Six hundred schools, counted at the door

Claim (verbatim)

The Hanshu Yiwenzhi (漢書藝文志, Hanshu Yiwenzhi) is the oldest systematic book-catalogue in the Chinese tradition - Ban Gu's abridgment of Liu Xin's Qilüe (劉歆 七略), which in turn condensed Liu Xiang's Bielu (劉向 別錄) - and both parent catalogues are themselves lost, so the Yiwenzhi is a survival census whose own frame did not survive. Its closing tally is explicit and countable: 六略三十八種，五百九十六家，萬三千二百六十九卷 - six divisions, thirty-eight categories, 596 authors/schools, 13,269 juan. The loss mechanism is that this is a Han bamboo-and-silk library catalogued before the paper transition and before any medieval canon-forming institution existed; anything not actively recopied through the Wei-Jin upheavals and the material shift to paper simply ceased to exist. So the survival rate of Yiwenzhi-listed titles, measured against any independent full-text corpus, should be catastrophically low - far below the survival rate of works that first appear in the later, paper-era catalogues. Prediction restated: fewer than one in six of the roughly 596 titles survive as substantially complete transmitted texts; the rest are quotation-fragments, Qing reconstructions, or nothing. The catalogue counts the books at the door; almost none of them came back out.

Prediction clause (verbatim)

Prediction: matching the Hanshu Yiwenzhi's roughly 596 listed titles against an independent full-text corpus, fewer than one in six will survive as a substantially complete transmitted text - a whole-survival share below 0.17 - with the remainder surviving only as quotation-fragments, Qing reconstructions, or not at all (primary clause: the <1-in-6 whole-survival rate; the verdict follows it); a title counts as extant only on an independent full-text witness, not a later re-use of its name, and the test voids for coverage if fewer than 500 catalogue entries resolve to a checkable title.

Kill-dataset (verbatim)

Kill: the Chinese Text Project (ctext.org) full text of the Hanshu Yiwenzhi (漢書藝文志) for the catalogue census, cross-matched title by title against ctext's own pre-Tang full-text library and the Siku quanshu union holdings for the survival determination.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

That the great majority of Yiwenzhi-listed works are lost is a commonplace of the field, and per-entry survival status is available in the annotated-catalogue tradition (Yao Zhenzong's 條理/拾補; Gu Shi's 講疏; Yu Jiaxi's transmission-loss framework). But the primary clause is a specific aggregate — a whole-survival share below 0.17 (fewer than one in six of ~596) computed against a fixed full-text corpus — and that fraction is the reader's own sum, not a stated published result. Materials in print, aggregate arithmetic un-run: adjacent by the conservative test. The qualitative answer is near-certain, but the thresholded survival ratio as posed has not been reported.

• 姚振宗 (Yao Zhenzong), 漢書藝文志條理 and 漢書藝文志拾補 (in 二十五史補編)
• 顧實 (Gu Shi), 漢書藝文志講疏 (Shanghai: Commercial Press, 1924)
• 余嘉錫 (Yu Jiaxi), 古書通例 (on the transmission and loss of ancient books)

Predictions

No prediction registered yet.