced to 50% when there is a drop in the number of top ranked correct words.
Discussion
The consistency of the behaviour of the syntax analyser given seriously degraded input is reassuring.
The quality of a handwriting recogniser can be extremely variable.
If a user is stored in the character database then very high performance can be achieved.
For a novel writer who may possess individual writing characteristics, unknown to the handwriting recogniser, performance can be low.
From the previous experiment, given a text in which half of the correct words are absent, it is still possible for the syntax analyser to select the correct word (if present) in more than 82% of the cases.
There are three main factors that may be responsible for this behaviour.
The first is the localised effect of the syntactic influences investigated (the trigram window only considers three words at a time) enabling the system to quickly recover from any incorrect decisions.
A second reason is the limited number of grammatical categories that are used and the frequency with which very common categories are used.
Hence, even if the correct word was not in the lattice, it is very possible that an incorrect word belonging to the same syntactic category may have been present instead.
Finally there is the behaviour of the pattern recogniser which tends to cluster its incorrect decisions.
If one word is incorrect then it is more likely that the surrounding words are also incorrect, possibly due to incorrect word segmentation.
Whatever the cause of the only slow degradation of performance the robustness of the analyser is impressive.
4.13.
Implementation Details
The syntactic recogniser has been designed to be used as part of a text recognition system.
For a practical application it is necessary to take into account the amount of memory required to run the system and the rate at which processing is carried out.
As part of a larger system the interface between the other modules is also of importance.
The processing for the probabilistic syntactic analyser has been carried out on the same hardware that was used for the rule-based investigations (i.e. a SUN Sparc 4/75 with 48 MBytes of memory).
The system has also been ported to an MS-DOS environment, an operation which would have been impossible for the rule-based system.
4.13.1.
Transition Matrix Storage
When dealing with levels of transitions higher than bigrams storage problems can arise due to the size of the transition matrix.
This is especially true within an MS-DOS environment.
The number of possible elements in a matrix is given by:
Not all of these transitions exist in the corpus.
The actual number of transitions found in the matrices derived from the LOB corpus was shown in table 4.8.
This showed that the bigram matrix is 53% full, whereas less than 5% of the trigram transitions are found in the LOB corpus.
Hence the bigram model is sufficiently non-sparse for full matrix storage to be sensible but the trigram matrix requires some means of sparse storage.
The method that has been used to store the bigram and trigram matrices relies on the fact that if no value for a trigram transition is available then the system backs off to the bigram value.
Fig 4.11 outlines the way in which the bigram and matrices are used in conjunction.
The method used to store the two matrices is as follows.
The bigram matrix is stored as a 1-dimensional array of length (109 * 109), reflecting the number of tags in the tagset.
(This arrangement is used to simplify access to the trigram array as will be explained.)
It is possible to calculate which element corresponds to bigram transition i, j using the formula 
An element of the bigram matrix consists of two fields.
The first is the score for the bigram transition [i, j]; the second is the index in the trigram matrix where the first trigram transition beginning with [i, j]is to be found.
The elements of the trigram matrix also consists of two fields.
The first field is the third index of the trigram (the k if the i, j, k transition).
The second is the value for the trigram, i, j, k.
In order to access the value for a trigram i, j, k the process is as follows.
The array index of the bigram, i, j, transition is determined.
This gives a value for the bigram transition i, j and the index in the trigram array of the first trigram beginning with that bigram.
The next element in the bigram array is also accessed to find the next start index in the trigram array.
This constitutes the stop index for the i, j trigrams.
(This is the reason for defining the bigram matrix as a 1-dimensional array.
It is easier to calculate the stop index without having to worry whether the end of the row in the bigram matrix has been reached.)
The trigram matrix is then accessed at the start index and processed until either:
1
The stop index is reached
or 2
The 3rd index is greater than the value of k.
or 3
The 3rd index is the same as the value of k
If the first or second conditions are satisfied then there is no value stored for the trigram i, j, k.
If the third condition is satisfied then the required value for the trigram is stored in the i--j--k--value at that index.
There is a significant reduction in memory requirements gained from the use of this technique.
Full matrix storage of the trigram matrix requires 2,600 KBytes.
By employing this technique, both the bigram and trigram matrices require only 200 KBytes.
4.13.2.
Interface
The syntactic analyser is only one module of a text recognition system.
Information generated by other modules (specifically the lexical lookup and compound processing modules) is required for the syntactic analyser to function.
Likewise the results generated by the syntactic processor must also be returned to the control process.
An interface structure has been produced which makes this possible.
As the character candidates are received from the pattern recognition process the combinations of characters are checked for validity in the lexicon.
When the possible words in a particular position have been determined, the information about these words is transferred from the lexicon to the interface structure.
The following information about a position is stored in the interface structure:
Both the syntax processor and the semantic processor (Rose, 1991) use a window-based approach.
That is, syntactic and semantic influences are only considered between words that are within a local neighbourhood of each other .
For the current system this neighbourhood is at most five words long.
Hence the syntactic and semantic relations between words further than this distance away in a piece of text are not considered.
The result of this approach is that it is not necessary to retain all of the information stored in the interface for all words in the text.
All that is required is that the information for five positions is retained at any one time.
Hence an array of five words positions is used as the communication structure (cf. fig 4.12).
A count is kept of the element in this array that corresponds to the latest word position.
The array is used as a circular list.
As a new word position is encountered it replaces the oldest information stored 
4.13.3.
Syntactic processing
When a new position is received by the syntax processor the relevant information from the interface structure is copied into a syntax structure.
This structure is similar to the interface structure, information being retained for the five most recently recognised words.
The structure is shown  diagrammatically in fig. 4.13.
As with the communication structure there is header information and a list of information about the candidates in the position.
Unlike the communication structure, the syntax structure is based around the possible grammatical tags rather than the actual words.
Hence the header contains a count of the number of different grammar tags that are possible in that position.
The advantage of copying the relevant information from the communication structure into a separate syntax structure is that it simplifies dealing with duplicate grammar tags in one position.
That is, if the same grammatical tag is found more than once in a position it is necessary only to know that the tag occurs in that position and the best scores associated with that tag.
Rather than continually allocate and reallocate memory as different words are processed, an array is used rather than a dynamic list for storing the information.
The advantage of this is that processing is simpler and more rapid.
The disadvantage are that it is necessary to set a maximum value for the number of different tags are allowed in any one position.
For each grammar tag used in each position the following information is stored:
The grammar tag
The highest LPF (recognition score) of the words having the tag in that position.
The GFF for the best recognised word having that tag in that position.
A window grammatical score.
A cumulative score.
Two grammar scores are kept in the list since the 'full window' method is used (as described in investigation 4 above).
The window-grammar score is the best score that is found for the grammar tag within the current window.
The cumulative grammar score is the sum of these window grammar scores as the window moves across the lattice.
4.14.
Discussion
The probabilistic analyser has been developed and tested out.
Favourable results have been obtained with this analyser.
Not only is it able to process any text, whether grammatical or ungrammatical (assuming such a distinction exists) but it is also able to do this processing quickly and efficiently.
Using a SUN Sparc 4/75 computer with 48 MBytes of memory, the syntax analyser requires just 0.4 seconds of CPU time (3.6 seconds real time) to process 521 written words, for which there were on average 10 alternatives for each word.
On the same system, an 808 word text (with 6.5 alternatives for each word) requires just 0.3 seconds of CPU.
This compares with up to 8 minutes CPU time for processing a single sentence using ANLT (see Table 3.3).
The majority of the work associated with the production of a probabilistic model lies in the development stages, with the production of a lexicon and the calculation of transition probabilities for the bigram and trigram transitions.
Unlike many generative systems which are developed from scratch, probabilistic systems tend to utilise existing resources.
The most important resource for probabilistic language processing is a good corpus.
Although corpora are costly to develop, once in existence they provide a valuable resource as evidenced by the number of different applications that utilise corpora.
These applications including corpus tagging, speech recognition, statistical parsing, studies of language development, machine translation, dictionary production and testing of language systems.
One of the most common reasons why NLP systems fail is due to the use of words that are not in the lexicon.
The absences may be either overt or covert (Zernik, 1989).
Overt unknowns are completely absent from the lexicon, for example a word such as Maastricht, despite the frequency of recent use, exists in few lexicons.
Covert unknowns arise from the use of a word in an idiomatic expression causing the word to behave in a different manner from its normal use.
The lexicon that has been developed is based on the vocabulary of a learners dictionary (the OALDCE) and supplemented by words from the LOB corpus.
The aim is to provide adequate, rather than exhaustive coverage of the language.
Complete lexical coverage is a goal that is impossible to achieve.
Nor is complete lexical coverage a goal that one would necessarily want to achieve since such a complete lexicon can cause problems for a recognition system.
As the lexicon storage increases, more processing is required at the lexical processing stage to validate character combinations.
There is a resultant increase in the number of word candidates in the word lattice.
With more words to decide between there is an increase in processing time for the higher level recognition stages.
Detection of recognition errors using lexical processing is made more difficult since there is a greater likelihood that an incorrect character combination will form an acceptable string for such a lexicon.
Furthermore an exhaustive lexicon of words may include words outside a normal user's vocabulary.
The proposal of such words to a writer can lead to disbelief about a word's existence and a lack of confidence in the system.
The development of the compound tree offers numerous possibilities for a text recognition system.
Firstly, the existence of compounds in text can be extremely difficult for higher level processors to deal with adequately for a number of reasons.
The lexical information for the constituent words is probably incorrect in the environment of the compound, leading to errors.
Even if one does know that a compound exists there is no guarantee that the lexical information will be available.
This is certainly the case if one is exploiting sublanguage processing.
Only through manual addition of the syntactic and semantic information can a suitable lexicon be achieved.
The 'opt out' method taken by the current system is not ideal but it is preferable for a system to admit it's ignorance than to submit an incorrect suggestion.
There are numerous improvements that can be made to the system.
A larger tagged corpus would allow the creation of a more accurate syntactic model.
The tagset could also be optimised.
The selection of the 109 tagset is rather arbitrary, being chosen simply because it offered the most distinctive tagset that could be formed from combining the Text710 tagset with the LOB tagset.
Investigations into the selection of an optimal tagset would be advantageous.
Such investigations may simply involve iterative selective alterations to the tagset used with monitoring of any improvement or degradation in performance.
Since the lexicon and corpus change each time the tagset is altered this is not a trivial task.
An alternative approach is to let the words decide their own tags.
This approach has been suggested by Atwell (1987) and taken by Finch and Carter (1992) and Brill et al(1990) among others.
The Brill method investigates transitions between words in a corpus.
The less frequent words are replaced by their grammar tag.
The more frequent words are then clustered together based on a similarity metric.
The tagsets produced (i.e. the groups of words) thereby contain not just syntactic but also semantic information.
Very large corpora and extensive processing are necessary to provide suitable information for a large lexicon using this method.
For sublanguage processing this may prove to be extremely useful if a large (raw) corpus is available, since the word classification will be specifically for that sublanguage type.
There will also be few omissions from the lexicon.
4.15.
Conclusions
A probabilistic syntax processor has been developed to assist in the selection of the correct words for a text recognition system.
A lexicon has been created which contains information for approximately seventy thousand words and transition matrices have been created from the LOB corpus.
A method of dealing with compound lexical forms has been developed which combines with the lexicon and allows the recognition of idiomatic phrases within the input.
The use of such a level avoids the syntactic processor from making incorrect syntactic judgements when dealing with idioms and provides an additional recognition stage.
The output produced by the syntax analyser is extremely basic.
No parse structure is produced, instead each candidate word is assigned a score indicating the goodness of fit with its possible neighbours.
There are a number of advantages of this system.
Firstly the system is robust.
No distinction is made between what is grammatical and ungrammatical.
The system simply gives preference to common grammatical combinations over unusual ones.
Secondly because the system is so basic in operation the processing occurs at a rate faster than a human writes.
Hence the system may be used for both dynamic and static recognition.
Because the system makes local optimal decisions rather than on a sentential basis, an incorrect decision early on in the sentence is unlikely to completely ruin the decisions later in the sentence.
With a rule-based approach an incorrect decision can have a devastating effect on the remainder of the analysis.
Chapter 5
Discussion
5.1.
Introduction
The problem of automatic text recognition requires information beyond the pattern recognition level.
Humans, the experts at text recognition, place relatively little importance on pattern recognition information once they become fluent readers.
(Indeed, a handwriting recognition system has been developed which is better at distinguishing between a hand-printed v and u than humans (Suen & Shillman, 1977)) The types of knowledge humans use include lexical, syntactic, semantic and pragmatic information although none of these are used in isolation.
The use of syntactic information for assisting in the selection of the correct word candidates in an automatic text recognition system has been investigated.
Although it can be shown that humans use syntactic information when processing language, the actual process involved has not yet been determined.
Syntax has been one of the most widely investigated areas of linguistics and human processing of language has been intensively studied.
Despite extensive research, how humans achieve their linguistic competence has still to be fully explained.
The problem has been approached from a number of different fields of research, and a large number of different theories have been proposed.
In computational linguistics the main approaches for implementing syntax can be broadly classified as either rule-based or probabilistic.
The use of both types of syntactic analysers has been attempted for helping to select the correct candidate words from the candidates suggested by text recognition systems in this thesis.
5.2.
Generative Systems
The interest in generative grammars over the last thirty years has resulted in a variety of theories for generative processing of language.
The aim of the majority of these systems has been language understanding.
For many of these theories computational systems have been developed using very small lexicons that have been sanitised of ambiguity and dealing with a conveniently simple set of grammatical constructions.
The resultant systems have been able to perform detailed analyses of small fragments of language.
Attempts to scale up these systems beyond their intended domain have met with little success.
One of the main barriers between the development of general purpose systems has been that few existing systems can be re-used.
Even systems sharing the same theoretical background tend to use different methods of representing the lexical information and ruleset.
Given that many of these theories require extremely detailed specifications of grammar rules and lexical entries this has for a long time formed an obstacle to the production of general systems.
In recent years the generative linguistic community has recognised this failing and made efforts to produce more general systems that can form the basis of further research.
A number of projects have been financed to produce re-usable tools.
The ANLT system, which has formed the basis of the rule-based work in this thesis, is one such product.
For many years the production of a lexicon to accompany a parsing system was viewed as a tedious task and most systems provided only illustrative lexicons, containing at most a few hundred words.
This situation has undergone a dramatic change in recent years.
Current linguistic theory places great emphasis on the importance of the lexicon and projects are currently underway to manually (supplemented by corpus processing) produce large lexicons (e.g. GENELEX (Normier & Nossin, 1990) and MULTILEX (McNaught, 1990)).
Perhaps the most significant lexicon under development is that of the EDR project in Japan which is developing bilingual resources for Japanese and English (Uchida, 1990).
The aim of this project is to obtain 200,000 words, term banks for 100,000 words and 400,000 concepts defined in terms of semantic networks.
The amount of work required to obtain this information may be inferred from knowing that this project will run for 9 years and cost 100 million US dollars.
5.3.
Corpus-Based Systems
For a long time the use of probabilistic information in linguistic systems has been frowned upon by the linguistic community.
It's use for tasks such as stylistic analysis was recognised but considered to be of only minor interest since it did not consider the 'important' questions, such as how humans are able to process language.
Recently there has been a large resurgence of interest in NLP systems based upon corpora.
A principal reason for this interest is the failure of rule-based linguistics to produce a general purpose, intelligent language processor capable of adequately performing a number of rather basic linguistic tasks (e.g. taking down unrestricted text, non-robotic reading aloud, making a precis of a text)(Sharman, 1990).
In contrast simple probabilistic models have been extremely effective in some speech and language tasks.
The increased importance of computers in daily life has led to an improvement in corpora.
There now exist huge volumes of text available for building corpora.
Once a million word corpus was considered to be a very large corpus, requiring a large amount of work to input the text to the computer followed by tagging and (possibly) annotation.
Current corpora under development aim to include a hundred million words.
Furthermore the availability of corpus tools means that the text in these corpora will also be tagged.
The approach employed by corpus linguists is well expressed by Levinson & Liberman (1981):
'The best design strategy is not to program a computer directly with the wealth of descriptive detail that constitutes a natural language but rather to give it the basic set of expectations and abilities that are needed to learn a language.'
This self-organising approach to language has been taken by Sharman (1992) who extracted a probabilistic grammar from a set of 2000 hand-parsed sentences.
The resultant grammar contained 3527 basic rules which were converted and extended to a set of 200,000 rules.
Each rule in the grammar was assigned a probability.
The probability of each parse was then calculated and the most probable parse selected.
The grammar was iteratively trained to improve the estimates of the rule probabilities.
Sentences from the associated press (AP) corpus and the Hansard corpus were parsed.
A parse was obtained for each of the sentences.
The AP data was correctly parsed for 73% of cases, the Hansard for 60% of cases.
These results are very encouraging when compared to the performance of more traditional, hand-built, grammars created without the benefit of a corpus to learn from.
5.4.
The Grammatical/ungrammatical Distinction
Central to the debate between generative and probabilistic approaches is the notion of grammaticality.
Generative linguists tend towards the view that language is a well-defined class of sentences.
Any string that is not a member of this class is ungrammatical.
However not all generative linguists take such a precise view of grammar, acknowledging the ability of humans to interpret a wide range of ungrammatical sentences.
Hindle (1983) distinguishes between unusual constructions and true ungrammaticalities.
For example the sentence 'That's the only thing he does is fight.'
is marginally grammatical, because it is often used although a precise grammar would reject it.
Truly ungrammatical sentences are regarded as not resulting from any regular grammatical process (e.g. 'I've seen it happen is two girls fight.').
The occurrence of such constructions is far less frequent.
Labov (1966) reported that less than 2% of everyday speech were ungrammatical in this sense.
Some generative systems have been designed to deal with deviant text.
One approach taken involves the relaxation of constraints in the parsing process.
This technique was used in the Linguistic String Project (Sager, 1981) where a failed parse was re-tried without the agreement constraints on syntactic features.
There is also the option of extending the grammar.
Marsh and Sager (1982) took such an approach with their analysis of a set of medical records written in compact text (i.e. terse form) again using the NYU Linguistic String Project parser.
Within their data they found recurrent ungrammatical constructions whose forms they characterised and included in the parsing grammar.
The justification used for this approach is that 'repetitive ungrammaticality is grammatical for the text set'.
Another approach ignores syntactic processing and instead concentrates on just using a semantic grammar (e.g. the PLANES system (Waltz, 1978)).
This approach allows very ungrammatical sentences to be processed but loses the advantages that can be obtained by applying syntactic restrictions.
Structural linguists question the existence of a clear-cut distinction between what is grammatical and what is ungrammatical.
Sampson (1987c) contends that there is a continuous gradient from very common to very rare constructions.
In support of this view Sampson provides an analysis of over eight thousand parsed noun phrases from the LOB treebank.
By his analysis almost two thirds of these noun phrase types are represented only once.
The conclusion Sampson draws from this is that it is extremely difficult to determine the boundary between what is grammatical and ungrammatical if such a high proportion of grammatical expansions are very rare.
Taylor et al(1989) dispute the claims made by Sampson.
Instead they believe that it is Sampson's generative grammar formulation that is at fault.
Their analysis provides a much more favourable view of generative grammars using the ANLT grammar system.
Sharman (1989) adopts the same stance as Sampson on this issue, believing that the question of the acceptability of a sentence is not a black-or-white judgement but a gradation of likelihoods.
McCawley (1976) asserts that sentences which are universally judged as grammatical are simply those for which no one has any difficulty in thinking of uses.
This point is illustrated with the 'sentence''Kissinger conjectures poached' which linguists would class as ill-formed.
However in the context of a reply to the question 'Does anyone know how President Ford likes his eggs?
Kissinger conjectures poached' becomes acceptable.
5.5.
Over-generation & Syntactic Ambiguity
Sentences are much more ambiguous than one would normally expect.
Church & Patel's investigation (1982) of this showed that there may be hundreds or even thousands of different parse trees for some very natural English sentences.
For large grammars this is especially the case since the interaction of many rules can cause the system to explode in a combinatorial manner.
Unfortunately it is essential to use a large grammar to achieve a large coverage of the language.
This problem has been borne out by use of the ANLT system.
For example, the sentence 'I enter orders until the markets close' resulted in 4,848 different parses being suggested.
A similar problem was encountered by Taylor et al(1989) when analysing the noun phrases from the LOB corpus using the ANLT system.
Their attempt at automatic application of the ANLT parser to the noun phrases failed due to'inadequacies of grammatical coverage and because of resource limitations with long and multiply-ambiguous NPs' which resulted in 'very high numbers of automatically generated parses'.
There are some generative grammars which resolve this ambiguity by use of a metric that ranks alternative parses (e.g. PEG (Jensen 1986)), or by the inclusion of statistical information into the grammar rules (e.g. METAL (Thurmair1990)).
Despite these modifications, syntactic ambiguity remains a large problem for generative systems.
Much of the ambiguity can be avoided by providing a less deep parse.
It is not always necessary to create very detailed parse representations.
Practical applications exist for text processors of limited parsing capability.
One such application involves text-to-speech synthesis.
For this application syntactic processing is required to determine exactly where in the output to correctly specify prosody.
The task does not require detailed syntactic processing, simply to know where to pause (the major syntactic boundaries), which words to stress (distinguishing content and function words) and whether the sentence requires the pitch to fall or rise at the end (is it a yes-no question?).
For a text-to-speech system only one parse can be accepted and failed parses are unacceptable.
Investigation showed that failures and many of the ambiguities resulted from combining phrase groupings at the clause level.
To avoid this the MITalk text-to-speech system (Allen et al, 1987) used a simple phrase-level parser to specify prosody.
In contrast to more detailed parsing systems this required few resources and was able to run in real-time.
A major issue in the decision between a generative and a probabilistic system is the robustness of the system.
There have been generative systems designed to achieve high degrees of robustness (e.g. the FRUMP system (Dejong, 1979)).
For this system the most frequent cause of failure to process the input was cited as 'missing vocabulary'.
The lexicon system provided with the ANLT is sufficiently large however to be of comparable size to that used in the probabilistic system.
The other issue of robustness is how to deal with sentences that do not conform to the grammar.
Jensen's PEG system 'is able to make some kind of syntactic sense out of any decent English sentence'.
Marcus (1990b) reports that a group from IBM informally surveyed a number of parsers that claimed to be 'broad coverage'.
The parsers were tested on a set of short sentences containing less than 14 words.
Of the parsers used, the best was able to correctly identify the best parse for these sentences with only 60% accuracy.
There are a number of textual features that tend to be ignored by generative parsing systems.
The most obvious example of this is punctuation.
Generative linguists consider punctuation to be simply prosody and of relevance for the language -- or they may simply believe it to be uninteresting.
Very few generative systems process punctuation at all.
Exceptions to this tend to be the simpler systems such as those used in text to speech applications.
The MITalk system (Allen et al, 1987) is one such system.
This system uses three classes of punctuation to distinguish whether the punctuation is a comma (,), a punctuation mark that is internal to a sentence (i.e.:;() and') or if it can be a sentence-final punctuation mark (i.e..!?).
Corpus-based systems are easily able to deal with punctuation.
Punctuation marks are classified as separate syntactic categories and grammars and transition matrices based around this assumption.
There are applications for which a generative grammar would be better suited than a probabilistic one.
The most obvious is text generation.
Such a system does not need to have access to a full-coverage grammar and lexicon.
Analysis is much more computationally expensive than generation.
Also, in comparison to knowledge-based approaches a HMM provides little information about the recognition process.
As a result it is often difficult to analyse the errors of a probabilistic system in an effort to improve its performance.
Briscoe (1991) concedes the importance of statistically-based methods for a number of practical applications such as part-of-speech tagging and the derivation of surface collocations.
However he argues that for the development of lexicons there is a need for parsers capable of phrasal analysis, requiring lexicons with reliable information about subcategorisation.
Hence although corpus analysis will play an increasingly important role, it is Briscoe's belief that it will not supplant others or render more theoretical work irrelevant.
Within sublanguage applications a generative approach may be of more use than a probabilistic method.
Lenhert (1991) contains details of a performance evaluation of text-analysis methodologies.
It is difficult to evaluate different systems because they are traditionally idiosyncratic depending on their particular area of specialisation.
To counter this effect, a number of NLP systems were developed to extract information about terrorist incident from a test suite of 100 previously unseen news articles.
Systems were evaluated in terms of how much correct information was extracted, how much of the extracted information was correct and how much of it was superfluous.
They found that text analysis techniques incorporating natural language processing were superior to the traditional information-retrieval techniques based on statistic classifications when the applications require structured representations of the information in the texts.
However the amount of work required to achieve satisfactory performance in this evaluation highlights the problem for such NLP systems.
Each site was allowed six months for system development for the domain in addition to time needed for corpus development.
The corpus that the systems were provided with contained 1300 texts with an average length of 12 sentences.
The preparation of answer keys for 100 texts required between two weeks and a month.
Hence it can be seen that tailoring a generative system to a particular domain can elicit high performance but there is a substantial effort required to tailor a system to a domain.
5.6.
Combining Syntax and Semantics
A central issue for many linguists is how do syntax and semantic information combine in language systems.
 Psycholinguists have analysed how humans combine information and a number of models have resulted.
Marslen-Wilson (1975) proposes an unstructured, fully interactive model of comprehension.
In this model all of the different categories of information interact in an ongoing manner to constrain the processing of a sentence.
Forster (1979) proposed a totally autonomous system in which syntax and semantic information are distinct components of a system with no communication between them.
Rayner et al(1983) propose that semantic and pragmatic information play no part in the initial syntactic choice, however if the syntactic result is ambiguous then semantic and pragmatic influences are used.
From a computational perspective the interaction of syntax and semantics can be important.
This is especially so for rule-based syntactic processing where the use of semantic analysis to remove improbable partial parses could lead to a reduction in the processing requirement.
Rule-based methods offer some method of combining the two together -- the grammar is able to provide information to the semantic analysis.
For example, a transitive verb such as 'find' is a two-place function requiring a subject and an object.
Therefore given the sentence 'John found the dog', a parse may be used to determine the subject and object of the verb and thereby create a semantic representation of who found what.
If the semantic representation cannot be created from the parse structure then the sentence is rejected.
However, there is even less agreement among linguists about how to apply semantics than there is about the use of syntax.
Nor is there currently available a semantic analyser capable of understanding unconstrained English.
One of the most commonly used arguments against the use of probabilistic grammars is that they do not take semantic information into account.
Sharman counters this by pointing out that conventional syntactic grammars do not take semantic information into account either.
A number of probabilistic systems employ word-word associations which Sharman claims is the implicit application of syntactic, semantic and pragmatic knowledge.
The semantic processor in the current text recognition system makes no effort to produce a representation of the meaning of the input.
Rather, the aim of the semantic processor is to give preference to words that are semantically acceptable over those that are not.
For such a task crude approaches to semantic processing are preferable to the detailed analyses found in understanding systems since they are computationally tractable, robust and are not restricted to a small language domain.
The methods used are based on collocations derived from text corpora and information extracted from machine readable dictionaries (Rose & Evett, 1992).
The issue of how to combine syntax and semantic processing in this system has not been completely resolved.
The results from the simple syntactic processor are of little use to the semantic analysis, being simply a rating of the 'goodness' of fit of a word given its local syntactic environment.
Similarly, the semantic processor operates in a very similar manner to the syntactic processor and simply provides a semantic rating.
With such limited information there is little to be gained by interleaving the two processors since neither can provide much assistance to the other processor.
Instead the two modules operate in isolation with the results being combined by a control processor.
The current control processor is very crude, using simply a combination of the scores provided by each of the recognition modules (i.e. lexical, compound, syntax and semantic).
A control system is currently under development for combining the output from the different modules in an intelligent manner.
5.7.
Further Work
Statistical processing provides a simple, efficient method for picking the correct words based on local neighbourhood.
However, although words considered in a local context may seem perfectly acceptable in the sentential context these combinations may be unlikely.
As long ago as 1957 Chomsky pointed out the weaknesses of n-gram techniques to provide an English grammar based on their inability to cope with long distance dependencies.
There are a number of possible remedies to this problem.
The first remedy involves the use of larger corpora which will become increasingly available in the next few years.
Portions of the ACL/DCI are currently being distributed with more to follow in what will eventually be a 100 million word collection.
Similarly, the British National Corpus will also be 100 million words of British English and will also be freely available.
Use of such corpora will improve the accuracy of the probabilistic models, allowing transitions beyond the trigram level to be investigated.
Although this will not provide sentential context it will be less localised than the current system.
There is also the possibility of using a simulated annealing approach (Sampson, 1986).
Simulated annealing uses random perturbations to shake the parameter values out of a local optimum so that globally optimal values may be found.
The major drawback being that such methods are computationally expensive.
Another approach that may be useful is a hybrid method.
Sharman et al(1990) suggests the use of a probabilistic ID/LP grammar.
The suggested method is to associate probability scores with the ID and LP rules.
Hence, the fact that a noun phrase usually contains a noun would be indicated in the ruleset.
This is extended so that for any given phrase there is a probability distribution over the objects that can be the immediate constituents.
Similarly there is also a probability distribution over the ordering of items in a phrase.
The probabilities for these rules can be obtained from corpora.
Speed up in the parsing process can then be achieved by eliminating low probability parses.
5.8.
Conclusions
The use of syntactic information as an aid for text recognition systems has been shown to be both feasible and useful.
Both a rule-based system and a probabilistic approach have been explored.
The rule-based approach proved to be too unwieldy and fragile to be of practical use given the current state of computer hardware and linguistic theories.
The probabilistic approach proved to be much better suited to the task.
The main conclusion is that, for the text recognition task probabilistic methods of syntactic analysis are more useful than rule-based ones.
The ease with which a probabilistic analyser can be created contrasts sharply with the immense amount of work required to create a comprehensive generative grammar.
The resulting analysis is also an order of magnitude faster for processing and does not suffer from the brittleness of a generative analysis.
When successful the ANLT system produce very detailed structural information for the input.
However the ANLT system is very fragile, computationally demanding and unable to deal with much naturally occurring language.
For a computational product the processing requirements are an important factor in the choice of system.
For a handwriting recognition system it is obviously desirable that a recognised word be displayed soon after it is recognised.
Rule-based parsers tend to work on a sentential basis which appears to be too much delay for users.
(Just how soon after writing a user should sees the recognised text is open to debate.
If writing on an electronic tablet, which combines the input and display devices, the user may find it disconcerting to see the handwriting changing under their pen.)
A straight comparison between the generative and the probabilistic approach is unfair for a number of reasons.
Firstly the probabilistic analyser was specifically constructed for the task whereas the rule-based analyser was simply a modification of an existing system.
Rule-based analysers are designed for certain(typed) input and the production of the parse tree for the input.
The recognition task is substantially different.
There is a much larger search space to be explored than for certain input.
The resultant parse is of only minor interest, being of use only to the semantic component in reaching its decision.
A syntactic recogniser rather than a parser would almost as useful and less computationally demanding.
Once one valid parse had been obtained the parsing process could cease.
However, in order to reach this stage it is still necessary to perform a large amount of parsing.
With the recognition problem not only are there a number of possible word combinations, the majority of the possible combinations are to be rejected.
The constraints applied by the rule-based methods are too rigorous.
It is not to be expected that each sentence written will obey grammar rules.
The search techniques to reject sentences need to be highly efficient so that invalid combinations can be quickly rejected.
There are points in favour of the rule-based method.
The generative system is able to operate on a sentential level rather than simply consider local word combinations.
Also, when successful, the rule-based system produced a much deeper analysis of the input than the probabilistic approach.
It must also be taken into consideration that the generative system is being used for an application alien to its originally design.
Nor were any attempts made to supplement the grammar or lexicon.
However it is unlikely that a set of rules to explain how general language is processed will be available in the foreseeable future.
One reason for this is the interaction of the various linguistic processes involved in processing language.
Syntax does not occur in isolation, there is also semantic, pragmatic and encyclopaedic knowledge contributing to the analysis.
The specification of each of these processors as well as a control system to combine them intelligently is currently far beyond any expectations.
Given the currently available linguistic theories and computers the only feasible way to build a system capable of dealing with unconstrained language is through corpus-based methods.
In the short term this brute force approach appears to be the best method of incorporating linguistic knowledge into computers.
The success with probabilistic methods is similar to that being found in other areas of NLP.
Specifically in the area of machine translation the lack of success with linguistic-based second generation methods is noticeable, only one of the commercially available MT systems uses generative techniques.
The failure of the rule-based systems has led some researchers to talk about third generation machines and others to return to the more basic systems from the 1960's.
However, the choice of rule based or corpus based methods is still contentious.
A number of considerations need to be made, and it is possible that a combined method may be preferable.
The final selection of which approach to take for the practical application of apply linguistic information is dependent on a number of factors.
3
Welfare Economics
15
Introduction to Welfare Economics
'We want to see an economy in which firms, large and small, have an incentive to expand by winning extra business and creating jobs....
Only a government which really works to promote free enterprise can provide the right conditions for that dream to come true.'
Conservative Election Manifesto
'The Tories say that 'competition' ensures that shoppers get a fair deal.
The customers know better.
Stronger legal safeguards are essential to protect customers...
We must rebuild...under a Labour Government working together with unions and managers to plan Britain's industrial development.'
Labour Election Manifesto [Italics added by the authors]
In Chapter 1 we pointed out that markets are not the only device by which society can resolve the questions, what, how, and for whom to produce.
Communist economies, for example, rely much more heavily on central direction or command.
Are markets a good way to allocate scarce resources?
What does 'good' mean?
Is it fair that some people earn much more than others in a market economy?
These are not positive issues about how the economy works but normative issues about how well it works.
They are normative because the assessment will depend in part on the value judgements adopted by the assessor.
Welfare economics is the branch of economics dealing with normative issues.
Its purpose is not to describe how the economy works but to assess how well it works.
The election manifestos of the Conservative and Labour parties fundamentally disagree about how well a market economy works.
But how are we to cut through the political rhetoric to see what lies behind the disagreement?
Two themes recur throughout our discussion of welfare economics in Part 3.
The first is the question of allocative efficiency.
Is the economy getting the most out of its scarce resources or are they being squandered?
The second is the question of equity.
How fair is the distribution of goods and services between different members of society?
15-1 Equity and Efficiency
We begin by defining what we mean by these terms.
Equity
Economists use two different concepts of equity or fairness.
Horizontal equity is the identical treatment of identical people.
Vertical equity is the different treatment of different people in order to reduce the consequences of these innate differences.
Whether or not either concept of equity is desirable is a pure value judgement.
Horizontal equity would rule out racial or sexual discrimination between people whose economic characteristics and performance were literally identical.
Vertical equity is the Robin Hood principle of taking from the rich to give to the poor.
Most people would agree that horizontal equity is a good thing.
In contrast, although few people believe that the poor should starve, the extent to which incomes or resources should be redistributed from the 'haves' to the 'have-nots'to increase the degree of vertical equity is an issue on which different people take very different positions.
Efficient Resource Allocation
By a resource allocation for an economy we mean a list or complete description of who does what and who gets what.
To emphasize that markets are not the only possible allocation devices, we begin by assuming that allocations are chosen by a central dictator.
Feasible or producible allocations depend on the technology and resources available to the economy.
The ultimate worth of any allocation depends on consumer tastes, which determine how people value what they are given.
Figure 15-1 shows an economy with only two people, David and Susie.
The initial allocation at point A gives David a quantity of goods Q D and Susie a quantity of goods Q S Are society's resources being wasted?
Suppose by reorganizing production it was possible to produce at point B to the north-east of A. If David and Susie assess their own utility by the quantity of goods they themselves receive, and if they would each rather have more goods than less, B is a better allocation than A since both David and Susie get more.
It is inefficient to produce at A if production at B is possible.
Similarly, a move from A to C makes both David and Susie worse off.
If it is possible to be at A, it is definitely inefficient to be at C.  
Figure 15-1 ALLOCATING GOODS TO TWO PEOPLE.
Provided people assess their own utility by the quantity of goods that they themselves receive, B is a better allocation than A which in turn is a better allocation than C. But a comparison of A with points such as D, E or F, requires us to adopt a value judgement about the relative importance to us of David's and Susie's utility.
But what about a move from A to E or F?
One person gains but the other person loses.
Whether we judge such a change desirable depends on how we value David's utility relative to Susie's.
If we think David's utility is terribly important we might consider a move from A to F a good thing, even though Susie's utility will be severely reduced.
Value judgements about equity, the fairness of the distribution of goods between people, have got mixed up with our attempt to make statements about waste or inefficiency.
Since different people will make different value judgements, there can be no unambiguous answer to the question of whether a move from A to D, E, or F is desirable.
It all depends who is making the assessment.
In an effort to separate as far as possible the discussion of equity from the discussion of efficiency, modern welfare economics uses the idea of Pareto-efficiency named after the economist Vilfredo Pareto whose Manuel D'Economie Politique was published as long ago as 1909.
An allocation is Pareto-efficient for a given set of consumer tastes, resources, and technology, if it is impossible to move to another allocation which would make some people better off and nobody worse off.
In terms of Figure 15-1 a move from A to B or A to G is a Pareto gain.
In either case Susie is better off and David no worse off.
If B or G is a feasible allocation which could be produced then point A is Pareto-inefficient.
How about a move from A to D?
David is better off but Susie is worse off.
The Pareto criterion has nothing to say about such a change.
Without bringing in an explicit value judgement about the relative importance of David's and Susie's utility we cannot evaluate this change.
Thus the Pareto principle is of only limited use in comparing allocations on efficiency grounds.
It only allows us to evaluate moves to the north-east or the south-west in Figure 15-1, but it is the most we can say about efficiency without becoming entangled in value judgements about equity.
Figure l5-2 takes the argument one stage further.
Suppose by all possible reorganizations of production, the dictator decides that the economy can produce anywhere inside or on the frontier AB.
From any allocation inside the  frontier it is always possible to achieve a Pareto gain by moving to the north-east onto the frontier.
Hence any point inside the frontier must be Pareto-inefficient.
It is possible to make one person better off without making the other person worse off.
But all points on the frontier are Pareto-efficient.
Beginning at a point such as C, it is possible to give one person more only by giving the other person less.
Since no Pareto gain is possible, every point such as C lying on the frontier must be Pareto-efficient.
Figure 15-2 THE EFFICIENT FRONTIER.
The frontier AB shows the maximum quantity of goods which the economy can produce for one person given the quantity of goods being produced for the other person.
All points on the frontier are Pareto-efficient.
David can only be made better off by making Susie worse off, and vice versa.
The distribution of goods between David and Susie is much more equal at point C than at points A or B.
Thus the dictator should never choose an inefficient allocation inside the frontier.
Which of the Pareto-efficient points on the frontier is most desirable will depend on the dictator's value judgement about the relative importance of David's and Susie's utility.
It is purely a judgement about equity.
15-2 Perfect Competition and Pareto-Efficiency
Will a free market economy find a Pareto-efficient allocation on its own, or must it be guided there by government intervention?
If we can answer this question, we shall have gone some way to understanding the different claims made in the Conservative and Labour manifestos.
Under certain conditions, soon to be elaborated, we can prove the following striking result: If every market in the economy is a perfectly competitive free market, the resulting equilibrium throughout the economy will be Pareto-efficient.
Formalizing Adam Smith's remarkable insight of the Invisible Hand, this result is the foundation of modern welfare economics.
Competitive Equilibrium in Free Markets
Suppose there are many producers, many consumers, but only two goods, meals and films.
Each market is a free, unregulated market and is perfectly competitive.
Suppose the equilibrium price of meals is 5 and the equilibrium price of films is 10.
Finally, we suppose that labour is the variable factor of production and that workers like equally the non-monetary aspects of jobs in the films and meals industries.
We now argue through the following stages:
1
The last film produced must yield consumers 10 worth of extra utility.
If it yielded less (more) extra utility than its 10 purchase price, the last consumer would buy less (more) films.
Similarly, the last meal purchased must yield consumers 5 worth of extra utility.
Hence consumers could swap 2 meals (10 worth of utility) for 1 film (10 worth of utility) without changing their utility.
2
Since each firm sets price equal to marginal cost, the marginal cost of the last meal must be 5 and the marginal cost of the last film must be 10.
3
The variable factor (labour) must earn the same wage rate in both industries in competitive equilibrium.
Otherwise there would be an incentive for workers to transfer their labour to the industry offering higher wages.
4
The marginal cost of output in either industry is the wage rate divided by the marginal physical product of labour.
Higher wage rates increase marginal cost, but a higher marginal physical product of labour means that less extra workers are needed to make an additional unit of output, thus reducing marginal cost.
5
Since wage rates are equal in the two industries and the marginal cost of meals (5) is half the marginal cost of films (10), the marginal physical product of labour must be twice as high in the meals industry as in the film industry.
6
Hence reducing film output by 1 unit and transferring the labour thus freed to the meals industry would increase the output of meals by 2 units.
The marginal physical product of labour is twice as high in meals as in films.
Feasible resource allocation between the two industries thus allows society to exchange 2 meals for 1 film.
7
Stage (1) says that consumers can swap 2 meals for 1 film without changing their utility.
Stage (6) says that, by reallocating resources, producers swap an output of 2 meals for 1 film.
Hence there is no feasible reallocation of resources that can make society better off.
Since no Pareto gain is possible, the initial position -- competitive equilibrium in both markets -- is Pareto-efficient.
Notice the crucial role that prices play in this remarkable result.
Prices do two things.
First, they ensure that the initial position of competitive equilibrium is indeed an equilibrium.
By balancing the quantities supplied and demanded, prices ensure that the final quantity of goods being consumed can be produced.
They ensure that it is a feasible allocation.
But in competitive equilibrium prices are performing a second role.
Each consumer and each producer is a price-taker and knows that he or she cannot affect market prices.
In our example, each consumer knows that the equilibrium price of meals is 5 and the equilibrium price of films is 10.
Knowing nothing at all about the actions of other consumers and producers, each consumer will automatically ensure that the last film he or she purchases yields twice as much utility as the last meal purchased.
Otherwise that consumer could rearrange purchases out of a given income to make himself or herself better off.
Thus by their individual actions facing given prices, each and every consumer will automatically arrange that 1 film could be swapped for 2 meals with no change in utility.
Similarly, each and every producer, merely by setting their own marginal cost equal to the price of their output, will ensure that the marginal cost of films is twice the marginal cost of meals.
Thus it takes society twice as many resources to make an extra film rather than an extra meal.
