>The Integrity Check

A signed commit is only as good as its signature. The compilers built a verifier that caught forged authorship the way a checksum catches a tampered file.

Chinese textual forgery was ancient and widespread. Attributing a text to a famous sage — Confucius, Laozi, the Yellow Emperor, the Duke of Zhou — gave it authority the real author could never claim. A medical text by an anonymous Song physician is just another file; the same text signed "Yellow Emperor" is the foundation of a tradition. A military manual by a minor official is forgettable; one signed "Jiang Ziya" is scripture.

By 1773 the accumulated forgeries numbered in the hundreds — some ancient and entrenched, their false signatures accepted so long that challenging them meant taking on centuries of consensus; others crude recent fabrications from the Ming's commercial publishing industry, where booksellers routinely bolted famous names onto unremarkable texts to move copies. The compilers approached this with professional enthusiasm. Verifying authorship wasn't a side task — it was central. The catalog's authority depended on correctly sorting genuine from fake.

The most powerful tool was linguistic analysis. Classical Chinese changed over time — not as fast as European languages, but enough that a trained reader could distinguish pre-Qin from Han, Han from Tang, Tang from Song prose. When a text claimed the Spring and Autumn period but used vocabulary that only appeared in the Han or later, the anachronism was a flag. Government terms, institutional names, argument formulae — all evolved, and a forger who didn't know the period's conventions left traces.

The Liu Tao (六韜) is the textbook case: it claims to record conversations between King Wen of Zhou and Jiang Ziya (eleventh century BCE), but the compilers found the rank "general" (將軍) — a title that "first appears in the Zuo Zhuan; it did not exist in the early Zhou." If the word for the rank under discussion didn't exist when the conversation supposedly happened, the conversation is fabricated. Simple, devastating: find the one detail the forger got wrong and the whole edifice collapses.

The dynasty system supplied a verification chain unavailable elsewhere: dynastic bibliographies. Each major dynasty compiled one (藝文志 or 經籍志) listing every known text — the Han Shu Yiwen Zhi, the Sui Shu Jingji Zhi, the Tang and Song catalogs — a chain of evidence from the first century BCE to the fourteenth century CE.

If a text claimed ancient origins but didn't appear in the relevant registries, the absence was strong evidence of forgery. The formula recurs across the catalog: 隋唐志皆不著錄 — "not listed in either the Sui or Tang dynastic bibliographies." The argument is probabilistic but powerful: the registries were compiled by officials with access to the imperial library and provincial collections, the most comprehensive information networks of their time. If a supposedly ancient text was unknown to every bibliographer for five centuries, the simpler explanation is that it didn't yet exist. Applied to the Guanshi Yizhuan (關氏易傳): 隋唐志皆不著錄，其為偽託明矣 — "not listed in the Sui or Tang bibliographies; that it is a forgery is self-evident."

The compilers tracked citation chains with forensic precision. If a text quoted a source that didn't exist when its supposed author was alive, the text was exposed. If it claimed antiquity but was first cited only much later, the gap was suspicious. If two texts cited each other in chronologically impossible ways, at least one was forged.

This caught forgeries vocabulary analysis alone might miss. A skilled forger could imitate archaic language — but imitating an entire intellectual ecosystem, knowing exactly which texts existed at a given moment and which didn't, was far harder. The compilers' deep command of the bibliographic tradition gave them an edge no individual forger could match.

Some texts claimed ancient origin, vanished for centuries, then resurfaced with no account of where they'd been. The compilers treated these transmission gaps as evidence demanding explanation. A text continuously cited Han-through-Song had a verified chain of custody; one supposedly from the Zhou but first mentioned in the Ming had a thousand-year gap to explain. Where was it? Who preserved it? Through what chain? No answers, and the gap itself was evidence.

This crushed the class of forgeries claiming to be "secret transmissions" (秘傳) — passed through private lineages outside the official record. If a text was so secret that no bibliographer, encyclopedia, or scholar in any surviving source ever mentioned it, the compilers concluded it was secret because it did not exist.

Their most exasperated investigation was the Zixia Yizhuan (子夏易傳), attributed to Confucius's disciple Zixia. It had been forged, debunked, re-forged, and re-debunked so many times that the review reads like a cold-case report with too many suspects: 未有如此書之偽中出偽，層層造作而不已者 — "none has generated forgery upon forgery, layer upon layer of fabrication without end, like this text."

The original, if it ever existed, was lost in the Han. Someone forged a replacement; it was detected. Someone forged another. A third forged yet another. Each successive forgery claimed to be the genuine ancient text and accumulated its own tradition of commentary. By the Qing the strata of fabrication needed a full paragraph of archaeology to untangle. The compilers didn't just declare it fake — they traced the entire history of its fakery: who forged what, when, and how each layer related to the last.

The methods anticipate Western textual criticism in striking ways. Systematic linguistic dating, cross-referencing bibliographic records, tracking citation chains, transmission-history analysis — all independently developed in European classical scholarship in the eighteenth and nineteenth centuries. Friedrich August Wolf on Homer (1795), Karl Lachmann on the New Testament — different materials, remarkably similar logic.

The comparison isn't exact. The Siku compilers had institutional limits: they wouldn't apply forgery-detection to the Yijing itself, or the Analects, or the Annals — those classics were assumed genuine and unquestionable. But within those limits the methods were rigorous and effective. They caught genuine forgeries that had fooled scholars for centuries, and two and a half centuries later their verdicts remain the starting point for any question of whether a classical Chinese text is what it claims to be.

Every dismissal rests on evidence; every forgery verdict is argued, not asserted. Fail any one test and the attribution is rejected. Fail all four and the review becomes a brief essay in textual forensics.