The catalogue outlived the books

Claim (verbatim)

The loss of hagiography did not end with the Middle Ages; the instruments just finally became good enough to watch it happen. In the late nineteenth century the Bollandists and the French cataloguing enterprise described hagiographic manuscripts library by library - the Catalogi codicum hagiographicorum, the Catalogue general des manuscrits des departements. Then the fires came anyway: Strasbourg under Prussian shells in 1870, the Biblioteca Nazionale of Turin in January 1904, Tours in June 1940, and on 26 May 1944 the municipal library of Chartres, taking one of the oldest great French legendaries, from Saint-Pere-en-Vallee. Because cataloguing preceded combustion, the repertorium still lists witnesses that no longer exist; and for some BHL numbers the destroyed witness was the only one ever recorded. These are the newest ghosts in the system: texts that survived Vikings, damp, and the abridgement economy for a millennium, only to die catalogued, in industrial fires, within living memory of the catalogue. Counting them is both an act of accounting and a calibration: the modern, fully observed loss event is the one hard data point for the medieval loss rates that every other item in this lane must infer.

Prediction clause (verbatim)

Prediction: filtering BHLms witnesses to the destroyed collections (the Turin Biblioteca Nazionale's 1904 fire losses, Tours BM 1940, Chartres BM 1944, Strasbourg 1870), at least 20 BHL numbers will prove to have had their sole recorded witness among the destroyed manuscripts (primary clause: the >=20 modern-ghost count; the verdict follows it); secondary clause: the ghosts skew local, their saints belonging to the burned libraries' own dioceses and provinces at more than twice the rate of the full panel of destroyed witnesses; guard: manuscripts salvaged, relocated, or reconstructed (the IRHT's Chartres recoveries) count as extant; the test voids if survival cannot be settled for more than a quarter of the flagged shelfmarks.

Kill-dataset (verbatim)

Kill (partly not yet built): the sole-witness join runs now from BHLms against the pre-fire catalogues (the Bollandists' catalogi in Analecta Bollandiana; the Catalogue general); the survival-settlement layer needs the IRHT's 'A la recherche des manuscrits de Chartres' reconstruction data and the published Turin and Tours loss inventories assembled into one concordance.

Novelty / leakage triage

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

Every component is documented - the four fires are dated history, the Bollandist catalogi and the Catalogue general preceded them, Philippart's 1977 census already marks destroyed legendaries as such, and the IRHT's Chartres reconstruction programme supplies the survival-settlement layer - but the accounting itself (>=20 BHL numbers whose sole recorded witness burned, with the local-saint skew) has never been assembled; the kill note concedes the concordance must be built. Adjacent: a fully framed, unrun register-before-compute join.

• G. Philippart, Les legendiers latins (Typologie des sources 24-25, Turnhout, 1977), whose census marks destroyed witnesses
• Societe des Bollandistes, Catalogi codicum hagiographicorum latinorum of French municipal collections, published in Analecta Bollandiana (1880s-1890s)
• IRHT, 'A la recherche des manuscrits de Chartres' reconstruction programme, as the survival-settlement instrument

Predictions

No prediction registered yet.