By rearranging production, transferring labour between industries, society can swap 2 meals for 1 film, exactly the trade off that would leave consumer utility unaffected.
Thus, as if by an Invisible Hand, prices are guiding individual consumers and producers, each acting only in their own self-interest, to an allocation of the economy's resources that is Pareto-efficient.
Nobody can be made better off without someone else worse off.
Figure 15-3 makes the same point in a simple diagram.
DD is the market demand curve for one of the goods, say films.
At a price P 1 a total quantity of films Q l will be demanded.
But we know that the last film demanded must yield consumers exactly P l pounds worth of utility; otherwise they would buy more films or less films than the quantity Q 1 .
Hence we can think of DD as showing also the marginal utility of the last unit of films which consumers purchase.
When a total quantity of Q l films is purchased the last film yields exactly P 1 pounds worth of extra utility to consumers.
In a competitive industry, the supply curve for films SS is also the marginal cost of providing films.
Variable factors (labour in our simple example) are paid not only their marginal value product in the film industry, but also their marginal value product in the meals industry.
Why?
Because labour mobility between industries ensures that wage rates are equated in the two industries.
Hence the marginal cost of producing the last film must be the value of the meals sacrificed by using the last units of labour to make films rather than meals.
Prices ensure that both industries are in equilibrium.
Figure 15-3 shows that at the equilibrium point E the marginal utility of the last film equals its marginal cost.
But the marginal cost of the last film equals the value of meals sacrificed, the price of meals multiplied by the quantity of meals forgone by using labour to make that last film.
However, the meals  industry is also in equilibrium.
An equivalent diagram for the meals industry shows that the equilibrium price of meals is also the marginal utility of the last meal purchased.
Hence the value of meals sacrificed to make the last film is also the marginal utility of the last meal times the number of meals sacrificed.
Figure 15-3 COMPETITIVE EQUILIBRIUM AND PARETO-EFFICIENCY.
DD is the demand curve for films.
At any output such as Q 1 , the last film must yield consumers P 1 , pounds worth of extra utility; otherwise they would demand a different quantity from Q 1 .
The supply curve SS for the competitive film industry is also the marginal cost of films.
If the meals industry is in competitive equilibrium, the price of a meal is also the value of its marginal utility to consumers.
Thus the marginal cost of a film is not only the market value of extra meals that could have been produced, but is also the value of the marginal utility consumers would have derived from those meals.
Hence at any film output below Q *; the marginal utility of films exceeds the marginal utility of meals sacrificed to produce an extra film.
Above Q *; the marginal utility of films is less than the marginal utility of meals sacrificed.
The equilibrium point E for films and the corresponding equilibrium point in the market for meals thus ensure that resources are efficiently allocated between the two industries.
No reallocation of resources could make all consumers better off.
Thus, provided the meals industry is in competitive equilibrium, the marginal cost curve for the film industry is simply the extra pounds worth of utility sacrificed by using scarce resources to make another film instead of extra meals.
It is the opportunity cost in utility terms of the resources being used in the film industry.
And equilibrium in the film industry, by equating the marginal utility of films to the marginal utility of the meals sacrificed to make the last film, guarantees that society's resources are allocated efficiently.
At any output of films below the equilibrium quantity Q *;, the marginal consumer benefit of another film would exceed the marginal consumer valuation of the meals that would have to be sacrificed to produce that extra film.
At any output of films in excess of Q *;, society would be devoting too many resources to the film industry.
The marginal value of the last film would be less than the marginal value of the meals that could have been produced by transferring resources to the meals industry.
Competitive equilibrium ensures that there is no resource transfer between industries that would make all consumers better off.
Equity and Efficiency
In the previous section, we saw that there are an infinity of Pareto-efficient allocations, each with a different distribution of income or utility between different members of society.
We have now discovered that a competitive equilibrium in all markets would generate one particular Pareto-efficient allocation.
What determines which of the possible Pareto-efficient allocations it picks out?
People have different innate abilities.
At any instant, people also have different amounts of human capital and financial wealth.
In Chapter 13 we saw that these differences allow people to earn very different income levels in a market economy.
They also affect the pattern of consumer demand.
Brazil, which has a very unequal distribution of income and wealth, has a high demand for luxuries such as servants.
In Sweden, where there is a much more equal distribution of income and wealth, almost nobody can afford servants.
Thus, the initial distribution of abilities, human capital and wealth, by affecting income-earning potential, determines the pattern of consumer demand in the economy.
Different patterns of demand imply different demand curves for individual goods and services and determine different equilibrium prices and quantities.
In principle, by varying the distribution of initial income-earning potential, we could induce the economy to pick out each and every one of the possible Pareto-efficient allocations at its competitive equilibrium.
Now we come to a very attractive idea.
The government is elected by the people to express the value judgements of the majority.
Ideally, we should like the government to ensure that the economy is on the Pareto-efficient frontier, shown in Figure 15-2, but take responsibility for making the value judgement about which point on this frontier the economy should attain.
Since every competitive equilibrium is Pareto-efficient, and since different Pareto-efficient allocations correspond to different initial distributions of income-earning potential in a competitive economy, it appears that the government can confine itself to redistributing income and wealth through income and inheritance taxes or welfare benefits without having to worry about interventions to ensure that resources are allocated efficiently.
Free competitive markets will automatically take care of allocative efficiency.
This seems like a pretty powerful case for the free enterprise ideal espoused in the Conservative Election Manifesto that we quoted at the beginning of the chapter.
The government should let markets get on with the the job of allocating resources efficiently.
Governments should not interfere by introducing a host of regulations, investigatory bodies, or state-run enterprises.
Nor need the free enterprise ideal be uncompassionate.
The government can make its value judgements about distribution or equality and can pursue its views about the desirable degree of vertical equity without impairing the efficient functioning of a free market economy.
The Conservative case is more than a political ideal: it can be backed up by rigorous economic arguments.
But surely the Labour Party must be aware of these arguments?
Before you conclude that the case of free markets has been convincingly made you should remember to read the fine print.
We began this section by stating that under certain conditions we could show that free enterprise or free markets led to a Pareto-efficient allocation.
It is time to study these conditions in more detail.
In so doing, we shall begin to understand the difference between the two views of how a market economy works.
The Conservative Party believes that these are minor qualifications that do not seriously challenge the case for a free market economy.
The Labour Party believes that these qualifications are so serious that they remove any presumption that the government can rely on a free market economy.
Accordingly, they believe that a considerable amount of government intervention is necessary to improve the way the economy works.
15-3 Distortions and the Second Best
Competitive equilibrium is Pareto-efficient because the independent actions of producers setting marginal cost equal to price, and consumers setting marginal benefits equal to price, ensure that the marginal cost of producing a good just equals its marginal benefit to consumers.
A distortion exists whenever society's marginal cost of producing a good does not equal society's marginal benefit from consuming that good.
It is simplest to begin straight away with an example of a distortion.
In the previous section we suggested that a government could use taxes and welfare benefits to redistribute income-earning potential and thereby enforce its value judgements about equity while leaving the market economy to take care of allocative efficiency.
We now explain why this apparently neat solution will not in fact be possible.
Taxation as a Distortion
Suppose the government wishes to subsidize the poor.
To pay for these subsidies the government must tax the rich.
It does not matter much whether the government taxes the incomes of rich people or taxes the goods that rich people buy.
Let us extend our example from the previous section where there are only two goods, films and meals.
Suppose everyone buys meals, but only the rich can afford to go to the cinema.
If the government wishes to raise tax revenue in order to subsidize the poor, it should levy a tax on films.
Figure 15-4 shows that tax on films makes the gross-of-tax price of films to consumers exceed the net-of-tax price received by producers of films.
The difference between the price to consumers and the price for producers is exactly the amount of the tax on each unit of films sold.
Consumers equate the gross price to the  value of the marginal benefit they receive from the last film, but producers equate the marginal cost of films to the lower net-of-tax price of films.
Figure 15-4 A TAX ON FILMS.
DD shows the demand for films and the marginal benefit of the last film to consumers.
SS shows the quantity of films supplied at each price received by producers and is also the marginal social cost of producing films.
Suppose each unit of films bears a tax equal to the vertical distance FE.
To show the tax-inclusive price required to induce producers to produce each output, we must draw the new supply curve SS '; that is a constant vertical distance EF above SS.
The equilibrium quantity of films is Q. Consumers pay a price P 1 , producers receive a price P 2 , and the tax per film is the distance EF.
At the equilibrium quantity Q the marginal consumer benefit is P l but the marginal social cost is P 2 .
Society would make a net gain by producing more films.
Hence the equilibrium quantity Q is socially inefficient.
Hence in competitive equilibrium the price system no longer equates the social marginal cost of producing films with the social marginal benefit of consuming films.
In this example, the marginal benefit of another film exceeds the marginal cost of producing another film.
The tax on films is causing too few films to be produced.
Producing another film would add more to social benefit than it would add to social cost.
When the other industry (meals) is untaxed and in competitive equilibrium, we showed in the last section that the marginal cost of producing films is exactly the value of the marginal utility sacrificed by not using the same resources to produce more meals.
But when films are taxed we have just seen that the  marginal social benefit of another film exceeds its marginal cost.
Hence the marginal value to consumers of another film exceeds the marginal value of the last meal currently being produced by the resources that would be necessary to make another film.
By transferring resources from meals into films, consumers of extra films could compensate meals consumers for the meals sacrificed and still have some extra utility left over.
Hence it would be possible to achieve a Pareto gain, making some people better off without making anyone else worse off.
A similar argument holds for any other commodity we try to tax.
A tax will cause a discrepancy between the price the purchaser pays and the price the seller receives.
Suppose there is an income tax.
Firms equate the marginal value of labour to the gross wage rate, but suppliers of labour equate the after-tax wage rate to the marginal value of the leisure that they sacrifice in order to work another hour.
The income tax ensures that the marginal value product of labour, society's benefit from another hour of work, now exceeds the marginal value of the leisure being sacrificed in order to work.
Again this is inefficient.
With another hour of work, society would gain more than sufficient to compensate workers fully for the value of the extra leisure forgone.
We can now pose the choice between efficiency and equity in a stark form.
If the economy is perfectly competitive, and if the government is quite happy with the distribution of income-earning potential that currently exists, the government need raise no taxes.
Perfectly competitive free market equilibrium will then allocate resources efficiently.
There will be no possibility of reallocating resources to make some people better off without making others worse off.
The economy is not wasting resources.
However, if as a pure value judgement the government considers that this distribution of income-earning potential is inequitable, the government will need to raise taxes from some people in order to provide subsidies for others.
Yet the very act of raising taxes introduces a distortion.
The price system, operating through  competitive free markets, will no longer equate the marginal benefits of taxed commodities with their marginal cost.
The resulting equilibrium in the economy will be allocatively inefficient.
Society will be wasting resources by producing the wrong output levels of different goods.
What should the government do?
The answer depends a good deal on the value judgements of the government in power.
The more the government dislikes the income distribution thrown up by the free market, a distribution reflecting differences in innate ability, human capital, and financial wealth, the more the government is likely to judge that the inefficiency costs of distortionary taxes are a price worth paying in order to secure a more equitable distribution of income and utility.
Conversely, the more the government feels able to tolerate the income distribution thrown up in a free market economy, the more the government can resist distortionary taxation and allow competitive free markets to allocate resources as efficiently as possible.
Thus one explanation for the differing attitudes to the market economy expressed in the election manifestos cited at the beginning of the chapter is a difference in value judgements about equity.
Caricaturing the argument so far, the Conservative Party supports a free enterprise economy because it considers the most important objective is to maximize the size of the national cake by allocating resources as efficiently as possible.
The Labour Party supports considerable state intervention because it considers that it is more important to divide the cake more fairly, even if this means having more allocative inefficiency and a smaller cake to share out.
But this is only part of the disagreement.
The pursuit of equity through redistribution taxation is not the only distortion that can lead to allocative inefficiency.
We shall consider only distortions in the next section.
Before leaving our simple tax example, there is one final point that it is important to make.
The Second Best
Thus far we have shown that when there is no distortion in the market for meals, a tax on films will lead to an inefficient allocation.
Because the market for meals is in competitive equilibrium we saw that the marginal cost curve for films also told us the marginal value of the meals being sacrificed to make the last film.
Hence, any tax on films, by driving a wedge between the marginal value of films and the marginal cost of films, drives a wedge between the marginal benefit of employing scarce resources in the film industry and employing these same resources in the meals industry.
Suppose, however, that there is a tax on meals.
What should we do in the film industry?
If we could abolish the tax on meals, then we could get back to free competitive equilibrium in both industries which we know is Pareto-efficient.
We sometimes call this the first-best allocation to remind ourselves that it is fully efficient.
Suppose, however, that we cannot get rid of the tax on meals.
The government needs some tax revenue to pay for national defence or its budget contribution to the European Community.
Given that there is an unavoidable tax on meals, do we allocate resources more efficiently by ensuring that at least there are no distortions in the film industry?
In a seminal article Professors Richard Lipsey and Kelvin Lancaster showed that the answer is 'No!'
Suppose both industries are in equilibrium but there is a tax on meals.
The marginal cost curve for films shows the opportunity cost to private producers of the resources they employ to make films rather than meals.
But this is no longer society's opportunity cost of these resources in the film industry.
Why not?
Because producers are valuing the resources in terms of the price producers could get by transferring them to the meals industry.
But the price of meals to consumers is higher than the price of meals received by producers.
To check you understand this, try drawing a diagram like Figure 15-4 for the meals industry.
Since consumers of meals equate the value of the marginal utility of the last meal to the tax-inclusive price that must be paid, society's valuation of the last meal exceeds the net-of-tax price to which producers of meals are equating marginal cost.
Hence the private producers' marginal cost curve for films reflects the market value to producers of using these resources to   make meals instead, but it no longer reflects the opportunity cost or utility valuation of forgone meals to society.
It understates the social valuation of the meals forgone in order to make films.
Figure 15-5 shows what happens in the market for films when there is a tax-induced distortion in the market for meals.
DD is the demand curve for films and reflects the marginal valuation of the last film by consumers of films.
MPC is the marginal private costs to film producers of using resources to make films.
It shows what the resources could have earned in producing meals on which meal producers would receive the net-of-tax price.
MSC is the marginal social cost of using these resources to make films rather than meals.
It exceeds the marginal private cost to producers because the consumer benefits of extra meals exceed the value of extra meals to producers.
Figure 15-5 THE SECOND BEST.
DD shows the demand curve and marginal consumer benefit of films.
MPC is the marginal private cost of films and reflects what the resources could earn for producers if they were employed in the meals industry.
With a tax on meals, the consumer price and marginal consumer benefit of meals will exceed the net-of-tax price received by producers of meals.
Hence the social opportunity cost of using resources in films, the value of the marginal consumption benefits of meals forgone, exceeds the private producers' opportunity cost of using these resources in films.
MSC shows the marginal social cost of the forgone utility by using resources in films rather than meals.
With no film tax, competitive equilibrium is at E ';.
At the output Q '; the marginal social benefit of films is less than the marginal social cost of films.
Q *; is the efficient output of films at which marginal social cost and marginal social benefit of films are equal.
A tax on films equal to the distance E *; F would induce a competitive film industry to produce Q *;.
Consumers would pay  and producers would receive .
With no tax on films, competitive equilibrium in the film industry would occur at E '; where the demand curve crosses the private marginal cost curve which is also the competitive supply curve of the film industry.
At this output Q '; the marginal social cost exceeds the marginal social benefit as given by the height of the demand curve DD.
To equate the marginal social cost of films and their marginal social benefit it is necessary to levy a tax FE *; on films.
Competitive film producers would then produce an output Q *;, receiving a net-of-tax price  equal to their marginal private cost at this output.
Consumers would pay the tax-inclusive price .
Marginal social cost and marginal social benefit would then be equated at the point E *;.
In contrast to the first-best allocation, when we achieve full efficiency by removing all distortions, we have now developed the principle of the second best.
The principle of the second best says the following.
Suppose we are interested only in allocative efficiency.
However, there is an inevitable distortion somewhere else in the economy that we are unable to remove.
It is inefficient to treat other markets as if that distortion did not exist.
Thus in the film industry it is inefficient to aim to equate private marginal cost and private marginal benefit, the efficient outcome in the absence of a meals tax.
Rather, it is efficient to deliberately introduce a new distortion to the film industry to help counterbalance the inevitable distortion in the meals industry.
In effect, the theory of the second best says that if there must be a distortion, for example if the government has to raise some taxes, it is a mistake to concentrate the distortion in one market.
It is more efficient to spread its effect more thinly over a wide range of markets.
In the example of Figure 15-5 a tax on meals leads to too few meals being produced.
The only place the resources can go is the film industry, so too many films are being produced.
A tax on films helps redress the balance.
Several applications of this general principle will be found in the ensuing chapters.
The real  world in which we live unfortunately provides us with several inevitable distortions.
Given their existence, the argument of this section implies that the government may increase the overall efficiency of the whole economy by introducing new distortions to offset distortions that already exist.
By now you will rightly be wanting to know the source of these inevitable distortions that the government should take action to offset.
We now list the most important ones.
15-4 Market Failure
We began by showing that in the absence of any distortions a freely competitive equilibrium would ensure allocative efficiency.
We use the term market failure to cover all the circumstances in which equilibrium in free unregulated markets (i.e., markets not subject to direct price or quantity regulation by the government) will fail to achieve an efficient allocation.
Market failure describes the circumstances in which distortions prevent the Invisible Hand from allocating resources efficiently.
We now list the possible sources of distortions that lead to market failure.
1 Imperfect competition
It is perfect competition that leads firms to set marginal cost equal to price and thus to marginal consumer benefit.
Under imperfect competition, producers set marginal cost equal to marginal revenue, which is less than the price at which the last unit is sold.
Since consumers equate price to marginal benefits derived from the last unit, in general marginal benefit will exceed marginal cost in imperfectly competitive industries.
Such industries will tend to produce too little.
Expanding output would add more to consumer benefit than it would to production costs or the opportunity cost of the resources used.
This idea forms the theme of Chapter 17.
2 Social priorities such as equity
We have already examined how redistributive taxation in the pursuit of equity will induce allocative distortions by driving a wedge between the price the consumer pays and the price the producer receives.
We study the principles of taxation more fully in Chapter 16.
3 Externalities
Externalities are things like pollution, noise, and congestion.
What they have in common is that one person's actions have direct costs or benefits for other people which that individual does not take into account.
Much of the remainder of this chapter is devoted to analysing this distortion.
We shall see that the problem arises because there is neither a market nor a market price for things like noise.
Hence we cannot expect markets and prices to ensure that the marginal benefits of making a noise are equated to the marginal cost of that noise to other people.
4 Other missing markets: future goods, risk, and information
These are further examples of commodities for which markets are absent or limited.
In Chapter 14 we saw how moral hazard and adverse selection tend to inhibit the setting up of insurance markets to deal with risk.
As with externalities, we cannot expect markets to allocate resources efficiently if the markets do not exist in the first place.
We pursue this theme at the end of this chapter.
But we begin by looking more closely at the problem of externalities.
15-5 Externalities
An externality arises whenever an individual's production or consumption decision directly affects the production or consumption of others other than through market prices.
Suppose a chemical firm discharges waste into a lake, polluting the water and directly imposing an additional production cost on anglers (fewer and smaller fish, which are harder to catch) or a consumption cost on swimmers (less pleasant swimming and a dirty beach).
If there is no 'market' for pollution, the firm can pollute the lake without cost.
Its self-interest will lead it to pollute until the marginal benefit of polluting (a cheaper production process for chemicals) equals its own marginal cost of polluting, which is zero.
It takes no account of the marginal cost its pollution imposes on anglers and swimmers.
Conversely, by painting your house you make the whole street look nicer and give consumption benefits to your neighbours.
But you paint only up to the point on which your own marginal benefit equals the marginal cost  of the paint you buy and the time you spend.
Your marginal costs are also society's marginal costs, but society's marginal benefits exceed your own.
Hence there is too little house-painting.
In both cases there is a divergence between the individual's comparison of marginal costs and benefits and society's comparison of marginal costs and benefits.
Free markets cannot induce people to take account of these indirect effects on other people if there is no market in these indirect effects.
Divergences between Private and Social Costs and Benefits
Suppose a chemical firm discharges a pollutant into a river, the quantity of pollutant discharged being in proportion to the level of chemical production.
Further down the river there are food-processing companies using river water as an input in making sauce for baked beans.
There are also farmers with agricultural land.
At small levels of chemical output, pollution is negligible.
The river can dilute the small amounts of pollutant discharged by the chemical producer.
But as the discharge rises the costs of pollution rise sharply.
Food processers must worry about the purity of their water intake and build expensive purification plants.
Still higher levels of pollution start to corrode pipes and contaminate agricultural land.
Figure 15-6 shows the marginal private cost MPC of producing chemicals.
For simplicity, we assume that marginal private costs are constant.
It also shows the marginal social cost MSC of chemical production.
The divergence between marginal private cost and marginal social cost reflects the marginal cost imposed on other producers by an extra unit of the production externality of pollution that the chemical producer disregards.
The demand curve DD shows how much consumers are willing to pay for the output of the chemical producer.
If that producer acts as a price-taker, equilibrium will occur at E and the chemical producer will produce an output Q at which the marginal private cost equals the price received by the firm for its output.
However, at this output Q the marginal social cost MSC exceeds the marginal social benefit of chemicals as given by the corresponding point on the demand curve DD.
The market for chemicals takes no account of the production externality inflicted on other firms.
Since at the output Q the marginal social benefit of the last output unit is less than the marginal social cost inclusive of the production externality, the output Q is not socially efficient.
By reducing the output of chemicals society would save more in social cost than it would lose in social benefit.
Reducing the output of chemicals and the corresponding amount of pollution would allow society to make some people better off without making anyone worse off.
Figure 15-6 THE SOCIAL COST OF A PRODUCTION EXTERNALITY.
Competitive equilibrium occurs at E. The market clears at a price P which producers equate to marginal private cost MPC.
But pollution causes a production externality which makes the marginal social cost MSC exceed the marginal private cost.
The socially efficient output is at E '; where marginal social cost and marginal social benefit are equal.
The demand curve DD measures the marginal social benefit because consumers equate the value of the marginal utility of the last unit to the price.
By inducing an output a in excess of the efficient output Q ';, free market equilibrium leads to a social cost equal to the area of the triangle E '; EF.
This shows the excess of social cost over social benefit in moving from Q '; to Q.
In fact the socially efficient output is Q ';, at which the marginal social benefit of the last output unit equals is marginal social cost.
E '; is the efficient point.
How much does society lose by producing at the free market equilibrium point E rather than at the socially efficient point  E ';?
The vertical distance between the marginal social cost MSC and the marginal social benefit as given by DD shows the marginal social loss of producing the last output unit.
Hence, by expanding from Q '; to Q, society loses a total amount equal to the triangle E '; EF in Figure 15-6.
This measures the social cost of the market failure caused by the production externality of pollution.
Production externalities lead to a divergence between marginal private production costs and marginal social production costs.
Similarly, a consumption externality leads to a divergence between marginal private benefits and marginal social benefits.
Figure 15-7 illustrates a beneficial consumption externality as when painting your house or planting roses in your front garden gives pleasure also to your neighbours.
Since there are no production externalities MPC is both the marginal private cost and the marginal social cost of making your house and garden look nicer.
It is the cost of the paint and plants plus the opportunity cost of your time.
DD is the marginal private benefit of house improvements and, comparing your own costs and benefits, you will undertake a quantity Q of improvements.
But you do not take account of the consumption value these improvements have for your neighbours.
Since these are benefits too, the marginal social benefit MSB is greater than your marginal private benefit.
In comparison with the free market equilibrium at E, the socially efficient quantity of improvements is Q ';.
At E '; the marginal social benefit and marginal social cost are equated.
Free market equilibrium leads to too few improvements.
Society could gain the triangle EFE ';, measuring the excess of social benefits over social costs, by increasing the quantity of improvements from Q to Q ';.
Alternatively, this same triangle measures the social cost of the market failure that leads free market equilibrium to a social inefficient allocation.
Figure 15-7 A BENEFICIAL CONSUMPTION EXTERNALITY.
With no production externality, marginal private cost and marginal social cost coincide.
DD measures the marginal private benefit and free market equilibrium occurs at E. The beneficial consumption externality makes marginal social benefit MSB exceed marginal private benefit.
E '; is the socially efficient point.
The consumption externality leads to market failure.
By producing the output a instead of the efficient output Q ';, free market equilibrium causes a social loss equal to the triangle EFE ';.
Property Rights and Externalities
Suppose your neighbour's tree grows into your garden, obscuring your light and giving you a harmful consumption externality.
If the law says that you must be compensated for any damage suffered, your neighbour will either have to pay up or cut back the tree.
What is the smart thing for your neighbour to do, and how much compensation will you get?
Your neighbour really likes the tree and wants to know how much it would take to compensate you to leave the tree at its current size.
Figure 15-8 shows the marginal benefit MB that your neighbour gets from the last inch of tree size and the marginal cost MC to you of that last inch of tree size.
At the tree's current size S 1 the total cost to you is the area OABS 1 .
You simply take the marginal cost OA of the first inch, then add the marginal cost of the second inch, and so on till you get to the existing size S 1 .
The area OABS 1 is what you require in compensation if the tree size is S 1 .
Your neighbour is about to pay up this  amount.
But he has a daughter who is studying economics.
She points out that at the size S 1 the marginal benefit of the last inch is less than the marginal cost to you, which is also the amount you must be compensated for that last inch on the tree.
It is not worth your neighbour having a tree this big.
Nor, she points out, it is worth cutting the tree down altogether.
The first inch yields a higher marginal benefit to your neighbour than the amount that you require in compensation to offset the marginal cost to you of that first inch.
A tiny tree has little effect on your light and makes the garden next door look nicer.
Figure 15-8 THE EFFICIENT QUANTITY OF AN EXTERNALITY.
MB and MC measure the marginal benefit to your neighbour and marginal cost to you of a tree of size S. The efficient size is S * where the marginal cost and benefit are equal.
Beginning from a size S 1 you might bribe your neighbour the value S * EDSI to cut back to S * .
Below S * you would have to pay more than it is worth to you to have the tree cut back further.
Alternatively, your neighbour might pay you the value OAES * to have a tree of size S *;.
Property rights, in this case whether you are legally entitled to compensation for loss of light to your garden, determine who compensates whom but not the outcome S * of the bargain.
The efficient allocation or tree size S * occurs where the marginal benefit to your neighbour equals the marginal cost to you.
Above S * it is worth cutting back the tree since the marginal cost (and hence the compensation) exceeds the marginal benefit.
Below S * it is worth increasing the tree size and paying the marginal compensation that is less than the marginal benefit.
At the efficient size S * your total cost is the area OAES&SUP*; and this is the total amount of compensation you will be paid.
Notice that since a larger tree benefits one party but hurts the other party, the efficient tree size and efficient quantity of the externality will not be zero.
Rather, it occurs where the marginal benefits equals the marginal cost.
Property rights, in this case the legal right for you to be compensated for infringement of your garden and light, enter in two distinct ways.
First, they affect who compensates whom.
They have a distributional implication.
Suppose there was no law requiring compensation.
Would you just sit there and allow your neighbour's tree to grow to a size S 1 that inflicts a huge cost on you?
Of course not.
You would bribe your neighbour to cut it back.
You would compensate your neighbour for the loss of his marginal benefit.
It would be worth you paying to have the tree cut back as far as S * but no further.
Beyond that size, you would be paying more in compensation for the loss of marginal benefit of another inch than you would be saving yourself in reduced cost of the externality.
So you would pay your neighbour a total of EDS 1 S * to compensate for the loss of benefit in cutting the tree back from S 1 to S * .
Who has the property rights determines who must compensate whom, but it does not affect the quantity that the bargain will determine.
It must pay to reach the point at which the marginal benefit to one of you equals the marginal cost to the other.
Property rights thus have a distributional implication -- who compensates whom -- but also act to achieve the socially efficient allocation.
They implicitly set up the 'missing market' for the externality.
The market ensures that the price equals the marginal benefit and the marginal cost, and hence equates the two.
Sometimes economists say that property rights are a way to 'internalize' the externality.
If people must pay for it they will take its effects into account in making private decisions and there will no longer be market failure.
Why then do externalities like congestion and pollution remain a problem?
Why don't private individuals establish the missing market through a system of bribes or compensation?
There are two obvious reasons why it may be hard to set up this market.
The first is the cost of organizing the market.
If a factory chimney dumps smoke on a thousand gardens nearby it may be very expensive to collect 1 from each household to bribe the factory to cut back to the socially efficient amount.
If cars joining a crowded road take no account of the extent to which they slow down other road users, it may be almost impossible to rush from car to car offering or collecting bribes!
Second, there is the free-rider problem.
Suppose someone knocks on your door and says: 'I am collecting bribes from people who mind the factory smoke falling on their gardens.
The money I collect will be paid to bribe the factory to cut back.
Do you wish to contribute?
I am going round 5000 houses in the neighbourhood.'
Whether you mind or not, you say: 'I don't mind and won't contribute.'
Provided everybody else pays, the factory will cut back and you cannot be prevented from getting the benefits.
The smoke will not fall exclusively on your garden merely because you alone did not pay up.
Regardless of what other people contribute, there is no incentive for you to contribute: you are a free-rider.
But everyone else will reason similarly; hence no one will pay, even though you would all have been better off paying and getting the smoke cut back.
15-6 Environmental Issues
When, for either of these reasons, there is no implicit market for pollutants, there will be an overproduction of pollutants.
Because private producers fail to take account of the costs they impose on others, in equilibrium the social marginal cost will exceed the social marginal benefit.
If the private sector cannot organize charges for the marginal externalities pollution creates, perhaps the government can?
By charging (through taxes) for the divergence between marginal private and social cost, the government could then induce private producers to take account of the costs inflicted on others.
This argument for government intervention through taxes is examined in the next chapter.
Pollution taxes, especially for water pollution, have been used in France, Italy, Germany, and the Netherlands.
But most policy takes a different approach, the imposition of pollution standards that regulate the maximum amount of allowed pollution.
Air Pollution
Since the Clean Air Act of 1956 governments in the UK have taken responsibility for designating clean air zones in which certain types of pollution, notably the smoke caused by burning coal, are illegal.
The number of designated clean air zones has increased steadily during the last 25 years, and Table 15-1 shows the dramatic reduction in smoke pollution in the UK.
Adding lead to petrol improves the fuel economy performance of cars.
However, lead emissions from car exhausts are an atmospheric pollutant that may be harmful to people's health.
Since 1972 the UK government has adopted a policy of progressive reductions in the quantity of lead permitted in petrol.
Table 15-2 shows that lead emission from petrol-engined vehicles has fallen since 1973 even though total consumption of petrol has risen by 20 per cent.
Water Pollution
Since 1951 governments in the UK have imposed increasingly stringent controls on discharges into inland waters.
Although we  tend to think of industrial effluent, sewage is a more important source of pollution.
During the 1970s regional water authorities in England and Wales spent (at 1989 prices) an annual average of 2.4 billion on water purification and sewage treatment.
Table 15-3 shows that this expenditure has been only moderately successful in reducing water pollution.
By the late 1980s an even more important problem has been recognized: water pollution from nitrates used as fertilizers on agricultural land.
The European Community has laid down tough standards for water purity which will take many years to achieve in countries like the UK.
Table 15-1 Smoke emission in the UK (million tonnes per annum)
Table 15-2 Petrol consumption and lead emission
Table 15-3 Water pollution in England and Wales (badly polluted mileage as % of total)
Evaluating Pollution Policy in the UK
Direct regulation of pollution has met with mixed success over the last 25 years.
In some cases, for example the smoke pollution which used to be especially acute in winter when smoke and fog mixed to produce dense and harmful 'smog', tougher standards have led to dramatic improvements in environmental quality.
Many rivers are also cleaner, and fish have reappeared.
In other cases, governments have tried to regulate but often been ineffective; it is hard to enforce regulations such as those that prevent ships discharging oil at sea.
In yet other cases there has been little attempt to intervene.
For example, coal-fired power stations continue to emit large quantities of sulphur dioxide, which the high chimneys are only partially successful in dispersing.
And ecologists continue to oppose government policies that allow new coal mines in previously green countryside, or permit the disposal of nuclear waste at sea.
Having described the consequences of anti-pollution policy in the UK, how should we assess its success?
Has the government been tough enough on polluters?
Recall from Figures 15-6 and 15-8 that the efficient quantity of pollution is not zero but rather the level at which the social marginal cost of cutting back pollution equals its social marginal benefit.
The fact that pollution still exists is not sufficient to establish that policy has not been tough enough.
What we can say is that where pollution control has been attempted it has usually taken a crude and simple form.
Calculations of the social marginal costs and benefits of cutting back pollution tend to be conspicuous by their absence.
In part this reflects the genuine difficulty in measuring marginal benefits.
For example, in considering how much to reduce lead emissions from cars it is not impossible to calculate the marginal social cost of producing cars with anti-pollution exhaust systems and the marginal social cost of cars that use more fuel per mile.
But even if doctors were unanimous about the effects of lead emission on health, how should society value a marginal increase in the health of current and future generations?
This is not merely a question of allocative efficiency, which we might answer by considering the resources tied up in looking after the sick or the extra output that healthier workers could produce.
It is also a question of equity, both within the current generation -- poor inner-city children may be most vulnerable to arrested development caused by inhaling lead-polluted air -- and across generations.
Devoting more resources to reducing lead pollution today may reduce the resources producing consumer goods for today's consumers, but will improve the quality of life for tomorrow's consumers.
Prices versus Quantities
Although it is difficult to reach a clear judgement about whether there is enough pollution control, we can discuss whether the current mechanism of pollution control is sensible.
We began this section by noting that, if free markets tend to overpollute, society can reduce pollution either by regulating the quantity of pollution (as it does) or by using the price system to discourage such activities by taxing them.
Would it be more sensible to intervene through the tax system than to regulate quantities directly?
Many economists believe the answer to this  question is yes.
One reason is that, if each firm were charged the same price or tax for a marginal unit of pollution, then every firm would pollute up to the point at which the marginal cost of reducing pollution was equal to the price of pollution.
Any allocation in which different firms have different marginal costs of reducing pollution is a socially inefficient way to reduce the total quantity of pollution.
By having the firms with low marginal reduction costs contract further and firms with high marginal reduction costs contract less, the same total reduction in pollution could be achieved at less cost.
However, there are two important qualifications to this argument.
First, it would be necessary to monitor the quantity of pollution of each firm in order to assess its tax liability.
Second, in order to assess the tax rate or charge for pollution it would still be necessary to   calculate the overall costs and benefits of marginal changes in the amount of pollution.
If the government has to make a decision on the socially efficient level of pollution anyway, it may be simpler to regulate the quantity directly.
Finally, there is the issue of the uncertainty of the effect of the legislation.
Suppose pollution beyond a certain critical level would have disastrous social consequences, for example irreversibly damaging the ozone layer above the earth.
By regulating the quantity directly it is at least possible to ensure that the disaster is avoided.
Indirect control through taxes or charges runs the risk that the government might do its sums wrong and set the tax too low.
Pollution will then be higher than intended, and possibly disastrous.
Thus, regulating the total quantity of pollution and conducting a series of spot checks on individual producers to see that they are not violating agreed standards is a relatively simple policy which may avoid the worst outcomes.
However, by failing to take account of differences in the marginal cost of reducing pollution across different polluters, it does not reduce pollution in the manner that is cost-minimizing to society, for that would involve the equalization across polluters of the marginal cost of reducing pollution.
There is no simple answer in an uncertain world where monitoring and enforcement also use up society's scarce resources.
Lessons from the United States
The United States has gone furthest in trying to use property rights and the price mechanism to cut back pollution in a manner that is economically efficient.
The US Clean Air Acts (19S5, 1970,1977) have established an environmental policy that includes an emissions trading programme and bubble policy.
The Acts lay down a minimum standard for air quality, and impose pollution emission controls to particular polluters.
Any polluter emitting less than their specified amount obtains an emission reduction credit (ERC), which can be sold to another polluter which wants to go over its allocated pollution limit.
Thus, the total quantity of pollution is regulated, but firms that can cheaply reduce pollution have an incentive to do so, and sell off the ERC to firms for which pollution reduction is more expensive.
In this way we get closer to the efficient solution in which the marginal cost of pollution reduction is equalized across firms.
When a firm has many factories, the bubble policy applies pollution controls to the firm as a whole rather than individual factories.
To achieve the specified overall reduction in pollution, the firm can cut back most in the plants in which pollution reduction is cheapest.
The bubble policy is efficient because it encourages equalization of the marginal cost of cutting back pollution at different plants within the firm.
Thus, the US policy manages to combine 'control over quantities' for aggregate pollution, where the risks and uncertainties are greatest, with 'control through the price system'for allocating efficiently the way these overall targets are achieved.
15-7 Other Missing Markets: Time and Risk
The previous two sections have been devoted to a single idea.
When externalities exist, free market equilibrium will lead to an inefficient resource allocation because the externality itself does not have a market or a price.
People take no account of the costs and benefits their actions inflict on others.
Without a market for externalities the price system cannot be expected to bring marginal costs and marginal benefits of these externalities into line.
In this section we discuss two other types of 'missing market', those associated with time and with risk.
The present and the future are linked.
People save, or refrain from consumption, today in order to consume more tomorrow.
Firms reduce current output by devoting resources to training or building in order to produce more tomorrow.
How should society make plans today for the quantities of goods to be produced and consumed in the future?
Ideally we should like to organize everyone's plans today so that the social marginal cost of goods in the future just equals the social marginal benefit.
In Chapter 14 we introduced the concept of a forward market, in which buyers and sellers made contracts today for goods to be delivered in the future at a price agreed today.
Suppose there was a forward market for delivery of copper in 1995.
Consumers would equate the marginal benefit of copper in 1995 to the forward price, which producers would equate to the marginal cost of producing copper for 1995.
A complete set of forward markets for all commodities for all future dates would lead producers and consumers today to make consistent plans for future production and consumption of all goods, and the social marginal benefit of every future good would equal its social marginal cost.
In Chapter 14 we explained why only a very limited set of future markets actually exists.
You can trade gold one year forward but not cars or washing machines.
Since nobody knows the characteristics of next year's model of a car or a washing machine, it is impossible to write legally binding contracts which could be enforced when the goods are actually delivered.
Yet without these forward markets the price system cannot be expected to ensure that the  marginal cost and marginal benefits of planned future goods are equal.
Similarly, there is only a limited set of contingent or insurance markets for dealing with risk.
In Chapter 14 we argued that people are typically risk-averse and dislike risk.
Risk is costly to individuals because it reduces their utility.
But does society undertake the efficient amount of risky activities?
A complete set of insurance markets would allow risk to be transferred from those who dislike risk to those who are prepared to bear risk at a price.
The equilibrium price or insurance premium would equate the marginal cost and marginal benefit of risk-bearing.
The equilibrium price of risky activities would include the insurance premium, and the price system would ensure that social marginal costs and benefits of risky activities were equated.
However, we saw that problems of adverse selection and moral hazard would inhibit the organization of private insurance markets.
If some risky activities are uninsurable at any price, again we cannot expect the price system to guide society to an allocation at which social marginal costs and benefits are equal.
Future goods and risky goods are examples of commodities with missing markets.
Like externalities, they induce market failure.
Free market equilibrium will not generally be efficient.
As we have seen, an important problem which inhibits the development of forward and contingent markets is the provision of information.
For example, it is the problem of acquiring the relevant information about purchasers of insurance policies that leads insurance companies to face problems of moral hazard and adverse selection.
We now study some practical examples of how informational problems affect the way in which markets work.
15-8 Quality, Health, and Safety
In the real world, information is incomplete because gathering information is costly.
This may lead to socially inefficient allocations.
For example, a worker who is unaware that exposure to high levels of benzene, as happens in some chemical plants, might cause cancer will be willing to work for a lower wage than she would if this information were widely available.
The firm's production cost will understate the true social cost and the good will be overproduced.
In most countries, governments have accepted an increasing role in regulating health, safety, and quality standards because it has been recognized that this is a potentially important area of market failure.
Although imperfect information would be sufficient to cause a divergence between private and social cost, in practice the argument for intervention is usually reinforced by externalities.
With better information, the chemical worker would ask for a wage that compensated for the danger of illness but would still neglect the marginal cost that illness would impose on society through expensive health care, which uses society's resources but may be free to the individual.
Examples of government regulation in the UK are the Health and Safety at Work Acts, legislation controlling food and drugs production, the Fair Trading Act which governs consumer protection, and the various traffic and motoring regulations.
The purpose of such legislation is twofold: to encourage the provision of information that will allow individuals more accurately to judge costs and benefits, and to set and enforce standards designed to reduce the risks of injury or death.
Providing Information
Figure 15-9 shows the supply curve SS for a drug that is potentially harmful.
DD is the demand curve when consumers are unaware of the full extent of the danger.
In equilibrium at E the quantity Q is produced and consumed.
With full information about the dangers, people would buy less of the drug.
The demand curve DD' shows the marginal consumer benefit with full information.
By moving to the new equilibrium at E' society avoids the deadweight burden EE 'F from overproduction of the drug.
If information were free to collect, everyone would already know the true risks.
From the social get EE 'F we should subtract the resources that society expends in discovering this information.
In saying that the free market equilibrium would be at E, we are really saying that it would not be worth while for each individual to check up privately on each and every drug on the market.
But it makes sense for society to have a single regulatory body that does the checking  and a law whose enforcement entitles individuals to assume that drugs being sold have been checked out as safe.
Figure 15-9 INFORMATION AND UNSAFE GOODS.
Consumers cannot individually discover the safety risks associated with a particular good.
Free market equilibrium occurs at E. A government agency now provides information about the product.
As a result, the demand curve shifts down and the new equilibrium is at E' where the true or full information valuation of an extra unit of the good equals its marginal social cost.
Providing information prevents a welfare cost E 'EF that arises when uninformed consumers use the wrong marginal valuation of the benefits of the good.
Certification of safety or quality need not be carried out by the government.
Sotheby's certify the quality of Rembrandts, and the private University College at Buckingham certifies educational attainments of its students.
People arrested on suspicion of drunk driving in the UK are allowed to send half their blood sample to a private certification agency to corroborate the results of the police analysis.
Nevertheless, two factors tend to inhibit the use of private certification in many areas of health and safety.
The first concerns the incentive to tell the truth.
