The Commentator Tells You Where He Stops Knowing

Thirty-four leaves into the sixty, the commentary stops commenting and starts talking about itself. It is the most interesting passage in the book, and it is about the reliability of everything that follows it.

聖歎曰證已往之事易推未來之事難然旣證已往不得不推及將來吾但願自此以後吾所謂平治者幸而中吾所謂不平治者幸而不中而吾或可告無罪矣

Shengtan says: to verify what has already happened is easy; to infer what is to come is hard. But having verified the past, one cannot avoid extending it. I only wish that from here on, what I call orderly may by good fortune hit the mark, and what I call disorderly may by good fortune miss — and then I may perhaps be held blameless.

Two crops, because the passage is printed across two columns and each crop shows its own half. The 1915 page runs it continuously; the break is the column edge, not a pause.

Read that again, because it is doing something unusual. He is not claiming foresight. He is conceding that everything he has done so far was the easy half — matching a verse to an event already in the histories — and that from this leaf onward he is doing the hard half, which is guessing. Then he asks to be forgiven for the guesses, and specifies the direction he would rather be wrong in: he hopes the disorder he foresees does not arrive.

The break is visible in the text, not just in the sentence

It would be a nice line and nothing more if his practice did not change at exactly that leaf. It does, and you can count it.

Classify each leaf by one mechanical rule: a commentary is speculative if it contains 疑 (it is suspected) or 或 (perhaps), and retrospective if it contains neither. That is a token test, not a judgement about whether anything is true — it asks only whether the commentator marked his own sentence as a guess. Across the sixty leaves: 46 are retrospective, 12 are speculative, 象34 is the hinge he declares himself, and 象21 carries no commentary at all.

Now split those by where they fall. Before 象34, across leaves 1 to 33: thirty-two retrospective, one with no commentary, and not a single speculative leaf. After it, across leaves 35 to 60: twelve speculative and fourteen retrospective. Every hedging token in the book falls after the leaf on which he says he is about to start hedging.

There is a second, sharper way to see it. Fifty historical persons are glossed in this corpus, and thirty-four of them appear literally in the printed commentary. The last leaf whose commentary names any of them is 象31. After that — for the final twenty-nine leaves — the printed gloss names nobody. He stops identifying people three leaves before he announces that he is going to.

What the first half looks like

Before the hinge, the commentary is not prophecy at all. It is annotation, and it is specific in the way a history note is specific:

象6: 至德二載九月廣平王俶郭子儀收復西京 — in the ninth month of the second year of Zhide (757 AD), Prince Guangping Chu and Guo Ziyi retook the western capital. 象17: 景德元年契丹大衆入寇寇準勸帝親征 — in the first year of Jingde (1004 AD) the Khitan invaded in force and Kou Zhun urged the emperor to lead the campaign in person. 象19: 熙寧初王韶上平戎三策 — at the beginning of the Xining era, Wang Shao submitted his three strategies for pacifying the Rong.

Reign era, year, month, names, outcome. These are events that were three to eight centuries old when this commentary was written, and every one of them is checkable in a standard history. That is what he means by verifying what has already happened is easy.

And what the second half looks like

After the hinge, 象36 in full: 聖歎曰此象疑一女子能定中原建都長安 — this image, it is suspected, concerns a woman able to settle the Central Plain and establish a capital at Chang’an. No name. No year. No dynasty. A 疑 at the front doing exactly the work he said it would do.

One honest qualification, because the counts invite an overstatement. Fourteen leaves after the hinge are still classified retrospective — the rule is purely the presence of 疑 or 或, and a short flat sentence with no hedge in it counts as retrospective whether or not the commentator had any idea what he was describing. So “after 象34 he only guesses” is false. What is true, and is enough: he never hedges before that leaf, and he begins to immediately after.

You can audit this

None of the above is an impression. Every leaf in the structured corpus carries a commentaryStance object holding the classification, the rule that produced it, and an evidence list naming the tokens that actually fired on that leaf — ["疑"] on 象36, an empty list on 象6. If you disagree with a verdict you can see precisely what it was based on and overrule it, without reading any code.

The leaf itself is at bookofchanges.academy/texts/tuibeitu/03-xiang-31-45, and the per-column transcription status for the whole commentary — column by column, with each one’s own rating — is at commentary-witness.json. That second file matters more than it sounds: the verse of this corpus was adjudicated against the scan by a three-way check, while the commentary is a two-recogniser transcription, and 72 of its 97 columns carry a status the corpus itself calls a candidate rather than an acceptance. The two halves of this book are not established to the same standard, and the popovers show you which is which.

Why this is the load-bearing article

A text that hedges is easy to dismiss. That is not what is happening here. What is happening is that the commentary draws a line through its own work, states which side of it each half stands on, and asks to be judged accordingly: I only wish … that what I call disorderly may by good fortune miss. The line is printed on the page, in the same type as everything around it.

Everything the next three articles examine sits on the far side of it: the dates that are not dates, the identifications added in ink, the question of what was written when. He told us the far side was different. He was right.