The catalogue that dates its own graves

Claim (verbatim)

The Suishu Jingji zhi (隋書經籍志, Suishu Jingji zhi, c. 656) is unique among the dynastic bibliographic treatises in that it timestamps its losses. Where a later cataloguer would simply omit a vanished book, its compilers worked with two lists open - the Sui imperial holdings and the earlier Liang court catalogue in the tradition of Ruan Xiaoxu's Qilu (阮孝緒 七錄) - and recorded the difference in situ, annotating hundreds of entries with the formula 梁有...卷，亡 ('the Liang had X juan; now lost') or 梁有...今亡. That marginal accounting brackets a specific catastrophe: the burning of the imperial library at Jiangling in 554, when Emperor Yuan of Liang put the traditional 140,000 juan to the torch as Western Wei forces took the city. So the loss annotations are not scattered noise but a dated stratum, and they should fall unevenly across the four divisions: the ji (集, belletristic collections) and zi (子, masters) divisions, whose contents lived by private prestige and had no classroom to recopy them, should carry the wang (亡) marks far more densely than the jing (經, classics), which the examination and teaching apparatus reproduced generation on generation. Prediction restated: the collections division loses at least three times the per-entry rate of the classics, with the classics the least-annotated division of the four. This is a loss-record whose stratigraphy is legible because a seventh-century bureau wrote its subtractions down.

Prediction clause (verbatim)

Prediction: counting the loss-annotation formulae (梁有...亡 'the Liang had it; now lost', and its close variants) across the Suishu Jingji zhi (隋書經籍志), the per-entry 亡-rate in the collections division (集部) will exceed the per-entry rate in the classics division (經部) by at least a factor of three, and the classics will be the least loss-annotated of the four divisions (primary clause: the >=3x collections-over-classics per-entry loss ratio; the verdict follows it); if fewer than 200 loss annotations are found in the whole treatise, the test voids as a coverage failure rather than resolving.

Kill-dataset (verbatim)

Kill: the Chinese Text Project (ctext.org) full text of the Suishu Jingji zhi (隋書經籍志) - string-match the 亡 and 梁有...亡 annotation formulae, assign each hit to the treatise's own division header (經史子集), and compare per-entry loss rates across divisions.

Novelty / leakage triage

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

The Suishu Jingji zhi's dated 梁有...亡 loss stratum is thoroughly described in the philological literature — Yao Zhenzong's 隋書經籍志考證 and the Kōzen–Kawai 隋書経籍志詳攷 annotate the Liang-versus-Sui subtractions entry by entry — so the annotation formula and its historical bracket (the 554 Jiangling burning) are established, not novel. What is not in print is the primary clause's arithmetic: the per-entry 亡-rate computed division by division, and specifically the >=3x collections-over-classics ratio with 經 as the least-annotated division. That is a new per-division computation of the discipline's flagged 'materials-in-print = adjacent' kind; the raw annotations are published but their differential rate across 經史子集 has not been tabulated as the clause pins it. The nearest prior art works the immediate neighborhood without running the test.

• 姚振宗 (Yao Zhenzong), 隋書經籍志考證, in 二十五史補編 (Beijing: Zhonghua shuju repr.)
• 興膳宏・川合康三 (Kōzen Hiroshi & Kawai Kōzō), 隋書経籍志詳攷 (Tokyo: Kyūko Shoin, 1995)

Predictions

No prediction registered yet.