If private certification firms are in business to earn profits, which they do by charging firms for a certificate attesting the quality or safety of a product, which then allows firms to sell more of the product at any given price, will not firms have an incentive to bribe the certifier to obtain the certificate?
And even if they do not, will people believe that they have not done so?
Firms issuing false certificates might be subject to lawsuits, but these are expensive.
Private individuals might not be able to afford to sue, and society might not wish to devote a large quantity of resources to court cases.
Second, a private certification agency would have to decide on standards.
What margin of error should be built into safety regulations?
How safe does a drug have to be before it receives a certificate?
These are questions on which society has views.
They involve externalities and may have important distributional implications.
Even where society relies on private agencies to monitor regulations, society will probably want to set the standards itself.
Imposing Standards
The public interest may be especially important when little is known about a product and where the consequences for society of any error may be catastrophic.
Few people would argue that safety standards for nuclear power stations can be adequately determined within the private sector.
In imposing standards, governments increase the private cost of production by preventing firms from adopting the cost-minimizing techniques that would otherwise have been employed.
Sometimes the justification is that the government has access to better information than the private sector and judges the true social cost to be in excess of the private cost.
The imposition of standards then increases private marginal cost, shifts private supply curves upwards, and reduces the overproduction of the good that occurred when the market ignored the divergence between private and social cost.
Frequently, however, the imposition of standards reflects a judgement that important externalities exist or is simply a pure value judgement based on distributional considerations.
One particularly contentious area in the field of health and safety is the valuation of human life itself.
Politicians often claim that human life is beyond economic calculation and must be given absolute priority whatever the cost.
An economist will raise two points in reply.
First, it is quite impossible for society to implement such an objective.
It is simply too expensive in resources to attempt to eliminate all risks of premature death, and in fact we do not do  so.
Second, in making occupational and recreational choices, for example being a racing driver or going climbing, people do take risks.
Society must ask how much more risk-averse it should be than the people it is trying to protect.
Thus, beyond a certain point the marginal social benefit of further risk reduction will exceed the marginal social cost.
It will take an enormous effort to make the world just a little safer, and the resources might have been used elsewhere to greater effect.
Zero risk does not make economic sense.
Economists have long been calling for safety regulations to be subject to cost-benefit analysis.
We need to know the costs of making the world a little safer, and we need to encourage society to make a decision about how much it values the benefits.
However society decides to value the benefit of saving human life, an efficient allocation would adopt health and safety regulations up to the point at which the marginal social cost of saving life by each and every means was equal to the marginal social benefit of saving life.
By shying away from the 'unpleasant' task of spelling out the costs and benefits, society is likely to produce a very inefficient allocation in which the marginal costs and marginal benefits are very different in different activities.
Suppose we assume that each regulation is enforced up to the point at which the marginal cost and marginal benefit of saving life are equal for that activity.
If we can measure the marginal cost directly, we can infer the implicit marginal benefit from saving life through that activity.
Economists frequently complain that such calculations reveal very different implicit marginal benefits across activities, which is unsurprising when those responsible for safety standards in building, motoring, medicine, and other areas make no attempt to reach a common view of the marginal social benefit from saving life.
For example, estimates for the implicit marginal social benefit from saving life in the UK range from 20 million in the case of building regulations introduced after the Ronan Point disaster to 50 for a rarely used test in pregnant women that might prevent some still-births.
Such wide disparities in the social marginal cost of life-saving suggest that society might achieve significant gains in allocative efficiency by adopting an integrated approach to a cost-benefit analysis of health and safety regulations.
Summary
1
Welfare economics deals with normative issues or value judgements.
Its purpose is not to describe how the economy works but to assess how well it works.
2
There are two concepts of equity or fairness.
Horizontal equity is the equal treatment of equals, and vertical equity the unequal treatment of unequals.
Equity is concerned with the distribution of welfare across people.
Decisions about the desirable degree of equity are pure value judgements.
3
A resource allocation is a complete description of what, how, and for whom goods are produced.
To separate as far as possible the concepts of equity and efficiency, economists use the concept of Pareto efficiency.
An allocation is Pareto-efficient if there is no reallocation of resources that would make some people better off without making some people worse off.
If an allocation is inefficient it is possible to achieve a Pareto gain, making some people better off and none worse off.
Many reallocations make some people better off and others worse off.
We cannot say whether such changes are good or bad without making explicit value judgements about the comparison of different people's welfare.
4
For a given level of resources and a given technology, the economy has an infinite number of Pareto-efficient allocations which differ in the distribution of welfare across people.
For example, every allocation that gives the maxi-mum attainable output to a single individual is Pareto-efficient.
But there are many more allocations that are inefficient.
5
Under strict conditions, competitive equilibrium is Pareto-efficient.
Different initial distributions of human and physical capital across people will generate the different competitive equilibria corresponding to each of the possible Pareto-efficient allocations.
When price-taking producers and consumers face the same prices, marginal costs and marginal benefits   are equated to prices (by the independent actions of individual producers and consumers) and hence to each other.
6
In practice, governments face a conflict between the objectives of equity and efficiency.
Redistributive taxation drives a wedge between prices paid by consumers (to which marginal benefits are equated) and prices received by producers (to which marginal costs are equated).
Free market equilibrium will not equate marginal cost and marginal benefit and there will be scope for Pareto gains.
Equilibrium will be inefficient.
7
Distortions occur whenever free market equilibrium does not equate marginal social cost and marginal social benefit.
Distortions lead to inefficiency or market failure.
Apart from taxes, there are three other important sources of distortions: imperfect competition (failure to set price equal to marginal cost), externalities (divergence between private and social costs or benefits), and other missing markets in connection with future goods, risky goods, or other informational problems.
8
When only one market is distorted the first-best solution is to remove the distortion, thus achieving full Pareto efficiency.
The first-best criterion relates only to allocative efficiency.
Governments caring sufficiently about redistribution might still prefer inefficient allocations with greater vertical equity.
However, when a distortion cannot be removed from one market it is not generally efficient to ensure that all other markets are distortion-free.
The theory of the second-best says that it is allocatively more efficient to spread inevitable distortions thinly over many markets than to concentrate their effects in a few markets.
9
Production externalities occur when decisions by one producer affect the production costs of another producer directly, as when one firm pollutes another's water supply.
Consumption externalities imply that one person's decisions affect another consumer's utility directly, as when one person's garden gives pleasure to the neighbours.
External effects shift indifference curves or production functions.
10
Externalities lead to divergence between private and social costs or benefits because there is no implicit market for the externality itself.
When only a few people are involved, a system of property rights may establish the missing market.
The direction of compensation will depend on who has the property rights, but the consequence would be to achieve the efficient quantity of the externality at which marginal cost and marginal benefit are equated.
The efficient solution is rarely to have a zero quantity of the externality.
Transactions costs and the free-rider problem may prevent implicit markets being established.
Equilibrium will then be inefficient.
11
When externalities lead to market failure the government could set up the missing market by pricing the externality through taxes or subsidies.
If it was straightforward to assess the efficient quantity of the externality and hence the correct tax or subsidy, and straightforward to monitor the quantities produced and consumed, such taxes or subsidies would allow the market to achieve an efficient resource allocation.
12
In practice, governments often regulate externalities such as pollution or congestion by imposing standards that affect quantities directly rather than by using the tax system to affect production and consumption indirectly.
Overall quantity standards may fail to equate the marginal cost of pollution reduction across different polluters, in which case the allocation will not be efficient.
However, simple standards may use up less resources in monitoring and enforcement and may prevent disastrous outcomes when there is uncertainty.
13
Moral hazard, adverse selection, and other informational problems prevent the development of a complete set of forward and contingent markets.
Without these markets the price system cannot equate social marginal cost and benefit for future goods or risky activities.
14
Incomplete information may lead to private choices which do not represent the best interests of individuals or society as a whole.
Health, quality, and safety regulations are designed both to provide information and to express society's value judgements about intangibles such as life itself.
By avoiding explicit consideration of social costs and benefits, government policy may be inconsistent in its implicit valuation of health or safety in different activities under regulation.
Key Terms
Problems
1
An economy has 10 units of goods to share out between two people.
(x, y) denotes that the first person gets a quantity x and the second person a quantity y For each of the following allocations say whether they are (i) efficient and (ii) equitable:(a)(10, 0)(b)(7, 2)(c)(5, 5)(d)(3, 6)(e)(0, 10).
What does 'equitable' mean?
If you were making the choice, would you prefer allocation (d) to allocation (e)?
2
Suppose the equilibrium price of meals is 1 and of films 5.
There is perfect competition and no externality.
What can we say about (a) the relative benefit to consumers of a marginal film and a marginal meal?(b) the relative marginal production cost of films and meals?(c) the relative marginal product of variable factors in the film and meal industries?
Hence explain why competitive equilibrium is Pareto-efficient.
3
In deciding whether or not to drive your car during the rush hour, you think about the cost of petrol and the time of the journey.
Do you slow other people down by driving in the rush hour?
Is this an externality?
Does this mean that too many or too few people drive cars in the rush hour?
Would it make sense for city authorities to restrict commuter parking in cities during the day?
4
Explain how an economist might defend laws making it compulsory to wear seat belts in cars.
5
In 1965, 200 people died when the steam boiler exploded on a Mississippi river boat.
This prompted Jeremiah Allen and three friends to form a private company offering to insure any boiler that they had inspected for safety.
The idea of boiler inspections caught on and boiler explosion rates plummeted.
(a) Would Jeremiah Allen's company have been so successful if it had certified boilers but not insured them as well?
Explain.
(b) Could this idea be carried over from boiler inspections to drug inspections?
If not, why not?
6
(a) Why might society wish to ban drugs that neither help nor harm the diseases they are claimed to cure?(b) It is sometimes argued that regulatory bodies will be blamed for bad things that happen in spite of the regulations (e.g. a plane crash) but not blamed so much for good things that are prevented (e.g. the quick availability of a safe and useful drug) by stringent tests and regulations.
Does this mean that regulatory bodies will tend to be too conservative and will over-regulate the activities under their scrutiny?
7
Why is it inefficient for different government departments to have different rules of thumb about the marginal value of human life?
8
Common Fallacies Show why the following statements are incorrect.
(a) Irresponsible firms discharge toxic waste with no thought for the damage inflicted on others.
Society should ban all such discharges.
It would be much better off without them.
(b) Anything governments can do the market can do better.
(c) Anything the market can do the government can do better.
ACID RAIN: A BITTER CONTROVERSY
Gases such as sulphur dioxide and oxides of nitrogen are discharged into the atmosphere, dissolved in water vapour, and fall as acid rain.
Acid rain poisons fish, destroys forests, and corrodes buildings.
The following table shows data for European emissions of sulphur ('000 tonnes a year).
It reveals the appalling pollution in Eastern Europe and the Soviet Union, mainly from power stations fuelled by low-grade coal.
In Western Europe, the UK is a big exporter of acid rain: prevailing winds blow it east, and Scandinavia is a big loser.
Installing and operating enough flue gas de-sulphurization plants to cut UK power station emissions of sulphur dioxide by 50 per cent would add 6 per cent to UK electricity prices.
Since much of the damage occurs in Swedish lakes and German forests, UK voters are unwilling to pay the extra cost.
It needs a concerted European policy, perhaps even with transfer payments between governments, to deal with externalities across national borders.
Europe is trying to agree a 20 per cent cut in sulphur dioxide emissions and a freeze on emissions of oxides of nitrogen (mainly from car exhausts).
But when different countries face a different marginal cost of pollution abatement, and inflict different amounts of marginal damage (polluting unpopulated areas is less costly than polluting densely populated areas), equal cutbacks for all is not the efficient solution.
For a given overall reduction, the efficient solution equates the marginal net benefit (damage reduction minus abatement cost) across different polluters.
The following table shows estimates by Professor David Newberry of Cambridge University of the efficient way to achieve a 30 per cent reduction in sulphur dioxide emissions in Europe.
BOX 15-2 A LOT OF HOT AIR?
Chlorofluorocarbons (CFCs) are gases that are useful in things like aerosols.
But they are thought to destroy the ozone layer, which protects the earth from the sun's rays.
Without this sunscreen, more people will get skin cancers.
As with acid rain, there is a major problem in co-ordinating the policy of different countries.
Each is tempted to act as a free rider: if other countries cut back on atmospheric pollution, everyone will enjoy the benefit.
But governments are beginning to get their act together.
The Montreal Protocol on Substances that deplete the Ozone Layer was concluded in 1987, and 46 countries, including most of the big polluters, have now signed.
The following table, based on estimates by the US Environmental Protection Agency, offers an optimistic assessment of the results of the Protocol:
DEPLETION OF THE OZONE LAYER (%)
A second type of atmospheric pollution is potentially of much greater significance, though as yet it remains the subject of considerable scientific controversy.
The greenhouse effect arises from emissions of CFCs, methane, nitrous oxide, and especially carbon dioxide.
Greenhouse gases are the direct result of pollution, and the indirect result of a reduction in the atmosphere's ability to absorb them.
Plants convert carbon dioxide into oxygen.
Chopping down forests to clear land for cattle, as world demand for hamburgers soars, may be good business in the short run, but its long-run effect on the accumulation of greenhouse gases is significant.
The consequence is global warming.
People in London and Stockholm may get better suntans, but in Africa the likelihood of drought and famine is magnified many times; and, as icecaps melt, the sea level will rise, flooding many low-lying areas.
In 1990 the Intergovernmental Panel on Climate Change forecast that by 2070 the temperature would increase by 3.5C, implying a rise of 45 centimetres in the sea level.
As with acid rain and ozone depletion, international co-ordination of government policy is required to tackle the free rider problem.
But progress is likely to be slow for two reasons.
First, greater scientific uncertainty about the greenhouse effect provides an excuse for the big polluters who have the strongest interest in delaying a change of policy.
Second, in many cases it is the poorer countries who will be most affected by the greenhouse effect.
It is hard for them to convince the rich countries to make sacrifices on their behalf.
16
Taxes and Public Spending: The Government and Resource Allocation
By the 1980s many people felt that the government had too big a role in the economy.
They believed that high levels of government spending were pre-empting resources that could have been used more productively in the private sector, that high taxes were stifling private enterprise, and that the abolition of the complex system of government regulations, interventions, and subsidies would unleash a new wave of private initiative and energy.
These were not quack ideas that never made it in practice.
On the contrary, electorates in many countries turned to the political leaders who promised to implement these new policies -- Mrs Thatcher in the UK, President Reagan in the United States, and Chancellor Kohl in West Germany.
This chapter is about the extent of government involvement in the economy.
How much should the government raise in taxation?
Are there good taxes and bad taxes?
If taxes are needed to pay for government spending, why do we need government spending in the first place?
We begin with three tables that provide some historical perspective.
Table 16-1 shows the scale of government spending in the UK over three decades.
It is important to distinguish government spending on goods and services schools, defence, the police, and so on-- from government spending on transfer payments, such as social security and state pensions.
Whereas spending on goods and services directly uses up factors of production that could otherwise have been employed in the private sector, transfer payments do not directly pre-empt society's scarce resources.
Rather they transfer purchasing power from one group of consumers, those paying taxes, to another group of consumers, those in receipt of transfer payments or subsidies.
Table 16-1 shows that between 1956 and 1976 there was a moderate increase in the share of national income and national resources directly pre-empted by the government through government spending on goods and services.
It also shows that the Thatcher government found it difficult to implement its objective of quickly reducing this percentage.
Nevertheless, the share of national income going to government spending on goods and services is now falling.
The second row of Table 16-1 shows that government spending on transfer payments has also risen faster than national income.
In part this reflects increasing expenditure on state pensions as more and more people live to a ripe old age.
However, the most important source of the large rise in transfer payments was the steady rise in unemployment until the mid 1980s.
The subsequent decline in unemployment has also been responsible for the fall in the share of transfer payments since 1984.
The last row of Table 16-1 shows the turnaround in total spending since 1984.
One reason for trying to reduce government spending is to make room for tax cuts.
Table 16-2 picks out the most controversial aspect of the tax system, the marginal rate of income tax.
The marginal rate of income tax is the percentage taken by the government of the last pound that an individual earns.
In contrast, the average tax rate is the percentage of total income that the government takes in income tax.
A progressive tax structure is one in which the average tax rate rises with an individual's income level.
The government takes proportionately more from the rich than from the poor.
A regressive tax structure is one in which the  average tax rate falls as income level rises.
The government takes proportionately less from the rich.
Table 16-1 Government spending as a percentage of UK national income*
Table 16-2 Marginal income tax rates in the UK (Tax rates on an extra pound of income)
Table 16-2 shows that, as in most countries, the UK has a progressive income tax structure.
Figure 16-1 explains why.
We plot pre-tax income on the horizontal axis and post-tax income on the vertical axis.
The line OG with a slope of 45 degrees would correspond to no taxes.
A pre-tax income OA on the horizontal axis corresponds to the same post-tax income OA on the vertical axis.
Now suppose there is an income tax with a tax allowance OA.
The first OA pounds of income are untaxed.
If the marginal tax rate on taxable income is constant, individuals face a schedule such a OBC.
The individual gets to keep only a constant fraction of each pound of pre-tax income above OA.
The higher the marginal tax rate the flatter the portion BC of the schedule.
How do we calculate the average tax rate at a point such as D?
We join up OD.
The flatter the slope of this line the higher the average tax rate.
Hence, even with a constant marginal tax rate and a constant slope of the portion BC of the tax schedule, the presence of an initial tax allowance makes the tax structure progressive.
If we join up OH we get a line with a flatter slope than OD, which in turn has a flatter slope than OB.
The higher an individual's gross income, the smaller is the tax allowance as a percentage of this gross income so the larger is the percentage of total income on which the individual is paying tax.
Figure 16-1 A PROGRESSIVE INCOME TAX.
The line OG with a slope of 45 degrees shows what would happen in the absence of any income tax.
A pre-tax income measured on the horizontal axis would convert into the same amount of post-tax income measured on the vertical axis.
An income tax plus an allowance OA implies that the first OA pounds of pre-tax income are still retained after-tax.
If income above OA is taxed at a constant marginal rate, the  individual is then on the schedule BC with a constant slope.
The slope is less than 45 degrees because for each extra pound earned the individual is only allowed to keep a constant fraction of it.
Higher pre-tax incomes move the individual up BC and imply that the government is taking a larger and larger fraction of total pre-tax income.
The individual is falling further and further below the no-tax schedule OG.
With a rising marginal tax rate, the schedule falls even further below OG.
But Table 16-2 shows that marginal tax rates also rise with income.
The tax schedule in the UK looks more like the schedule OBDEF.
As individuals move into higher tax bands they pay higher marginal tax rates and move on to even flatter portions of the tax schedule.
The average tax rate now rises sharply with income.
The line joining OF has a much flatter slope than the line joining OD.
Table 16-2 shows that the first Thatcher government was able to reduce marginal tax rates substantially, especially for the very rich.
A millionaire paying an 83 per cent tax rate on all taxable income except the first 54 000 1978-79 was paying only 40 per cent per cent in 1990 91.
Table 16-3 Income tax reform 1975 90 (Marginal tax rates, %)
Table 16-4 Expenditure and revenue of UK central and local government 1987
Table 16-3 shows a worldwide move to cut tax rates, especially for the very rich.
Were the tax cuts designed to make the rich richer?
Or was their purpose to revive hard work and enterprise?
If so, will they work?
These questions go to the heart of the current debate and form the background to much of the discussion of this chapter.
16-1 Taxation and Government Spending
Table 16-1 shows that government spending, and the taxation that finances it, are now running at over 40 per cent of national income.
Table 16-4 shows the composition of government spending and revenue in 1987.
Table 16-4 shows that in 1987 82.4 billion, almost half of total government spending, went on transfer payments such as unemployment benefit and debt interest.
Of the remaining 92.1 billion spent directly on goods and services, the most important spending categories were the National Health Service, defence, and education.
Why is the government directly involved in providing defence, schools, and health services?
How much of each should be provided?
Would it make sense for these activities to be provided by the private sector in the same way as haircuts and cars?
If refuse collection can be 'privatized', why not defence?
To deal with these issues in democratic decision-making, we shall need a large dose of economics and a fair helping of political science.
Table 16-4 shows that most government spending is financed through taxation.
The most important taxes are income tax and expenditure taxes such as value added tax (VAT).
Since state provision of retirement pensions is included on the expenditure side under transfer payments, the pension contributions under the National Insurance Scheme must be included on the revenue side.
Against this background, we begin by discussing the reasons for government spending.
Then we ask how spending should be financed.
Are there good and bad taxes?
The answer depends on the criteria of efficiency and equity that we developed in the last chapter.
16-2 The Government in the Market Economy
In this section we consider the argument that can be used to justify government spending in a market economy.
We begin with public goods.
Public Goods
A private good is a good that, if consumed by one person, cannot be consumed by another person.
Ice cream is a private good.
If you eat an ice cream it prevents anyone else from eating the same ice cream.
For any given supply of ice cream, your consumption reduces the quantity available for others to consume.
Most goods are private goods.
A public good is a good that, even if consumed by one person, can still be consumed by other people.
Clean air and defence are examples of public goods.
If the air is pollution-free, your consumption of it does not interfere with our consumption of it.
If the Royal Navy is patrolling Britain's coastal waters, your consumption of national defence does not affect our quantity of national defence.
In fact, for a pure public good we must all necessarily consume the same quantity, namely, whatever quantity is supplied in the aggregate.
We may of course get different amounts of utility if our tastes differ, but we all consume the same quantity.
The key aspects of public goods are (1) that it is technically possible for one person to consume without reducing the amount available for someone else, and (2) the impossibility of excluding anyone from consumption except at a prohibitive cost.
A football match could be watched by a lot of people, especially if it is televised, without reducing the quantity consumed by any individual; but exclusion is possible -- the ground holds only so many, and the club can refuse to allow the game to be televised.
The interesting issues in economics arise when, as with national defence, exclusion of certain individuals from consumption is effectively impossible.
Free-Riders
In the last chapter we introduced the free-rider problem when discussing why bribes and compensation for externalities might not occur.
Public goods are likely to be especially vulnerable to the free-rider problem if they are supplied by the private sector.
Since you get the same quantity of national defence as everyone else, whether or not you pay for it, it would never be in your interest to purchase national defence in a free private market.
Everybody else would adopt similar reasoning, and no defence would be demanded even if we all wanted defence.
Public goods are like a very strong externality.
If you buy defence everyone else gets the benefits.
Since marginal private and social benefits diverge, private markets will not produce the socially efficient quantity.
There is a case for government intervention to make sure marginal social cost and marginal social benefit are equated.
The Marginal Social Benefit
Suppose the public good is the purity of the public water supply.
The more infected the water, the more likely it is that everyone will be hit by an epidemic of cholera or some other disease.
Figure 16-2 supposes there are two people.
The first person's demand curve for water purity is D 1 D 1 .
Each point on the demand curve shows what the individual would pay for the last unit of purer water.
It shows the marginal benefit  to the individual.
D 2 D 2 shows the marginal benefit of purer water to the second individual.
The curve DD gives the marginal social benefit of purer water.
At each output level for the public good, we vertically sum the marginal benefit of each individual to get the social marginal benefit.
Thus at the output Q the social marginal benefit is .
We sum vertically at a given quantity because everyone consumes the same quantity of a public good by definition.
Figure 16-2 A PURE PUBLIC GOOD.
D 1 D 1 and D 2 D 2 are the separate demand curves of two individuals and show the marginal private benefit of the last unit of the public good to each individual.
What is the social marginal benefit of the last unit to the group as a whole?
Since both individuals consume whatever quantity of the good is produced, we must add up vertically the price each is prepared to pay for the last unit.
At the output Q the marginal social benefit is thus .
The curve DD showed the marginal social benefit and is obtained by vertically adding the demand curves of the two individuals.
If MC is the private and social marginal cost of producing the public good the socially efficient output is Q *; at which social marginal cost and social marginal benefit are equal.
Figure 16-2 also shows the marginal cost of producing the public good.
If there are no production externalities the marginal private cost and the marginal social cost of production will coincide.
The socially efficient level of production of the public good is at Q *;, where the marginal social benefit equals the marginal social cost.
What would happen if the good were privately produced and marketed?
Person 1 might pay a price P 1 to have a quantity Q produced by a competitive supplier pricing at marginal cost.
At the output Q the price P 1 just equals the marginal private benefit which person 1 derives from.
the last unit of the public good.
Would person 2 be prepared to pay to have the output of the public good increased beyond Q?
The answer is, 'No'.
Because it is a public good, person 2 cannot be excluded from consuming the output Q which person 1 has commissioned.
But at the output Q, person 2's marginal private benefit is only P 2 , which is less than the current price P 1 .
Person 2 would certainly not pay the higher price necessary to induce a competitive supplier to expand production beyond the output Q. Person 2 is thus a free-rider enjoying person 1's purchase Q. And the total quantity privately produced and consumed in a competitive market lies below the socially efficient quantity Q *;
Revelation of Preferences
By constructing the marginal social benefit curve DD, the government can decide how much of the public good it is socially efficient to produce.
But how does the government find out the individual demand curves that must be vertically added to get DD?
If people's payments for the good are related to their individual demand curves everyone has an incentive to lie because of the free-rider problem.
People will understate how much they value the good in order to reduce their own payments, just as in a private market.
Conversely, if payments are divorced from the question of how much people would like, people will overstate their private valuations.
We are all for safer streets if we do not have to contribute to the cost.
In practice, democracies try to resolve this problem through elections of governments.
Different parties offer different quantities of public goods together with a statement of how the money will be raised through the tax system.
By asking the question, 'How much would you like, given that everyone will be charged for the cost of providing public goods?' society can come  closer to providing the efficient quantities of public goods.
However, since there are only a few parties competing in the election and many different aspects of government on which they are offering a position, this can be only a very crude way to elicit people's view of how much of any particular public good should be provided.
Government Production
The economist's definition of public goods relies solely on the fact that everyone consumes the same quantity.
We have seen that the free-rider problem implies that private markets will not produce the socially efficient level and that there is a case for government intervention on efficiency grounds.
But this merely says that the government must determine how much is produced.
It does not imply that the government must produce the goods itself.
Public goods are not necessarily the goods the government happens to produce.
For example, in the UK, as in most countries, national defence is a public good and is also produced largely within the public or government sector.
We have few private armies.
On the other hand, street-sweeping, though a public good, can be subcontracted to private producers, even if local government determines its quantity and pays for it out of local tax revenue.
Conversely, state hospitals in the National Health Service involve public sector production of private goods.
One person's hip replacement operation certainly prevents the busy surgeon from doing something else at the same time.
In the next chapter we examine why the public sector may wish to produce private goods.
Whether public goods need be produced by the public sector depends not on their consumption characteristics, on which our definition of public good relies, but on their production characteristics.
There is nothing special about street-sweeping, and it can as easily be produced by the public or the private sector.
In contrast, armies and navies rely on discipline and secrecy.
Generals and admirals may believe, and society may agree, that offences against these regulations should receive unusual penalties which would not be generally sanctioned in private firms.
Few people believe that insubordination is an important offence for street-sweepers and should be punished by incarceration or even death.
Hence it may make more sense for soldiers to be in the public sector than street-sweepers.
Where such considerations do not arise, for example in the production of uniforms, it is more likely that the production of defence goods will take place in private firms.
Transfer Payments and Income Redistribution
The government spends money on public goods because there is a market failure when public goods are left entirely to private markets.
Thus the motivation for this type of intervention is social efficiency.
In contrast, government spending on transfer payments is primarily concerned with equity and income redistribution.
By spending money on the unemployed, the old, and the poor (who in the UK are entitled to supplementary benefit if their total income from whatever source falls below a certain minimum level), the government seeks to ensure that the distribution of income and welfare that a totally free market economy would otherwise have produced is at least truncated: there is a minimum standard of living below which no citizen should fall.
The specification of this standard is of course pure value judgement.
Where does the money come from to pay the poor and the disadvantaged?
Primarily from those who can most afford to pay.
Table 16-2 shows that the income tax system in the UK is progressive.
Increasing marginal tax rates on income ensure that each individual's average tax rate, the proportion of total income paid in taxes, increases with income.
Taken as a whole, the tax and transfer system takes money from the rich and gives to the poor.
The poor receive not merely the direct financial transfer in the form of transfer payments such as supplementary benefit, but also the consumption of public goods that have been paid for by income taxes raised from the rich.
As we pointed out in the last chapter, not only is the amount of redistribution to be undertaken by the government a pure value judgement on which different individuals and different political parties will disagree, but there is an inevitable trade-off between the competing objectives of efficiency and equity.
To undertake more redistribution the government will have to increase tax rates, thereby driving a  larger wedge between the price paid by the purchaser and the price received by the seller of the good or service.
Since the price system achieves Pareto efficiency by inducing each individual to equate marginal cost or marginal benefit to the price received or paid, and hence to one another, taxes that imply that buyers and sellers face different prices ensure that the marginal cost to a seller no longer equals the marginal benefit to a buyer.
Taxes are generally distortionary and tend to reduce efficiency.
In Table 16-2 we saw that the Thatcher government succeeded in reducing marginal tax rates, especially for the very rich.
Opponents of the government argued that the objective as well as the consequence of the legislation was to increase the after-tax incomes of the rich at the expense of the poor.
The government argued that reducing distortions in the labour market by cutting income tax would lead to efficiency gains that would far outweigh the valuation that society should put on a more equal income distribution.
If society's resource could be used to make more output, even the poor might be better off in the long run.
Merit Goods and Bads
Merit goods (bads) are goods that society thinks everyone ought to have (ought not to have) regardless of whether they are wanted by each individual.
Examples of merit goods are education and health.
Merit bads are products such as cigarettes.
Since society places a different value on these goods from the value placed on them by the individual, it follows that individual choice within a free market economy will lead to a different allocation from the allocation that society wishes to see.
There are two distinct reasons for designating merit goods.
The first is a version of the externality argument we examined in the previous chapter.
If more education raises the productivity not merely of an individual worker but of all other workers with whom this worker co-operates, there is a production externality that the individual does not take into account in choosing how much education to purchase.
If individuals demand too little education, society should encourage the provision of education.
Free schooling to ensure a minimum level of education, communication, and social interaction might be one way to achieve this.
Conversely, if people take account of the costs to themselves but not the burden on the National Health Service in deciding whether or not to smoke and damage their health, society may regard smoking as a merit bad that should be discouraged.
We shall shortly see how the tax system, in this case a tax on cigarettes, may be used to offset externalities that individuals fail to take into account.
The second aspect of merit goods is where society believes that individuals are no longer acting in their own best interests.
Addiction to drugs, tobacco, or gambling are obvious examples.
Economists rarely subscribe to the value judgement of whole-scale paternalism.
The function of government intervention is less to tell people what they ought to like than to allow them better to achieve what they already like.
However, the government will sometimes have more information or be in a better position to take a decision.
Much as some people hate going to school, they will frequently be glad afterwards that they were made to do so.
Thus the government may spend money on compulsory education or compulsory vaccination because is recognizes that, left to their own decisions, individuals will act in a way they will subsequently regret.
16-3 The Principles of Taxation
This section is in three parts.
First we consider the different kinds of taxes through which the government can raise revenue.
Then we consider again the equity implications of taxation.
Finally, we examine the efficiency implications of taxation.
Variety of Taxes
Governments can raise tax revenue only if they can identify the activities on which the tax rates apply.
Before sophisticated records of income or sales were ever kept, governments raised most of their revenue through customs duties and road tolls, the two places where transactions could be easily monitored.
Income tax in peacetime was not introduced in the UK until the 1840s, and VAT -- a general tax on goods and services (with a few specified  exemptions such as good and children's clothing)-- was not introduced until the 1970s.
We briefly outline the main taxes shown in Table 16-5, grouped under three headings: taxes on income, or direct taxes; taxes on expenditure, or indirect taxes; and taxes on assets, or wealth taxes.
Direct Taxes
Individuals pay income tax on earnings from labour, rents, dividends, and interest.
In Chapter 14 we saw that the return on an asset is not just the dividend or interest payment but also the capital gain.
Although many economists would argue that capital gains, as for example when ICI shares are purchased for 2 and subsequently sold for 3, are as much income as the dividend component of the return on an asset, in practice the Inland Revenue assesses and taxes capital gains separately.
National insurance contributions by individuals are also a form of direct personal taxation.
Companies pay corporation tax calculated on their taxable profits after allowance for interest payments and depreciation.
They also make a national insurance contribution on behalf of their employees.
Indirect Taxes
Indirect taxes are taxes levied on expenditure on goods and services.
The most important source of indirect tax revenue is value added tax (VAT), which is effectively a retail sales tax.
Whereas a sales tax is collected only at the point of final sale to the consumer, VAT is collected at different stages of the production process.
Suppose a firm mines iron ore and converts it into 200 worth of high-grade steel, which is then sold to a car producer.
The car producer converts the steel into a car costing 3200.
A simple sales tax levied at 15 per cent would raise the cost to the consumer to 3200 + 480 (15 per cent of 3200) or 3680.
In contrast, VAT works as follows.
The steel firm has a value added or net output of 200 on which it pays 15 per cent or 30 in tax.
Passing the tax on to the car producer, the steel is sold for 230.
The car producer has a value added or net output of 3000 and pays 15 per cent or 450 in tax.
Since the car firm paid 230 for the steel, the final price to the consumer is 230 + 3000 + 450 = 3680.
As far as the consumer is concerned this is just the same as a 15 per cent sales tax.
This example makes it seem that the consumer price is raised by the full amount of the tax.
But a higher consumer price will reduce the quantity demanded.
In turn this will move producers back down their marginal cost curves and alter the net-of-tax price producers require.
Later in this section we show how to analyse these induced effects to determine how the burden of the tax is ultimately divided between producers and consumers.
Revenue from VAT is supplemented by other indirect taxes including special duties on tobacco and alcohol, licence fees for motor cars and televisions, and customs duties on imports.
Wealth Taxes
In the UK there used to be two taxes that tax wealth per se rather than the income that is derived from wealth.
The first was the tax on property values, which formed the main source of revenue for local government.
That is the tax that was replaced by the poll tax (a simple flat-rate tax per person) in the reform of UK local government finance during 1988-90.
The second, which still exists, is capital transfer tax, which applies to transfers of wealth between individuals, whether as gifts during life or as inheritances after death.
How does the UK tax structure compare with that in other countries?
Table 16-5 shows data for several advanced countries in 1983.
The most notable feature of the UK tax system appears to be its low reliance on social security taxes for state pension and unemployment provisions.
Table 16-5 also suggests that the UK relies quite heavily on indirect taxes rather than direct taxes.
Tax revenue is necessary to pay for government expenditure.
We now assess the UK tax system against our two welfare criteria, equity and efficiency.
How To Tax Fairly
In the last chapter we introduced two notions of equity: horizontal equity, or the equal treatment of equals, and vertical equity, the redistribution from the 'haves' to the 'have-nots'.
In Table 16-2 we showed that income tax is progressive.
In taking proportionately more from the rich than from the less well off, income  tax reflects the principle of ability to pay.
There are two reasons society might think it fair that the rich should pay more.
First, society may wish to take from the rich in order to give to the poor.
Second, if money has to be raised to pay for public goods, society may wish to avoid taxing those whose incomes are already low.
The principle of ability to pay thus reflects a concern about vertical equity.
Table 16-5 Sources of tax revenue in 1983 (percentage of total taxes)
A second principle is sometimes applied in discussing the extent to which unequal people should be treated unequally.
The benefits principle argues that people who receive more than their share of public spending should pay more than their share of tax revenues.
Car users should pay more towards public roads than people without a car should pay.
And to some extent they do.
Car users pay heavy duties on petrol and must pay licence fees for running a car.
However, the benefits principle often conflicts directly with the principle of ability to pay.
If people who are most vulnerable to unemployment must pay the highest contributions to the government unemployment insurance scheme, it becomes very difficult to achieve a significant redistribution of income, wealth, or welfare.
If the main objective is vertical equity, the ability to pay.
principle must usually take precedence.
Although Table 16-2 shows that the income tax system in the UK is progressive, it is the entire structure of taxes, transfers, and public spending that we must examine before we can judge how much the government is effectively redistributing from the rich to the poor.
We have already mentioned two factors that make the entire structure more progressive than an examination of income tax alone would suggest.
First, transfer payments actually give money out to the poor.
The old get pensions, the unemployed get unemployment benefit, and, as a final safety net, anyone whose income from whatever source falls below a certain minimum is entitled to supplementary benefit.
Second, the state provides public goods that can be consumed by the poor, even if they have not paid any taxes to finance these goods.
In addition to pure public goods, such as defence, the state also makes free provision of certain goods, such as parks and swimming pools, which have part of the characteristics of a pure public good.
Although the whole population cannot squeeze into Hyde Park, quite a few people can enjoy its amenities without spoiling the enjoyment of others.
And since the rich tend to sit in their own gardens, public parks help redistribute enjoyment towards the poor.
As against these progressive elements of the tax, transfer, and spending structure, it should be noted that there are some important regressive elements that take proportionately more from the poor.
Beer and tobacco taxes are huge revenue-earners for the government.
Yet the poor spend a much higher proportion of their income -- in some cases even a larger absolute amount -- on these goods than do the rich.
Such taxes reduce the effectiveness of the tax, transfer and spending structure in redistributing from the rich to the poor.
Tax Incidence
The incidence of a tax measures the final tax burden on different people once we have allowed for the indirect as well as the direct effects of the tax.
The ultimate effect of a tax  can be very different from its apparent effect.
Thus to get a really good idea of the extent to which taxes (or subsidies) alter people's spending power and welfare, we need to examine the issue of tax incidence in more detail.
Figure 16-3 A TAX ON WAGES.
In the absence of a tax, free market equilibrium is at E and the wage is W. A wage tax makes the gross wage paid by firms higher than the net wage received by workers.
Measuring gross wages on the vertical axis, the demand curve DD is unaltered by the imposition of the tax.
Firms still choose the quantity of labour demanded to equate the gross wage to the marginal value product of labour.
SS continues to show labour supply, but as a function of the net wage.
To get labour supply in terms of the gross wage we must draw the new supply curve SS ';.
At each quantity of hours, SS '; lies vertically above SS by a distance reflecting the tax on earnings from the last hour worked.
The new equilibrium is at E ';.
The gross wage paid by firms is W '; but the net wage received by workers is W';.
The vertical distance  shows the amount of the tax.
Whether the government collects the tax revenue entirely from firms or entirely from workers, the incidence of the tax is the same.
It falls partly on firms, who must pay a higher gross wage W', and partly on workers, who receive the lower net wage W';.
The area of pure waste  will shortly be discussed in the text.
Figure 16-3 shows the market for labour.
DD is the market demand curve for labour and SS is the supply curve for labour, which we assume slopes upwards.
Thus a higher wage rate increases the supply of hours of work, but reduces the demand for hours of work.
In the absence of an income tax (a tax on wages), the labour market will be in equilibrium at point E.
Now suppose the government imposes an income tax.
If we measure the gross wage on the vertical axis, the demand curve DD is unaltered since it is the comparison of the gross wage with the marginal value product of labour that determines the quantity of labour demanded by firms.
Workers' preferences or attitudes are also unchanged, but it is the wage net of tax that workers compare with the marginal value of their leisure in deciding how much labour to supply.
Thus, although SS continues to show the labour supply curve in terms of the after-tax wage, we must draw in the higher schedule SS '; to show the supply of labour in terms of the gross or pre-tax wage.
The vertical distance between SS '; and SS measures the amount of tax being paid on earnings from the last hour's work.
Since DD and SS '; now show the behaviour of firms and workers at any gross wage, the new equilibrium will be at the point E ';.
The new equilibrium gross wage is W '; at which firms demand a quantity of hours L ';.
The vertical distance between A '; and E '; measures the tax being paid on earnings from the last hour of work.
Thus the after-tax wage is W'; at which workers are happy to supply a quantity of hours L ';.
Relative to the original equilibrium wage W, the imposition of the tax on wages has raised the pre-tax wage to W ';, but lowered the after-tax wage to W';.
It has raised the wage that firms must pay but lowered the take-home wage for workers.
The incidence of the tax has fallen on both firms and workers even though, as a matter of administrative convenience, the tax may be collected by the government directly from workers.
The lesson from Figure 16-3 is an important one: the incidence or burden of a tax cannot be established by looking at who actually hands  over the money to the government.
Taxes usually alter equilibrium prices and quantities and these induced effects must also be taken into account.
However, we can draw one very general conclusion.
The more inelastic the supply curve and the more elastic the demand curve, the more the final incidence will fall on the seller rather than the purchaser.
Figure 16-4 TAXING A FACTOR IN INELASTIC SUPPLY.
If the supply curve SS is vertical, a tax A '; E per unit leaves the quantity L unaffected.
Since the demand curve DD is unaltered, the tax has no effect on the pre-tax wage rate.
The full incidence of the tax falls on workers whose after-tax wage is reduced by the full amount of the tax.
Figure 16-4 depicts the extreme case in which the supply curve is completely inelastic.
In the absence of a tax, equilibrium is at E and the wage is W. Since the vertical supply curve SS implies that a fixed quantity of hours L will be supplied whatever the after-tax wage, the imposition of a tax on wages leads to a new equilibrium at A ';.
Only if the pre-tax wage is unchanged will firms demand the quantity L that is supplied.
Hence after-tax wages fall by the full amount of the tax.
The entire incidence falls on the workers.
To check you have grasped the idea of incidence, try drawing for yourself a market with a relatively elastic supply curve and a relatively inelastic demand curve.
Show that the incidence of a tax will now fall mainly on the purchaser.
Taxation, Efficiency, and Waste
So far, we have been considering the equity implications of a tax.
But we must also think about the efficiency implications of a tax.
We can use Figure 16-3 again.
Before the tax is imposed, labour market equilibrium is at E. The wage W measures both the marginal social benefit of the last hour of work and its marginal social cost.
The demand curve DD tells us the marginal value product of labour, the extra benefit society could have from extra goods produced.
The supply curve SS tells us the marginal value of the leisure being sacrificed in order to work another hour, the marginal social cost of extra work.
Before the tax is imposed, the labour market is in equilibrium at E. Since marginal social cost and benefit are equal, this initial position is socially efficient.
When the tax is imposed, the new equilibrium is at E ';.
We have already discussed the incidence of the tax on firms and workers.
The tax A '; E '; increases the wage to firms to W '; but reduces the after-tax wage for workers to W';.
But there is an additional tax burden or deadweight loss that is pure waste.
