各详名目，并题卷第

Eleventh of February, 2034. Starting a job I should have started last October — when Jia Dan said that even one man's sentences come in two colours, and I wrote in the notebook: four years recording who said it, and not once recording whether he had done it. Then November was the opening of the cells, December the accounts, January the celestial element. In February I took the two hundred and forty-five cards out of the cupboard and spread them on the floor. Two hundred and forty-five, twenty-one classes. A new box on the back of every card, headed: did he do it.

Ten days, twenty-odd cards a day, two questions each. Who said this — easy; the name is under every card. Has the person who said it done this himself — not easy. Day one: twenty-three cards, nine answerable. Day two: twenty-six, eleven. Ten days, and of two hundred and forty-five I could answer one hundred and thirty-one. The other hundred and fourteen I do not know.

Wang Tao came on the twenty-first. Fifties. A civil official's ordinary robe, not new but very neat, and a thin film of dust at the cuffs — the kind you get from years among stacked books. No medicines, no book: a bundle of bamboo slips, fine and long, all the same length, with very small writing on them. He looked at the floor and said: half of what you've got there has no head on it.

His own book first. Forty scrolls. How many prescriptions? Over six thousand. All yours? Not one of them. Flatly: all copied. And: I was in the Hongwen library over twenty years. I read those books one by one, and copied out whatever was of use. Did you add anything of your own? No. And: the preface has six characters: I have not dared to add or discard by my own opinion.

So is the book yours? No. I never took it for mine. He raised the bundle of slips. This is what I did. — Every time I copied something, I wrote beside it which book it came from and which scroll. And, and I picked up the pen: each with its exact title, and the scroll it came from. And: those eight characters are what I did with my life.

It's laborious. Very. Copying one prescription is a quarter of an hour. Finding which book and which scroll it came out of can take a day. Then why do it? At the time I didn't know why. I only felt that without the source written down it becomes something I said. And: I didn't dare. And: the use came later. What use? Of the fifty or sixty houses I copied from, most are gone. Slowly: the books are gone. All that is left is those few lines in my book, with a title and a scroll number after them. And, six times into the book: that those books ever existed is proved by those few lines.

I moved the hundred and fourteen aside and said I could not attribute them. He squatted and read a few. These are not without a source. You simply didn't ask at the time. Can they still be filled in? Some. And: where the person is alive, go and ask. And: when you ask, don't ask "have you done this." What then? Ask what year he last did it. And: a man who has done it gives you a year. And: a man who heard it says "everybody does it that way." And, five times into the book: between a year and an "everybody," there is a whole working life.

Last, the thing he could not do. There is one thing in my book I cannot supply. Which? Who originally thought of the prescriptions I copied. And: I can only get back as far as the book. Who compiled the book, I can write down. But which physician first used that prescription, in what year, on whom — not one of those can I write. Putting the slips away: I can only trace back one layer. At the door, turning: one layer is better than none. And: that you cannot get to the bottom is not a reason not to start. — M.

Memo to file. Chengdu, 11 February – 6 April 2034. Provenance, and the book got thinner.

1. The audit. Two hundred and forty-five cards, ten days, two questions each. Who said it: 245/245. Has that person done it himself: 131 answerable, 114 not. Four years of recording attribution and never once recording whether the attributed person had first-hand experience of the thing.

2. The question that works, and it is not the obvious one. Do not ask have you done this — it invites yes. Ask what year he last did it. A man who has done it produces a year. A man repeating what he heard produces "everybody does it that way." Applied 22 February – end of March to the five people still able to answer (Chang 66, Luo 63, Imin 68, Rehmanjan 49, Gul 29). 47 of the 114 filled in: 31 gave a year; 16 gave an "everybody."

3. What happened to the 16. Not discarded. Stamped 来路：传闻 — source: hearsay — and moved to a separate section. They may be perfectly true. I simply do not know, and now the book says so on its face.

4. And the 67 that cannot be filled. Of those, 41 were said by people who are now dead — Tursun, Grandpa Du, Master Qin, the old bookkeeper Zhou, and Abdul Rahman, who is alive and cannot speak. Those 41 went into one envelope, labelled: these forty-one are all that is left. And under it, after Wang Tao's line about the lost books: that they did it is proved by these forty-one cards — and on those forty-one cards it does not say whether they did it.

5. Third edition, and the new rule. Two colours as before, plus three compulsory fields per entry: who said it; what year he last did it; what year it was recorded. Field two unfillable means the entry drops to the hearsay section. Second edition: 47 entries (44 black, 3 red). Third edition main text: 31. Sixteen demoted. In the book: the third edition is thinner than the second. The sixteen that came out are not wrong — I don't know whether they are right. Until now those two were the same thing to me.

6. Luo's four, and the exchange on 5 April, which I would not have had the nerve to force. Of his twelve clauses, four dropped, including clause eleven — who is going to enforce these twelve — the one he added himself and the one I have quoted most often. He objected: you can't cut eleven. I said I had asked him what year he last did it, on 14 March, and his answer was this isn't something I did, it's something I thought. Silence. Did I say that? Yes, on the fourteenth. Then cut it. Quickly. And: go by your own rule. And: the rule is yours; break it once for me and there is no rule. And, written down the same day: I thought about clause eleven for thirty years, and I have genuinely never done it. Those are two different things. And: I'm sixty-three and I have only just learned they are two different things.

7. The last page of the third edition, which I titled twice. First: things nobody has done yet. Struck out — untrue and self-flattering. Replaced with: someone has done these; the person recording them didn't ask. The "person recording them" is me. The last page of this book is a list of what I failed to ask in four years. And underneath: if anyone ever fills in the provenance of one of these sixteen, move it to the front — and write your own name beside it.

"Each with its exact title, and the scroll it came from."
The compiler of a forty-scroll medical collection containing something over six thousand prescriptions, asked whether any of it was his own, would have said no. He had spent more than twenty years with the run of an imperial library, reading through the collections one by one and copying out whatever looked useful, and his stated policy was that he had not dared to add or discard by his own opinion.
What he did add was the attribution. Every borrowed item carries the title of the book it came from and the number of the scroll. By his own account copying a prescription took a quarter of an hour and establishing its source could take a day — sixty or more times the effort, spent on the part that contained no medical content whatsoever.
He is candid that he did not, at the time, have a good reason. He simply felt that an unattributed line becomes a line he is asserting himself, and he did not want to assert it.
The use arrived afterwards, and not in his lifetime. Of the fifty or sixty collections he had drawn on, most were subsequently lost — fires, wars, the ordinary attrition of manuscript transmission. What survives of several of them survives nowhere else: a handful of lines in his book, each followed by a title and a scroll number.
That those books ever existed is proved by those few lines.
There is a limit he states himself, and it is worth keeping because it prevents the practice becoming a fetish. He could trace a prescription back to a book. He could not trace it to the physician who first used it, in what year, on whom. I can only trace back one layer.
And the answer to the objection that follows from that:
One layer is better than none. That you cannot get to the bottom is not a reason not to start.