Are the Verse and the Commentary the Same Age?

The question the 推背圖 invites is when was it written, and it has no answer, because it is the wrong shape of question. A transmitted text is not an object with a date. It is a lineage of copies, each of which could add, drop, reorder or rewrite, and the thing you hold at the end has as many dates as it has layers.

What we have is a witness: a printing from 1915, whose own date is not in doubt. That fixes when this object was made. It fixes nothing about when its contents were composed — and article 1 showed the contents are at least three distinguishable layers.

So the better question is the one in the title, and this article does not resolve it. It sets out two observations that sit oddly together, and stops.

One: a commentary that sounds late

象33’s commentary closes on a line that is hard to read neutrally:

觥觥漢族對之得毋有愧

Sturdy Han people — set against this, can they be without shame? The leaf is about the Manchu entry through the passes, and the commentary has just observed that after the Liao and the Jin, Hu peoples have now twice ruled the Central Plain.

That is a Han-nationalist register, and it fits a late-Qing or post-Qing setting particularly naturally: 漢族 is used here as an ethnic self-designation, and the sentence puts a duty of shame on it.

And here the argument has to stop, deliberately. Register is not a date. To read this line as necessarily post-1911 you would have to know when such language could and could not be printed, who was printing it, and under what constraints — and that is publication history, which this corpus does not contain. Every claim in this series about the 1915 witness is traceable to a file you can open; this one would not be. So it is not made. The observation is that the line reads late. Whether it is late is a question for evidence we do not have.

Two: a verse that fits, and a commentary that misses it

象34’s verse, both halves:

頭有髮 · 衣怕白 · 太平時 · 王殺王

太平又見血花飛 · 五色章成裏外衣 · 洪水滔天苗不秀 · 中原曾見夢全非

Read that with the Taiping in mind and it is almost uncomfortable. 太平 is the movement’s own name — 太平天國, the Heavenly Kingdom of Great Peace — and it occurs in both halves. 頭有髮, the head has hair: the Taiping refused the Qing queue and were called 長毛, long-hairs, for it. 王殺王, king kills king: the Tianjing Incident of 1856, in which the Taiping kings destroyed each other. 洪水滔天, flood-waters reaching heaven, where 洪 is the surname of Hong Xiuquan.

Four hits on one leaf. If you were building a case that this verse was written after the fact, this is the leaf you would build it on.

Now read the commentary printed on that same leaf. After the hinge passage, it says: 疑遭水災或兵戎與天災共見此一亂也 — it is suspected there was a flood disaster, or warfare and natural calamity appearing together; this is one disorder.

A flood. Perhaps. The commentator, glossing a leaf that says 太平 twice and puts 洪 in front of the flood, took 洪水 literally and guessed at weather.

What that asymmetry is worth

The temptation is to turn this into an aphorism about forgery. Resist it; as a universal rule it does not hold. The argument that does hold is narrower and is about incentives.

Suppose verse and commentary were composed together, after 1864, by a redactor maximising the appearance of fulfilled prophecy. Such a redactor had every reason to make the identification explicit on this leaf — it is the strongest match in the book, and the commentary elsewhere names people freely. Instead the gloss guesses at a flood and moves on. The commentator’s failure to identify weighs against the simplest theory, which is that the two layers were manufactured as one piece after the event.

It weighs against that theory and nothing else. In particular it says nothing against layered composition, which is precisely the hypothesis space it opens:

• older commentary attached to verse that was later altered;
• older verse receiving commentary long afterwards;
• a compiler preserving material he had inherited and did not fully understand;
• several editorial hands at different removes, none of them coordinating.

Each of those accommodates both observations — a commentary whose vocabulary sounds late, and a commentary that fails to see what a post-1864 author could not have missed. None of them is established by anything above, and this article does not pick one.

Ending unresolved, on purpose

It would be easy to close this series with a verdict. The material does not support one, and manufacturing one here would repeat, in slower motion, exactly what article 4 documented happening in blue ink: taking a page that withholds, and supplying the part it withheld.

What the five articles do establish is smaller and more durable. The verse of this witness names no historical person. Its commentary names people constantly until 象34, where the commentator states that he is switching from verifying to inferring, and after which he names nobody and dates by cyclical branches that recur eight times a century. Five columns of this particular copy carry later handwriting or marks in a measurably different ink; three of the five carry a reading record, at three different strengths, and two are not identifications at all. Every one of the five sits on the far side of that same line.

The real object is stranger and more interesting than the legend, and every claim above is one you can go and check.