It is the triangle A '; E '; E. By reducing the quantity of hours from L to L ';, the tax causes society to stop using hours on which the marginal social benefit, the height of the demand curve DD, exceeds the marginal social cost, the height of the supply curve SS.
By driving a wedge between the wage firms pay and the wage workers receive, the tax induces a distortion which destroys the efficiency of free market equilibrium.
Must Taxes Be Distortionary?
Government need tax revenue to pay for public   goods and to make transfer payments to the poor.
Must taxes create distortions and lead to the waste or inefficiency which Figure 16-3 suggests?
Figure 16-4 showed what happens when a wage tax is levied but the supply of labour is completely inelastic.
Although the tax reduces the take-home pay of workers, there is no change in the gross wage or the equilibrium quantity of hours.
Since the quantity is unchanged, there is no distortionary triangle or deadweight burden.
The equilibrium quantity remains the socially efficient quantity.
We can make this into a general principle.
When either the supply or the demand curve for a good or service is very inelastic, the imposition of a tax will lead only to a small change in quantity.
Hence the deadweight burden triangle must be small.
Given that the government must raise some tax revenue, the smallest amount of total waste will be achieved when the goods that are most inelastic in supply or demand are taxed most heavily.
This principle finds practical expression in the UK tax system.
The three most heavily taxed commodities are alcohol, tobacco, and the oil being extracted from the North Sea.
For these commodities tax rates range from 50 to 90 per cent.
Alcohol and tobacco are generally assumed to be products with a very inelastic demand.
North Sea oil is in inelastic supply.
Having spent large amounts of money on exploration and drilling, oil companies are quite keen to recoup their investment, even if the government is taking a big slice off the top.
So far, we have discussed the taxes that would do least harm to the allocative efficiency of the economy.
Sometimes the government has the opportunity to levy taxes which will actually improve efficiency and reduce waste.
The most important example is when externalities exist.
Figure 16-5 TAXES TO OFFSET EXTERNALITIES.
Given private demand DD and supply SS free market equilibrium is at E with a quantity a With a negative consumption externality, the social marginal benefit is DD '; lying below DD.
E *; is the socially efficient point at which output is Q *;.
At this output the marginal externality is E *; F. By levying a tax of exactly E *; F per unit, the government can shift the private supply curve from SS to SS '; leading to a new equilibrium at F at which the socially efficient quantity Q *; is produced and the deadweight burden of the externality E *; HE is eliminated.
Cigarette smokers pollute the air for other people but take no account of this in deciding how much to smoke.
They give rise to a harmful consumption externality.
Figure 16-5 shows the supply curve SS of cigarette producers.
Since there are no production externalities, this marginal private cost curve is also the marginal social cost curve.
DD is the private demand curve showing the marginal benefit of cigarettes to smokers.
Because there is a harmful consumption externality, the marginal social benefit DD '; of cigarette consumption is lower than DD
In the absence of a tax, free market equilibrium is at E, but there is over-consumption of cigarettes.
The socially efficient quantity is Q *; since marginal social cost and marginal social benefit are equated at E *;.
Suppose the government levies a tax, equal to the vertical distance E *; F, on each packet of cigarettes.
With the tax-inclusive price on the vertical axis, the demand curve DD is unaffected, but the  supply curve shifts up to SS ';.
Each point on SS '; then allows producers to receive the corresponding net-of-tax price on SS.
After the tax is introduced, equilibrium is at the point E. The socially efficient quantity Q *; is produced and consumed.
Consumers pay the price P '; and producers receive the price P'; after tax has been paid at the rate E *; F per unit.
Only the particular tax rate E *; F per unit will guide the free market to the socially efficient allocation.
A lower tax rate (including a zero tax rate) leads to too much consumption and production of cigarettes.
A higher tax rate than E *; F will move consumers further up their demand curve and lead to under-consumption and under-production.
Why must the tax rate be exactly E *; F if the efficient quantity is to be achieved?
Because this is exactly the amount of the externality on the last unit when the efficient quantity Q *; is produced.
By levying a tax at precisely this rate, the government raises the price to the consumer above the price to the producer by the amount of the externality.
Consumers are induced to behave as if they took account of the externality, though in fact they take account only of the after-tax price.
Whenever consumption or production externalities induce distortions in the free market equilibrium allocation, the government can improve efficiency and reduce waste by levying taxes.
The fact that alcohol and tobacco have farmful externalities provides another reason for taxing them heavily.
16-4 Taxation and Supply-side Economics
We began the chapter by noting that many Western countries have become disenchanted with the extent of government involvement in the economy.
In part, it was felt that governments were spending too much.
Resources used to produce goods and services for the government cannot be used to make goods in the private sector.
We shall have more to say about this in the next chapter.
However, the major objection to high levels of government expenditure seems to have been associated with the need for correspondingly high levels of revenue collection.
Table 16-3 reminds us that some government expenditure is financed by borrowing.
In the UK this is known as the public sector borrowing requirement (PSBR).
In Part 4 we shall examine the argument that a high PSBR leads to high inflation, high interest rates, or both.
For the moment we ignore government borrowing and consider the argument that high taxation to pay for high levels of public spending necessarily strangles the economy.
We have already seen that in order to pay for public goods and redistribution the government must raise tax revenues, which typically introduces allocative distortions and leads to a dead-weight burden.
Suppose the government adopts a less ambitious spending programme and is therefore able to reduce income tax rates.
What will be the consequences?
First, by spending less on goods and services, the government will free some resources which can now be used by the private sector.
If it were true that the private sector uses resources more productively than the public sector, the transfer of resources might directly produce more output.
The total supply of goods and services would rise.
Whether or not the private sector does use resources more productively on average than the government remains a contentious issue.
What about the effects of lower income tax rates?
Figure 16-3 suggests that income taxes introduce a distortion that leads to a level of work that is socially inefficient.
With lower taxes and a smaller distortion there would be a lower dead-weight burden.
Since the distortion leads to a level of work that is lower than the socially efficient amount, cutting income taxes would also increase the amount of work done in the economy.
How large could this effect be?
It all depends on the elasticity of labour supply.
The more inelastic the labour supply, the lower is the distortion introduced by any particular income tax rate.
When labour supply is completely inelastic as in Figure 16-4, income tax does not induce any distortion at all and there will be no allocative gain in reducing income tax rates.
In Chapter 11 we showed that an increase in the after-tax wage (as for example when income tax rates are cut) will have a substitution effect, tending to make people work longer hours, but an income effect, tending to make them work fewer hours.
With higher after-tax wages it takes fewer hours to earn any given target  income.
Hence we argued that, for people already in work, changes in after-tax rates have only a small effect on hours of work supplied.
Then we showed that increasing the after-tax wage would encourage labour force participation by those not currently in the labour force.
Hence, taking hours and participation together, the supply curve of labour input (hours times people) will not be completely vertical.
Cutting income tax will increase the supply of labour input, chiefly by attracting new workers into the labour force.
But the total effect on labour supply might not be as large as some proponents of tax cuts believe.
In contrast, the tax cut enthusiasts believe that income tax is a major distortion and labour supply is very elastic.
The socially efficient quantity of labour input would then be much larger than the equilibrium level under current tax rates.
One illustration of this view is the famous Laffer curve, named after Professor Arthur Laffer, one of President Reagan's most influential economic advisers.
Suppose, for example, that all government tax revenue was raised through income tax.
Figure 16-6 shows that with a zero tax rate the government would raise zero revenue.
At the opposite extreme, with a 100 per cent income tax rate, there would be no point working and again tax revenue would be zero.
Beginning from a zero rate, a small increase in the tax rate will yield some tax revenue.
Initially revenue rises with the tax rate, but beyond the tax rate t* higher taxes have major disincentive effects on work effort and revenue starts to fall.
Figure 16-6 THE LAFFER CURVE.
The Laffer curve shows the relationship between tax rates and tax revenue.
Moderate tax rates raise some revenue.
Beyond t *;, higher tax rates reduce revenue because disincentive effects greatly reduce the supply of the quantity being taxed.
At 100 per cent tax rate, supply and revenue will be zero again.
Professor Laffer's idea was that many 'big government-big tax' countries are now at tax rates above t *;.
If so, tax cuts would be the miracle cure.
Everybody likes a tax cut but the government would actually raise more revenue by cutting taxes.
By reducing the tax distortion and increasing the amount of work a lot, lower taxes would be more than compensated by the extra work and incomes to which the tax rates were applied.
It is not the shape of the Laffer curve that is in dispute.
Rather, what many professional economists in the UK, the United States, and other Western countries have disputed is that these economies do in fact have tax rates above t *;.
Most economists' reading of the empirical evidence is that our economies lie to the left of t *;.
Figure 16-3 implies that cutting income tax rates may eliminate some of the deadweight burden of distortionary taxation, but governments should probably expect their tax revenue to decline if such policies are put into effect.
Hence, if governments do wish to reduce tax rates without adding to government borrowing it is essential that they reduce their spending.
16-5 Local Government
Thus far we have been chiefly interested in the principles of central government.
In this section we examine the economics of local government.
Local government expenditure may cover a variety of things, from sweeping the streets to providing local schooling.
In turn this must be  financed through taxes.
Some of these taxes will be local, but some will come from central government revenue raised through the national tax system.
Finally, local government is responsible for some types of regulation, for example land use or zoning laws.
Economic Principles
Why don't we make central government responsible for everything?
Two arguments are usually used.
First, diversity matters.
People are different and they don't want to be treated the same.
Civic pride is necessarily local.
Second, people feel that central government is remote from their particular needs.
Even if central government wished to pay attention to local considerations, it would find it hard to do so efficiently.
We turn now to two important models of local government.
The first is the Tiebout model.
This model emphasizes diversity.
Some people want a lot of local expenditure on public services and are prepared to pay high local taxes; others want to pay lower local taxes even though this means lower public services.
If all local governments are the same, everyone will be unhappy with the compromise.
The Tiebout model is sometimes called the invisible foot: people will cluster together in the area providing the package of spending and taxes they want.
The invisible foot brings about an efficient allocation of resources through competition between local governments.
In practice, the invisible foot is sometimes a very imperfect incentive structure.
First, it may be hard to move between local authorities.
For example, being born in a neighbourhood may entitle you to a higher place in the queue for housing provided by that local authority.
Second, if much of local authority revenue comes from central government, the levels of spending and taxes may be insensitive to the wishes of local residents.
We discuss this more fully for the UK in Box 16-1.
Earlier in the chapter, we stressed the distinction between efficiency and equity.
Even if the invisible foot led to efficiency, it might also lead to inequity.
The rich are likely to cluster together in suburbs.
Then they pass zoning laws specifying a minimum size for a house and its garden.
This makes it impossible for the poor to move to that neighbourhood.
By forming an exclusive club, the rich have ensured that their tax contributions do not have to go to supporting the poor in their neighbourhood.
And the poor get stuck with one another in inner-city areas whose governments face the biggest social needs but the smallest local tax base.
The Tiebout model assumes that residents mainly consume the public services provided by their own local authority.
But when each unit of local government is responsible for a small geographical area, this may be a poor assumption.
If an inner-city supplies free art galleries, financed out of taxes on inner-city inhabitants, the rich still come in from the suburbs to make use of these facilities.
Conversely, inner-city inhabitants spend their Sundays enjoying countryside facilities supported by taxes raised out of town.
In these cases, provision of public services in one area confers a beneficial externality on neighbouring areas.
Economic theory suggests the right answer to this problem.
Unless the externality can be priced (charging suburban users but not city dwellers for entry to subsidized galleries and opera houses), the most efficient solution is to widen the geographical area of each local government until it includes most of the people who will use the public services it provides.
Thus, for example, it may make sense to have an integrated commuter rail service and inner-city subway, and to subsidize it to prevent people driving through congested streets; but only a local government embracing both the suburbs and the inner city is likely to get close to the efficient policy.
These two theories of local government pull in opposite directions, and the right answer is likely to lie somewhere in between.
The assumptions of the Tiebout model favour a lot of small local government jurisdictions to maximize choice and competition between areas.
The model emphasizing externalities across areas suggests larger jurisdictions to 'internalize' externalities that would otherwise occur.
Summary
1
In industrialized economies, government revenues come mainly from direct taxes on   personal incomes and company profits, indirect taxes on purchases of goods and services, and contributions to state-run social security schemes.
Government spending comprises spending on goods and services and transfer payments.
2
Government intervention in a market economy should be assessed against the criteria of distributional equity and allocative efficiency.
A progressively tax and transfer system takes most from the rich and give most to the poor.
The UK tax and transfer system is mildly progressive.
The less well off do receive transfer payments and the rich face the highest rates of income tax.
Although some necessities, notably food, are exempt from VAT, other goods intensively consumed by the poor, notably cigarettes and alcohol, are heavily taxed.
3
Externalities and public goods are classic cases of market failure where intervention may improve allocative efficiency.
By taxing or subsidizing goods that involve externalities, the government can induce the private sector to behave as if it takes account of the externality, thus eliminating the deadweight burden arising from the misallocation induced by the externality distortion.
4
A pure public good is a good for which one person's consumption does not reduce the quantity available for consumption by others.
Together with the impossibility of effectively excluding people from consuming it, this implies that all individuals consume the same quantity, although they may attach different utility to this consumption if their tastes differ.
5
A free market will undersupply a public good because of the free-rider problem.
Individuals need not offer to pay for a good that they can consume if others pay for it.
The socially efficient quantity of a public good equates the marginal social cost of production to the sum of the marginal private benefits over all people at this output level.
Diagrammatically, this implies that individual demand curves are vertically added to get the social demand or marginal benefit curve.
6
Except for taxes designed to offset externalities, taxes are generally distortionary.
By driving a wedge between the selling price and the purchase price, they prevent the price system achieving the equality of marginal costs and marginal benefits.
The amount of the deadweight burden is higher the higher is the marginal tax rate and the size of the wedge, but it also depends on supply and demand elasticities for the taxed commodity or activity.
The more inelastic are supply and demand, the less the tax will change equilibrium quantity and the smaller will be the deadweight burden triangle.
7
The incidence of tax describes who ultimately pays the tax.
The more inelastic is demand relative to supply, the more a tax will fall on purchasers as opposed to sellers.
8
Rising tax rates initially increase tax revenue but eventually lead to such large falls in the equilibrium quantity of the taxed commodity or activity that revenue starts to fall again.
Cutting tax rates will usually reduce the deadweight tax burden but might increase revenue if taxes had initially been sufficiently high.
Most Western economies do not appear to have reached this position.
If governments wish to reduce the deadweight tax burden and balance spending and revenue, it is necessary to reduce government spending in order to cut taxes.
Key Terms
Problems
1
Which of the following are public goods?(a) the fire brigade;(b) clean streets;(c) refuse collection;(d) cable television;(e) social toleration;(f) the postal service.
Explain and discuss alternative ways of providing these goods or services.
2
Why does society try to ensure that every child receives an education?
Discuss the different ways this could be done and give reasons for preferring one method of providing such an education.
3
How would you apply the principles of horizontal and vertical equity in deciding how much to tax two people, each capable of doing the same work, but one of whom chooses to devote more time to sun-bathing and therefore has a lower income?
4
Whereas a progressive tax takes proportionately more of a rich person's income, a regressive tax takes proportionately more of a poor person's income.
Classify the following taxes as progressive or regressive. (a) 10 per cent tax on all luxury goods;(b) taxes in proportion to the value of owner-occupied houses;(c) taxes on beer;(d) taxes on champagne.
5
There is a flat-rate 30 per cent income tax on all income over 2000.
Calculate the average tax rate (tax paid divided by income) at income levels of 5000, 10000, and 50000.
Is the tax progressive?
Is it more or less progressive if the exemption is raised from 2000 to 5000?
6
(a) Suppose labour supply is completely inelastic.
Show why there is no deadweight burden if wages are taxed.
Who bears the incidence of the tax?(b) Now suppose labour supply is quite elastic.
Show the area that is the deadweight burden of the tax.
How  much of the tax is ultimately borne by firms and how much by workers?(c) For any given supply elasticity show that firms bear more of the tax the more inelastic is the demand for labour.
7
Common Fallacies Show why the following statements are incorrect:(a) The only reason for taxation is to provide for shirkers by penalizing people who do an honest day's work.
(b) It is obvious that the government is supplying too many goods and services.
In a free enterprise economy it would be profitable only to supply a fraction of the amount to which the government is currently committed.
(c) The government spends all its revenue.
Taxes cannot be a burden on society as a whole.
BOX 16-1 POLL POSITION: UK LOCAL GOVERNMENT REFORMS
Until the late 1980s, local government in the UK was financed from three sources.
First, domestic households paid rates, a property tax assessed on rateable values or hypothetical house prices.
Second, local firms paid business rates on their property.
The third, and much the largest source, was from central government through the rate support grant, which took some account of the needs of the area.
The Thatcher government believed this system has a bias towards overspending by local authorities.
Many poor households were exempt from domestic rates but still had a vote in the local election; they had an incentive to vote for high spending programmes.
Firms paid rates but had no vote at all in local elections.
Only about 20 per cent of local government revenue was being raised from households actually required to pay rates.
A subsidiary problem was a reluctance to raise rateable values, the basis of the property tax, in line with market prices.
Mrs Thatcher believed that greater local democracy would lead to a voter rebellion against high spending local authorities, forcing them to be less ambitious and more efficient.
The reform of local government, which took effect in Scotland in 1989 and the rest of the UK in 1990, made three changes.
First, education in state schools (the largest component of local expenditure) was moved from local to central government.
Second, local business rates were replaced by a uniform business rate, a single tax rate nationwide.
Simultaneously, the rateable values for business were moved more into line with market values, leading to enormous (and justified!) increases in the tax assessment for firms in the south-east, where property prices had risen substantially since the previous assessment of rateable values years earlier.
Whereas previously each local authority had access to its own business rate revenue, now the central government collected all this revenue and redistributed it to local authorities in proportion to the local population.
The consequence was to redistribute money from the rich south, where property prices were high, to the poorer north.
Third, domestic rates were replaced by the community charge or 'poll tax', a flat-rate tax per head with some partial relief for the poorest households.
The logic of the reforms was to move local government closer to the Tiebout model, making it more transparent that 'you get what you pay for'.
Greater transparency is almost certainly a good idea.
In the text, we discuss the pros and cons of the Tiebout principle itself.
Whatever its merits, this principle should be clearly separated from the government decision about how progressive the community charge should be.
It was violation of the ability-to-pay principle that caused much of the public hostility to the poll tax.
If future policy decides to abandon the flat-rate community charge, this reopens the question of how a more progressive local tax should be designed.
This year's income may be a poor indicator of a household's true economic spending power, and there is something to be said for basing the tax assessment on property values.
But this creates difficult administrative problems about how often official valuations of property values are revised.
Some countries do use property taxes in this way.
Others, such as the United States, prefer a progressive local income tax, which may be simpler to administer.
17
Competition Policy and Industrial Policy
What do Durex, Valium, and Cornflakes have in common with electricity and coal?
The answer is that the London Rubber Company, Hoffman LaRoche, Kelloggs, the Central Electricity Generating Board, and the National Coal Board have all been the subject of investigations by the Monopolies and Mergers Commission.
Chapters 17 and 18 are about the very large firms that exist when economies of scale are very  important .
This chapter is about firms in the private sector; the next chapter about public sector firms.
When economies of scale are important, industries will not be perfectly competitive.
There will be a few large firms in the industry.
In an unregulated free market, each firm will maximize profits by setting marginal cost equal to marginal revenue, which is less than price and marginal consumer benefit.
Since marginal cost is less than marginal consumer benefit, society would gain by expanding the output of such industries.
Imperfectly competitive industries are a source of market failure because free market equilibrium is no longer Pareto-efficient.
Different governments adopt very different solutions to this problem.
In the United States the Anti-Trust laws have been used to break up existing private monopolies and prevent the formation of new monopolies through mergers of existing companies.
European governments have tended to take a more lenient view of merger activity and have sometimes actively encouraged it.
Natural or inevitable monopolies such as electricity generation enjoy huge economies of scale.
In some countries they are private industries subject to state regulation; in other countries they are state-run or nationalized industries.
Whether nationalization improves the efficiency of resource allocation, or whether we would do better to privatize existing public corporations, is an issue we take up the next chapter.
This chapter analyses government policy towards private sector firms that are necessarily imperfectly competitive.
In some cases there are no clear and simple answers.
This is an area of economic as well as political controversy.
But at least we can sort out the issues involved.
We begin by discussing the social cost of monopoly and other forms of imperfect competition.
17-1 The Social Cost of Monopoly Power
In Chapter 15 we saw that, in the absence of externalities and other sources of market failure, competitive equilibrium would be Pareto-efficient.
Each firm's private assessment of its marginal cost would coincide with society's assessment of the social marginal cost of the resources employed.
Each firm's private assessment of the marginal benefit of output, namely the output price received, would coincide with society's view of the marginal value of output to consumers.
Competitive equilibrium would ensure that each industry expanded its output up to the point at which price equals marginal cost and therefore social marginal benefit equals social marginal cost.
No resource reallocation would make consumers as a whole better off.
When an industry is imperfectly competitive we say that each firm in the industry enjoys a degree of monopoly power.
Equating marginal cost and marginal revenue, each firm will produce an output at which price exceeds marginal cost.
This excess of price over both marginal revenue and marginal cost is a convenient measure of the firm's monopoly power.
When firms have monopoly power, price and social marginal benefit of the last output unit exceed the private and social marginal cost of  producing that last output unit.
From society's viewpoint the industry is producing too little.
Expanding output would add more to social benefit than to social cost.
How should we measure the social cost of monopoly power and inefficient resource allocation?
Later we take up the case of natural or pure monopoly where huge economies of scale allow only a single producer to survive in an industry.
We begin with a more general discussion of all forms of imperfect competition and monopoly power.
As we saw in Chapter 10, intermediate forms of imperfect competition require some economies of scale to limit the number of firms an industry can support.
Nevertheless, to introduce the idea of the social cost of monopoly power it is convenient to ignore economies of scale altogether.
We thus ask the following question: what would happen if a competitive industry were taken over by a single firm which then operated as a multi-plant monopolist?
Figure 17-1 shows how this question may be answered.
Under perfect competition LMC is both the industry's long-run marginal cost curve and its supply curve.
With constant returns to scale, LMC is also the long-run average cost curve of the industry.
Given the demand curve DD, competitive equilibrium is at B. The competitive industry produces an output Q C at at price P C .
When the industry is taken over by a monopolist, the monopolist recognizes that marginal revenue MR is less than price at each output.
The monopolist produces an output Q M at a price P M thus equating marginal cost and marginal revenue.
The area P M P C AC shows the monopolist's profits from selling Q M at a price in excess of marginal and average cost.
The area of the triangle ACB shows the deadweight burden or social cost of monopoly power.
Why?
Because at Q M the social marginal benefit of another unit of output is P M but the social marginal cost is only P C .
Society would like to expand output up to the competitive point B at which social marginal benefit and social marginal cost are equal.
The triangle ACB measures the social profit or excess of benefits over costs from such an output expansion.
Conversely, by reducing output from Q C to Q M the monopolist imposes a social cost equal to the area ACB.
For the whole economy the social cost of monopoly power is obtained by adding together the deadweight burden triangles such as ACB for all industries in which marginal cost and marginal revenue are less than price and social marginal benefit.
An early study by Professor Arnold Harberger suggested that the social cost of monopoly power, as measured by these deadweight burden triangles, was considerably less than 1 per cent of national income.
If social costs are really this small, perhaps there are more important things for the government to worry about in designing economic policy.
Professor George Stigler, a recent Nobel Prize winner in economics, has argued that 'Economists might serve a more useful purpose if they fought fires or termites instead of monopoly.
Figure 17-1 THE SOCIAL COST OF MONOPOLY.
The industry has horizontal long-run average and marginal costs.
A perfectly competitive industry produces at R, but a monopolist sets MR = MC to produce only Q M at a price P M .
The monopolist earns excess profits P M P C CA, but there is a social cost or deadweight burden equal to the triangle ACB.
Between Q M and Q C social marginal benefit exceeds social marginal cost and society would gain by expanding output to Q C .
The triangle ACB shows how much society would gain by this expansion
However, other economists believe that the social costs of monopoly cannot be ignored.
Professor F. M. Scherer of Yale, whose work on industrial structure we cited in Chapters 8-10, has argued that in the United States the social cost of monopoly is large enough 'to treat every family in the land to a steak dinner at a good restaurant'.
Moreover, Professors Keith Cowling and Dennis Mueller have argued that the social cost of monopoly could be as high as 7 per cent of national income.
Why is there such disagreement about the cost of monopoly?
First, the area of the deadweight burden triangle in Figure 17-1 depends on the elasticity of the demand curve.
In calculating the size of deadweight burden triangles under monopoly, different economists have used different estimates of the elasticity of demand.
Second, the welfare cost of monopoly is greater than the deadweight burden triangle itself.
Since monopoly may yield high profits to the firm, it is likely that firms will expend large quantities of resources in trying to acquire and secure monopoly positions.
In Chapter 10 we saw that existing oligopolists would have an incentive to advertise too much, not because advertising provides additional consumer information about the product, but because it raises the fixed cost of being in the industry, thereby making it harder for new firms to enter the industry.
How much should society view this advertising expenditure as a waste of resources?
Similarly, firms may devote large quantities of resources trying to influence the government in order to obtain favourable judgements which will enhance or preserve their monopoly power.
They may also deliberately maintain extra production capacity so that potential entrants can see that any attempt at entry will be matched by a sharp increase in production by existing firms, forcing price reductions which in the short run will be unprofitable for all but which may bankrupt the entrant first.
From the economy's viewpoint, resources devoted to lobbying the government or maintaining deliberate over-capacity may also be largely wasted.
For these reasons, the precise extent of the social cost of monopoly remains a subject of continuing controversy.
Different economists will continue to disagree about its measurement and extent.
Nevertheless, few governments believe that the social cost of monopoly is sufficiently small that it can safely be ignored.
We will shortly examine the policies which have been adopted to restrict the degree of monopoly power exercised by large firms.
Our discussion relates only to the efficiency losses arising from imperfect competition.
Society might also have views on two other aspects on monopoly performance: the amount of political power that large companies are in a position to exert, and the distributional issue of fairness in relation to the large supernormal profits that a monopolist can earn.
The Distribution of Monopoly Profits
In Figure 17-1 the area P M P C CA shows pure monopoly profits after all economic costs.
Should society tolerate such privately collected taxes?
Whether we think the high price P M charged by a monopolist is a rip-off or the just reflection of what consumers are prepared to pay is a pure value judgement about equity.
In passing, it is worth remarking that the ultimate recipients of monopoly profits are the monopolist's shareholders.
Since a large fraction of the stock market is held by pension funds and insurance companies which will eventually make payments to workers, monopoly profits may indirectly pay income to some relatively poor people.
Nevertheless, in addition to any efficiency argument against monopoly, society may decide that it dislikes monopoly profits purely on the grounds of equity.
Suppose the government imposes a profits tax on a monopolist: what effect would this have on the monopolist's output decision?
The simple answer is that it would have no effect!
Why not?
Because, whatever the tax rate (assuming it is less than 100 per cent), the way to maximize after-tax profits is to maximize pre-tax profits.
Provided the government does not take all the extra pre-tax profit in taxes, increasing pre-tax profits must always increase post-tax profits.
Hence the monopolist will produce exactly the same output as in the absence of a profits tax and, facing the same demand curve, will charge the same price as before.
Since it is always open to the government to tax away a monopolist's excess profits, it is the allocative inefficiency of monopoly on which economists have focused their criticisms.
17-2 Regulating Private Monopolies in the UK
Suppose it is recognized that economies of scale may lead to monopoly or other forms of imperfect competition that tend to misallocate resources.
Government intervention may take one of two forms.
Either the government can order large firms to be split up into smaller independent companies, which it is hoped will act more competitively (the so-called 'structural approach'), or the government can leave monopoly firms intact but seek to control their performance, for example by monitoring prices and profits and ordering price reductions when firms appear to be exerting their potential monopoly power.
Figure 17-2 MONOPOLY WITH COST REDUCTION.
Facing the marginal cost schedule LMC, a competitive industry produces at E. If a monopoly can produce at the lower cost LMC '; it will equate this to MR to produce Q M at a price P M .
The deadweight loss to consumers from lower output and higher prices is measured by the triangle A, but society gains because output is now p produced using less resources.
The rectangle P measures the cost savings and it could outweigh the triangle A if cost savings are sufficiently large.
In the UK policy has typically followed the second route, seeking to regulate rather than remove monopolies.
One reason why this might be a sensible approach is that, if the original monopoly arose from technical economies of scale, breaking up large firms will not change technical know-how or remove the advantages to large scale.
Under these circumstances, market forces will tend to re-establish large firms again.
Indeed, this suggests that there may be some compensating advantages of monopoly, in particular the ability to operate at larger scale and with lower average costs.
Figure 17-2 shows how these advantages should be compared with the disadvantage of the deadweight burden caused by the failure to equate marginal cost and price and hence marginal cost and marginal benefit.
Suppose a competitive industry of small firms has the supply curve and long-run marginal cost curve LMC whereas a monopolist, by taking advantage of scale economies, would face the long-run marginal cost curve LMC ';.
The competitive industry would produce Q C to be in equilibrium at E, whereas the monopolist, equating MR to the lower curve LMC ';, would produce Q M to be in equilibrium at E ';.
Which outcome is better for society?
Begin from the output Q M produced by the monopoly.
If the same output were produced by a competitive industry, society would lose the light green rectangle B shown in Figure 17-2.
That is the quantity of extra resources that a competitive industry would use because it has higher average and marginal costs.
But a competitive industry would produce at E, not at E ';.
Using the long-run marginal cost curve relevant to the competitive industry, this move from E '; to E would produce a social gain equal to the area of the triangle AE '; E. Thus, the more a monopolist can achieve economies of scale and lower production costs, the more likely it is that the area of the rectangle B will exceed the area of the triangle A. On balance the cost savings arising from monopoly would then more than offset the deadweight burden triangle, and society would actually gain from monopoly.
Thus in the UK monopoly policy is relatively pragmatic.
Although the existence of monopoly power may provoke a government investigation, each investigation is a cost-benefit analysis attempting to identify the costs and benefits on a case by case basis.
Some investigations conclude that on balance the existence of a large firm is actually in the public interest.
This approach is fundamentally different from the approach adopted in the United States, where the law tends to assume that the very existence of monopoly power is against the public interest.
Competition Law in the UK
Legislation has been steadily extended since the Monopoly and Restrictive Practices Act of 1946.
Restrictive practices, which we discuss shortly, have been separately examined since the establishment of the Restrictive Practices Court in 1956.
Monopoly policy was comprehensively reassessed in the 1973 Fair Trading Act and has been amended in the 1980 Competition Act.
The 1973 Act introduced a Director-General of Fair Trading to supervise many aspects of competition and consumer law including the regulation of quality and standards discussed in Chapter 15.
The Director-General is responsible for a monitoring company behaviour and, subject to a ministerial veto, can refer individual cases to the Monopolies and Mergers Commission (MMC) for a thorough investigation.
A company can be referred if it supplies more than 25 per cent of the total market.
The Commission can also be given cases where two or more distinct firms by implicit collusion operate to restrict competition.
There is no presumption that monopoly is necessarily bad, and the Commission is charged to investigate whether or not the monopoly acts against the public interest, a brief that may be widely interpreted though in recent years there has been increasing emphasis on the 'maintenance and promotion of effective competition'.
The Restrictive Practices Court examines agreements between firms supplying goods and services in the UK, for example agreements on collusive pricing behaviour.
All agreements must be notified to the Director-General of Fair Trading, who will refer them to the Court unless they are voluntarily abandoned or judged of trivial significance.
The Court will find against these agreements unless they satisfy one of eight 'gateways' or justifications, for example that their removal would cause serious and persistent unemployment in the area.
Thus for restrictive practices the burden of proof lies on the companies to show that they are acting in the public interest: in contrast, the legislation on monopolies is more open-minded, requiring the MMC to make the case that companies are acting against the public interest.
The UK is now subject to the monopoly legislation of the European Community as well.
Article 85 of the Treaty of Rome is rather similar to the UK legislation on restrictive practices.
Agreements have to be notified and they are likely to be outlawed.
Article 86 bans the abuse of a 'dominant position' as a monopolist, but exactly what this means is rather vague.
The UK legislation, allowing any firm with more than 25 per cent of the market to be referred to the MMC, is more clear-cut.
Monopolies Policy in Practice in the UK
The MMC has wide powers to make recommendations, and the Secretary of State to act on these recommendations, yet on only a few occasions have companies been formally penalized as a result of MMC investigations.
More frequently, the MMC has relied on informal assurances that criticized behaviour will be discontinued.
In fact, it is only a slight exaggeration to say that the main deterrent effect of monopoly policy has been not the threat of what changes might be required as a result of the MMC investigation, but the threat of having to tie up a large quantity of senior executives' time to argue the firm's case in the event that its activities attract a reference to the MMC.
The Commission has investigated a wide range of cases, from beer to breakfast cereals and from contraceptives to cross-Channel ferries.
Because the MMC is charged to investigate each case with an open mind, its judgements have tended to stress different aspects of behaviour in different cases.
Certainly a high market share has not been sufficient to attract an unfavourable judgement.
For example, Pilkington's Glass (making car windscreens) and Rank Xerox (making copiers) were honourably acquitted in spite of huge market shares and healthy profits.
In both cases the MMC held that these companies were efficiently run and had contributed to substantial cost savings, the Schumpeterian view of monopoly to which we referred in Chapter 9 and illustrated in Figure 17-2.
On the other hand, cost reduction has not been sufficient to avoid censure by the MMC.
Hoffman LaRoche was praised as 'a highly competent organization with a product range of high quality', but its enormous profits, sometimes as high as 60 or 70 per cent on capital employed, were held to be unjustified and the MMC recommended that the price of both Librium and Valium be halved.
In only one instance, the tobacco industry, has the MMC recommended companies actually be split up; it recommended that Imperial Tobacco should sell its 42.5 per cent share in Gallaghers, but the Secretary of State did not accept this recommendation.
One other case deserves special mention.
As we noted in Chapter 10, oligopolists have an incentive to carry out socially unproductive advertising in order to make it harder for new entrants to meet the fixed costs of breaking into the industry, thereby leaving more of the market and the profit for existing firms.
In 1966 the MMC judged that.
Proctor & gamble and Unilever, the two giants of the detergent industry, were guilty of this practice, and it recommended that advertising be reduced and product prices cut.
In 1981 the Conservative government withdrew this restriction on advertising, partly in the belief that competition from 'own brand' soap powders sold in supermarkets would provide an effective check.
Advertising spending by Proctor & gamble took off immediately.
Restrictive Practices
Since restrictive practices legislation was first introduced, almost 5000 agreements have been registered, the vast majority of which were abandoned even before they were taken to the Court.
Most explicit price-fixing has gone.
Although these facts look impressive, they may overstate the success of policy against restrictive practices.
First, they may simply have forced collusive agreements underground.
Where oligopolistic market structures remain it seems likely that some firms will resort to informal agreements and the other collusive devices examined in Chapter 10.
An important aspect of recent legislation and policy has been to tighten up on 'information agreements', which could form the basis of secret collusion.
Second, the various 'gateways' have permitted some agreements to be ratified, and the wisdom of some of these ratifications has been challenged.
Finally, it is possible that tighter control of restrictive practices agreements between firms is one of the factors that provide an incentive for mergers, a subject we take up in the next section.
By formally merging, companies could continue their old practices within the merged company and take their chance if they got investigated by the MMC.
Assessing Competition Policy in the UK
To assess UK competition policy we need ideally to compare the evolution of the UK economy under the policy with the evolution that would have occurred under some different policy, for example a policy of laissez-faire or complete non-intervention.
Such an assessment would be a major undertaking.
At a more modest level, we can say that legislation on restrictive practices has eliminated many cases of blatant anti-competitive behaviour and that the Director-General of Fair Trading now has powers to promote the provision of better consumer information (e.g. the Trade Descriptions Act) and to monitor general company behaviour.
And the MMC has identified some practices, such as wasteful advertising, that most people would regard as undesirable.
If these seem relatively modest benefits, it is perhaps more useful to consider whether a more radical and comprehensive anti-monopoly policy would have been better for society.
We began this section by noting that, unlike the United States which takes a structural approach to anti-trust policy, in which the possession of monopoly power is itself regarded as objectionable, the UK has taken a more open view of the benefits of promoting competition.
Large scale may be necessary to achieve minimum efficient plant size, and the cost of breaking up large companies may be considerable, an argument we develop in Section 17-4.
Large scale may also promote better management, co-ordination, and research, a version of the Schumpeterian argument in favour of monopoly.
Co-ordination via cartels or single ownership may facilitate better planning when different products are close complements in production.
As we stressed in Chapter 9, the Rover Group may produce a large share of the motor cars manufactured in the UK but this does not necessarily mean it has significant market power.
When tariffs on imports are low and transport costs are moderate, domestic producers may face severe international competition.
Without knowing the size of the relevant market in which firms are competing, large size cannot immediately be equated with uncompetitive behaviour.
Nor is it obvious that the government should always oppose large profits.
As we pointed out in Chapter 14, profits are the carrot that encourages firms to take risks in a market economy.
Many firms take risks that do not come off.
But if potential risk-takers are assured that success will immediately invite investigation by the MMC and an order to cut prices and eliminate excess profits, there will be less risk-taking in the economy.
Society has to decide how much risk-taking it wishes to encourage and to allow a proper return to risk-taking as an economic cost against accounting profits.
These doubts about the merit of a blind pursuit of perfect competition lie behind the UK policy and the judgements of the MMC.
Thus policy has considered each case on its merits.
Notice, finally, that this approach suggests that monopoly policy should not be independent of other aspects of government policy.
Large firms make more sense when the UK is competing within a large European market.
We might wish to be tougher on large firms if the government is pursuing a policy of protection with high import tariffs.
17-3 Mergers
Two existing firms can join together in two different ways.
First, one firm may make a takeover bid for the other by offering to buy out the shareholders of the second firm.
Managers of the 'victim' firm will usually resist since they are likely to lose their jobs, but the shareholders will accept if the offer is sufficiently attractive.
In contrast, a merger is the voluntary union of two companies where they think they will do better by amalgamating.
When one firm is sufficiently large it is difficult for another firm to raise the money to buy out all the first firm's shareholders in a contested takeover bid.
Hence very large firms created by amalgamation are usually the result of mergers.
In deciding to get together the two firms reveal that they think this union will be in their private interest, but we must ask whether mergers are in the public interest.
Since mergers help to create monopoly power the discussion of this section is largely an extension   of the analysis of competition policy in the previous section.
It is important to distinguish three types of merger.
The production process typically has several stages.
For example, the first stage might be iron ore extraction, the second stage steel manufacture from iron ore, and the third stage production of cars from steel.
By a horizontal merger we mean the union of two firms at the same production stage in the same industry, for example the merger of two steel producers or two motor car manufacturers.
By a vertical merger we mean the union of two firms at different production stages in the same industry, as when a car manufacturer merges with a steel producer.
Finally, there are conglomerate mergers, where the production activities of the two firms are essentially unrelated.
For example, a tobacco manufacturer perceiving that the cigarette market is in long-term decline might decide to join forces with a perfume company or a sugar producer.
What do firms think they stand to gain by merging?
In horizontal mergers it is possible that a merger will allow exploitation of economies of scale.
One large motor car factory may be better than two small ones.
(Notice that this requires that each of the original companies were producing at an output below minimum efficient scale.)
In vertical mergers it is often claimed that there are important gains to co-ordination and planning.
It may be easier to make long-term decisions about the best size and type of steel mill if a simultaneous decision is taken on the level of car production to which steel output forms an important input.
Since conglomerate mergers involve companies with completely independent products, these mergers have only small opportunities for a direct reduction in production costs.
Two other factors are frequently mentioned as potential benefits of mergers.
First, if one company has an inspired management team it may be more productive to allow this team to run both businesses.
Managers of course are very fond of this explanation for mergers.
Economists have tended to be more sceptical.
Second, by pooling their financial resources, the merging companies may enjoy better creditworthiness and access to cheaper borrowing, enabling them to take more risks and finance larger research projects.
There may also be some economies of scale in marketing effort.
These managerial and financial gains could explain why mergers make sense even for companies producing completely distinct products.
If companies achieve any or all of these benefits they will increase productivity and lower the cost of making any specified output level.
These private gains are also social gains, since society can use less resources to achieve the same output.
If these were the only considerations, merger policy might be confined to ensuring that companies formed accurate assessments of the costs and benefits of mergers.
Social and private calculations would coincide.
However, there are two important reasons for private and social assessments to diverge.
First, the merger of two large firms will give them the immediate monopoly power that derives from a large market share.
The merged company is likely to restrict output and increase prices, providing private profits but a dead-weight burden to society as a whole.
Second, the merged company may be able to use its financial power as distinct from its power derived from current market share.
This danger is especially apparent in conglomerate mergers.
A car producer and a food manufacturer cannot merge to gain economies of scale in production or any direct reduction in costs; but they can use their joint financial resources to start a price war in one of these industries.
Because they have extra financial resources they will not be the first company to go bust.
By forcing out some existing competitors, or merely holding this increased threat over potential entrants, they may be able to increase their market share in the long run, deter entry, and charge high prices for every more.
In framing merger policy, the government must therefore decide whether the potential social gains from reduced costs and more efficient production are outweighed by the social costs of monopoly power and inefficient resource allocation between firms and industries that might arise.
Concentration and Mergers in the UK
British industry is becoming increasingly concentrated in the hands of a few producers, and oligopoly seems to be the commonest market 
structure in manufacturing industry.
By 1970 the 100 largest firms produced nearly half of the output of British industry.
The concentration of production in a few firms was considerably more marked in 1970 than it had been 15 years earlier, and many economists attribute around half the increase in concentration in the 1960s to the merger boom during those years.
From 1964 onwards many mergers were actively encouraged by a Labour government seeking to reorganize British industry into large modern units capable of taking on foreign competition.
Mergers have continued since 1970.
Table 17-1 shows the average number of mergers per annum of industrial and commercial companies, and the value of the assets involved.
The table shows a new UK merger boom starting in 1985.
Notice that it reflects less a dramatic change in the number of mergers and more a spectacular increase in the size of companies involved.
Nowadays not even the largest companies are safe from a takeover.
One reason for this change is the greater competition among banks and the more aggressive lending policies that ensue.
Several of the most recent supermergers have seen one company borrow large amounts in order to be able to afford to acquire another company.
