The later the net, the more ghosts it holds

Claim (verbatim)

If anthologists sample the poetry in circulation, then the fraction of their named poets who are already textually dead should climb with the anthology's date, because each passing century drowns more of the tail. Line up the anthologies chronologically, Vidyakara's Subhasitaratnakosa (c. 1100), Sridharadasa's Saduktikarnamrta (1205), Jalhana's Suktimuktavali (1257), Sarngadhara's Sarngadharapaddhati (1363), and the share of poets who have no surviving work should rise monotonically across the series. Sarngadhara, quoting at the far end, is quoting from a corpus that has lost two and a half more centuries of its members than Vidyakara's had, so his roster of the otherwise-lost should be visibly the fullest. This makes the anthology series a stratigraphic core: reading the loss-rate up the column dates the drowning, and turns a pile of medieval quotation-books into a time-series measurement of literary extinction. The instrument is the chronological ordering itself, which no single anthology can supply but which the sequence exposes as cleanly as tree-rings.

Prediction clause (verbatim)

Prediction: scoring each anthology's named poets for independent-work survival, the fraction of poets with no surviving work will rise monotonically from the Subhasitaratnakosa through the Saduktikarnamrta to the Sarngadharapaddhati, with the latest anthology's lost-poet fraction exceeding the earliest by at least fifteen percentage points; the monotone rise across the ordered series is the deciding clause. Survival is scored identically for every anthology against one fixed control (the New Catalogus Catalogorum), shared canonical poets are counted in each anthology they appear in, and the test voids if any anthology yields fewer than 100 datable attributed poets.

Kill-dataset (verbatim)

Kill (not yet built): the kavi-indexes of the four printed anthology editions (Kosambi and Gokhale, Banerji, Krishnamacharya's Gaekwad's Oriental Series 82, and Peterson) with survival scored uniformly against the New Catalogus Catalogorum and the poet identities controlled by Sternbach's Descriptive Catalogue of Poets Quoted in Sanskrit Anthologies and Inscriptions.

Novelty / leakage triage

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

No printed study lines the anthologies up as a dated series and scores their lost-poet fractions uniformly against one control; the ingredients - the four kavi-indexes, Sternbach's Descriptive Catalogue, the New Catalogus Catalogorum - are all published, but the time-series itself has not been assembled. The monotone-rise clause is genuinely at risk, because the Saduktikarnamrta's heavy sampling of near-contemporary Bengalis could hand the 1205 anthology a higher ghost-fraction than the later Deccan and western collections, breaking monotonicity in the middle of the series. An open outcome over published materials: adjacent.

• L. Sternbach, Subhasita, Gnomic and Didactic Literature, A History of Indian Literature IV.1 (Wiesbaden, 1974)
• L. Sternbach, A Descriptive Catalogue of Poets Quoted in Sanskrit Anthologies and Inscriptions, 2 vols. (Wiesbaden, 1978-1980)
• New Catalogus Catalogorum, ed. V. Raghavan et al. (Madras, 1949- )

Predictions

No prediction registered yet.