The margin outlived the book

Claim (verbatim)

The loss of Byzantine books did not stop in 1453; it is still running, and the epigram corpus can measure its modern tail. DBBE anchors each occurrence to a physical manuscript through the same diktyon identifiers Pinakes uses, and Pinakes keeps records for manuscripts that no longer exist: Greek codices of Turin's Biblioteca Nazionale burned on 26 January 1904 but catalogued by Giuseppe Pasini in 1749; Escorial books lost in the fire of 1671; codices recorded olim in old inventories and never seen since. Wherever an editor or cataloguer transcribed a book epigram before its carrier perished, the verse now survives with zero physical witnesses: the margin outlived the book. A join of DBBE occurrences to Pinakes status fields therefore yields something rare - a countable set of Byzantine poems whose every material support is ash, preserved purely by the accident of early description - and a measured rate of post-medieval attrition on the epigram-bearing stock, which skewed deluxe and was therefore concentrated in exactly the famous collections the fires found. The operation also closes this wave's loop: the same instruments that count the medieval losses (items above) here count their own ongoing losses, the database becoming the last witness of its witnesses.

Prediction clause (verbatim)

Prediction: joining DBBE occurrences to Pinakes/diktyon manuscript status, at least 20 occurrences will lie in carriers recorded as destroyed or untraced (deperditus, olim), and at least 5 DBBE epigram types will have all of their occurrences in such carriers - poems surviving only as records of destroyed books (primary clause: the conjunction of >=20 destroyed-carrier occurrences and >=5 sole-carrier types; the verdict follows it); coverage guard, stated honestly: if DBBE ingested no pre-destruction catalogue transcriptions at all, the join returns zero by construction and the test voids for coverage rather than resolving; disambiguation: carrier status follows Pinakes' own fields, not external lists.

Kill-dataset (verbatim)

Kill: the DBBE open data export joined on diktyon numbers to Pinakes manuscript status fields - a destroyed-carrier occurrence count and a sole-carrier type census. Kill (partly not yet built): the fallback operation - hand-collating Pasini's 1749 Turin catalogue against DBBE types - is a build, not yet data.

Novelty / leakage triage

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

The components are printed - Pasini's 1749 Turin catalogue and the 1904 fire, de Andres's catalogue of the Escorial Greek codices lost in 1671, Pinakes' own deperditus and olim status conventions - but the join is new: no one has counted DBBE occurrences whose carriers are destroyed or untraced, nor censused epigram types surviving solely in records of dead books, and the item's own guard honestly concedes the count may void on DBBE's ingestion policy. A genuine open computation over the two databases.

• G. Pasini et al., Codices manuscripti Bibliothecae Regii Taurinensis Athenaei (Turin, 1749)
• G. de Andres, Catalogo de los codices griegos desaparecidos de la Real Biblioteca de El Escorial (El Escorial, 1968)
• Pinakes: Textes et manuscrits grecs (IRHT, pinakes.irht.cnrs.fr), manuscript status fields

Predictions

No prediction registered yet.