Little fish can now swallow bigger fish.
Table 17-2 shows the composition of mergers in the UK since 1965.
The numbers refer only to mergers of large companies where assets worth over 5 are acquired through the merger.
These of course are the mergers most relevant to the
formation of even larger companies with potential monopoly power.
It can be seen that vertical mergers are relatively rare.
Whereas horizontal mergers, the type most likely to yield production economies of scale, were the vast majority in the 1960s, there has been a steady trend away from horizontal mergers and towards conglomerate mergers more recently.
Table 17-1 UK takeovers and mergers, 1972 88 (annual averages)
Since 1970 industrial concentration has continued to increase and mergers have continued to play a major role in this process.
Professor Keith Cowling and his research team have argued that mergers continue to explain half of the increase in industrial concentration in the UK.
Table 17-2 Proposed UK mergers: percentage by type
Merger Policy in the UK
This significant increase in industrial concentration through merger activity would not have been possible if the government had been operating a tough anti-merger policy.
In Section 17-2 we saw that the United States takes a structural approach to monopoly, believing that concentration and monopoly power are undesirable per se.
This has the natural corollary that the United States also operates rigorous controls on mergers that promote new monopolies.
In contrast, the UK takes a more neutral view of monopoly, requiring the Monopolies and Mergers Commission to demonstrate that a monopoly is acting against the public interest, and the same principle carries over to the assessment of prospective mergers.
Indeed, it is only since 1965 that mergers have been subject to public scrutiny at all.
There are now two grounds for referring a prospective merger to an investigation by the MMC:(1) that the merger will promote a new monopoly as defined by the 25 per cent market share used in deciding references for existing monopoly   positions, or (2) that the merger involves the transfer of at least 30 million worth of company assets.
Since the legislation was introduced in 1965, only about 3 per cent of all merger proposals have been referred to the MMC.
Thus for much of the period government policy has been to consent to, or actively encourage, mergers.
The reason for this may be understood using Figure 17-2.
By creating a larger firm with more monopoly power, a merger will tend to produce a dead-weight burden since the new firm will use this power to restrict output and drive up the price.
However, the merger may achieve a cost reduction and saving on scarce resources that more than offset this social cost.
In believing that the benefits would outweigh the costs, British merger policy reflected two underlying assumptions.
The first was that the cost savings from economies of scale and more intensive use of scarce management talent could be quite large.
The second was that the UK was effectively part of an increasingly competitive world market so that the monopoly power of the merged firms, and the corresponding social cost of the dead-weight burden, would be small.
Large as they were, the merged firms would still be small in relation to European or world markets, and would face relatively elastic demand curves which gave little scope for raising price above marginal cost.
Nevertheless, it would be wrong to suppose that all mergers were approved.
Most of the mergers actually referred to the MMC were found to be against the public interest, and the effects of the legislation went beyond the cases actually referred.
Investigation was a lengthy process taking many months, a delay during which company share prices could move considerably and upset the original negotiations about the terms on which the relative shares of the companies should be valued.
In practice, even the threat that a merger might be referred to the MMC was often sufficient to induce the companies to drop their merger proposals.
Were Mergers Successful?
Merging companies believed they could achieve significant gains in productivity and profits, and government policy assumed that these would offset any adverse effects from an increase in monopoly power.
Yet recent investigations of the post-merger record of these firms suggests that few of these gains materialized.
Even where post-merger firms did achieve productivity gains, they did little better than the productivity performance of the industry as a whole.
In contrast, mergers led to a significant increase in industrial concentration and monopoly power in the UK.
In terms of Figure 17-2, it is highly questionable whether the cost saving rectangle B was larger than the dead-weight burden triangle A. 
17-4 Regulating Natural Monopoly
In the theory of market structure developed in Chapter 10 we saw that the key relation was the size of economies of scale relative to the size of the market.
Most of the firms in private industry to which monopolies and mergers legislation is relevant are in fact oligopolists.
They do not control the entire market and are not free from the threat of entry from new firms or foreign competitors.
We turn now to the analysis of natural monopolies, the industries that really do have such enormous economies of scale that only one firm can survive in them.
These industries have falling long-run average cost curves, as for example in British Telecom, where the cost of transmission lines dwarfs the marginal cost of providing an extra phone call for a user already connected to the system.
Nor does the industry have to worry about imports or the world market.
The Problem of Natural Monopoly
Figure 17-3 shows an industry with steadily decreasing long-run average costs reflecting the technological advantages of producing a very large output.
In Chapter 10 we explained that only one private firm could survive in such an  industry.
With many firms, the firm that expands output will always be able to reduce costs and undercut its rivals.
Facing a demand curve DD and the corresponding marginal revenue curve MR, the resulting monopolist will produce Q M and earn profits P M CBE.
Figure 17-3 NATURAL MONOPOLY.
The socially efficient point E '; occurs where long-run marginal cost LMC equals marginal benefit DD.
A private monopolist sets MR = MC, produces Q M , and earns profits P M CBE.
The dead-weight loss under private monopoly is therefore the area AEE ';.
Since LMC passes through the lowest point on LAC, a natural monopolist with steadily falling LAC must face LMC below LAC.
If by law the monopolist was forced to charge a fixed price P C , the monopolist would face a horizontal demand curve P C E '; up to the output Q ';.
Since P C would then also be marginal revenue, the monopolist would produce at E '; where the marginal revenue and marginal cost coincide.
Although Q '; is the socially efficient output, society cannot force the monopolist to produce here in the long run.
Since E '; lies below the corresponding point on LAC at the output Q ';, the monopolist is making losses and would rather go out of business.
At this output the social marginal benefit P M exceeds the social marginal cost at A. The monopolist produces too little.
Social marginal cost and marginal benefit are equal at the output Q '; and the efficient point for society is E ';.
The private monopoly creates a dead-weight burden AEE ';.
Suppose you sat on the Monopolies and Mergers Commission and were investigating this monopolist.
What are your options?
If you split the firm up you will have a lot of small firms each producing at higher average cost, a waste of society's resources.
You could order the firm to produce at the socially efficient point E ';.
Then you will get the desired output Q ';.
However, the price P C will be less than the firm's average costs at Q '; so it will be making losses.
You cannot force private firms to make losses.
They will shut down instead.
Few countries allow unregulated natural monopolies to produce at E and impose the full dead-weight burden AEE '; on society.
The first solution is a regulatory body such as OFTEL, which regulates British Telecom.
The aim is to get as close as possible to the socially efficient allocation E '; while allowing the monopolist to break even after allowing a proper deduction for all economic costs.
For example, by ensuring that the monopolist produces Q at the price corresponding to average cost at this output, the social cost of the dead-weight burden can be reduced from AEE '; to GHE';.
An even better solution is to allow the monopolist to charge a two-part tariff.
A two-part tariff is a price system where users pay a fixed sum for access to the service and then pay a price per unit which reflects the marginal cost of production.
Thus the aim of a two-part tariff is to use fixed charges to pay for fixed costs and then to levy marginal charges to cover marginal costs.
It acts as a lump-sum tax on users of the system.
Hence in Figure 17-3 the monopolist can be instructed to charge P C for each unit of the good.
Consumers will demand the socially efficient quantity Q ';.
Since the monopolist is now a price-taker at the controlled price P C it will be loss-minimizing for the monopolist to produce Q ';, at which price and marginal revenue and equal marginal cost.
If the regulatory body does its sums correctly, it will then allow the monopolist to levy the minimum fixed charge necessary to ensure that the monopolist breaks even after allowing for all relevant economic costs.
The two-part tariff is not always a feasible solution.
If the fixed charge has to be very high it may induce people to abandon consumption of the commodity altogether.
Moreover, whereas it may be easy to collect fixed charges from consumers with telephone or gas installations, it is harder to enforce a fixed charge for the right to travel by rail and a fare per journey   reflecting marginal cost.
The costs of enforcing such a system might be enormous.
The third solution to the natural monopoly problem is to order the monopolist to produce at the socially efficient point E '; and the corresponding price P C but to provide a government subsidy to cover the losses that this will inevitably imply.
However, the government will wish to be closely concerned with the operation of the company to ensure that this does not provide a blanket guarantee to underwrite whatever losses the company makes through its own stupidity or inefficiency.
It is still socially desirable to produce the efficient output Q '; in the cost-minimizing way.
Thus, where the subsidy solution is adopted there is a pressure for the government to take over the entire running of the industry so that all operations can be carefully monitored.
Where this has been done in the UK these natural monopolies are called nationalized industries, and we discuss them more fully in the next chapter.
However, it is important immediately to dispose of one popular fallacy.
We have just demonstrated that natural monopolies cannot both survive as profit-making industries and produce the socially efficient output by pricing at marginal cost unless they have very favourable opportunities for levying two-part tariffs.
If such industries are nationalized in order that they can produce closer to the socially efficient output, it is inevitable that they will make losses and require a subsidy from the government.
The fact that nationalized industries make losses is not sufficient to prove that they are not minimizing costs or  that they are producing the wrong output from society's view-point.
Two problems recur in attempts to adopt any of the above solutions to the problem of natural monopoly.
The first is that it is hard to ensure that the industry really does minimize costs.
Unnecessarily high costs can be passed on under average cost pricing (solution 1), can result in a higher fixed charge to ensure break-even under a two-part tariff (solution 2), or can require a larger subsidy (solution  3).
In each case the regulatory body has the difficult task of trying to ensure that the management of the natural monopoly is as efficient as possible.
One practical attempt to overcome this problem of regulation is the 'RPI -- x' formula, first introduced in the UK when OFTEL was established to regulate the newly privatized British Telecom (BT).
The government believed that BT could reduce the real price of phone calls in subsequent years for two reasons: unusually good opportunities for cost-saving technical advances, and a previous lethargy as a public industry which it was hoped privatization would dispel.
OFTEL was instructed to ensure that, on average, the nominal price of BT phone calls rose x per cent per year less than the inflation rate of the retail price index.
Initially, x was set at 3 per cent.
The second practical problem is regulatory capture.
Regulatory capture implies that the regulator gradually comes to identify with the interests of the firm is regulates, eventually becoming its champion, not its watchdog.
Clearly, the regulated devote considerable time, effort, and money to lobbying and otherwise trying to influence the regulator.
A sense of public duty is the only reason for the regulator to resist.
It can become an unequal contest.
More subtly, the regulated firm has all the inside information about its own activities, information which it is the whole purpose of the regulatory to try to acquire.
Of necessity, regulators have to build up contacts with the regulated.
At the end of long conversations, it is unsurprising that the regulator can feel quite sympathetic to the problems as perceived by the regulated.
Recap
In this chapter we have discussed aspects of competition policy, the process through which governments seek to prevent large private firms from abusing their potential monopoly power.
Some issues remain to be discussed, most notably the questions of whether some industries should be nationalized or whether existing public sector industries should be privatized.
This we take up in the next chapter.
We conclude this chapter by discussing industrial policy in a wider context.
17-5 Industrial Policy
Competition policy and industrial policy are closely related.
In Chapter 15 we set out the basic case for allowing free markets to allocate resources.
We argued that this allocation would be efficient only if market failures did not occur.
Competition policy aims to offset market failures arising from scale economies and market power.
When marginal revenue is less than price, profit-maximizing firms no longer set price (and marginal consumer benefit) equal to marginal cost.
Industrial policy is designed to offset other sources of market failure which arise in the production process.
In this section we highlight some of the issues that arise in industrial policy.
Inventions and the Patent System
In Chapter 14 we saw that information is a very special economic commodity which frequently causes indigestion in freely competitive markets.
It is hard to trade information: the buyer needs to see the information before being willing to offer a price, and having seen the information then has no incentive to pay for it!
Inventions -- the discovery of new information about the production process -- are a particular example of this general theme.
Suppose a company develops a product in secret, and then markets it.
If other firms can quickly imitate the new invention, competition will rapidly compete away the profits on this new product.
Since everyone can foresee that this will occur, few resources will be devoted to searching for inventions, even though they are socially valuable.
Inventions are an example of a public good, which we discussed in Chapter 16.
The problem arises because the inventor cannot privately appropriate the benefits since imitators cannot be excluded.
The solution to this market failure is a patent system, which confers a temporary  legal monopoly on the inventor who registers or patents the invention.
The temporary monopoly provides, before the fact, the assurance that if the search for a new discovery is successful the inventor will be able to cash up after the fact.
In the language of Chapter 10, it is a credible pre-commitment.
Why is it important that the legal monopoly should be only temporary?
Otherwise, successful inventors would have an entrenched entry barrier which would prevent competition from other firms or new entrants for all time.
The trick in designing a successful patent system is to provide a big enough incentive for invention, but not such a large and long-lived cushion that the benefits of competition are suppressed for ever.
The New Industrial Economics, which we introduced in Chapter 10, suggests that this is not a trivial problem, especially when considerations of strategic competition are introduced.
For example, there are documented cases of pre-emptive patenting by incumbent firms.
The incumbent may discover a new process or product, patent it, but not actually introduce it.
Potential entrants are aware that any attempt at entry will be met with the launch of this new product, which will place the entrant at a disadvantage.
Thus, pre-emptive patenting can be a highly effective strategic entry barrier.
This example shows that intelligent industrial policy and competition policy must work hand in hand.
Evidence of pre-emptive patenting is the type of information the Monopolies and Mergers Commission seeks in evaluating whether incumbents are abusing their market power.
Research and Development (R&D )
In many countries, including the UK, one of the chief aims of industrial policy is to promote R&D .
In Western Europe, the United States, and Japan, R & d accounts for between 2 and 3 per cent of national output.
Table 17-3 shows UK government expenditure on R & d in recent years.
Why should governments, even the Thatcher government, which is strongly committed to allowing market forces to work, spend several billion pounds of taxpayers' money promoting R&D ?
Table 17-3 UK government expenditure on R & d (% of GDP)
Roughly half the spending shown in Table 17-3 relates to military projects in the field of defence.
But what about the other half?
This seems to indicate widespread agreement that there are market failures in R & d which the patent system alone is insufficient to offset.
Economics can provide several insights into what may go wrong with market forces in R&D .
First, large projects can be very risky for an individual company.
Nowhere are these projects larger than in the development of a major new commercial airliner, and the chief executive of Boeing, the largest plane manufacturer in the world, has described each major new project as 'betting the company'.
In other words, failure on one new project could threaten the very existence of the company.
In Chapter 14 we described why private individuals may be risk-averse, and this applies even to executives of large corporations.
Consequently, private firms may undertake less R & d than is socially desirable.
This implies that the social return on such projects exceeds the private return to those making the decisions.
If the private decision-makers require a large risk premium, or on average a high expected return, before being willing to assume such risks, why should society demand any less?
Essentially for two reasons.
First, the government may be able to pool the risks across a large number of projects in its portfolio.
Second, even if projects go wrong, the government can spread the burden very thinly across the population: 2 per cent on everybody's income tax rate for a year should cover even the biggest disaster.
Thus, the population as a whole should require only a small risk premium, much smaller than that required by executives in an individual company who may face personal disaster if the project turns sour.
And these arguments have generally  been found persuasive by governments.
Hence their provision of public subsidies for R & d expenditure.
It should also be noted that no patent system can be watertight.
Indeed, as we remarked above, to make it so would effectively be to suppress all future competition indefinitely.
In these circumstances, a further argument for public support of R & d may be because private firms realize they will not be able to appropriate for themselves all the benefits of their efforts.
Some imitation will occur, but it will be insufficiently clear-cut to guarantee that a law suit will not be protracted and expensive.
Moreover, other inventors may be stimulated by what they see to make a breakthrough in an entirely different area.
New breakthroughs build on past discoveries.
These provide additional motives for R & d support as part of an effective industrial policy.
Strategic International Competition
In Chapter 10 we discussed how individual firms might try to erect strategic entry barriers to preserve and enhance their market power.
Such considerations also apply in international competition between large firms.
For a concrete example, consider again the commercial airliner industry.
Effectively, there are three large airliner firms left in the world market.
Boeing is much the largest and has an entire product range, from the relatively small twin-engined 737 to the massive 747, the Jumbo.
Another American firm, McDonnell Douglas, has a smaller product range, and by the mid-1980s was wondering whether to get out of the industry entirely rather than compete in the next generation of civil airliners.
The European consortium Airbus Industrie, in which British Aerospace (BAe) has a stake, is a relatively recent entrant, still building up its product range.
Airbus Industrie asked the governments of its member producers (Germany, France, the UK, and Spain) for launch aid, a grant or loan on favourable terms to help with R & d on a new type of aircraft.
In addition to the standard arguments for R & d support, what extra issues does  international competition raise?
What must the U K government consider in deciding whether to put in taxpayers' money?
First, will Airbus succeed even without government support?
If the answer is yes, public subsidies are simply a transfer payment to Airbus shareholders, for which there is no strong economic rationale.
Second, will other European governments support Airbus even if the U K does not?
If so, the U K government may be able to act as a free-rider.
However, while it may be wise to attempt the free-rider strategy occasionally, governments that systematically try to do this will eventually get a bad reputation.
Suppose the U K government has decided that it cannot free-ride on other governments: either it pays its share of launch aid or the whole project collapses.
What are the benefits of providing launch aid?
If Airbus pulls out, McDonnell-Douglas may still survive, or it too may pull out of the industry, leaving Boeing the sole producer without fear of competition.
In that event, Boeing will surely cash up, raising the price of aircraft.
British airlines and ultimately British consumers will pay high prices, and Boeing can earn monopoly profits.
It may well be worth preventing this.
If the U K could be sure that McDonnell-Douglas would stay in, and provide effective competition to Boeing, it might be worth allowing Airbus to fold: European consumers would still get the benefit of cheaper planes.
But this strategy is risky.
And there are further considerations.
First, launch aid may be seen as a pre-commitment by European governments not to allow Airbus to be bullied out of the industry.
With such a commitment, McDonnell-Douglas may then decide to pull out if it concludes that the industry is not big enough for three profitable producers.
If so, launch aid will have directly enhanced the market share of Airbus.
Second, in the presence of such a pre-commitment, Boeing may conclude that there is no point attempting a price war to try to force Airbus out.
If European governments can display the credible threat to back Airbus if necessary, Boeing shareholders are only going to lose by an unsuccessful price war.
Hence, the pre-commitment may prevent a price war which might otherwise have occurred.
This example draws on the ideas of the New Industrial Economics of Chapter 10 to show how strategic international competition can provide a rationale for strategic industrial  policy.
Just as industrial policy is related to competition policy, so it is closely related to international trade policy, an issue to which we return in Chapter 32.
Sunrise and Sunset Industries
The final aspect of industrial policy which we examine concerns dynamic change, the rise and fall of industries and the firms within them.
Sunrise industries are the emerging new industries of the future, such as those in hi-tech.
Sunset industries are those in long-term decline.
Currently, sunrise industries include information technology and genetics.
Sunset industries in Western economies include the old heavy industries such as steel and shipbuilding which are now suffering from massive excess capacity as these industries have been undercut by more efficient producers in the Pacific basin.
Why not leave such changes to market forces?
What are the market failures that might justify government intervention through industrial policy?
We begin with the sunrise industries.
Three types of market failure have sometimes been put forward to justify the case for intervention, though it should be stressed that those who believe in the efficiency of market forces would argue that it is easy to exaggerate the importance of these factors.
First, there may be imperfections in the market for lending to new companies and new industries.
Banks and other lenders may be too risk-averse, or too unfamiliar with the new business, to lend the money needed through the early loss-making years.
Second, the market may be slow to provide the relevant training and skills: Catch 22 (until the industry exists, people won't perceive the need for developing such skills; but without the skills, the industry cannot exist).
Third, there may be important positive externalities when similar producers locate in the same place, as in the concentration of information technology firms in Silicon Glen in Scotland or Route 128 round Boston in the United States.
These arguments suggest that it may be possible to provide a rationale for an industrial policy to subsidize sunrise industries.
But certain questions must be answered satisfactorily.
First, why are markets so short-sighted and uninformed?
If existing lenders or trainers are getting it wrong, why don't new firms come in and do a better job?
If the answer is entry barriers, this again demonstrates the close relation between industrial policy and competition policy.
Second, even if markets get it wrong, can the government do better?
The strategy of trying to outguess the market by 'picking winners' is now highly discredited.
It seems implausible that civil servants or politicians can do better than trained analysts in industry and finance.
Rather, if such industrial policy is to be attempted, it seems preferable to diagnose the cause of the market failure and provide a generalized incentive which market decision-makers then take into account when undertaking their professional analysis.
Sunset industries present different problems.
For example, the government may or may not attach importance to local unemployment when industries with a heavy geographical concentration are allowed to go under all at once .
It may be desirable to spend what could otherwise be dole money on temporarily subsidizing lame ducks to ease the transition.
Sometimes, however, a sharp shock is required to signal the extent of the adjustment eventually required and the government's commitment to seeing that adjustment is actually made.
Strategic considerations may be important here too.
Suppose for example there are two remaining producers in an industry which has now contracted to the point when it can profitably support only one firm.
Each firm would like to be the one to survive.
They are playing an exit game of chicken.
It seems plausible that one of two things may happen, neither of which is socially desirable.
First, the industry survives with two firms for much longer than is socially efficient.
Second, the firm with the smaller financial backing will be the first to crack, even though it may be able to produce at slightly lower cost than its richer rival.
In such circumstances, an industrial policy that seeks faster and more efficient rationalization of the sunset industry may be advantageous.
Strategic international considerations may also apply.
In the late 1980s, the European steel industry had an enormous over-capacity, partly because it had been undercut by Korean, Japanese, and other producers in the Pacific, partly because of a wave of added Italian steel capacity in the 1970s.
Clearly, the European governments were engaged in a game to see who would close capacity to leave a profitable market share for the survivors.
In such circumstances, each country's steel industry needs effective government representation, and laissez-faire industrial policy may be a poor policy.
In fact, in the first half of the 1980s, British Steel achieved a larger reduction in its capacity than the steel industry in any other major EC country.
Far from being a free-rider, British Steel actually eased the adjustment problem for other European steel producers.
Summary
1
The social cost or dead-weight burden of imperfect competition and monopoly power arises because marginal cost is set equal to marginal revenue, which is less than price and marginal consumer benefit.
The social cost is measured by the cumulative difference between the value that consumers place on the lost output and its marginal production cost.
The social cost is higher still if oligopolists waste society's resources lobbying for monopoly power or incurring unnecessary expenditure to deter entrants.
2
For the UK the highest estimate of this social cost is 7 per cent of company output, but most economists think the cost is much lower.
Estimates refer only to the cost of allocative inefficiency.
The distributional implication of high monopoly profit is less important, since it could always be taxed away.
3
In the UK any firm with more than 25 per cent of the market can be referred to the Monopolies and Mergers Commission, which must then consider whether or not the monopoly is against the public interest.
The MMC takes account of a wide range of factors in making a judgement.
4
Anti-competition agreements between firms, such as collusive price-fixing, must be notified and are generally outlawed.
The Restrictive Practices Court does recognize certain gateways through which such agreements may be ratified as being in the public interest.
5
Mergers may be horizontal, vertical, or conglomerate.
Conglomerate mergers have the smallest scope for obvious gains through economies of scale but have become increasingly common in recent years.
Mergers account for about half of the increase in industrial concentration in the UK.
In spite of this, most studies agree that mergers have not on average improved the subsequent performance of the merging companies.
6
In principle, mergers can be referred to the MMC if they will create a firm with a 25 per cent market share or if they involve assets of over 30 million.
In practice, a very small number of mergers satisfying these criteria are actually referred to the MMC.
In part this may be justified because the UK competes in large world markets where the firms will have little monopoly power, but, given the documented lack of success of merged companies, successive governments may have encouraged too many mergers.
7
A natural monopolist with large economies of scale may continue to make profits by severely restricting output below the socially efficient level.
But when the average cost curve is falling the marginal cost curve lies below average cost.
Pricing at marginal cost might equate marginal cost and benefit but would entail losses.
Average cost pricing or two-part tariffs may come close to the socially efficient output while allowing the monopolist to survive as a private company.
Alternatively, the industry may be nationalized.
8
Regulators find it difficult to ensure that the regulated really do produce on the lowest possible cost curve.
The 'RPI -- x' formula is one attempt to solve this problem.
Regulatory capture is also a danger.
9
Industrial policy seeks to offset market failures in production which do not arise directly from scale economies in the domestic market and the imperfect competition to which these give rise; offsetting the latter is the object of competition policy.
The two are frequently related.
10
Patents provide a temporary legal monopoly for successful inventors, and hence an incentive to look for inventions.
Otherwise the incentive for inventors would be low since they could foresee that profits on successful inventions  would quickly be competed away by imitators.
11
Many governments believe that the social return on R & d exceeds its private return.
In part, this stems from society's ability to handle risk better than the private individuals on whom it falls, and who therefore require large profits in compensation.
Some benefits of R & d may also spill over to other firms, creating an externality.
12
When international competition is strategic between superfirms, the industrial policies of national governments towards their own 'national champions' may be an important pre-commitment which affects the bargaining power of their firms in the international market.
13
Sunrise and sunset industries may involve other market failures.
Before taking this as a general licence for an active industrial policy to manage change, governments must ask why the market is not doing a better job, and whether intervention can itself improve on the existing situation.
Generally, picking winners has not been a success, but decentralized incentives may be effective if their rationale has been clearly identified.
Key Terms
Problems
1
With constant AC and MC equal to 5, a competitive industry produces 1 million output units.
Taken over by a monopolist, output falls to 800000 units and the price rises to 8.
AC and MC are unchanged.
How would you calculate the social cost of monopoly?
What is it?
2
Explain the difference between UK and US policy towards monopolies and mergers.
3
Why do sports clubs have both an initial membership fee and an annual subscription for people who are already members?
4
In 1986 there was a major debate about whether Westland Helicopters should be allowed to merge with the US helicopter giant Sikorsky, or whether it should be forced into a European partnership.
What arguments can be used to support each of these viewpoints?
5
Compared with other countries, a relatively large fraction of UK R & d expenditure by the government is devoted to defence projects.
Is this necessarily economically undesirable?
6
Quite apart from aid to sunrise industries, UK policy provides substantial tax breaks for those investing in small firms.
Do you think the issues discussed in this chapter can be used to justify such a policy?
7
Common Fallacies Show why the following statements are fallacious. (a) Monopolies make profits and must therefore be well-run companies.
(b) Monopolies create social waste.
Society should prohibit any firm from having more than 20 per cent of the domestic market (c) It does not matter what other governments do; ours should not get involved in industrial policy.
(d) Mergers are obviously beneficial; otherwise companies would not bother merging.
BOX 17-1
AIRLINE DEREGULATION IN EUROPE
Although the charter market for package holidays is very competitive, scheduled flights in Europe are heavily regulated.
The extent of airline regulation illustrates many of the principles we are studying in Part 3.
It is also of interest because the US airline deregulation provides a live case study of the effects of completing the internal market.
We discuss 1992 and the completion of the single market in Europe in Chapter 34.
Lessons from the United States
Internal US flights were deregulated in 1978.
With the removal of legal barriers to entry, the industry quickly became more competitive.
By 1984 the number of airlines had risen from 36 to 120, fares were down 30 per cent, and passengers use was up 50 per cent .
Planes were fuller so airline costs fell.
And a more frequent service meant greater passenger convenience.
Free marketeers rejoiced as their predictions came true.
But the story after 1984 was rather different.
Cutthroat competition led to a lot of bankruptcies and mergers.
By 1989 only 27 airlines survived, and the top 12 controlled 97 per cent of the market.
Fares were rising again.
Powerful incumbent airlines found three ways to erect strategic entry barriers and consolidate their market power.
First, they moved to a hub-and-spoke system.
Each airline had a single airport (the hub), out of which long-haul flights operated.
They operated a system of feeder services (the spokes) to get you to the hub.
The hub-and-spoke system made it harder for small airlines to mount an effective challenge to major networks.
Second, the big airlines also owned the computerized reservation systems (CRS) which travel agents use to locate empty seats.
By programming the computer to ensure that their own flights appeared first on the screen, incumbent airlines put new entrants at a disadvantage.
Finally, they offered 'frequent flier discounts'.
Passengers clocked up bonus points by flying with a particular airline, but these could be cashed in only with the same airline.
These tactics illustrate how strategic behaviour can be used to consolidate existing market power.
And the US authorities operated a very lax merger policy: essentially, they approved all proposed mergers.
Thus some of the benefits of deregulation had been eroded by 1990.
Regulation in Europe
European scheduled air travel has been the subject of extensive regulation by government agreement.
Until the late 1980s, flights between two European capitals were restricted to one national carrier from each of the two countries (e.g. British Airways and Sabena on the London-Brussels route).
And competition was often nominal.
In some cases, the two airlines actually agreed to divide their joint revenue equally.
Fares were strictly controlled by international agreement.
Scheduled air travel was specifically exempt from EC Competition Law, Articles 85 and 86 of the Treaty of Rome.
What would you predict that the consequences of such intensive regulation would be?
High fares.
It is alleged that London-Brussels is the most expensive journey per mile in the world.
And there is a second implication.
In Chapter 12 we argued that, when firms have extensive product market power, trade unions will succeed in grabbing a big slice of these excess profits for their members.
Hence the lack of competition shows up in abnormally high airline costs.
The following table shows how beautifully this theory fits the facts.
Apart from the UK, which has begun to deregulate the airline industry, all continental airlines pay workers substantially more than in the United States, even though countries like Spain and Portugal have much lower living standards in general.
The future of regulation in Europe
In 1987 the European Commission reached a deal with airlines in the European Community extending the block exemption from EC Competition Law until 1991 in exchange for some small steps towards immediate increases in competition.
The EC clearly intends to keep up the pressure.
For example, when the UK government approved the merger of Britain's two largest airlines, British Airways and British Caledonian, in 1988, the EC Commission exercised its power to override national policy and insisted on some tough additional requirements (e.g. BA giving up some of its routes from Gatwick) before approving the merger.
What principles should inform European airline regulation in the future?
First, we should distinguish general externalities from things that directly affect market structure and the degree of competition.
Externalities requiring the regulation of safety and noise can be treated separately from competition policy.
The US experiment indicates that, contrary to popular belief, existing safety regulations were quite adequate: the US industry's safety record did not deteriorate after deregulation.
Second, in a completely free market, the US experiment indicates that existing airlines can exert too much market power, less through scale economies and innocent entry barriers than through strategic barriers erected by predatory behaviour.
Some regulation is still required to promote competition.
For example, unlike the United States, mergers should be very closely scrutinized.
And in Europe there is one additional consideration that is relatively unimportant in the United States.
Many of Europe's major airports are heavily congested.
This means that the allocation of 'slots' the time flights can take off or land is extremely important.
At present, existing airlines have the good slots.
If they are allowed automatically to keep them, this becomes a major barrier to new entry.
The right solution is to auction off slots to the highest bidder.
If the reform of Europe's airline regulation is carried out intelligently, the gains will be substantial.
Cheaper flights, greater efficiency, and lower airline costs as sheltered workers are forced to settle for more realistic wages.
But the US experience warns us that, unless we do it right, deregulation will not automatically yield substantial benefits.
18
Nationalization and Privatization
'In every great monarchy in Europe the sale of the Crown lands would deliver a much greater revenue than any which these lands ever afforded to the Crown...
When the Crown lands had become private property, they would, in the course of a few years, become well improved and well cultivated.'
Adam Smith, The Wealth of Nations (1776)
This chapter deals with the boundary between the public sector and the private sector.
As suggested by the quotation above from Adam Smith, the father of modern economics, the controversy is more than 200 years old.
State intervention can take many forms.
In previous chapters we have discussed tax incentives, competition policy, and industrial policy.
The main focus of this chapter is whether these policies are sufficient, or whether it is necessary for the state to have direct control of certain industries through public ownership.
Conversely, if public ownership has in the past been tried and found wanting, should the government now return these public sector corporations to private ownership?
Nationalization is the acquisition of private companies by the public sector.
Privatization is the sale of public sector companies to the private sector.
In the UK, the great wave of nationalizations -- industries such as rail and steel -- occurred during the Attlee government elected in 1945.
By the 1960s, most countries in Europe had a significant sector of industrial production under public ownership and control.
The tide has truly turned.
Major privatization programmes are under way not only in the UK but also in countries as different as France, Japan, Taiwan, Mexico, Poland, and Hungary.
In this chapter we analyse the nationalized industries, explain how they have been run, and assess their performance.
Then we discuss the case for privatization.
Finally, we examine the UK privatization programme in practice.
18-1 Nationalized Industries
In Chapter 16 we distinguished between government production of public goods such as defence and government production of private goods such as steel.
The nationalized industries are basically the part of government production that covers the provision of private goods for sale through the market place.
Thus the nationalized industries include British Rail and British Coal, but not the provision of defence or the provision of social services such as education or housing, which are not sold commercially.
Nationalized industries tend to be much more capital-intensive than the rest of the economy, and it is precisely the presence of these large capital costs that generates the economies of scale that make many of these industries natural monopolies.
Nationalized industries or other forms of public ownership are not confined to the UK.
Other Western countries face the same problem of natural monopoly in these industries.
The precise extent and nature of public involvement differs from country to country (for example, France had a very large degree of state involvement in industry), but the basic pattern is essentially the same.
Whereas European countries tended to acquire public ownership of the assets of natural monopolies, the United States preferred to handle the same problems through public regulation of industries whose assets were left in private ownership.
Regulatory agencies set prices and specify quality and quantity of output.
Hence, when countries concluded that the public sector was too involved in the economy, the initial policy priorities differed in Europe and the United States.
Having few nationalized industries, the US emphasis was immediately on deregulation, which began with airlines in 1978.
In Europe, privatization came first, but deregulation is now also under way.
Reasons for Nationalization
We distinguish three reasons why governments may wish to nationalize industries.
The first is the natural monopoly problem, which we examined in the previous chapter.
Large economies of scale mean that marginal cost lies below average cost.
Social efficiency requires that prices be close to marginal cost but this will imply losses for the natural monopolist.
Public subsidy may therefore be desirable, but the public commitment to pick up the bills requires public monitoring to ensure that the monopolist continues to minimize costs and produce efficiently.
Public management may be the simplest solution.
Since private shareholders cannot be expected to subsidize the wider goals of society as a whole, public ownership may then be inevitable.
Second, externalities may be important.
The social gains from an efficient road or rail network may exceed the private benefit for which direct users are prepared to pay.
In Chapter 16 we argued that taxes and subsidies could sometimes be used to offset externalities and guide market equilibrium to the efficient allocation.
But taxing road use may be administratively difficult and expensive, and direct regulation may be preferable.
Third, important judgements of equity or distribution may be involved.
A private profit-maximizing railway would close most rural railway lines.
Society might judge that this severely reduced the welfare of citizens in remote areas or that it promoted regional dissent which reduced the sense of national unity.
Again, society faces two options.
It can order a private supplier to provide these services, perhaps with the carrot of an explicit bribe or subsidy, or it can directly take over production to run the industry in the interests of the nation as a whole.
We now consider the principles by which nationalized industries should be run.
We distinguish investment policy and pricing policy, though the two are closely connected.
Decisions on pricing will affect the rate of return on investment and the incentive to carry out new investment, but past investment, by affecting the current capital stock, will affect current marginal costs to which prices will be related.
Investment Decisions
In Chapter 13 we discussed the demand for capital and the investment decision by a private firm.
We saw that we can think about this on a stock basis at a point in time or on a flow basis per period of time.
Using the stock concept, at a point in time an investment project is profitable if the present value of net operating benefits the stream of future operating profits discounted at the interest rate at which firms must borrow funds -- exceeds the initial purchase price of the new capital good.
Thus investment or additions to the capital stock will proceed up to the point at which the present value of the future profits equals the initial purchase price of new capital goods.
Equivalently, on a flow basis, firms compare the rate of return on a new investment with the interest rate at which they must borrow to finance the project.
Investment is profitable if the rate of return exceeds the interest rate.
Hence investment will proceed up to the point at which the rate of return just equals the interest rate at which firms can borrow.
Exactly the same principles carry over to social investment decisions provided we are careful to use social rather than private measures of costs, benefits, and the interest or discount rate.
We discuss these in turn.
The Initial Cost of the Capital Good
In the absence of any distortions, the private cost or market price of a capital good also measures its social cost, the opportunity cost of the resources used to make the capital good, or the value of the goods these resources could otherwise have made.
There are two reasons why society might consider that the social cost of a capital good is less than its private cost.
The first is unemployment.
If British Rail builds a railway line in South Wales by employing construction workers who would otherwise have been unemployed, the cost includes the wages of construction workers, but these do not represent a social cost if they would not otherwise have produced any goods.
Their  social opportunity cost may be close to zero.
Governments are sometimes suspicious of this argument because they are more accustomed to thinking as accountants about financial payments than as economists about production and distribution of goods and services that the price system is merely a mechanism to facilitate.
Nevertheless, you should be able to convince even an accounting-oriented government that it will save on unemployment payments if some of the unemployed are put to work.
A second reason why private and social costs may diverge is a production or consumption externality arising from the production of the capital good.
For example, the production of Concorde may generate advances in technical knowledge that would be of wider benefit to society.
Whether these are reflected in the market price that a private supplier of capital goods would charge depends on how well the patent system allows the private producers to reap the benefits of these advances.
If society gains more than the private supplier would gain, the social cost of the capital good is less than the market price that a private supplier would have to charge.
Valuing the Stream of Future Benefits and Costs
Whereas a private firm would simply use market prices to determine the stream of private operating costs and benefits, and hence the stream of future profits, to which the new investment gives rise, society will wish to value the social marginal costs and benefits of future output.
It will have to take into account externalities and distributional effects.
For example, the social profit in investing in rural transport almost certainly exceeds the private profit.
We shall shortly explain how prices should be set to reflect marginal social costs.
When a nationalized industry is committed to pricing in this way, it should use these prices to calculate the stream of social profit and the social rate of return on the investment project.
Choosing the Discount Rate
Private firms should invest if the rate of return on the project exceeds the market interest rate at which they can borrow funds.
That is the private opportunity cost of the funds tied up in the project.
Should nationalized industries use the same interest rate as private firms?
To assess the social opportunity cost of tying up resources in a project in the nationalized industries, we need to ask what would have happened to the resources had the project not been undertaken.
Suppose first that investment in the nationalized industries simply displaces private investment; resources used by British Rail are resources no longer available for investment projects in ICI or Shell.
Unless the nationalized industries use the same interest rate as private firms, there will be misallocation of investment resources between the private and the public sector: society can gain by reallocating resources to the sector with the higher rate of return.
This is a powerful argument.
Nevertheless, there are two reasons why society may wish to use a lower discount rate in public sector investment projects, thus requiring them to earn a lower rate of return than in private industry.
First, there is the question of risk.
In Chapter 14 we saw that risky activities on average will need to earn a higher rate of return to compensate for the higher risks involved.
But social risk can be reduced by spreading it thinly across a large number of people.
Private shareholders in Concorde would have required a high expected return to entice them into such a risky project.
A private Concorde producer could have borrowed only at a very high interest rate.
And as it turned out, private investors in Concorde would have lost a lot of money and minded terribly.
Even a disastrous outcome in the public sector can be financed by a fraction of a penny on everyone's income tax, which does not have nearly such disastrous consequences.
Hence, on average, risky projects in the nationalized industries do not require as high a rate of return as such projects in private industry.
The second reason arises from the fact that public projects do not displace just private sector investment projects.
Where public projects are financed by taxation, their effect is to reduce private household consumption by reducing the after-tax incomes that households can spend on goods and services.
When this happens, the social opportunity cost of the resources tied up in the public project is the return that households could have obtained on the same resources.
How do we measure this return?
Households will so arrange their affairs that the utility  return on future as opposed to present consumption is just equal to the real interest rate at which households can lend out money.
Why?
Because if the marginal utility of future consumption relative to present consumption differs from the rate at which households can convert current spending power into future spending power by saving and lending this money out at interest, households could improve their long-run utility by saving and lending a different amount.
This argument is spelled out in detail in Chapter 19.
Hence the rate of return on household savings is the relevant measure of the social opportunity cost of public sector resources that could have been used in private consumption.
But whereas firms borrow at the gross of tax interest rate to finance investment projects, households must pay income tax on interest earnings as a result of saving and lending.
It is the lower after-tax interest rate that measures the social opportunity cost of resources that could have been devoted to private consumption.
These two reasons -- the public sector's ability to spread risk more thinly and the lower after-tax interest rate relevant to resources displaced from private consumption -- justify the use of a public sector discount rate that is a little lower than the interest rate inclusive of tax and risk at which private firms must borrow.
The public sector should undertake some investment projects whose rate of return would be too low to justify the project in a private industry.
Pricing Decisions
If there are no distortions elsewhere in the economy, nationalized industries should set prices at social marginal cost.
Private marginal cost should be adjusted for any production or consumption externalities that arise, thus ensuring that society equally values the marginal cost and marginal benefit from the last unit of output.
Social marginal cost also takes account of any distributional value judgements the government wishes to impose.
Thus, for example, the price of a rural railway line should be less than the private marginal cost if society judges it a good thing to protect the living standards of people in remote areas or to foster a sense of national cohesion.
Conversely, bus fares should take account (among other things) of the social cost of the damage inflicted as a result of pollution from lead emissions from their exhaust systems.
We now consider some aspects of nationalized industry pricing that deserve special attention.
Short-run or Long-run Marginal Cost Pricing?
We have said that nationalized industries should price at social marginal cost, but should this be short-run marginal cost (SMC) or long-run marginal cost (LMC)?
In the short run capital or plant capacity is fixed but in the long run it is variable.
Thus LMC must also include the cost of providing plant capacity.
The once-and-for-all cost of building a plant is a stock concept.
The cost is incurred at the point in time when the plant is built.
But LMC is a flow concept, relating to the cost per period of producing output.
In Chapter 13 we explained how stocks and flows are related.
At the discount rate r the present value PV (a stock concept) of a permanent flow of c k per period is given by the perpetuity formula .
Conversely, the per period cost c k of an initial outlay PV can be expressed as .
The flow cost is the opportunity cost of the money or social resources tied up in the plant and is measured by the initial capital cost multiplied by the discount rate r.
In the absence of any divergence between private and social costs, it would simply be the interest payments on the money originally borrowed to build the plant.
Society will wish to evaluate this cost as the social cost of building the plant multiplied by the social rate of discount which we discussed above.
Figure 18-1 shows the short-run marginal cost SMC of producing output.
For simplicity, we assume that existing plant has a maximum capacity of Q units of output and the SMC is constant at the level c up to full capacity and then becomes vertical.
