The footnotes outweigh the history

Claim (verbatim)

Chen Shou's Sanguozhi (陳壽 三國志, c. 289) is a terse history; the reason we know the Three Kingdoms in depth is Pei Songzhi's commentary (裴松之, presented 429), which does not gloss the text so much as flood it with rival sources - quoting, by title and at length, well over 150 distinct works to supplement, correct, and contradict Chen Shou. Of those cited works - the Weilue (魏略), the Wei shu (魏書) of Wang Shen, the Jiangbiao zhuan (江表傳), the Shiyu (世語), the Han Jin chunqiu (漢晉春秋) and scores more - the overwhelming majority no longer exist as independent texts. The mechanism is a historian's version of the leishu filter: Pei Songzhi excerpts the vivid and the contentious, and once the parent histories fall out of copying his footnotes become their only tomb. The commentary is thus heavier, as surviving evidence, than the classic it annotates - a quotation reservoir whose sources died around it. Prediction restated: more than 85% of the distinct works Pei Songzhi cites have no independent full-text survivor, and one Weilue- or Wei-shu-class lost history supplies a large share of all citation instances. The apparatus meant to serve the history has outlived almost every book it was built from.

Prediction clause (verbatim)

Prediction: extracting the distinct works Pei Songzhi cites in the Sanguozhi commentary (三國志裴注) and testing each for independent survival, more than 85% of the distinct cited titles will have no surviving independent full-text witness, and the single most-cited lost history (a Weilue- or Wei-shu-class title) will account for a large share of citation instances (primary clause: the >85% lost-among-distinct-cited-titles rate; the verdict follows it); a work counts as surviving only on a full-text witness outside the commentary tradition, and the test voids for coverage if fewer than 100 distinct cited titles can be extracted.

Kill-dataset (verbatim)

Kill: the Chinese Text Project (ctext.org) full text of the Sanguozhi with Pei Songzhi's commentary (三國志裴松之注), extracting attributed citation titles from the commentary and testing each against ctext's full-text library for independent survival.

Novelty / leakage triage

already answered in the literature

The discipline flags this correctly: Pei Songzhi's sources are counted in print. The enumeration of the works cited in his Sanguozhi commentary and their overwhelmingly lost status is catalogued — Shen Jiaben's 三國志注引書目 and de Crespigny's study of the Records of the Three Kingdoms lay out the cited titles and that the great majority survive only through the commentary. The primary clause (>85% of distinct cited titles have no independent full-text witness) is therefore guaranteed by the published catalogues, which already register the lost status work by work; only the exact percentage against ctext remains a recomputation. Leaked.

• 沈家本 (Shen Jiaben), 三國志注所引書目 (in 沈寄簃先生遺書)
• Rafe de Crespigny, The Records of the Three Kingdoms (Canberra: Australian National University, 1970)

Predictions

No prediction registered yet.