No matter how much is spent, output cannot be increased beyond Q given the existing level of plant or capital capacity.
Since SMC measures the marginal social opportunity cost in the short run of the resources used to produce this good, price should be set at SMC to equate marginal social cost and marginal social benefit.
Thus if  demand is low the industry should produce Q 1 and will operate at less than full capacity.
Given the higher demand curve D 2 D 2 , short-run marginal cost pricing will lead to a price P 2 and the industry will produce at full capacity.
Since output cannot be increased above Q, a price P 2 higher than c is required to ensure that the quantity demanded is no larger than the maximum available quantity that can be supplied.
Figure 18-1 MARGINAL COST PRICING.
For simplicity we assume that short-run marginal cost is constant at c until full capacity is reached, after which short-run marginal cost curves become vertical.
Thus SMC corresponds to plant with a maximum capacity Q ' and SMC 2 to plant with a maximum capacity Q 2 .
Suppose SMC is relevant and demand is D 1 D 1 .
In the short run the efficient output is Q 1 and price should be set at c.
If demand is D 2 D 2 , pricing at short-run marginal cost leads to a price P 2 on the vertical part of SMC.
LMC also includes the capital charge c k .
Beginning from the plant size with maximum capacity Q 0, in the long run society can gain the triangle ABE by expanding plant capacity until LMC equals social marginal benefit at E. At this plant size Q 2 the price is P 1 and still lies on the short-run marginal cost curve which is now SMC 2 .
Short-run marginal cost pricing equates immediate marginal costs and marginal benefits.
In the long run it is by making efficient investment decisions that capacity is adjusted to ensure that this price also reflects long-run marginal cost.
Suppose in the long run the industry can build more identical factories.
For example, with a slightly higher capital stock the industry will then face the short-run marginal cost curve SMC 2 with constant marginal cost c up to the new level of full capacity Q 2 , after which marginal costs become vertical.
But LMC also includes the per-period opportunity cost of the resources tied up in the capital stock, what we might call the capital charge c k , whose calculation we described above.
Figure 18-1 assumes that this capital charge c k is given by the vertical distance .
Thus we draw the LMC curve as a horizontal line at the height .
The long-run marginal cost of another output unit is both the marginal operating cost and the capital charge c k per unit of output.
Suppose the demand curve, reflecting social marginal benefit, is given by D 2 D 2 and that the industry begins with the capital stock for which Q is the maximum output and SMC the relevant short-run marginal cost curve.
Pricing at short-run marginal cost, the industry will charge a price P 2 .
Although this is the socially efficient output in the short run it is not efficient in the long run.
Once capacity can be varied society will wish to produce at point E, where marginal social benefit equals long-run social marginal cost LMC.
By increasing the capital stock until its maximum output is Q 2 society can gain the area of the triangle ABE, which reflects the excess of marginal benefit over marginal cost when output is increased from Q to Q 2 .
We can now see how pricing and investment decisions are interconnected.
In the short run output should be priced at SMC, which is the social opportunity cost of the resources employed.
However, in the long run society should invest if the present value of future benefits exceeds the cost of adding to the capital stock.
Instead of this stock evaluation of investment decisions we can use the equivalent flow evaluation.
The capital stock should be increased if the marginal social benefit exceeds the  long-run marginal cost inclusive of the capital charge.
Thus, although prices are set according to SMC, in the long run investment will change full capacity and shift the SMC curve.
If price exceeds LMC it will be socially profitable to increase capital capacity, and this process will continue until price and marginal benefit equal LMC.
Hence, although prices are set according to SMC, in the long run investment will adjust the capital stock and full capacity output until price also equals LMC.
In the long run the efficient allocation is at point E in Figure 18-1, where marginal benefit, SMC, and LMC coincide.
Peak Load Pricing
In Chapter 9 we argued that a private monopolist would make higher profits if it were possible to price-discriminate, charging different prices to customers whose demand curves were effectively distinct.
The problem with price discrimination in goods is that a secondary market is likely to be established in which low-price customers resell to higher-price customers, thus tending to equalize the prices customers actually pay.
Resale is not a problem when the commodity is a service (train journeys, electricity supply), which must be consumed as it is purchased.
Consider the problem faced by a producer of electricity.
It faces a high demand for electricity at breakfast time and dinner time, moderate demand throughout the rest of the day, and very little demand at night.
To cater for demand at peak times it has to build extra power stations, which are then idle during times of day when demand is lower.
Thus in the long run, taking account of the capacity cost of building the extra power stations, the marginal cost of supplying peak users is very high.
It does not make sense to charge all users the same rate, for users at different times impose very different marginal costs on society.
It makes sense to charge peak users higher prices to reflect the higher marginal costs they impose.
Peak load pricing is a system of price discrimination whereby peak time users pay higher prices to reflect the higher marginal cost of supplying them.
Peak load pricing has two attractive consequences.
Not only are peak users paying for the high marginal costs they impose, but also those users who would not mind consuming at a different time (e.g. households with night storage heaters, who can use electricity at a time when marginal costs are low) are induced by cheaper prices to switch to consuming at off-peak times.
By spreading total daily consumption more evenly, society reduces peak demand and has to devote less resources to building power stations whose number is determined by peak usage.
In terms of Figure 18-1, the efficient pricing policy makes prices vary with short-run marginal cost.
Off-peak users, with the lower demand curve D 1 D 1 , do not exhaust full capacity and pay the marginal operating cost c exclusive of capacity charges for the marginal plant.
Peak users with the higher demand curve D 2 D 2 , pay a higher price P 2 on the vertical part of the SMC curve through which their demand curve passes.
Given this efficient pricing structure, what is the efficient quantity of investment?
Investment should increase capacity up to the point at which the once-and-for-all cost of a new plant equals the present value of operating profits when the efficient pricing structure is in operation.
Equivalently, on a flow basis, the average daily price (in Figure 18-1, the average of the prices P 2 and c, weighted by the fraction of the day for which the demand curves D 1 D 1 and D 2 D 2 are relevant) should cover long-run marginal cost inclusive of capacity costs.
Thus in Figure 18-1 the efficient level of capacity might be Q if the weighted average of P 2 and c is equal to .
To sum up, prices should be set at short-run marginal cost.
This reflects the actual opportunity cost of the resources to society as a whole.
When there is a daily or seasonal pattern in demand, peak load pricing reflects the different short-run marginal costs of supplying different customers at different times of day or times of year.
When these principles are implemented, the amount of total demand met on the vertical part of short-run marginal cost curves will depend on the level of total capacity.
At low capacity, prices will frequently be high since they are set on the vertical part of the short-run marginal cost curves.
In the long run, investment and the efficient level of capacity should be determined to ensure that, when pricing at short-run marginal cost, average  daily or yearly prices equal long-run marginal cost.
Electricity Pricing in the UK
The following example combines the principle of peak load pricing with the idea of a two-part tariff that we introduced in Section 17-4.
(In the UK electricity generation used to be a nationalized industry known as the Central Electricity Generating Board (CEGB).
We expect the underlying pricing structure to remain rather similar when the electricity industry is privatized, but closely regulated, in the early 1990s.)
Marginal costs are highest during peak hours, both because LMC is high, reflecting the fact that peak demand requires additional capacity which is costly, and because SMC is high.
Figure 18-1 explains one reason why SMC is high during peak use.
At full capacity SMC is vertical and high prices are required to choke off demand.
However, Figure 18-1 is simplified by assuming constant average and marginal short-run operating costs up to full capacity.
In practice, the CEGB had a range of different power stations and used the less efficient higher-cost stations only when demand was at its peak.
When these stations are in use SMC is higher than during off-peak hours.
The CEGB operated peak load pricing in two distinct ways.
It used capacity charges to directly recoup costs of building plant, and then it sold electricity by the kilowatt hour.
In 1986-87 the CEGB charged area boards 10 for every kilowatt of basic demand plus 23 charge per kilowatt demanded during peak hours.
These capacity charges were intended to cover the annual cost of the interest payments (at the social rate of discount) on the money used to build the plant.
Peak users paid more because they force extra plant to be built.
In addition, there were energy charges intended to cover short-run marginal costs of operating the plant.
These too followed peak load principles.
In 1986-87 the CEGB had a complex structure of energy charges, ranging from 1.4 pence per kilowatt hour (p/kWh) for use at night during weekends to 3.8 p/kWh for use at breakfast time on a weekday.
In addition, there was a further surcharge of 2.5 p/kWh during the hour and a half of heaviest demand during the day, at whatever time that occurred.
Thus, the CEGB had explicitly tried to implement the pricing principles we have been discussing.
18-2 Nationalized Industries and Government Policy
When industries were first nationalized, many of them immediately after the Second World War, the government had two concerns.
First, that nationalized industries should broadly break even, and second, that they should be run not directly by ministers but through boards with a considerable measure of managerial independence.
In 1967 the government accepted many of the economic principles we have set out in the last two chapters.
Nationalized industries were directed to adopt marginal cost pricing.
The government fixed the social rate of interest or test discount rate at 8 per cent and urged nationalized industries to undertake explicit present value calculations for investment decisions using this discount rate.
This rate was considerably closer to the discount rate being used in private investment decisions than to the after-tax interest rate faced by households.
The government believed it more important to balance public investment against private investment by keeping their discount rates in line than to balance overall investment against consumption by using a test discount rate closer to the rate at which consumers would swap current for future consumption.
Finally, the government accepted the distinction between private and social costs.
Where, for example, British Rail wishes to maintain a loss-making rural railway for social reasons, the government attempted to value the divergence between private and social cost and made an explicit subsidy payment to British Rail.
Thus commercial considerations in running the railway could be kept distinct from the value judgements the government wished to adopt.
The 1967 policy was organized along classic first-best principles of Pareto efficiency.
Gradually, however, governments came to feel that second-best considerations could not be ignored.
In particular, they became increasingly worried about two things.
First, to the extent that nationalized natural monopolies made losses because marginal costs lay below average costs, the government became increasingly concerned  about the social cost of the distortions introduced elsewhere by the taxes required to finance these losses.
Second, governments became increasingly convinced that the commitment to meet any residual losses was leading to sloppiness and a failure to minimize costs.
These concerns partly explain the switch in policy in 1978 where first-best requirements for marginal cost pricing were partially replaced by a return to an emphasis on average costs.
The 1978 changes were (1) the introduction of a required real rate of return of 5 per cent, not on new investment projects, as the test discount rate implied, but on existing capital assets, and (2) the introduction of cash limits for each industry, a target profit or loss specified by the government in the light of the industry's circumstances.
In general these cash limits were tighter than the losses industries had previously been making.
The cash limits were intended to reduce the social cost of distortionary taxes elsewhere, to impose 'financial discipline' in the hope of attaining a performance that was closer to that which was cost-minimizing for any output, and to increase the bargaining power of management in wage negotiations.
Workers could no longer rely on wage increases that would simply increase the industry's losses and the government subsidy required.
The switch in policy in 1978 was reinforced by the 1980 Competition Act wherein the Monopolies and Mergers Commission was given powers to investigate nationalized industries as well as private companies.
These investigations were intended as 'efficiency audits' to check up on management performance.
These investigations such as the 1981 report on the CEGB have been quite critical of the way nationalized industries have been run.
Assessing Nationalized Industry Performance
Many people have strong views on whether or not the nationalized industries have been a success story.
Sometimes these views are based on reasoning that an economist would judge fallacious.
For example, unregulated private monopolists should have no difficulty earning profits.
Conversely, natural monopolies in the public sector which at times have been instructed to use marginal cost pricing will almost inevitably make losses.
Simple profitability is not the right yardstick by which to assess performance.
Similarly, at various times nationalized industries have been instructed to pursue social objectives or to help implement incomes policy by holding their price increases to less than the general inflation rate.
Assuming that the government has plenty of other policy measures at its disposal -- tax incentives, subsidies to private activities, competition policy, industrial policy, and so on the key question in reaching an assessment is surely a simple one: are enterprises more efficient in the public or the private sector?
Remember that economists mean two things by efficiency: first, that firms are on the lowest possible cost curve (absence of slack and waste); and, second, that the balance of activities in the economy is Pareto-efficient: resource reallocation would not increase social welfare.
If, and some would say it is quite a big if, other policies can be used to get the right balance of activities in the economy, we can confine our attention to the narrower questions: are firms more likely to attain the lowest possible cost curve in the public or the private sector?
Many economists have tried to investigate this question.
Before reading further, think about the last few chapters and decide what you expect their empirical studies have found.
Evaluating the results of 30 studies conducted in different industries in different countries, George Yarrow of Oxford University reached the following conclusion.
Private sector firms tend to be more efficient than public sector firms, provided both operate in markets facing strong competition.
When effective competition is absent, or when other market failures are very important, the evidence is much less clear-cut, and sometimes favours public sector firms.
Was that what you expected?
We now develop this theme in more detail.
18-3 Private or Public: The Issues
Firms do not become efficient by accident.
It requires sustained effort and leadership by management.
We need to think about the differing incentives for managers in the public and private sectors.
Incentives for Private Managers
In theory, private managers' performance is monitored by actual and potential shareholders.
If managers do badly, the company's directors may be voted out of office at the annual general meeting of shareholders.
Moreover, if bad management is perceived by the stock market, share prices will be lower than they might have been and a takeover raider may see an opportunity to buy up the company, install a better management, improve profits, and hence make capital gains when share prices subsequently rise.
Together, these threats are supposed to discipline managers and keep them on their toes.
In practice, these threats are weak and not very credible.
In the first place, individual shareholders face a free-rider problem: provided other shareholders monitor the management, everything will be fine.
So everyone tends not to bother.
Second, takeover raids typically bid up the share price significantly, so again the incumbent management has considerable leeway before it is likely to get into trouble and be out of a job.
Finally, incumbent managers have a lot of insider information about the true state of the firm, information that is not available to either existing shareholders or potential raiders.
Because shareholders know all this, they try to give managers a direct incentive to care about profits and cost reduction.
Senior executives get large profit-related bonuses.
These are a carrot, for those who like carrots; but shareholders have no accompanying stick, which effective two-handed discipline would require.
In these circumstances, we cannot guarantee that private management will be efficient unless it is subject to effective competition and challenge, either from other competitors in the market-place or from a tough and watchful regulatory body.
When the private firm is very large, and therefore subject to little domestic competition, and also produces in a sector sheltered from imports and foreign competition, it is quite likely to be a relatively sleepy monopolist unless it has regulators to whom it must account.
And these regulators must be capable of resisting the regulatory capture we discussed in Chapter 17.
Incentives for Public Managers
Whereas private managers sometimes face a weak market mechanism for discipline and control, public sector managers face no market mechanism at all.
Now everything depends on the effectiveness of the government as a watchdog.
In principle, the government might do this at least as effectively as private markets.
It too can offer performance-related bonuses, and it does not face the free-rider problem which confronts individual shareholders.
On the other hand, its civil servants probably have less training in business evaluation than do private sector analysts, and, of greatest importance, there is the overwhelming political temptation for the government of the day to hijack the nationalized industries and make them an instrument of whatever is the pressing problem of the day.
Until the 1980s, nationalized industries were often instructed to cut prices to help control inflation, increase prices to help raise revenue for the Chancellor, take on workers to alleviate unemployment, and sack workers to curtail government expenditure.
Whatever else, privatization does act as a pre-commitment that there will be less political tampering with these industries.
On average, this seems likely to make it easier to run them more efficiently.
Thus we are sympathetic to Yarrow's conclusion.
Provided either competition or effective regulation provides adequate discipline on private managers, on balance, companies will tend to achieve lower cost curves in the private sector than in the public sector.
If, also, market failures are relatively unimportant, or can adequately be offset by other policies, then on balance, privatization is likely to be desirable.
But both caveats are important.
Privatization is not everywhere and always the best policy.
The Importance of Ownership
It follows from the arguments above that the transfer of ownership from the public to the private sector is not in itself the most important issue.
Neither public nor private owners happen to be great at monitoring managements.
If ownership transfer to the private sector has any direct benefit, it is only the one we cited above; namely, that it makes it less easy for governments to use industries to meet the requirements of other political dictates.
On the other hand, it should not be imagined that private industries are immune from such considerations.
If you doubt this, talk to any private oil company that has been involved in the North Sea.
Such companies have at times faced a petroleum revenue tax rate of over 90 per cent, considerably in excess of anything they were led to expect when they first went into the North Sea.
Thus, most economists agree that the key issue is not ownership itself but rather the severity of the market competition, or its substitute government competition policy, which the industry faces.
Sheltered monopolies will be sleepy and slack no matter who owns them.
Conversely, the major benefits of privatization are likely to stem not from literally returning ownership to the private sector, but rather from associated measures which make it clear to managers that they are expected to compete effectively and efficiently.
The litmus test of this assertion is the fact that so many UK nationalized industries managed to improve their performance in the run-up to privatization, while they were still in public ownership.
And industries such as British Steel dramatically improved their performance in the early 1980s, while they were nationalized industries for whom privatization at that stage was still a distant prospect.
Selling the Family Silver?
Some people worry that selling off state assets mortgages the country's future.
Others think privatization makes the state better off.
Is there any simple way through the confusion in the public debate?
You already know enough to work out the answer for yourself.
The market price of an asset is simply the present value of the income stream to which it entitles the owner.
Suppose a fair price is 100.
If we sell you the asset, we get 100 in cash but you get an asset worth 100.
Neither of us is better or worse off than before.
Of course, if we put too low a price on the asset we are selling, perhaps because we want large queues of buyers for this asset so we can claim how popular it is, we make a loss on the sale and you make a profit.
That is one way the government could be worse off by privatization.
How could the government be better off by privatization?
Essentially for the reason first advanced by Adam Smith in the quotation at the start of this chapter.
If nationalized industries previously had bad performances, they would have been worth little had they been sold at that stage.
If management policy is changed prior to privatization, the present value of future profits rises at this point, and it is this that is the source of improvement to the public finances, not the subsequent change of ownership.
Only if one believes that the prospect of privatization is the only way to improve management performance in nationalized industries can one attribute the benefit to privatization itself.
The British Steel example suggests that this belief is at best doubtful.
Finally, notice that the question of mortgaging the future hinges crucially on what is done with the privatization revenue.
If it is invested in physical capital for the public sector or used to retire outstanding government debt, it is not at all imprudent.
Only if it is 'blown', for example on a tax-cut-financed consumer boom, can the government be said to be piling up financial trouble for future governments.
18-4 Privatization in Practice
In the UK, privatization began not with sales of companies but with the sale of public sector council housing to incumbent tenants.
Between 1979 and 1983, local government sold almost 600000 houses or flats, and at the peak were raising 2 billion a year in revenue from this source.
Table 18-1 shows the principal privatization of public sector companies in the UK.
Note that revenue from sales was relatively small until the British Telecom flotation in 1984.
From 1986 onwards, the Thatcher government raised 4-6 billion a year by selling off some of the major public corporations -- British Gas, British  
Airways, and British Steel.
And following the successful privatization of water supply, the early 1990s will see the privatization of electricity supply; the CEGB will be broken up into two companies (National Power and Powergen), from which the sale proceeds will dwarf the proceeds of previous privatizations.
Table 18-1 The UK privatization programme
Table 18-2 shows the price at which some of the early privatizations were offered for sale and compares this with the free market price established on the first day of trading on the stock market.
While it is of course very difficult to fix an offer price several weeks or even months in advance who knows how share prices in general will change in the meantime?-- the table does seem to confirm our earlier claim that, on average, the offer price was set below the free market price.
Table 18-2 Offer prices and prices on first day of trading
Earlier, we stressed the importance of the extent of competition that newly privatized companies would face.
We discuss this under three headings.
Significant Competition and Little Market Failure
Examples in this category include Cable and Wireless, National Freight, Sealink, and Jaguar.
Generally, these companies have been highly successful since privatization, even after account has been taken of nationwide trends affecting all companies.
Special mention should be made of the National Freight Corporation, which was exceptionally successful after privatization.
This company was very special in the extent of both worker participation in management, and employee ownership.
It was not a typical privatization, since all shares were sold to employees -- what we call an employee buyout.
Partly Competitive/Partly Regulated Industries
Examples of this kind of industry are oil, aerospace, and air transport.
Within this category, Britoil, Enterprise Oil, and British Aerospace had already been privatized by 1986, with British Airways and the British Airports Authority soon to follow.
It is too soon to reach any definitive judgement, given the relatively few companies privatized and the extreme movements of oil  prices which make it hard to compare that industry with others.
However, we can say that to date privatized companies in these categories, when measured against all other companies, have not displayed any dramatic improvement since privatization.
Market Power and Regulation
Having privatized most of the smaller companies first, by 1984 the government turned to sales of the giant corporations which are close to natural monopolies.
Whereas water and electricity were broken up into different private companies between which there might be a degree of competition, British Telecom (BT) and British Gas were essentially privatized intact.
We have considered at length why natural monopoly leads to socially inefficient outcomes: too little output and too high a price in that industry.
Moreover, in Chapter 10 we discussed how large firms enjoying a significant but not prohibitive entry barrier could erect strategic entry barriers to consolidate their market power.
In short, the privatizing of giant firms not subject to severe international competition is unlikely to be attractive unless it is accompanied by regulation.
Before discussing regulation in detail, we must make one obvious point.
Competition policy is the device already in place to combat market power.
At a time of large-scale privatization, one should be seeking to strengthen competition policy.
Privatization should not be seen as the signal for the government to withdrawn from the economy entirely.
The regulation of British Gas followed the general principles laid down for BT two years earlier, so it will suffice to concentrate on the latter.
As we discussed in Chapter 17, the first element in the regulation of BT was the 'RPI -- x' formula, to apply over an initial five-year period and then be reviewed, and with x equal to 3 per cent.
Second, BT was required to guarantee certain public services such as public telephone boxes and the 999 emergency service.
Third, one small competitor -- Mercury, a subsidiary of Cable and Wireless -- was licensed to compete over part of BT's activities.
OFTEL was established as a regulatory body to supervise these arrangements.
Critics of this framework have argued that it is quite inadequate to cope with a firm with the market power of BT.
Because BT was not restructured or broken down into more manageable units which might compete with one another(the US solution to their phone giant AT & t ), competition between giant BT and tiny Mercury is unlikely to constrain BT much.
All the responsibility has been thrown on to the regulatory body OFTEL, when it might have been possible to help the market work better.
APRIL, 1402 TO JULY, 1403
1
The girl had chosen to take her seat in the one spot from which the opening of the inner door would afford her brief glimpses into the prince's audience-chamber.
She sat erect in her black, sombre and motionless on the bench against the tapestried wall.
Thomas Prestbury, abbot of Shrewsbury, had given over his entire lodging to the prince's party, since there were ladies among them, and the castle, tightly-garrisoned and well-supplied as a vital base against the rebellious Welsh, had little comfort to offer the Lady Percy and her women.
For his own part, the prince would cheerfully have bedded down in the cramped military quarters he normally used on his periodical visits, but he was punctilious in providing every amenity for his guests, and the greater space and grace of the abbot's apartments made approach to his own person easier, and brought more petitioners in search of his favour, which at once satisfied his thirsty sense of duty, and wore him out into childish sleepiness by nightfall.
He was fourteen years old, intelligent, forceful, capable of listening attentively to his ministers and then overruling them and going his own way, capable, even, or so they said, of arguing a case strenuously and sensibly against the king himself in Westminster, though he seldom won his way there; but he was still a boy, unpractised, with little experience yet of living.
The girl in black was nineteen, five years older than the boy to whom she was bringing her petition.
She had watched others  come and go, and seen the hall empty round her, until only she and her opponent remained; and she had felt no impatience, only a growing resolution.
And all this while, from behind her mourning veil, she had fixed her eyes on the inner door, and watched for revelations from within.
The narrow chink presented to her vision showed her the spot where the prince sat.
She found him surprisingly ordinary, a solemn-faced boy in plain, clerkly brown clothes, long-legged and angular like everybody's young brother, with a fierce, cleft chin and huge, attentive hazel eyes.
But at least his concentration never flagged.
She was encouraged, because it seemed she could rely absolutely on that devoted attention, but discouraged, too, because he looked and was so young, and what could he know of marriage and widowhood, and the things that happen to women?
He might will well to her, and yet be too green to do her anything but harm.
She never moved or relaxed her watchfulness; but in her very stillness there was something of violence, as though a touch might cause her to spring into startling and daunting life.
When the man who waited with her crept to her shoulder and whispered in her ear, as he did several times between his nervous pacings about the room, she made him no answer, and never seemed even to be aware of him, though her braced tension made it plain that nothing that passed in this apartment escaped her instant notice.
'You'd do well to think better of it, and come home.
Do you think I won't make it worth your while?'
She gave him no sign.
He ranged about the room uneasily and leaned to her ear again: 'Waste of time!
He won't receive you!'
But he knew and she knew that the prince denied access to no one who ventured to appeal to him.
Those who had no case had no such courage, either.
'Is it likely he'll listen to you, against me and my house?'
The answer to that she did not yet know, but she had her own answer already sworn, and she would not go back on it.
'Come, be reasonable!
Listen to me, girl -- I'll make your  fortune!
I mean you nothing but good, why should we quarrel?
Give over this fool plea, and be wise for yourself and your sire!'
The outer door was thrown open with haste and ceremony, to admit a cloaked and booted gentleman who swept through the anteroom on a gust of chill April air, shedding a knot of servants and gallants at the threshold, and hurling all doors open before him with an alacrity that spoke to her of royalty.
She turned her head towards the servant who had just hurried to let him in to the audience-chamber, and put out an imperious hand to arrest his attention.
'Who was that?
He that just went in?'
She had seen little but the outline of him, and the walk, which was individual enough to be remembered, once seen; rapid and vehement, with a long stride that barely lit upon the earth before leaving it again as vigorously.
And a passing glimpse of a profile clear as bronze, and at the moment of passing as aloof and serene.
The page was disdainful.
'Do you not know the Lord Henry Percy, the prince's governor?
He's just ridden in from London, and his lady's here to meet him.
He's newly made the king's lieutenant here in North Wales, now we're as good as at war.'
'Are we as good as at war?' she asked, and there was no way of knowing whether she laughed or was alarmed behind the widow's veil.
'With this Glendower rampaging round Wales as free as a bird, and threatening Ruthyn every time Lord Grey turns his back?
And urging on Ireland and Scotland to his help?
Can you doubt it?'
Her face was quite still behind the shrouding veil, giving away nothing.
She had said all she had to say.
So that sudden presence was the great Hotspur, the most celebrated, the most gallant, perhaps the last, knight-errant of the age.
A strange man -- or perhaps a plain man lost in a world where most other men had grown strange -- collecting superlatives to himself as Saint  Sebastian collected arrows in the wall-paintings.
Something had blown through the room with him, a gust of exhilarating air trapped in the folds of his garments, leaving a breath of his own vigour behind.
'You can still withdraw,' whispered the wheedling voice at her shoulder.
'Come, be wise in time!
You shan't regret it.'
The door of the audience-chamber had opened.
She saw the prince's chamberlain lean out and speak to his waiting page.
She felt the burden of their eyes upon her, and rose from her bench silently, waiting.
'Mistress Hussey -- His Grace will receive you now.
Master Hussey, he begs you also be in attendance.'
She walked into the prince's presence, her anxious enemy treading hard on her heels.
And she thought as she crossed the threshold: He is still there.
You have more audience, Julian, than a mere royal child, whatever his goodwill.
You can address yourself to a man, and at least hope for a quick ear and an open heart!
'There is still a lady without,' the chamberlain had reminded them respectfully, 'who has a petition to your Grace.
And a gentleman who desires to speak in answer to it.'
They broke off their colloquy at once, postponing all that they had to ask and to answer.
'I'm sorry!' said Hotspur.
'We have time, and indeed I did mark this lady waiting in your anteroom.
Yes, surely have her in.
If she has far to go, the evening will soon be setting in.'
'You'll stay with me?
I should appreciate it.
I have no notion what her case may be.'
He was quite without knowledge of women, but granite in the acceptance of his responsibility.
'Admit the lady,' he said, and sat down again in his chair of state; though it was but a rather uncomfortable chair, not raised by a brace of steps like the abbot's own judgment seat.
She came in with a light, wary step, made a deep obeisance just within the doorway, and then advanced to the prince's chair and  sank at his feet, touching her lips to his proffered hand.
The boy took her cold fingers in his, and raised her.
'Madam, you have a petition to us.
I pray you speak out, and we shall listen.
You, sir, are a party to this lady's plea?'
'I desire to speak in answer to it, your Grace,' said the man, stooping obsequiously to the extended hand.
He was a handsome person in his florid, full-fed way, ruddy and brown-haired and aware of his consequence.
The woman was a mystery, tall and slender in her black, straight and steely as a boy, and thus far silent.
Silent women are always formidable, and always mysterious.
'Madam, I see by your habit that you are in mourning, and for that I am sorry.
How may I help you?
And what is your name?'
The girl raised her hands to put back the veil from her face.
Her youth blazed at them suddenly like a torch kindled, a thin, bright, deeply-moulded face all pearl-tinted skin over abrupt, burnished bone, with a wide, firm, full-lipped mouth, and dark eyes.
On either side her head gleamed coiled braids of dark-gold hair, almost pale copper in the subdued light of the room.
The intensity of that face turned her mourning and her stillness into the mere dark casing of a lantern.
'Your Grace is kind.'
Her voice was guarded, mellow and low, a well-schooled boy's voice.
'My name is Julian Hussey.
I was born Julian Parry, only daughter to Rhodri Parry, a merchant of this town, dealing in wool and woollen piece goods, and married by him a year and a half ago to Master Nicholas Hussey, who held lands here north of the town.
My husband died two months ago, and we had no issue.
By my husband's will, and the custom of his house, all his manor and lands go to his nephew and heir, and I am without purpose in the household longer.
For me there is no function left but to return to my father's house, and care for his old age.'
She lifted eyes like gem-stones, ruby-bright in the light of the torches, black, surely, in full daylight.
They looked at the prince, and passing by him, fastened with intent upon  Hotspur's watching face, as yet impassive.
'Your Grace, my father is but a merchant, and to marry me into this noble family he gave me a noble dowry, eight hundred marks.
Now he desires, as is but right, that my dowry should be returned with me.
But my husband's heir, my nephew by marriage, will not repay what is due.
And if he receive not his right, my father will not receive his daughter.
I ask justice of your Grace, for unless your Grace do me right, I am without redress.'
She spoke to the prince, she even looked at the prince, but what she said was addressed to the man who sat withdrawn at the prince's elbow.
'Your Grace,' said Hussey, bent reverently double, and eyeing his widowed kinswoman from the corner of one eye, 'if your Grace will but hear me...'
'You may speak to the matter.
You are the heir?'
'Yes, your Grace.
I am Edward Hussey.'
'And what this lady says, is true?
You have no quarrel with it?'
The prince leaned back in his chair, and waited, and wondered.
'None, your Grace, so far as it goes.
All is as my most valued kinswoman has told you, but your Grace will comprehend that money matters are none so simple.
My lord, my inheritance is indeed enough, but my immediate resources in money are limited.
I would with goodwill repay Master Parry his daughter's dowry, but at this moment it is out of my power.
I have not the sum to hand, and cannot raise it even by loan.
Yet I have made provision,' he said fatly.
'There need be no dissension.
Until I can repay the dowry, I have set aside apartments in my household, where my kinswoman may keep her own establishment however she may dispose.
Her keep shall be a charge upon my estate, and for company she shall be assured of the society of my own wife, her close kin and most attentive servant and friend.
I cannot offer more, nor with better will.'
'It would appear,' said the prince mildly, 'a fair offer.
What say you, madam?'
'Your Grace must consider,' said the girl, lowering her eyes, 'that I am still young.
My father may well wish for me a second marriage, and to that end I must be at offer soon, and with a proper dowry.
It is well I should marry again from my father's house.
And that would, indeed, be my wish.'
'Nor would I stand in my kinswoman's way, my lord,' said Hussey eagerly.
'So soon as may be, Master Parry may make such disposal as he thinks fit, and I will never say the loath word.
It is my grief that I am unable to repay at this time the money that is due.
Within a year there should be no such restriction upon me.
And Master Parry himself is willing to agree to the offer I make.'
It seemed that with every exchange the man was growing more confident, and the girl, for all she maintained her fiery calm, a little more pressed and on the defensive.
'Your Grace,' she said quickly, 'even a year may deprive me of my best prospects.
And what should I do in another woman's household, who have been used to managing my own?
Your Grace knows that two women in one hall is not good sense.'
'Yet, your Grace, with all the goodwill in the world, I cannot repay this year, or not until after the harvest.
There is not so much ready cash in my treasury.
And it is but right that until I can make restitution, my kinswoman's expenses should fall upon me.
I don't seek to escape my duty.'
He was, perhaps, a thought too complacent on the subject.
The girl flashed one brief look in Hotspur's direction, and for an instant the glitter of her eyes seemed to him hunted and wild.
He leaned forward to the arm of the prince's chair.
'Yet -- with your Grace's permission?-- if this eight hundred marks was paid over no more than a year and a half ago, surely it cannot all have been used or turned into goods so soon,' he said.
'Unless your estimable uncle had expensive amusements, Master Hussey, you must surely have come in for this very money along with the rest.'
The girl was quick to catch at the hint, drawing breath gratefully.
But so was the man; he must have come prepared even for this, and there was surely something unnatural in his readiness.
'My lord, you say truth.
My uncle was already old, and not given to rash spending, indeed he carried his carefulness too far, and was something of a miser.
And what he has done with such ready money as he kept about his house neither I nor his clerk can tell as yet.'
'He did not trust his clerk?'
Hotspur asked negligently; and it was at the girl's wry face that he looked.
'He trusted no one,' she said with sudden muted violence, and paled at the bitter sound of her own voice.
A year and a half ago, Hotspur thought, this fierce faun was surely no more than seventeen years old, and married off, like many another, to an old miser three or four times her age, for the sake of a noble name and a set of paltry quarterings, and the hope of a grandson set up in the landed estate.
And here she is, so short a time after, widowed and childless, with nothing gained and much lost.
'The better reason for feeling certain that his eight hundred marks are still unspent, and still safely bestowed somewhere about his household.
It will be needful to find them, Master Hussey.
You may very well find it possible yet to send them home again in the very minted pieces in which they left home with the lady.'
'So I hope, too, my lord, for then our whole dispute is solved.
But if my fair kinswoman will but be patient and abide in my house until we have brought all into order, and made proper search for this money...'
Even a few weeks would do for him, thought Hotspur, and caught the girl's eyes fixed upon him in silent desperation and appeal; though indeed it was so imperious as to be more of a demand.
But for God's love, he thought, half-intrigued and half-exasperated, if she has anything to charge against him why does she not speak?
Why has she not spoken long ago in the right quarter?
For it seems she has a father!
He leaned to the prince's ear, and said in a rapid whisper: 'There's more in this!
Call an adjournment until tomorrow and have the merchant summoned to attend.'
The prince, at a loss with the complexities of women, was quick enough to pick up an offered lead, and not infrequently bettered the prompting, as he did now.
'It seems to me, Mistress Hussey, that we should hear you further on this matter, and that we have need of more certain information than is available here.
I will hear the case again tomorrow, at three in the afternoon, and I desire that Master Parry shall also appear then to speak for himself.
It is not enough for one party or the other to tell us what his mind is, that he must do in person.
And further, Master Hussey, it would be helpful if you would bring with you your manor clerk, to speak to the value of your own holdings and this inheritance.
We cannot make a judgment without knowing what your resources are.'
That tasted bitter, thought Hotspur, watching Hussey swallow it down perforce, for on the face of it it was reasonable, and minor lordlings from the fringes of Wales do not argue with the prince of Wales.
And sweet!
For the girl's eyes, which he was beginning to read as he read his own children's, had flared briefly in vindictive joy, and again veiled themselves.
There goes a woman who would dance on her enemy's grave, he thought, curious and thoughtful.
And lie down in her friend's, too, if need be!
He understood instinctively the nature that deals in extremes.
They had both made their reverences, and were withdrawing, markedly separated by three feet or more, and shrinking fastidiously from approaching more closely in the doorway, when Hotspur called the lady back.
'Madam, I misdoubt that you should be abroad after dusk alone, either here or in the town.
With his Grace's goodwill, I would offer you a night's lodging here in the safety of the abbey hospice, and tomorrow you shall have escort to bring your father to the audience.
My lady and her women are lodged in the  guesthouse, you need have no fears in joining their company.
If you accept, his Grace's page will conduct you, and commend you to my wife.'
She had risen from her deep curtsey, and stood for a long moment gazing steadily into his face.
This was the moment when she elected him, with her eyes and her heart wide open, knowing what she did.
She never turned back from it after; nor was it her habit, any more than his, to repent of what she did.
'My lord,' she said, 'I know of no greater honour you could offer me, nor any that I would more joyfully embrace.
With all my heart I will go to your lady.'
Elizabeth Mortimer, Lady Percy, was thirty-one years old, and had everything woman could wish for, royal blood in her veins, wit and spirit and beauty, a husband the envy of all his peers, a little son the budding image of him, and a baby daughter on whom he doted.
She had also a gallant and generous heart wide-open to affection.
She received the unknown and unexpected widow like a welcome cousin, asking no questions and extorting no confidences, but offering on her own part enough warmth to make the early April evening glow.
She talked of her children, far away in the north at Alnwick castle with their household, and of the late Spring when she would take her husband home to them for a brief visit.
He was newly come today from Eltham, the king's favourite manor, where he had been a witness at his Majesty's proxy marriage to the widowed Duchess Joan of Brittany.
The king had been many years a widower.
She spoke with courteous compassion of the widowed, and for her guest's sake did not dwell on their sorrows.
And well might one so gloriously married feel pity for all those less happy than herself.
They were still sitting together when Hotspur came from the prince's apartments.
Julian rose as he came in, and bowed herself unnoticed from the room; but from beneath her lowered lids she saw them meet, and the private radiance that lit their faces was  still a dazzle in her eyes as she closed the door upon them.
She saw him cross impetuously to his wife, lift her bodily by the waist, and kiss her heartily; and she knew that before the latch clashed into place they were in each other 's arms.
Proof, she thought, astonished, that men and women do love.
For she had seen little enough evidence of it in her life so far.
She watched them in hall, from her place among Elizabeth's ladies, and they were two bright lanterns burning with sparks of laughter and joy and prodigal kindness, for pure pleasure of being together after an absence.
They lit the whole of the high table, startling the prince's grave face into gaiety.
Julian watched them, and every moment of watching only confirmed his election, though it set him another league away from her.
Nevertheless, he had felt her need and advanced to fill it.
She did not suppose there would ever be another such moment to hope for, yet few could ever have had so much.
It was unlooked-for grace that after supper he should send his page to ask Mistress Hussey to be kind enough to come and speak with him in the small chamber the prince was using as a study.
And a prodigy that he should receive her there alone.
'Now that we are private,' he said, looking her in the eyes, 'you may speak out openly what for some reason you did not wish to speak out before the prince.
Why are you so urgent to get away from your husband's house and back to your father's?
Oh, the wish I well understand.
You're young and handsome, and your family clearly wealthy enough, you have a life before you.
But if that were all, would it be so insupportable to stay a few more weeks in Hussey's household?
And you yourself made it clear to me that you would not stomach so much as one more day, if we could but be induced to open a way of escape for you.
If we had not provided you this delay, and this opportunity, what would you have done?'
His voice was quiet, matter-of-fact and kind, and his eyes  smiled at her with interest and curiosity, she thought, beyond the mere charm he must use by native grace upon all who came near him and were weaker than he.
She debated for a moment whether to answer, and then how to answer.
'My lord, I have good reason to know that it was you who said a word for me in his Grace's ear, and gave me this day for thought, and even offered me arguments I was not wise enough to find for myself.
If you had not, and the prince had consigned me again to that man's house, I should hardly have known what to do, or where to turn.
Though I should have discovered a tolerable way,' she said, with force and finality.
'Needs must, and I have had some practice.'
'And why was it so urgent to you to go?
And to him that you should not go?'
'He wants me to remain under his roof,' she said deliberately, 'because I am young and handsome -- it was you, my lord, who called me so -- and his wife is neither.
Widowed and childless and penniless, a woman is at the mercy of her husband's kin -- or so he reasoned.
He offered me a fat life as his whore, and I laughed in his face.
Since then, twice already he has attempted me by force.
In his house I should never be safe.'
'Then why have you not spoken out and accused him?' demanded Hotspur warmly.
'If he has done you such injury this fellow must be held to account.'
She smiled, a small smile as bitter as rue that just curled the corners of her lips.
'He has done me no injury -- yet.
It is I who have left my mark upon him.
Your lordship may see the print of my affection scored across the back of his left hand and wrist, if you care to make him hand you his manor roll tomorrow.
But give him his due, he does not give up for a scratch or two, and not even a dagger could hold him off for ever.'
'No!' said Hotspur, flushed with anger.
'This cannot be allowed to pass without redress!
You should have spoken out boldly before the prince.
Why should you hesitate?
Do you think he would not see right done to you?'
'With all my heart, my lord, I believe he would, but it would be a costly redress to me and mine.'
She leaned forward, flushing in her turn.
'I do not want right done.
All I want is silence, and to go back to my father's house.'
'And how much silence will there be when your father hears how you have been used?'
'He will not hear,' she said vehemently.
'Never from me, and never from you, my lord, if you regard my good.
What I have told you here I have told you in confidence, whether I exacted any promise from you or no, and you cannot violate my trust.'
'God forbid!' said Hotspur, baffled and frowning.
'But why should you be willing to let so gross an injury pass?
For I think it is not out of any maiden meekness in you,' he said, with a sudden blazing smile that turned the stricture into a compliment.
'Because he is English, and noble, though he may be only the small cousinly fry of his house.
He is distant kin to Arundel himself.
And I am the daughter of a Welshman, a settler in this town for many years, but still Welsh.
Your lordship of all people knows what that means now.
We Welsh are all suspect since the Lord Owen raised his banner in mid-Wales two years ago.
We cannot own property, or hold office, we must give hostages for our good behaviour -- we are prisoners and outcasts in our own town.
My father was to have been bailiff, now that and all other honours are out of his reach.
He had a great and flourishing trade with the Welsh weavers, bringing their wool and piece goods into England here, and transporting them to London.
Now all trade with Wales is forbidden.
My father is a proud man.
If he knew of the insult offered me he would want revenge -- and against the Husseys he cannot possibly speed.
I want my father living, not dead.
What use is revenge to me if it means the destruction of what life we have left?
No, my lord, I beg you keep this thing secret.
I need only one thing from you and his Grace, to return to my home and take part at least of my dowry with me.
And if you can give me that relief, I will be grateful.'
He thought it over, to judge by his frowning face with something of a struggle, and gave way reluctantly.
'Well, you shall have it as you wish, though it goes against the grain with me to see this fellow go free.
But you must give me leave to confide in the prince, and I make the same promise for him.
Tomorrow's audience must be managed plausibly, he will need to know what he is about.'
She hesitated.
'And can you so promise for him, my lord?'
'Never fear for that.
His Grace is wise beyond his years.
I pledge his secrecy as freely as my own.'
He sat silent for a long moment, considering her soberly from beneath knitted brows.
She felt that the interview was over, and yet his stillness held her still, and the solemnity of his regard filled her with a curious sense of freedom and enlargement, as though she enjoyed the very fashion of intimacy with him that he might have shared with a man and his peer, even with the prince himself.
Those who were his companions in arms must have known such moments.
He was at peace with her while he reasoned and thought before speaking; and the lengthening silence had neither weight nor tension, but lay between them gently like the comfortable warmth of a fire.
She had never before known how to be still and wait; there had never been anything as worth the waiting for as this was.
'Lady,' said Hotspur, 'there is yet something you may do for me, if you will.'
'I will,' she said, without haste or hesitation.
'Gently, you do not know yet what it is.
You said well, no one knows better than I what it means to be Welsh now in these border towns such as Shrewsbury, with trade forbidden, and restrictions bearing down hard on every man with Welsh blood.
I know that for two years now Welsh labourers and students and journeymen have been stealing away out of England, to make their way back to Wales and enlist under the banner of their self-styled prince.
I know the manner of this same Lord Owen's  disaffection, and I tell you honestly, though I must and will fight with him wherever I may, yet I think his cause not all empty and not dishonourable, and there have been faults committed against him foolishly and grievously, which a man may well resent.
It is gross waste that a man with such qualities should be turned into a traitor, when I think he never willed to be any such monster, but wanted only his native right.
And if I could bring him to reconciliation with the king, without more bloodshed and without revenge, I would count it a good deed both for England and Wales.
You see, I speak with you as with my own conscience.
I have already attempted something in this vein, but there are those who deal distantly with such realities, and the prince and I, both, suffer from them hardly less than Owen does.
Nevertheless, it is our purpose to keep a channel open by which we may still talk with Glendower.'
'I have already heard,' said Julian, 'how in the council of last November you spoke for a parley with the Lord Owen, and told the king's ministers that he would be willing to talk peace.
And how there were some there who thought no shame to recommend that you should invite him to a meeting, so that he could be taken unawares and murdered there.
And I have heard,' she said, 'how you answered.'
The eyes that dwelt upon her had opened wide in surprise, perhaps even in amusement; but the mind behind them was held and deeply exercised.
How came she to know so much?
'And how did I answer?'
'You said it was hardly in accord with your rank and honour to make use of the oath of fealty to lure a man treacherously to his death.
And the prince,' she said, watching him narrowly, 'approved you.
And so the matter lay.'
His eyes were very bright now, though the constancy of their regard never wavered.
Looking down into them, she could see clean into his soul.
If ever this man needed to deceive, she thought, dismayed, he might as well bare his neck for the axe, for  he could not save himself.
He has no mask to cover his intent.
He is like a naked light.
And his lady is no different.
Does that come from living and loving together?
Or from choosing after his own kind?
He said drily: 'You are very well-informed concerning the workings of the king's council.'
'The Welsh are all wizards,' she said.
'And the Lord Owen is the greatest wizard of all.
Have you not heard as much?'
'Lady, I study to learn.
There are more secrets than those belonging to the English, and well I know it.
Your father, as I hear, was the main channel through which the Welsh woollens reached England.
Parliament in its wisdom or folly has banned all such trade, true enough.
But parliament is far away, and the border of Wales is very near, and is there a soul in these parts who does not know that smuggling goes on day by day, and that life here would be impossible if it did not?
Almost half of Shrewsbury has Welsh blood, or at the least friends in Wales.
How can such a border be blockaded?
It cannot, nor it is not.
The Severn, even in spate, is not impassable.
Now supposing, lady, supposing, I say, that there are traders who have kept their contacts across the water.
Then such men can be of use both to England and to Wales.
Peace is in the interests of both.
You believe me?
You trust me?'
'I believe you,' she said.
'I trust you.'
She did not say: 'I love you.'
'Good!
You have spoken openly with me, and I speak openly with you.
No penalties!
I am neither a bailiff nor a tax-collector.
What you say to me has been said only to me.
Is there a reliable man who goes back and forth freely into Wales?'
'Yes,' she said, 'there is such a man.'
She had said it, and she was committed, and had committed others along with herself.
The thought did not daunt her, but she had great need of a moment of silence, to take breath and consider how much she dared tell.
He had dealt with her honestly, not as with a mere woman, one who must necessarily be only on the fringes of her menfolk's concerns.
He had asked her only for what he needed, though there must be many blunter questions he could have put to her had he been so minded.
She trusted him, for her own part, without reserve or doubt; but she was trusting him for those who crossed the river and took the risks.
Moreover, he himself had loyalties of his own to preserve, and it was well for him that he should not know too much.
'Yes,' she said, picking her way delicately for his sake, 'the cloth still finds its way in, though not so freely and not so profitably.
Yes, the money still finds its way out to pay for the cloth.'
'And letters can find their way both in and out,' he said, 'by the same route?'
'Yes,' she said, 'they can.'
Can and do, she thought, ever since that September of two years ago, before I was given in marriage, when Owen fired Ruthyn, and his men plundered Oswestry and Welshpool, and the king came storming through here and across the border with all his army, but never found any enemy to fight.
The Welsh had all vanished into the hills, as they always vanish, and Henry marched his army back through Shrewsbury empty-handed.
The despatches that mapped every move he planned went into Wales ahead of him by this same route.
And that was only the beginning.
'Will you bring me in touch with this messenger?'
She hesitated, pondering means.
'You must meet him?'
'I think it may be even more important that he should meet me.
Will he take your word for my good faith?'
'Do you take my word for his?' she said.
He smiled.
'Come, then, let's be plain.
I have warrant from the king to deal-to keep open, if I can, a means of communication with Glendower, any honest way of continuing the debate that may yet stop this fighting, and let tradesmen and students and friars move freely about their business again.
I want pacification.
I want an open border, as you do.
I want Owen back lawfully on  his own lands and in good odour again with his king.
What I do in approaching him thus I do with the king's goodwill.
But well you know, it seems, that one such attempt has been ruined and rejected by the council, and it may take us a weary while to find a fair agreement and a council sobered enough and sensible enough to abide by it.
Yet I think the trying well worth while.
If I move as privately as possible, it is to hold off the fools and rogues until the thing has a chance of success.
And if I handle all with my own hands, and keep the prince clear of it -- though he knows my mind, and it is his mind, too -- I do so to preserve him from harassment by those who will hear of nothing less than Owen's head on London Bridge.
The letters your man will carry will be in this hand!
He has a right to come face to face with me, and judge for himself whether he can honourably deliver them, and never fear that he is helping to lure a brave man to his death.'
'My man knows your reply to the council,' she said, 'as well as I do.
But your lordship says right, it is better there should be as few hands in the chain as may be.
No one between you and Prince Owen but one man.'
'Will you bring me to meet this man?' he asked.
'He cannot well come here.
Or to the castle.'
He nodded assent, and did not ask for a reason.
'I would rather come to him.
There must be no suggestion of the prince being involved.
If your father will receive me in his house, I should be glad.
I would not have him think I have taken advantage of his daughter, and shun facing him.
Nor that I will to borrow his messenger without his knowledge.
I need not ask if he can be secret.'
'He can be the most secret man alive.'
'Will you be my go-between to him?'
'Yes,' she said, 'I'll be answerable for my father.'
There was nothing he could have asked of her that she would not have done for him.
She had never been afraid to look straightly at whatever fortune sent her, to map its every feature, and acknowledge it for what it was.
She had examined with analytical precision her empty and disgusting marriage with an impotent old man, and her candid delight at the death that had put an end to it.
She had even considered the possibility that the old fool's frantic efforts to match a young wife had been the cause of his death, and the idea had not caused her to turn her eyes away and evade the issue.
She had rejoiced at her childlessness, and even found the heart to laugh at the lamentable end of her father's fond dynastic ambitions.
If there was one thing not ugly and absurd about that marriage, it was a small but irreversible change in her own situation.
She was no longer an unmarried daughter, but a widow, and widowhood represented status, and liberation not merely from spinsterhood, but also from any expectation of fulfilment in marriage.
Through that illusion she had walked with blessed speed, and out beyond it into a world of other possibilities.
Eventually the experience might have to be repeated, in her own defence, but she could not hope to find much more in it than she had found in this first venture.
There were other relationships.
Whatever she could look for in the future must be looked for outside marriage.
What was here being offered to her she took with both hands, roused and grateful.
She had not yet even recognised what it was, but she knew it for better worth than she had ever yet been given.
'You will go home with him tomorrow --'
'If he gets his money,' she said with a rueful smile.
'He'll get his money, or part of it, at least.
How soon can you bring me to meet the man you speak of?
If he goes and comes, it may be days yet?
Or weeks?
I cannot afford weeks.'
'He is here in Shrewsbury,' she said.
'It can be tomorrow.
Give me but one day, and by night he shall be waiting for you at our house.'
'I will come there after dark.
Tell me how to find the place.'
'It is a house in a court behind St Chad's church.
You may come there by the alley from the town wall, and leave the church  on your right.
There is a sign over the gateway, the Fleece.
Or if you will, I will come to the abbey gate, and bring you there by the least frequented way.'
'I would not so burden you.
No, I will come, and alone.
You and your house shall take no risk by me.
Oh, I know,' he said, 'the pains of being Welsh, and too close neighbour to the English, when the standards are out.'
He took her hand, not to kiss, but as he might have taken a man's hand who had met him fair and done him honour.
'I am ashamed that I have kept you from your rest, after so troublous a day.
Go now and sleep, and have no fear for tomorrow.
I will take care that you shall not be persecuted further.'
She crossed the chill, moonlit parclose of the abbey of St Peter and St Paul, the faint smoke of frost from her breath going before her.
The great church was mute and dark, for compline was long over, since the routine of the house clung as yet to its winter timetable, and the monks were in their beds until midnight should rouse them for matins.
She walked alone in the silence, hearing her own muted footfalls like echoes of past or future, she could not distinguish one from the other.
Faithful to his orders, she had no fears for the morrow.
There was no man but one who could trouble her rest ever again, and whatever disorder or ordeal he cast into her path she knew she would go gladly, and gather like flowers.
She entered the warmth of the hospice as Lady Percy was leaving the hall to go to her own apartment, attended by a single demure damosel.
Elizabeth saw her, and stretched out a generous hand:
'Mistress Hussey, I missed you!
Margaret has orders for your comfort.
Sleep well, and be sure they will see right done to you.
Good-night!'
Julian looked after her, the high, exultant step, the reared head under its coronal of brown hair, the lofty joy in every movement.
And she thought: What must it be like to go to one's marital bed with delight, instead of disgust?
And to such a man, instead of an old fool with the shakes, and sweating like a pig in an ague?
And how strange, she thought, that I have come from him, and you are but now going to him, and yet I do not envy even you, the most enviable of women!
For God's sake, what is it I have got from him, that it sets me so high?
2
She suffered one paroxysm of doubt, the first and the last, and a matter of shame to her as often as she remembered it after, when the hour of noon came and passed, and no one sent for her to go into the town and fetch her father to the audience; and when she ventured to enquire, she was told that one of his Grace's clerks had already gone to summon Master Parry, and she need not concern herself in the matter.
That struck her hard; but the shock passed as quickly.
It seemed the prince would not have a witness influenced, even by his own daughter.
She had relied on having time and opportunity to confide in him, and set his mind at rest about the unexpected summons.
She knew him, and could imagine his state now, half surly defiance, half anxious and defensive fear.
He was not a brave man, and often said so, brandishing his supposed nervousness like a banner.
If they did not let her see him until they both appeared before the prince, what wild errors might he not commit in his insecurity?
Yet she could not, once that first convulsion was past, feel any unease.
Hotspur had promised her a fair deliverance, vouching for the prince no less than for himself, and in his promise she believed as in the mass.
So she waited for the summons to the prince's presence, and went with a demure step and a high heart when she was called at last.
Her father was in the anteroom, waiting for her with a dour face and uneasy eyes, but so closely attended by page and  chamberlain that barely a word beyond her submissive greeting and his mumbled acknowledgement, phrased as a blessing but uttered like a malediction, was able to pass between them.
Then they were ushered together into the presence-chamber.
She looked round for Hotspur, but he was not there.
She was not to be trapped into suspicion again; if he knew his power so well that his presence was unnecessary, that was enough for her.
The prince was seated, not in his chair of state, but between two of his clerks at a trestle table, with a quantity of papers and parchments spread before them; and his treasurer stood at his shoulder, ready to advise if requested, but looking on so impartially that it seemed to her he had already done his share.
Before the table Edward Hussey stood hunched in defensive composure, very plainly dressed -- had he set out to demonstrate the modesty of his means?-- and with his countenance fixed in an expression of resigned and dutiful benevolence.
It was somewhat of a disappointment to her vengeful mind to consider that he might have taken yesterday's omen to heart, and prudently drawn in his horns, resigning his pretensions on her rather than venture even token opposition to the fiat of the prince and his governor.
It was, she supposed, a possibility.
The girl was well enough -- all the better for hating and fighting him!-- but not worth that risk.
There were plenty more to be had cheaper!
'Mistress Hussey,' said the prince as she entered, 'we beg your forgiveness for keeping you waiting a little beyond the time we appointed.
Master Parry, pray pardon also a summons at such short notice, but the case we have to judge concerns you nearly, and I make no doubt that your daughter's wellbeing is your first anxiety.
We have pleased to reserve this hearing until we had had an opportunity of acquainting ourselves with all the circumstances.'
He lowered his eyes for an instant to the parchments that littered the table.
The hovering clerk at Hussey's back was watching them narrowly every moment, as though one of them might elude him when he came to gather them up again.
'The  matter at issue, as you no doubt know, is the return to your household of your daughter, now widowed and without children, and the repayment of her dowry.
Madam, it is your personal wish that you should so return, is it not?'
Julian inclined her head and veiled her eyes.
'To the end that you may make a second and happier marriage from your father's house -- is that so?'
This time she raised her face, blazing with candour, and said: 'Yes!' boldly.
She was not one of those who cannot lie with wide-open eyes and angelic faces when needful.
'Such an intent we find wholly commendable,' said the child gravely, monumentally sure of his shaky ground, 'and we confide that your father must feel with you, as is but natural.'
He is acting, she thought, touched and elated, he is prompting my father; he has learned his lesson well.
But then she looked into his eyes, which dwelt upon her in huge solemnity, and knew that he was burningly sincere.
If he delivered her, with her endowment secured, as she had been promised he should, he did so of his own will, because he was convinced of her need and the justice of her complaint.
And he had been left to conduct this hearing on his own, secure that he would make the decision both Hotspur and she needed.
Never had he surrendered, or been asked to surrender, his independence of action.
Had Hotspur even confided to him all that she had urged and confessed, yesterday evening?
But yes, surely he had.
He had given the boy all the evidence, dumped it in his lap without ceremony, and left him to examine all, and act as the prince he was.
That argued a very profound knowledge of the royal mind on Hotspur's part, and an even deeper confidence in its infinite will to justice.
On what grounds, she reflected now, enlightened, had he vouched for the boy?
He had never said: 'The prince will do as I tell him', but simply: 'The prince is wise beyond his years.'
Why, after all, should that cause her any surprise?
The boy  had been in his tutelage now for two and a half years, closer far to him than to his own Lancaster kin.
'Very well!
We have now examined into Master Hussey's means, and we are satisfied,' said the prince, very gravely and courteously, 'that even though no exact inventory has yet been made of all the property passing to you, Master Hussey, by your uncle's will, yet you have certainly acquired assets which must be disposable, and of such a nature as to be very readily disposable.
You are already possessed of a substantial household, and have here been visited with a second.
We are taking into account that both manors must be properly manned and maintained.
But there is still a handsome balance of advantage to you.
And it is our judgment that you can and should pay at once a portion of Mistress Hussey's dowry, so that her maintenance may be assured, and she may return to her father's house from this court, as is her wish.
We cannot feel that such an arrangement is in any way unjust to you.
Even if we were to order the repayment of the whole sum at once, the amount would be less than if you were paying tax of a lawful fifteenth of your movable goods.
Which could well happen,' said the Prince -- was it possible that he was capable of a strain of malicious humour?--'whenever parliament meets.
But we are not demanding the whole.
One half of the dowry, four hundred marks, you will pay to our treasurer here by this day week, and we ourselves will see it conveyed to Master Parry.
The remaining four hundred you will pay through the bailiff of our town of Shrewsbury within six months from today.
We have given our judgment,' said the prince formally, and sat back in his chair with an authority and finality that no one cared to challenge.
There was nothing for Hussey to do but bow before the wind with as good a grace as he might, profess his resolve to do all that was required of him-at whatever penal cost to himself, his martyred countenance implied -- reverently kiss the prince's hand, and withdraw to his plain, melancholy wife and his two fat manors up-river.
Julian made her reverence in her turn.
Over the extended hand she looked up into the prince's eyes, and saw there the same candid regard she had seen in his model; yet the shafts that pierced into this boy's inmost being were somewhere shuttered close, standing off all communion.
There was no obliquity, no deceit; neither was there any revelation.
The boy had learned what the man would never learn.
She watched dispassionately, concealing a faintly malevolent smile, as her father's inflexible knee forced itself to bend before the wrong prince of Wales.
The return of his four hundred marks was a strong inducement, nevertheless it went against the grain with him to do homage to an Englishman.
'Master Parry,' said the prince, innocent of his offences as of the benefits that spoke loudest for him, 'we will make it our business to see this money duly paid to you.
Madam, I pray you may enjoy a more fortunate marriage hereafter.'
They were blessed and dismissed.
They went out into the precinct of the abbey church, and the showers had passed, and April wore its radiant face.
Distant across the river the towers of the castle rose against the sky, straddling the only land approach into Shrewsbury.
Pale, rushing clouds danced across a blue almost as pale.
'And what the plague,' demanded Rhodri Parry irritably, grasping her arm as they crossed towards the guesthouse doorway, 'did you want with starting such a frantic legal bother, without a word of warning to me, without a hint of your purpose?
Could you not have sat tight in the fellow's house for a few weeks longer?'
'Not a day longer!
I could not stand another hour of the man himself or his bleating sheep of a wife,' she said tartly.
'And moreover, could I know how long the prince would stay here?
It might have been no more than two or three days.
I had to strike now or never.
And you had best be grateful to me, for if you had left it to the little men of law he could buy better and shiftier than you, and you would never have got your money at all.'
The old man -- she was the child of a second marriage, and he had been well past forty before she was born -- snorted his disbelief, but she knew he was doing little more than vent the nervousness of the past hours, now that the suspense was over.
'He could not have denied me.
It was due.
Or else he could have fed and kept you for the rest of your life!'
'Then there would have been bloodshed, for I tell you I could not be in the same house with them and keep my temper.
And as for letting go of what he had, he would have gone to every possible shift first.
It might have cost you more to get it than the gold was worth.
You should never have been so eager to pay so much for so poor a privilege.'
'Poor?
Do you know the value of that manor of his?
It is a noble name and a noble family.'
'Very like, and I make no doubt there are even some noble members somewhere within the clan, but my husband was none, nor is this nephew of his.
I suppose if he had been wifeless you would have had me married off to him as soon as I was out of mourning!'
There had never been a time when they had not wrangled, and yet after his fashion he was fond of her.
'You might have done worse.
But let it go -- the man is married, and there's an end of it.
And I'm glad enough to have you back, since you could not give me the grandson I hoped for --'
'I could,' she said fiercely.
'Will you have me prove it?
On what stock shall we graft?
You should have bought me a man, and not a threadbare purse.'
'In God's name, girl,' he protested, shaken, 'what devil has got into you, to talk to your father so?
You were not wont to be so loud and bold.'
'I was not wont to be a married woman, and now a widowed woman.
The degree loosens the tongue.
' And she meant to use it to the full, though perhaps not in this wasteful way.
In one unwelcome marriage you learn much about the means of evading  a second, and still retaining the consequence gained by the first.
It had been her only gain; she did not mean to let it be whittled away.
'Wait but a few moments here for me,' she said more gently, 'for I must make my farewell to the Lady Percy, who has been more than kind to me.'
Elizabeth gave her a warm, vivid smile, and her hand, and good wishes to go home with her.
Her faith in the prince's justice perhaps larger than that, in God's -- had never admitted any question of the outcome, assured that a plaintiff her husband favoured must be in the right.
Julian went out from her strangely uneasy for creatures who walked through the world so openly and confidently.
It is too simple a matter to damage those rare few who are too brave, too scornful and too trusting to put on armour.
'Lady Percy!' said her father, musing, as they walked side by side towards the gatehouse.
'That's Harry Percy's wife?
He they call Hotspur?
He was not there with the prince at the audience?'
'No,' she said, 'he was not there.'
'A pity!
Since you needs must drag me here, I should have liked to see this Hotspur men talk so much about.'
She said nothing to that.
It might be needless caution, yet she kept silence by instinct in all public places upon all that touched Hotspur's affairs, and more because they were his than because they affected the state, and such gravities as peace and war.
All the way home she held her tongue, answering only in monosyllables to her father's habitual complaints and strictures, which never were meant to be taken too deeply to heart.
It was not so long a walk, though it led her back in twenty minutes through a year and a half of her life, and was quick with memories both sharp and sweet.
The sun had come out fully over the abbey mills and the narrow bridge of Meole brook, and in the foregate there was bustle enough.
This English gate into Shrewsbury was guarded less stringently than the Welsh bridge on the further side, and therefore used more freely, and the  drawbridge was lowered from earliest dawn.
The river was high and sparkling, piling light debris of branches and leaves against the piers of its four stone arches; and beyond, the walls of the town rose, and the tunnel of the open gate.
Down the steep slope from the walls to the shore the narrow terraces of the abbot's vineyard ranged like a staircase, the vines like charred black stumps as yet barely showing the first shadowy tint of green.
Reflected light shimmered upwards from the rapid water, and rippled along the stone of the ramparts.
A fair city, something dishevelled after uneasy times, and hampered and straitened now -by the loss of the thriving Welsh trade which was half its life, but still capable of living on its own fat for some while yet, and still hard to take and invaluable to hold.
'You'll not find it easy to get the keys from old Joanna,' Rhodri remarked with malice, as they passed in through the archway and continued along the town walls.
'She's had her own way too long now.'
She cared less than nothing for the privileges of the housewife; the old woman could have kept the keys for ever, and Julian would have been indifferent.
Yet her homecoming and remaining at home would have to be justified, and there was no other immediate way except by assuming the direction of the household.
'Leave me to fend for myself,' she said.
'What's my due I shall have, and today.'
She had not tried, as yet, to see beyond today; she sensed that what she saw when she did lift her eyes might well be a void, and as bleak as winter ice.
Another marriage -- probably as chill and loveless as the last -- or a comparable prison in her father's house.
Some women found at any rate a sanctuary behind the veil, but a convent was more likely to be a hell of rebellion and constraint to her.
And yet, she thought, as they picked their way gingerly along under the stooping eaves of the alley that led to the rear of St Chad's church, to avoid the running kennel thawed and filled by the morning showers, the finger of God had intervened  in her life only yesterday, and might again lean down to point out for her an acceptable and fruitful way.
What she wanted was not a sanctuary, but a battlefield.
But nobody less than God was ever likely to offer her one.
The house of the Fleece was timbered and dark and beetle-browed on the side next the street, with jutting upper storeys and shuttered windows.
There was an arched cartway into the yard, and a narrow wicket let them in through the thick oak portal to the cobbled court, ringed round with stables and storehouses.
Another prison, indeed, but at least this one had a visitation promised.
Not until they were within, the door closed after them, and the silence of the thick walls like a seal against the world, did she tell him what manner of guest she had invited to his dwelling, and how soon he was to have his wish.
It was past nine o'clock when he came, late enough to have emptied the streets.
The house of the Fleece was fast shuttered by then, the wicket in the yard door closed but not barred; and Julian was waiting in the doorway of the undercroft, her ears pricked for every footfall that passed along the alley.
She knew him when he came; though he trod quietly his step was unmistakably light and long and confident, and he was one of the few who came that way by night alone, yet not furtively.
She was at the wicket before him, and because the yard was dark, put out a hand to guide him within.
The implications of the contact she honestly had not considered; she would have done as much for any who came by night.
He accepted the service as naturally, closing long, hard fingers on hers to read the hints they gave him.
Her veins ran fire, flashing back like a powder-train to the heart, and there bursting in a brilliance and violence such as she had never experienced or dreamed of.
She contained it and gave no sign, for she was enlarged, as the night is by the moon.
She even detached herself composedly when he was within, to have  both hands free to drop the heavy wooden bar into place and fasten the door, and then took him by the hand again to bring him safely to the door of the undercroft, across the uneven stones.
Only when they were within the house, in the timber-scented darkness of the staircase, did she halt to kindle a light.
The tinder caught and glowed, the candle billowed, a small orb of yellow light between their two faces, and they were looking intently into each other 's eyes across the flame.
'You keep close and careful house here,' he said, with a small, grim smile.
'The Welsh in Shrewsbury do not invite notice, especially by night, my lord.
And my father is a timid man.
These precautions are not all for you.'
She had answered him, as he had spoken, in an undertone.
He cast a glance about him, noting the half-empty spaces of the undercroft, and the faint gleam of light from above-stairs.
'Are there servants in the house?' he asked in the same low voice.
'One old woman -- deaf, and in bed and asleep in the garret long since.'
'Your father has no journeymen living with him?'
'Two, but they sleep across the courtyard, above the storerooms.
You may be private enough, my lord.
The house is sealed.'
She held the candle steady until it burned tall and pale, and the light swelled and smoothed the mellow wood of the walls, calling gaunt shadows out of empty air.
Then she lit him up the stairs, and went before him into the panelled solar, where Rhodri rose from a tall chair by the fire to receive him.
'Father, here is Lord Henry Percy.'
She was quiet, muted, moving before her sire with downcast eyes and dutiful voice, and withdrawing into the shadows as soon as her errand was done, as though she had no part in what was still to do, unless to wait on their requirements.
For a moment Hotspur almost believed in her extinction, her relegation to the servant's role which was the lot of daughters in their parents' households; and even for a moment  it grieved him to believe in it.
But the face into which he had gazed across the candle-flame had been neither tamed, nor troubled by any foreboding.
She did what she chose to do, and was as for the time being she chose to be.
'Master Parry,' he said, 'I must thank you for granting me this interview, so strangely requested.
No doubt your daughter will have told you why I am here.'
'You are welcome to my house, my lord,' said Rhodri.
'For your kindness to my girl I am in your debt, as she is.
Yes, she has told me.'
Hotspur had felt some curiosity about this father of hers, for she was not a woman whose antecedents could easily be guessed at.
What he saw was a man of about sixty years, older than he had expected, but still hale, and of a powerful frame.
He stood no taller than his daughter, probably she exceeded him by an inch or two, but he had the shoulders of a bull, and a great head of brindled brown hair laced with grey, like his short, square beard.
He leaned and peered a little, but not from any weakness of the eyes, rather out of a fixed suspicion that caused him to study with narrow attention all who came near him, and especially strangers.
His gown was rich and sombre; he knew cloths, and had the means to buy the best.
He valued ceremony, too, perhaps as a barrier, negotiable when desired, but inestimable as a means of maintaining distance during a parley.
On the heavy oak table beside him there were good silver goblets set out, and a flagon of wine.
'Be seated, my lord!
Put off your cloak and draw near the fire.
You'll drink a cup of wine with me?'
Hotspur put up a hand to the furred collar of his cloak, and let it slide from his shoulders; and like a silent and attentive valet the girl came gliding out of the shadows and took it from him.
He was plain and sombre, dark brown from head to foot, to pass in the night unnoted though undisguised.
As soon as he was seated at the table she poured and handed wine, and he marked the  breadth and strength of the shapely hand that offered the cup, and realised suddenly, noting the blue cloth sleeve of her gown, that she had already shed her mourning.
'Master Parry, I have so much confidence in the good offices of your daughter that I have brought with me the letter of which I spoke to her.
Is he here that should deliver it?'
'He is here,' said a soft, deliberate voice from the darker side of the room, remote from the fire.
The narrow inner door had been invisible in the uniformity of the panelling, and its latch had made no sound as it was lifted; but suddenly there was a man framed in the doorway, a lean, wiry, lightly-built creature, stepping out of the wall with a conjuror's aplomb and a deer-hound's lanky grace.
Hotspur had swung round in his chair to face the voice, which had an aloof, noncommittal sweetness of tone, promising nothing.
It did not disturb him that he had been under observation; he was not accustomed to deprecate what he was, or to undervalue it.
He looked up with interest, candidly returning the inspection, at Rhodri Parry's agent.
The young man -- perhaps not quite so young after all, he might have been as much as thirty -- came forward into the room, closing the door behind him.
He wore the usual faded, dun-coloured every-day clothes of the peasant and labourer, coarse woollen chausses and short homespun tunic, with a capuchon pushed back from his head and dangling at his back; but the belt that circled his hips was of finely-tooled leather, and had straps to attach both sword and dagger, though he wore neither; and his boots were knee-high, and also of soft leather, no doubt hand-worked somewhere in Wales, from native deerskin.
His face was long and gaunt, tanned by outdoor living in all weathers to a deep olive tone, but his eyes, deeply-set beneath black brows that flared upwards like wings, were unexpectedly blue and cool and far-looking.
He wore a short, close-trimmed beard that scarcely veiled the shape of a wide, mobile, quirky mouth like a jester's; and though his thick crop of short, curling  hair was black, there were twin streaks of mingled rust-red in the beard's blackness, sharpening and tapering his long chin to a fiery point.
The corners of his mouth bit inward deeply; it was sometimes difficult to know whether he smiled, or had a wry taste on his tongue.
He stood unmoving to be examined, in no way disconcerted by the length of the scrutiny, and feeling no need to break the silence, until Hotspur said at length, with slow consideration:
'Somewhere I have known you before.'
'It may well be, my lord.
At any rate, your lordship will know me again.'
'I think so.
You might be all too easily remembered.
Yet I cannot call to mind where I have seen you.'
'As good a gift as being easily forgotten,' said Rhodri Parry drily.
'Your lordship desired to meet the man who can carry messages freely into Wales, and out again.
Here he is, and for his ability I can vouch.'
'I pray you -- with your leave, Master Parry!-- sit down with us.
It is fair you should have time to consider well.
You are vouched for,' said Hotspur, with the large simplicity that was warp to the weft of his equally vast pride, 'but no one has vouched for me.'
The young man sat down readily at the table, leaning his homespun elbows at ease; and Julian, without being bidden, came forward noiselessly and filled a cup for him.
Rhodri watched them with his hooded, wary eyes, and said nothing.
'He has avowed only that I can.
Not that I will,' said the young man coolly.
'That he can hardly promise.
Only you can do that.
May I know your name?'
'My name is Iago Vaughan.
I am a Welshman from under the Berwyns, and distant kin to the Tudors of Anglesey.'
Hotspur smiled, for the sons of Tudor ap Goronwy were first cousins to  Owen Glendower, prominent in his counsels, and in his warbands, too.
'Does your lordship require to know more of me?'
'Not even that, if you had not pleased to tell me.
But I like a face to have a name.
And it is a fair exchange, for mine you know before I name it.'
He turned suddenly to Julian, standing silent and attentive in the shadows, and said to her, with a hand outstretched: 'Hand me my cloak!'
But he did it with a warming smile and a ready assumption of her allegiance and willingness, more as if he had asked a small current courtesy of his wife than given an order to a servant.
And when she brought it to him, he did not take it from her, but only felt in the deep pocket stitched into its lining, and drew out a parchment rolled and sealed, which he reached across the table and laid before Iago Vaughan.
'You can read?'
'In four languages, my lord.
I was brought up in the cloister, and have Latin and French as well as English and Welsh.'
'Faith, I wish I could say as much, for I never mastered Latin, and have no Welsh.
But I write a tolerable hand in English.
I trust the superscription is clear?
I have not named the place where he is to be found, since I do not know it, and a week hence it may be very far from where he bides today.
Yet I make no doubt he can be found.'
By one of the Tudors, he thought, a finger could be laid on him any day of the year, I'll swear.
But he asked nothing; it was not his habit to woo men from their clan allegiance, or try to make dishonest use of a man he wished to employ honestly.
He leaned back in his chair, leaving the letter in Iago's hands.
'Will you take Owen my  message And bring me back his answer?
Will you undertake as much again, if this bears fruit?'
He looked at the old man, peering darkly under his down-drawn brows; and there was one who would have questioned and writhed and wondered, pondering long before he would have given any answer, and then, most likely, regretting the answer he had given, whatever it chanced to be.
But he looked at Iago Vaughan, and was suddenly aware that his motives would not be  questioned nor his matter suspected.
The thin, clever, spatulate fingers -- they could have been an apothecary's or a musician's -- held the roll of parchment delicately; the light blue eyes studied Hotspur's face without disguise and without wavering.
They saw a man curiously like and curiously unlike himself; like, in that he was no man's man but his own, and what he pledged, he would perform; unlike in his innocence, pride and primitive simplicity.
A man without a glimmer of serpentine wisdom about him, for all the sword-sharpness of his mind, a man who thought with his blood and his bowels, for good or evil.
A dangerous man, and a man who lived eternally in danger.
He wrote English vehemently, scoring deep into the vellum:
'To the most excellent lord, Owen ap Griffith, lord of Glyndyfrdwy and Cynllaith.'
From one haughty and courteous prince to another.
Only a few months ago, after this Lord Henry Percy had withdrawn to his other urgent command on the Scottish borders, Owen had run wild over most of North Wales, and made himself master of the counties of Carnarvon and Merioneth; and while the woollier heads in King Henry's council had seethed and talked bloody war, Hotspur had come swooping back to hold the balance so sturdily that he had been allowed, on the king's warrant, to approach the Welsh prince, and attempt to bring him back to his allegiance, on promise of honourable terms.
There were still Welsh grooms and servants and even lawyers about Westminster, to listen and observe and send word.
And there was no man in the kingdom who knew better than Owen what Hotspur's disdainful answer had been to the council's shameful proposal of murder in place of magnanimity.
This new approach, it seemed, was not to be made so publicly, not to be exposed to the expedient treason of little devious minds far removed from the battlefields on which honest men met, and contended, and killed  one another without malice.
Yet the girl had said that what he did, he did with the king's warrant, and the prince's approval.
They had learned, apparently, not to let the pack near the scent too soon.
'I will do your errand, my lord,' he said, 'and I'll bring you word again from the Lord Owen.
And for the future, so long as there is hope that this may speed, I will be your go-between.'
'God speed both it and you!' said Hotspur, and sat back with a short, sharp sigh of satisfaction.
'When do you set out?'
'Tonight.
But for your lordship's visit I should have been gone before this.'
'And how soon can you be again in Shrewsbury?
No, never tell me more of your goings and comings than is my due, I need only to know, as nearly as you may judge, how long I must bide here to wait for the answer.'
'Within a week, unless the Lord Owen move too often and too fast for me, I shall be here again.
How may I come in touch with you?'
'Why, that's no great problem while we remain at the abbot's lodging, since half of Shrewsbury and a good part of the shire goes in and out freely at the abbey, and you may ask an audience whenever you will, and always find yourself one of three or four, various enough to keep any man in countenance.
But it's well,' he said seriously, 'that you should have a token about you that will get you in to me or the prince at need, and stand you in good stead if ever you should fall foul unawares of any of his officers or mine.
I need your discretion, yes, but I will not have you brought into needless suspicion or danger upon this account.
Here, wear this!'
He plucked a heavy silver ring from the middle finger of his left hand, tugging it over the knuckle impetuously, and held it out across the table in his palm.
The single stone with which it was set, opaque in browns and golds, passed from hand to hand like a glowing eye.
'It is known to be mine, and I will give my chamberlain orders that it shall admit the bearer at any time.
If ever you have word when I am not by, it will bring you to the prince in my place.
I'll see to that.'
He smiled, seeing his pledge lie in the other man's palm, as yet only half-accepted.
'Put it on, and wear it.
It's yours, whether we speed or no, and some day you may need it.'
'I take it, then,' said Iago Vaughan slowly, 'since there may be a time when there'll be little leisure for persuasion.
Are you not afraid, my lord, that I shall use it upon my own occasions? which may not always be yours?
A princely warrant to pass where I will could be a godsend to such a man as I am.'
'Unless I have lost my judgment,' said Hotspur bluntly, 'you would not be beholden.
Though so it serve our purpose, I would not quarrel with a little license.'
Iago slid the ring on to his finger, and admired the deep sheen of the polished stone.
'If I have not lost my cunning, as you have not lost your judgment, it will never be used but to come to you with that which belongs to you.
What is it, this stone?'
'The Scots call them simply pebbles.
I got it while I was prisoner there, after Otterburn.'
He smiled a little ruefully, remembering his captivity and his costly ransom, fourteen years past now, before he married.
'And take this also, for you will have charges to meet on your journey.'
He had not asked, nor would he ever ask, what provision was normally made for these frequent and illicit journeys, the relays of horses, the hire, perhaps, of boats, transport for the cloth.
That was Rhodri's business, and if it was illegal, it was also inevitable and understandable.
He laid upon the table a drawstring purse of soft leather, that chinked faintly as it shifted and settled.
He saw Iago's lean and secret face stiffen, the blue eyes shrink to points of steel, and the linked hands draw back from touching.
'Ah, never look so, man!
I have not bought nor bribed you.
I am asking a service of you, and I will pay the expenses, as is but right.
You need not feel your hands tied -- if you can bring that  prince of yours and his armies victorious into Shrewsbury over my body in fair fight, go do it, and I'll never cavil.
You are as free a man as I am, we do but agree over the hire of a courier.
So you give me that service, there are no debts between us, and no obligations.'
Iago looked up at him over the wine with a face suddenly bright, astonished and disarmed, and burst into a muted crow of laughter.
'By God, my lord, I think you are a man after my own heart!
And I pledge you my word, and I take my fee.
I shall use it only on your business, and if your business is done and myself discharged before your gold is spent, you will take back the balance, or I will break your teeth with it.
And if I play you false, you may break mine.
Within the week I will bring you back your answer.'
He reached for his cup, and there was Julian with the flagon lifted, ready to refill both his and Hotspur's, so silently and impassively that they might almost have dreamed her into the fringes of their conference, but for the fourth cup which had appeared beside theirs, and which she was also filling to the brim.
Her hand -- that firm, forceful, boy's hand of hers -- lifted the goblet as they lifted theirs.
She drank with them as they pledged each other.
Her face was tranquil and still, but her eyes flashed like arrows from one face to the other, observing and remembering, before her large, creamy eyelids veiled their light.
Iago could not recall that he had ever truly noticed her until that moment.
She had a tension about her like a strung bow, and every bit as lethal, and she had a piercing beauty -- why had he never marked it, he who had known her nearly three years? that made his heart contract as he looked at her.
He could afford to study her, for she was not looking at him with any but surface attention.
She recorded everything that passed in the room, but she cared for only one person.
Nor was that one altogether unaware of her, or unappreciative.
But on what curious terms, only God knew.
He had never before seen man and woman  regard, consider and touch each other as man with man.
No, even that fell short of truth; there are degrees that cannot be accurately plotted.
They made their own terms of reference; she, perhaps, with knowledge and calculation; the man, after his kind, by impulse and the blind brilliance of his own nature.
He was on his feet, roused and content, and looking round for his cloak; and she was there with it, lifting it to his shoulders.
His eyes encountered hers, and smiled in pleasure and gratitude; hers smiled, too, but mutely, making no marked acknowledgement.
Rhodri Parry thrust himself stiffly out of his chair.
'I shall look to hear from you,' said Hotspur, fastening the clasp of his cloak, 'in a week, if all goes well.
God speed you, and bring me a hopeful answer -- both for England and Wales!'
'So I pray, too, my lord.'
'Master Parry, I owe you my thanks.
Should I ever be able to serve you, I pray you let me know.'
'We are still in your lordship's debt,' said Rhodri formally.
Hotspur had withdrawn a step or two towards the door when he turned again to look at Iago.
'Carry my respectful greetings and compliments to your kinsmen, Gwilym and Rhys ap Tudor and their brothers, and say I still owe them a shrewd knock for their taking of Conway on Good Friday of last year.
It cost me a deal in time and labour and money to get my castle back from them, and if we had not had such a pious garrison -- all but the lame and bedridden in church!-- they would never have prised their way into the place.
I look for an adjustment some day.'
'I make no doubt they will accommodate your lordship,' said Iago.
'Yet if I speed too well you may have no remedy.'
'I'll bear that as the yoke of God.
Mistress Hussey -- Master Parry -- Master Vaughan -- I bid you good-night.'
'I'll light you down to the gate,' said Julian, the candle ready in her hand.
The door of the solar closed behind them; their feet felt a way silently down the staircase.
At the house door she left the candle  burning upon a shelf within, and took him by the hand to lead him across the stones of the court to the wicket gate; but he halted her suddenly, drawing her back within the shelter of the doorway.
In the soft, still light she felt his eyes earnestly and faithfully searching her face.
'Madam, are you content?'
She knew he did not mean with their night's work.
Evenly she said, in the same undertone he had used: 'All is very well with me, my lord.'
'I would I could be sure of it.
I have in some sort made myself a party to your situation, and it will go hard with my conscience if you find yourself no better blessed than in your old condition.
If ever you should be in need of a place of refuge, you have a resort in my wife's household.
You need only come to her.'
She offered her thanks with composure.
In her heart she thought: Kind as she has been to me, and much as I respect her, your wife, my lord, is the last lady living to whom I am likely to apply.
Nor is it a place of refuge I want, among the women of a countess's retinue.
I do not know as yet what it is, but I know it is not that.
'Well, bear it in mind.
In case of need, it will always be open.
God be with you!'
'And go with you, my lord,' she said.
His hand withdrew from hers, and he was gone, stepping silently through the narrow wicket into the darkness of the street.
But the warmth of his touch remained with her, in spite of the chill of the night, as she dropped bolt and bar into place, and fastened the house door.
She let herself into the comfort and glow of the solar to hear her father's querulous voice complaining, in terms in which surely he himself did not believe: 'My mind misdoubts me we have done wrong to have any part in this.
How if he is sending you with a bait to bring Owen to his death under cover of a parley, as the  king's council urged?
Our lord would not be the first to come to bargain and stay to bleed.'
'You trouble needless,' said Iago, undisturbed.
'You know as well as any what answer he made to their urging and how fast they dropped it, at least in his hearing.
Do you think they would send such a lure through such a palpably honest man?'
His eyes, fragments of pale, clear sky, and yet so impenetrable, were on the doorway, waiting, she thought, for her.
They did not leave her as she closed the door and drew the curtain over it, and came forward into the room.
He saw his own words sink deep into her mind like water into a secret thirst; but her face was motionless and indifferent.
'Fool,' said Rhodri impatiently, 'would they be likely to send it by one palpably dishonest?
There's hardly a man in England but this one to whose lure Owen would stoop now.
And perhaps his father!
And why?
They're well-disposed enough, all the Percy tribe, because their own holdings are in the north, and they have nothing to lose here on the Welsh border.
It would be a different tale if they were in the king's own shoes, for he's a marcher lord himself by reason of his Bohun marriage, and when Wales is in question he thinks like a marcher lord, and there's an end of it.
No, if they want to flush him out of hiding, there's hardly a man they could use but Hotspur.'
'I would stake what I have,' said Iago, still watching the girl, 'that this Hotspur is not an easy man to use.
And in November at the council he let see how much trust he was ever again likely to put in the little law-givers at Westminster.
Never be deceived by his simplicity and honesty.
In defence of that same honesty I think he could be shrewd enough, and ruthless enough, too.
And as for deliberately lending himself -- why, he would cheerfully hew off Owen's head in fair fight in the field, and never lose a night's sleep for it, though he'd grieve for the loss of a grand fighter.
But as for putting poison in his wine, or setting a pitfall under his feet at a hunt -- no, he'd hew off the head of any man  who tried to put him up to it.
What they call policy nowadays he'd still call by its old and uglier names.
And he has a sense for it, as the cleanly know by instinct how to avoid filth.'
The girl had lifted her head and turned her face towards him, though he could not flatter himself that she was looking at him; rather at the image he drew before her, that spare portrait of the departed visitant, sketched in so few lines on the firelit air.
He saw her breath quicken ever so slightly, and her chin lift, and the light glittered in her half-hooded eyes, red as the embers.
It caused him to look back in some compunction, in case he had lied to discover what he wanted to know; but he could find no lies.
Hotspur was as he had painted him.
He seldom had to meet a man twice to be certain of his ground with him.
Women were another matter.
How was it that it had taken him so long to see Parry's daughter thus clearly?
'In any case,' he said, rising and stretching lazily before the fire, 'you lose your pains if you trouble on Owen's account, unless you think him easily gullible.
A letter in a man's own hand is evidence to be read, along the lines and between the lines, and Owen can read as well as any, and better than most.
You think he will move unless he's sure of his ground?
And now I'm off to pick up my cloak and pack, ready to shift before dawn.'
He had put away the letter, somewhere inside his ample tunic.
He moved towards the inner door by which he had entered, no long time ago.
Before he reached it he was aware that Julian's eyes had shortened their focus, and were fixed with sharp intelligence upon his face.
She said nothing, but she met his gaze fully and did not veil her own.
Rhodri was still muttering, unwilling to give up his customary pessimism, but he might as well not have been in the room with them, he counted for so little at this moment.
As little as she did to him, Iago thought, for now that he was to have half his money back, that he had staked on her ennoblement and his grandchild's inheritance, he scarcely  noticed her.
She filled her place, she fetched and carried for him, but any tame girl would have done as well.
He had no need of this mewed, motionless falcon, waiting now only for the moment and the means to shake off her jesses.
In the world into which she intended to soar there was little room for Rhodri Parry.
But there might, he thought, if he knew how to wait, be room some day for Iago Vaughan.
3
In the foothills of the Clocaenog forest, snugly folded within the pleats of thickly-wooded ground, the camp was invisible from all sides at any distance, covered on one flank by an upland bog, and guarded on the other by a line of outposts.
Long before an enemy could get near enough to distinguish any glint of arms, the whole company could fold their belongings and slip away into the mountains at their back.
They needed little and carried little.
They were expert at vanishing silently and reappearing suddenly in some unexpected place; and in case of need the Lord Owen's main stronghold of Glyndyfrdwy was no great distance away, mound and manor guarded by a curve of the Dee, down there to the south in the close confine of the valley.
The weather that mid-April had turned fair and mild; the trees were coming into delicate leaf, and ladysmocks fluttering over the marsh meadows.
It was pleasant to live in the open, and easy to provision both men and ponies; and the courtier and man of law who had lived a high life in the London Inns and colleges, and been in the king's own service, was nonetheless a hardy Welshman, well able to campaign in the hills winter or summer, and never complain of a hard bed or a scanty meal.
He was the master of most of North Wales and part of the central lands, but a swathe of bracken and heather covered by a skin rug and his own cloak was bed enough for him, and he ate what his men ate, and wanted no more.
He had the roads to Ruthyn and Denbigh  under his eye from this eyrie, and Mold was not too far for a raid if the weather and the omens were good; but since his active autumn of last year he had contented himself with holding and consolidating, and swooped down in the occasional raid along the border only to keep his hand in for greater things if the season should indicate the necessity.
He sat on a couch of deerskin, under the awning of his tent, a long, sinewy man in the prime of his powers, forty-eight years old, black of eye and black of hair, but for the first frostings of grey at temple and lip.
He was changed since his days at Lancaster's court; with all his polish and scholarship, which neither time nor place could tarnish, he had nevertheless shed all the cramping tensions of city life, and moved like a young stag, long-stepping in motion and magnificently abandoned in repose.
His armour was piled not three yards away, arrayed ready to be donned at short notice.
Everything he bore in hand was but half-achieved and for ever in the balance; yet if at this moment there was a prince in Wales, his name was Owen, and Owen knew it.
Outside, on the grass stippled with the bright embroidery of light and shadow under the trees, Iago Vaughan sat clasping his little travelling harp.
Of the prince's bards he was the least, the stray; but his touch on the strings was no less sure than that of Owen's court poet.
'And he said he had known you somewhere before?'
'He said so.'
'I doubt it may have been somewhere in a circle of exiles, plucking that familiar of yours.'
For in England every bard was an incendiary, with however much deceptive mildness he chose his songs; he was the voice calling the Welshman home, and for what purpose except to join the golden dragon in arms?
There had been disaffection among the Welsh in the universities for more than eighteen months now, and many a wandering musician had been thrown into prison for stirring up sedition with his tribal songs, and more than one had been put to death.
Welsh labourers, however indentured, however bound, had somehow found a means to slip away, and England well knew where, and for what purpose.
Iago shrugged and smiled, muting his vibrating strings with a flattened palm.
'It's true he was in Oxford while I was there, and the schools were no very safe place to be singing about great Llewelyn in his seven-foot grave, or making verses after the manner of Cynddelw.
God knows we were not always as discreet as we might have been.
But more likely it was some time in London.
He has his own town house in Bishopsgate Street.
Does it matter?' he said, lazily watching the dew distil into vapour as the sun drank it.
'Even if he remembers, it's a year and more ago now, and he'll remember to forget.'
'You at least have confidence in him,' Owen said, thoughtfully frowning down at Hotspur's vehement hand.
'Yes.
Confidence in his will to end this war, even upon terms not ungenerous.
By comparison with that, what is it to him if one bard goes free?
I'm of more use to him, if I carry his letters faithfully, than all the statutes and limitations and restrictions they've clapped on the Welsh trade.
He knows them for folly, and has no patience with the little, grudging, timorous minds that made them.'
'Will you hear what he writes to me?
Listen, then, you've seen and spoken with the man, and have some insight into his mind.'
The prince unrolled the scroll, and read aloud:
'' To the most noble and puissant Owen, lord of Glyndyfrdwy and Cynllaith, Greeting and Respect!
By this it will have come to your lordship's ears what little success my attempt to put forward your terms for negotiation achieved in the council.
There are those among the ministers and members who have little knowledge of Welsh affairs, and are not well-disposed to proffer any concessions.
But I beg you to believe that there are also men  of a wiser and more experienced sort, who by no means decline all consideration of negotiating terms.
For my part, I promise you I will continue at all times to have this possibility in mind, and to take every opportunity of bringing it to the minds of those who can best move in the matter.
The same guarantee I can offer for all my house, whose will to you is as mine.
And I am in haste to get this word to you, that you may know your affair is not in abeyance, and that the undeclared truce which has held good, but for some small brushes, since the council met in November may continue unbroken.
This circumstance of restraint on your part is the most favourable argument they can have, who are your well-wishers here.
Every day that passes without further raiding speaks for you, the more confirming those who give their voice for reconciliation, and by little persuading those who were against.
But if you again burn and provoke, for every enemy you slay you raise up in England a score of enemies, and do but increase the odds against your cause.
Which, as I esteem it, is a cause not all unjust, and not to be distorted by ill-judged action.
'' I charge you, therefore, for the present abjure all fighting but that is forced upon you, when no man can blame if you do valiantly in your own defence.
But forswear all attacks upon cities and towns, upon travellers going their way without ill-thought towards you, all provocation of all kind against the borders, and avoid, so far as ye may, any meeting with any English soldiery.
You well may do so, as I know, where every valley and hill and track is known to you.
And in return, I promise you that if I get from you the reply for which I hope, I will be about your business presently.
'' But further, my lord, one stipulation I make, for your own protection: do not upon any consideration come to any meeting, or respond to any advance, however seeming honest, that does not come to you by my hand, and by this  same messenger.
For notwithstanding I trust to bring you off happily, with the goodwill of our lord the king and all who best speak for this land, yet I do know there are some who may have other thoughts concerning you.
Therefore bide your time and refrain from all action, until I send you word that you may come with my warranty to the council table.
Which warranty, when I have given it, I will make good with my life.
'' I trust to have word from you by this messenger, and delay only to know that you wish me to proceed.
Thereto I pray all good both to you and to Wales, again reconciled soon, I trust, to the king's Grace.
'' Given by my hand at Shrewsbury, the seventh day of April, this year of our Lord fourteen hundred and two.
'' Henry Percy, Knight.''
'By God, he goes a degree beyond even what I had thought,' said Iago, roused and vindicated, 'to warn you not to put your trust in princes and councillors.
Surely he knows you're well warned already, but for his honour he cannot keep from underscoring it three times.'
'It was he laid my terms before them, and put the idea into their heads,' said Owen with a wry smile.
'He may well feel the need to scare me off.
But I grant you there are not many would have gone to the trouble.
You'll need a fresh horse, Iago.
Go and see to it.
Einion will find you whatever you need.
And ask Philip, and Griffith Fychan, to come to me here.
I have a letter to write.'
'You'll remember, my lord,' said Iago, rising with alacrity from the grass, 'that he is but an earl in the making and has no Latin.'
'It shall be in English,' the prince promised him drily.
'And favourable?'
'God granting, everything shall be as he wishes.
I will keep my hands from the English -- any and all but one,' he said grimly.
Iago slung his harp over his shoulder, where it carried snugly under his cloak on horseback, and hunched one shoulder slightly under the cape of his capuchon when he went afoot in England.
He had made no more than half a dozen strides towards the heart of the encampment when Owen called him back suddenly, in a sharp, changed voice; and when he looked back in surprise:
'What was that you said of him?-- of Hotspur?
'He is but an earl...''
'An earl in the making, I said, but it was a foolish saying.
In what does he fall short of an earl now?
Why, my lord, what is it?'
Owen's black eyes, deep-set and far-seeing, stared blindly inward, narrowed after some vision they had almost captured, and yet let slip.
His face was honed bright, like carved ivory.
'Nothing!
I cannot be sure now.
The fire kindled, Iago, when you spoke.
Now it's gone.'
Colour came back into his weathered cheeks; the volcano of prophecy that was known to burn in him had cooled and crusted over.
There would be no further prodigy.
'I have it in me, Iago, that this Hotspur, whatever he be, will never be an earl.
Strange!
Less and more I see him, but never that.
Never Northumberland!'
Iago ambled down out of the forest in mid-afternoon, on a Welsh mountain pony with a barrel like a butt of wine, a cross-grained temper, and a turn of speed no one would have credited from her build.
He wanted no more showy mount; for all his notably individual looks, he could jog like a pedlar and fade anonymously into any background when he chose.
He had Owen's letter in the breast of his tunic, and his harp tucked away behind his shoulder, and three days of his promised week left for getting back into Shrewsbury.
He could have moved directly south into the valley of the Dee, but instead he chose to head eastwards towards the vale of Clwyd, to take a cautious look at the borders of Lord Grey's domain before he turned south to cross the mountains to Valle Crucis.
The truce, after all, did not depend all on one man's goodwill.
The sun was high and bright as he dropped gently out of the hills towards the vale, faintly misted with vapour, and saw in the far distance before him the mole-hill of Ruthyn, hunched and veiled in the smoke of its house-fires, a delicate blue flower in the sparkling folded green, with the giant hogback of Moel Famau towering beyond.
And he saw, too, narrowing eyes that were used to singling out detail at great distance, the betraying glitter of sunlight upon arms, below him in the copses of the valley.
A crackle and sparkle of steel, spitting light and vanishing into shadow, but to reappear by spasmodic flashes thereafter, moving up towards him.
They were no threat to him; they were far away, and he was in tree-shade, and had nothing bright about him to catch the light.
He was invisible; and at need he could better their speed.
He loitered, untroubled but curious, for they were no small company, and by the line of their march they had come from Ruthyn.
Reginald de Grey reckoned every Welshman a thief and an outlaw, and had his borders patrolled as though against the entire army of France, in great measure creating the animosity and disorder he saw everywhere; and this company might be no more than a routine patrol meant to impress and intimidate on his usual terms.
Yet they moved with more than usual purpose towards the hills; and it was always a possibility, however remote, that some vagabond poacher or time-expired soldier living wild had hit upon Owen's outposts without being detected, and thought it worth his while to carry a tale to Ruthyn.
He did not take it too seriously, but nonetheless he wheeled his pony and made off at speed, back towards the fringes of Clocaenog, where he had passed the last of the prince's watch.
The man looked down at him from his perch in a beech-tree above the track, and laughed.
'You think we're asleep, up here?
A runner went to the prince half an hour since -- by now he knows better than you.
Get on your way, and watch how you cross them, for it's Grey's livery they're wearing.'
It was true enough, he had good need to take thought for his own safe passage, for his course must somewhere cross that of the armed company.
To avoid their notice, and give them time to get clear of the folded valley before he ventured it, he turned on a contour course towards Ruthyn, and kept in the fringes of the forest, watching the glint of steel come and go on the track lower down the slope, drawing steadily nearer to him, but some half-mile below.
And having found a vantage-point where he had a clear view of the meadows and was himself sheltered, he halted his pony and stood to watch, narrowing his eyes to single out coat-armour, and number the forces in the English party.
There were archers with them, but not a great company, and some three-score men-at-arms, all mounted; and a knot of bright devices he could not quite read at that distance, though their colours did almost as well.
Four knights at least, all Grey's men; and a rugged, thickset figure in half-armour, whose seat in the saddle was familiar, even if the black horse under him had not been so signally ornamented with his blazon.
Reginald de Grey himself was on the move, with a strong and well-mounted party in arms.
They passed, and left the valley free for him to cross.
He waited until they were lost to sight beyond a fold of ground and a belt of trees, and then made good speed down to the little river, splashed through it where the banks were level and firm, and climbed the slope on the other side.
Clocaenog village he left at a distance on his left hand, and wound his way up into the hills again.
But his mind was not easy.
It was an ill omen that Grey should appear on this day of all days, so close to where Owen lay hidden, newly resolved to take Hotspur's advice, bide his time for peace-making, and forbear aggravating the English further.
Any and all but one!
And that one had to appear, like a spirit raised by necromancy, suddenly almost within grasp of his hand.
And for what purpose, with such a force, unless he had some word of power to lure him out of his castle?
The old quarrel, sprung from a tract of land in dispute, had been fomented by many acts of hostility since.
When King Henry had summoned his muster for Scotland, the year after his coronation, Grey had been charged with delivering the summons to his Welsh neighbour, and had withheld it until too late, so that the lord of Glyndyfrdwy was exposed to the charge of being a traitor.
Out of this personal feud had burned up, like the sudden flaring into splendour of a bush fire, the old, old quarrel that belonged not solely to Owen, but to Wales.
After so long of acquiescence, the Welsh felt themselves Welsh again, a nation with a prince and a prophet of their own.
And yet the personal bitterness still rankled in the heart of the fire, and as it had kindled it, so might it sour it and put it out.
His uneasiness grew, and yet he could not tell why, for the prince was warned, and could very well deal with this matter.
Avoidance would be easy, the camp could dissolve into the hills like mist within half an hour.
More likely, if the force from Ruthyn seemed to be passing without ill intent, they would merely sit still and let the enemy go.
Yet Iago suddenly wheeled his pony again, and made for the highest point of the ridge, where he could look back over the valley, and see as far as the scattered outer copses and the rim of the forest.
Far below him the river was a silver thread, curling and twining through meadows freshly green in sunlight; and beyond it the folded hillocks rose plumed with clumps of trees, heaving and falling in a series of green bowls all along the flank of the dimpled ridge that soared to the dark green of woodland above.
He saw, as though some wall painting had come to life before his eyes, the glitter of steel and the minute clusters of rainbow colours just moving over the crest of one rise, to descend into the next bowl; and riding towards them, negligently like men out hawking, he saw a smaller group, no more than half a dozen mounted men, who had been until this moment hidden from them by the lie of the land.
There was an instant when both parties halted at gaze, no more than a quarter of a mile apart; and though they were so far from him across the valley, he felt the shock of confrontation and recognition quiver through his own body as they measured each other.
Then the handful of riders wheeled their mounts in wild haste, and rode back by the way they had come, and after them in headlong pursuit streamed Reginald de Grey and his knights and men-at-arms.
The quarry must needs have ridden steeply uphill if they were to gain the cover of the trees, and he saw that they were not even attempting it, but climbing only very obliquely towards shelter, preferring to gain distance on the level.
And as the half-dozen lengthened into a line, he watched them like a file of horsemen on a hanging tapestry, each separate, and the last flagging.
He saw the tall grey horse stumble, or seem to stumble, and recover but lamely.
The pursuers saw it, too, and lingeringly across the valley, long after they had launched it, he heard their shout of triumph.
The Lord Owen was taken at a disadvantage, surprised in the open, hunted like a hart, and his horse fallen lame.
The men of Ruthyn had abandoned all caution, spurring their horses furiously, lengthening out in their turn into a long frieze parallel with the edge of the forest, every man mad to be the first to lay hand on the arch-enemy.
Iago felt in his blood the coming of the climax, the moment when Owen had drawn them, with his body for bait, exactly where he would have them, with all eyes on him, and never an archer ready to string bow, or a lookout to shout an alarm.
There had never been a more insolent ambush.
He did not know whether to laugh or to weep.
Suddenly the grey horse was lame no more, but picked up his heels and leaped ahead with stretched neck and lunging shoulders, the prince lying forward over his neck and thrusting with him, as though he and his beast were one flesh.
The gap between him and his pursuers widened; and at the same moment, though they were invisible and their volley could not be followed by eye or ear, the  Welsh archers deployed all along the rim of the forest loosed their shafts together.
It was like corn falling before the scythe.
They had every man his mark, and they loosed at leisure; not at the horses -- good horses never came amiss, and certainly never were wasted -- but at the men.
The mounts, suddenly lightened and without hand on the rein, wheeled and circled curiously in the heaving bushes and trampled grass, more at a loss than frightened.
Without a sound the men of Ruthyn, more than half of those leading the pursuit, fell with the impetus of the arrows that pierced them, heeling out of their saddles like a breaking wave, downhill from the forest.
Some were dragged by a foot still caught in the stirrup, round in a circle in the turf.
Some shook themselves clear, and even rose again, but many lay threshing, and some lay still.
And before those following could rein in and look for cover, or dismount and string their bows, or drive headlong into the trees from which their death was launched, the second volley followed the first.
At that range, Welsh arrows could shear through plate-armour and fine mail shirts within, and these were riding half-armed.
Owen and his half-dozen were in the trees by then.
They took breath for a few moments before they emerged, after the third volley, to finish what they had begun, the Welsh swordsmen boiling out of the bushes joyously on their heels.
Iago watched that fight to its end, and saw the survivors haled away into the forest and silence.
A few, those who had been last in the line, turned their horses in time and rode for Ruthyn to carry the news, and were not hindered in their going.
All the wilds of the Cambrian mountains were at Owen's back, there were plenty of places where Reginald de Grey could safely be hidden, long before any party ventured out of Ruthyn to collect and bury the dead.
Iago dug his heels into his fat pony's ribs, and took the shortest way down into the Dee valley, riding hard for Shrewsbury.
And still he did not know whether he should be laughing or weeping.
He came into Hotspur's presence in the abbot's lodging at Shrewsbury abbey, still stained and dusty from the road, a thin brown packman with some plea about a permit to carry his goods to Chester.
His extraordinary eyes he veiled with lowered lids and humility, and only the satirical curve of his long lips, accentuated by those twin russet flames that forked upwards through his short black beard, caused the chamberlain who admitted him to look at him a second time.
Both chamberlain and clerk accepted it without question when they were dismissed from attendance.
There were no other petitioners waiting, and they had routine work to do.
This fellow's matter was simple enough.
'You've made good speed,' said Hotspur when they were alone.
'Better even than you promised me.
You have a letter for me?'
'My lord, I have.'
But he held it in his hands still, not yet proffering it; and his eyes were unveiled now, two slivers of clear sky, but a winter sky.
'My lord, I entreat you to believe that what you find in this letter was honestly written and honestly meant.
I pledge my own honour for it.
But there has that happened since that may well have changed all.'
Hotspur sat very still, watching his visitor's face.
'What has happened?
There has been no news here.'
'Not yet.
But there will be.
I have come straight from the event, and as your lordship sees, I have wasted no time in applying to you.'
'I am content,' said Hotspur quietly, 'that you have fulfilled all terms, and done everything you undertook to do.
What more has happened cannot be of your doing.
But I need to know.'
'Read the letter,' said Iago.
Hotspur broke the seal in silence, and unrolled the parchment.
Owen had written to him in his own hand, a fine and scholarly hand.
He read it through, while Iago watched his always eloquent face.
'To the most noble and excellent Sir Henry Percy, Knight, Greeting and Love!
'For your lordship's letter, duly come to hand, I send you my thanks and my grateful sense of your lordship's kindness and good feeling towards me.
I have, as I ever had, the fullest confidence in your honour, and am willing in all things to deal with you as man with man.
But I confess that I have not the same trust in some your peers.
Nor does the issue of life and death rest only with your estate, as we have well seen in the fate of those earls lately in dispute with the king's Grace who fell into the hands of certain lawless gatherings of commoners, and were shortly done to death.
I will well, therefore, that you should, as you have said, deal for me as you may, and when you summon me to conference with your own warranty, I will not hesitate to come.
But for no other will I come in confidence, unless it be for Prince Henry, whose mind I conceive as noble, and his word as his bond.
'And to the end that you may deal for me without hindrance, to bring about this peace, I undertake that I will not henceforth, at least until I do withdraw this word, have any ado of my own willing with any English company in arms, but will forbear them as you counsel.
And for the rest, if any do challenge me, and I cannot but defend myself, you shall hold me justified and excused.
Though whether any other of your part will, neither you nor I can well determine before the event.
'I greet your lordship well, in the hope of a good deliverance for us all, and look to hear from you again, God willing.
'Given under my hand, this eleventh day of April, the year of our Lord fourteen hundred and two.
'Owain Glyn Dwr, lord of Glyndyfrdwy and Cynllaith.'
He looked up from the scroll, thoughtful and faintly frowning: 'You know what the Lord Owen writes to me?'
'I know what he gave me to understand he would write.
And I know it was written in good faith.'
'That I never doubted.
So what has befallen to make this of none account?
And how came you to know of it, if it befell, as it seems, after you had left the Lord Owen?'
'I myself witnessed it, my lord, though from a distance.'
He told it, exactly as he had seen it, not concealing that there were curious and doubtful points in it which even witnessing could not make plain.
'Sure I am that until this party was sighted, the Lord Owen had no thought in his mind of any such happening, and no plans to provoke it.
But when the word reached him, and especially when it was seen that Lord Grey himself was with the party, thereafter I cannot be sure.
I tell what I saw.
The English came over a crest, face to face with the Lord Owen in open field.
And they spurred forward to pursue and take him, no doubt believing it a happy chance for them, and the Lord Owen caused his horse to appear to drop lame, and so encouraged and led them until they were spread all along the field in open order, within close range of the bowmen in the woods.
When he had them so placed, he spurred ahead and drew dear, and the archers cut them down like corn.
Only a few of the rearmost broke away and escaped back towards Ruthyn.
But whether they came out knowing of the Welsh camp, and with some plan of attack, which the Lord Owen by his own stratagem forestalled, or whether they were on other business and would have passed by but for this lure, I tell you honestly, I do not know.
And even if they meant him no threat, how could the Lord Owen be sure of it?
And with all his force to guard, how dared he assume it?'
'You argue well,' agreed Hotspur, watching him keenly, and with a sudden remote spark in his eyes that looked like involuntary laughter.
'And whatever the way of it, had I been in the Lord Owen's place, with my chief enemy thus presented  naked into my hand, I doubt if I could have resisted the temptation.
Certain it is, if he had more than three-score armed men with him, he was not on his way to church!
Well, tell on to the end.
The archers cut them down -- a few broke back for home unscathed.
Some, no doubt, made their way back later with their hurts.
Were there prisoners taken?'
'My lord, there were.
I saw two or three of the knights haled away into the woods.
And after them,' he said with deliberation, 'Lord Grey himself.
If he was hurt at all, it was but a minor hurt -- he walked where he was led.'
He caught the wide, levelled eyes watching him with the first faint shadow of doubt and disquiet, almost distaste, and laughed shortly.
'Oh, never wonder about me, my lord!
You are the first of the king's officers to know that Lord Grey of Ruthyn is carried off prisoner into Wales, and if this moment you turned out the muster of every shire between here and Denbigh, and loosed them into Clocaenog forest, do you think you would find hide or hair of a Welshman there?
I would not have told you place or time if I did not know that every man of them is far into the mountains and out of your reach long before this.
And I tell you now in order that you may reckon well what chance is left of keeping any hope of peace alive, after this skirmish.
For I do believe you honest in desiring it, and so it deals fairly with the Lord Owen, I desire it, too, and will still be your instrument in pursuing it.
But I tell you plainly, I count the chances as low enough.'
The shadow broke like a cloud, and was blown away in a gust of rueful laughter.
'And so do I, Iago, so do I!
What can follow now but renewed war, and hotter than before?
Can I argue and persuade for moderation, when every baron along the march will see himself in Grey's shoes?
I am sorry, Iago, that it is so soon over.'
'For this time,' said Iago, and slid the ring from his finger.
'There is also money which is yours.'
'  No. keep it -- keep both.
I hope there will yet be occasion to  make use of them again in the same cause, even if we must wait now for a better opportunity.
Or a verdict in the field,' he said, abruptly flashing fire, 'for if you have occasion to speak with the Lord Owen again, you may tell him I will not spare to do my uttermost against him.'
'That he knows,' said Iago, 'and would expect of you.'
'Yet I would rather a resolution less wasteful.
So, Iago, keep the ring.
And if you ever have word for me that may bear fruit, come and ask entry to me wherever I am.
And should I need you, can I find you or get in touch with you at Rhodri Parry's house?'
'They may not always know where to find me.
But they will always know the times when I shall be here in Shrewsbury, and I will see to it that whenever possible they shall have word of my moves between.'
He thought of Julian in her drab housewife's gown, with her still, tense body and her hungry eyes, and he said, hardly knowing why, or whether it was mischief in him or mercy: 'If you need me, apply to the girl.
She keeps at home now, and she can be as secret as any man.'
But into that flight, though he looked at him thoughtfully and searchingly, Hotspur would not follow him.
And so they parted.
When the prince heard the story, over a conference table littered with notes, despatches and letters, he first opened eyes and mouth wide with shock and disbelief, and cried: 'Never say so!' and then as suddenly laughed aloud, crowing: 'A judgment!', looking, for once, a year or so less than his age and capable of mischief; and then he looked very grave indeed, and sat staring moodily at his table-full of papers, and said, dismayed: 'The king will be out of himself with anger.
He values Grey.'
'Too well!' said Hotspur grimly.
There was no third present, and they spoke openly as they always spoke together.
'I don't say the man could not be of the highest value, if he did not ruin everything he touches with his implacable spleen.
He fights well, he mans and maintains his castles well -- God knows not all in the  marches do so!-- and he has a good grasp of tactics in the field.
And yet he is the man who first made this needless quarrel, and now inflames it even when Glendower is disposed to be reasonable.
He would never in life agree willingly-or let the king agree -- to any settlement but a total victory over the Welsh, and the hanging of all their leaders into the bargain.'
'To be fair,' the prince pointed out generously, 'in this case it seems to have been Owen who took the offensive.'
And briefly, before resuming his burden of responsibility with a resigned sigh, he laughed again.
'It must have been a rare sight!
I wish I'd been there to see.'
'Faith, and so do I, but we're like to pay dear for it.
And what do you think Grey was doing, skirting the forest with such a force?
No, he had his information, no question -- only it seems it was none too accurate.
And the upshot is, there will be war on hotter terms than ever, and no more listening to counsels of peace.
And we had best get our fences in order, you and I.'
'We must get back to Chester,' the prince said, 'and call a council at once.
We can better keep care of Denbigh and Mold and Flint from there, and I must see to it that Ruthyn is properly garrisoned, now that Grey's gone.
Where do you suppose they have taken him?'
Hotspur laughed shortly.
'Where no one but a Welshman is likely to be able to track them.
The king had his fill of trying to find the Welsh in their own mountains, a year and more ago.
The most we can do is expect Owen everywhere, and be strong enough to match him wherever he strikes.
For strike again he will, now the die's cast.
There's no going back from Grey's capture -- not until time has dulled the sting, at least, and made it possible to mention peace without being called a traitor by some city haberdasher in the commons.
I've already written the news to my uncle of Worcester in Cardigan.
Who knows, the next foray may be into the south.
And, Hal, from Chester I must go north to the march as fast as I may, for Walton sends me word  there are new raids threatening, and it's his belief and mine there are French knights serving there with Douglas.'
'France has declared its intent to maintain the truce,' the prince objected.
'To send a force with King Charles' official blessing is one thing,'Hotspur agreed with a hollow smile, 'to finance small parties of adventurers and let them slip away privately to Scotland is another.
It's cheaper than out-and-out war, and they can be disowned if things grow too difficult.
But trust me, they're there.
And both France and Scotland are receiving Glendower's letters, and finding them tempting too.
We may yet find ourselves fighting a war on three fronts, and all one war.'
'I know it is a possibility my father has much in mind,' the boy admitted soberly.
They did not speak of what lay behind France's bitter enmity, though it was always present in their minds, a spot too sore to touch on lightly.
Charles of France might shrink from fomenting a direct war, but he would be glad to use every oblique weapon against the upstart king who had deposed his son-in-law, and sent his little widowed daughter back in clumsy state, but without her dowry, which had been fed of necessity into King Henry's treasury to keep it solvent during his first year of kingship.
'Dunbar is there in the north,' said the prince, offering what even he felt to be dubious reassurance.
'The more reason I should be there, too,' said Hotspur tartly, 'for a man who can turn his coat once can turn it again as readily.
I'll take my wife home to Bamburgh, Hal, and go north to Berwick myself for part of the summer.
If you need me, I'll be in Chester within three days.
But to say truth, I trust this border to you with a far lighter heart than I trust the east march to Dunbar.'
He hungered for the north, too, the prince knew that.
It was his country, and campaigning across those noble moors under the Cheviot was his true life, as natural to him as to the hawks hovering on languid, sinewy, expert wings above the heather.
He  did his work here well and thoroughly, but he hankered, every so often, for the rough grey seas and painted, cloud-dappled hills of Northumberland, and his children, and the soil that knew his step and warmed under the sole of his foot, like a caress.
'I'm faced with this business of the Danish marriages,' the boy said without enthusiasm.
'I shall have to go to London, perhaps in May, to appoint proctors.
I suppose I must at least be civil to the Danish envoys, and offer them some entertainment.
I shall not linger.'
Hotspur forgot his preoccupations for a moment, and looked more closely and with quickening affection at his friend.
To be thinking of marriage, at this age, to a girl he had never seen, who might be ugly, stupid and inert, where he was handsome, intelligent and almost excessively alive!
He felt a wave of almost incredulous pity for princes.
'Are you happy about this proposed match, Hal?'
The boy shrugged, raising his brows with a mild affectation of surprise that it could be thought to be important; but the stillness of his face and the steadiness of his eyes on Hotspur went some way towards betraying him.
The three northern countries had recently agreed by treaty to unite under one king, the fifteen-year-old Eric, and Eric had sent envoys to propose a marriage for him with King Henry's second daughter, Philippa, and as an opportunist gesture by the way, a second match between Eric's sister Catherine and the prince of Wales.
The boy had lived through the negotiations, rather less sordid than most of their kind, for his sister Blanche's marriage, and had got the tune of these affairs very well off by heart now.
It was no shock to him that he should be marketed in his turn.
And yet the Danish princess was no great catch for the heir to the English throne.
He knew that, too.
By this time there was very little he did not know about being a prince; and long before he came to it he would know more than most men born to it about being a king.
'Why, it's nothing yet but the beginning of talks.
They'll play  with it for two or three years yet, and in the end very probably nothing will come of it.
After all, Philippa is not yet eight years old.
I was but an afterthought, and I doubt if they're bidding high enough for me.'
His voice was cool, even a little cruel, in its effort to be adult and civilised.
For when it came to the point, he would probably do what was expected of him, whatever that might be.
Marriages were an acknowledged part of the to-and-fro of barter and bargaining that royal children were born to.
(But there was always the sudden stab at his heart when he reasoned thus, because he had not been born to it!)
And that thought brought him sharply into collision with the one marriage that stuck most obstinately in his throat.
He had not mentioned it in all this week that Hotspur had spent with him at the abbey, had asked no questions but the most current politenesses about his stay and his journey, and had shown no interest at all in the ceremony from which he had come.
But now suddenly he came out with it violently, almost in the manner of Hotspur himself over-riding some constraint that tied his tongue:
'Harry, what does my father hope for from this marriage of his?'
And as Hotspur turned to face him, in mild but sympathetic surprise: 'He cannot suppose that allying himself to the duchess of Brittany will either placate or frighten the French.
It will not even give him any power in Brittany, for if she comes here she loses whatever sway she has there.
He neither gains an ally nor sweetens an enemy, and say she brings but a token household with her, yet it will cost him dear to keep them.
And you know how willing the commons will be to grant an aid for a foreign queen!
They starve him of funds even for paying his soldiers.
No one knows it better than you -- we've both pledged our own valuables before now to keep our archers from deserting.
What does he want,' said the boy, pale and passionate in resentment, 'with a new expense?
What does he want with a queen?
He's lived content enough these eight years since my mother died.'
He had put a finger too accurately on the true cause of his indignation, and flinched away from it hurriedly.
'And if it is not some political advantage he is after, what else is there?
Why, he can have seen the duchess no more than once, and that at least four or five years ago, when she was a wife.'
Wife to an elderly duke, he could have added, and his third wife at that; it did occur to him fleetingly that she, perhaps, had something to gain, a brief recapture of life and youth before it was too late.
'So what is it he hopes to get out of it?'
'True enough he's seen her but once,' Hotspur said gently, 'and by the same token he must have liked well what he saw, for he's been in correspondence with her ever since he was crowned.
And but that he had many things to occupy his mind, and she no less, I think something would have come of it before now.'
It was hard to urge forbearance and sympathy with the father on the son, all the more in face of that bitter resentment that was all for poor Mary Bohun, mother of six young children and dead at twenty-four.
More than likely his younger brothers and sisters would welcome a new mother, but he was too nearly a man to take kindly to any woman set with so little warning in his own mother's place, especially when this new incumbent stood to gain a crown as well as a ring.
The earl of Derby had never offered Mary a crown.
'You must not think,' gently said the man who had married for love in his late twenties, 'that the king has always reasons of cold policy for what he does.
What does he hope for?
A little happiness, perhaps, Hal, nothing stranger than that.
A little happiness, while there's still time.'
4
It was a grim summer that year.
There were torrential rains, rivers burst their banks and flooded standing crops, churches were struck by lightning in heavy thunderstorms.
After the first fair flush of spring, nothing went right.
Like the weather, the fortunes of the time were soured.
Nothing but bad news came in from every frontier.
The king withdrew in great weariness and exasperation of mind from his son's manor of Kennington, where he had presided over an anxious council on Wales, and took refuge in mid-June in his castle of Berkhamsted, with only his intimate household about him.
Strange how wide a gulf he found between these old retainers of Lancaster and the full council of the realm, let alone the unpredictable vagaries of parliament.
Nothing could shake the steadiness of such men as Hugh Waterton or John Norbury, who had been in his service from the time when he had been merely Henry of Bolingbroke, earl of Derby, and even that title borrowed by courtesy from his father.
On these, on the Leventhorpes and Rempstons and Erpynghams, he could lean when he would, and they would not let him fall.
But the council of England turned in his hand.
Parliament crossed him, always with the greatest respect but implacably, criticised his use of the council to levy an aid for the marriage of his elder daughter without consulting them, doubted if there was a precedent recent enough to justify the aid, and periodically and obstinately  restated to him the principle that the king should live 'of his own', without demanding that parliament should raise money by taxes for his expenses.
Had not John of Gaunt been reckoned the richest man in the kingdom?
And so might his son have been and remained, with only a duchy to administer, but a kingdom was a different thing.
How different, he had never dreamed until he made the assay.
They had declared at his accession, they had repeated often since, that they desired him to reign upon the selfsame terms as his predecessors; and yet they made him aware, whenever it was needful to ask for a grant of money, that in fact he stood upon ground subtly changed, and must ask as a favour what had been Richard's unquestioned right.
But there were never any open words expressing the inflection, never anything to which he could raise objection.
Only the feeling of mute resistance, the chill sense of acquiescence so grudging as to give pain.
It was, perhaps, partly his own fault.
In that first parliament of his, immediately after his coronation, he had refrained from asking for money, had even prided himself on his princely forbearance, and believed it had won him friends and trust.
Fool, he should have known that that was his one chance to strike, and make known his mettle, and assert his right once for all.
It would have been time to win them with clemency later, when they knew his power and will to dominate, and could be stunned into love by the unexpected mercy.
Now it was too late.
They had his measure.
But was this indeed his measure?
He knew he was no such man!
What had gone awry, that he should have been led to this pass, and even now he felt himself following, perforce, the twists of his fortune, headlong as a fall, when he should have been steering his own course and bearing them strongly with him?
He had been king for two and a half years, and he was aged by ten.
When he peered into his mirror he saw himself already a little stooped in the shoulders, a little heavy in the body, the full  cheeks beginning to hang, their old ruddy colour grown muddy and pale, strands of grey in the short, forked beard and at the high temples, and above all, that permanent, aching double pleat between the long, thin brows, scored a little deeper every day.
He was thirty-six years old, and his youth was gone, and even his prime was passing.
He came from hearing vespers in his chapel, and shut himself early into his private chamber.
The wind tugged at the banner-pole that carried his standard, outside at the turret, and made a dolorous creaking sound that accompanied his steps along the chilly corridor, and whined faintly in his ears even after he had closed the door and shut out the sound of the rain.
Such a night for a ten-year-old child to be out on the North Sea, as by now she must be, if contrary winds had not driven the ship back into port.
Blanche, born an ordinary little noblewoman, and now a princess, and bound for Cologne to meet her bridegroom there, Louis, son of Rupert, king of the Romans, duke of Bavaria and Count Palatine.
She was small and fair like her name, shy like her mother, and his favourite child, dearer even than Thomas, for daughters are more fragile and vulnerable than sons, and she was his first daughter.
And though she had been excited and proud about all the frantic arrangements for her state departure, when it came to the point she had been frightened, and sad at going, though she would not complain.
He prayed that the boy might be fine and gentle and kind, and take good care of her.
At least she had her uncle Somerset to watch over her on the journey, and the countess of Salisbury to be a mother to her until she was handed over at Cologne to Louis.
Even this, in its way a successful transaction, had done no better than limp lamely to its achievement.
He flushed with anger when he remembered how the legal aid he had levied to furnish her to her wedding had brought in only miserable trickles of money on the date appointed, and how he had been forced to  send out letters to all and sundry requesting loans to help to pay for her clothes and dowry, and even to borrow abjectly from the City of London and some of its richest citizens, with all the members of his council pledging themselves for repayment, so low was his own credit fallen.
Even so, only a negligible part of Blanche's forty-thousand-noble dower left England with her.
Even her departure had been delayed for weeks for want of the funds necessary to fit out her ship and escort.
And had there been any real need to send the child away to her bridal so soon?
True, Rupert had proposed the match, and sent envoys a full year previously to treat for Blanche's hand, and the alliance was not one to be despised.
But had he not fallen in with it too readily and too rapidly simply because it was a testimony to his secure tenure, a declaration before the world that he was a king indeed, and his progeny fit mates for the royalty of Europe?
Was he clutching too eagerly at every such evidence?
To flourish before whom?
Charles and his quarrelsome relatives in France?
King Robert and his dangerous regent Albany in Scotland?
Or Henry of Lancaster, here solitary and discouraged in Berkhamsted?
Did he need Rupert's reassurance to prove to him that he was indeed king of England?
And was he to be as abjectly grateful for proffers even from young Eric in the north lands?
For only a month ago he had seen his eldest son and his younger daughter appoint proctors to treat in the matter of their proposed Danish marriages.
And the little one, Philippa, barely eight years old!
He was not committed, of course.
The discussions would move languidly enough, the parties being so young, and there was time to extricate either of them, or both, at whatever stage he found it desirable.
Yet the first step had been taken, and an inexpressible sadness closed in upon him, as if he had stripped himself wantonly of the children who were his own flesh.
zzz