The largest corpora hang by single threads

Claim (verbatim)

Survival size and survival depth come apart spectacularly at the top of the troubadour roster. The largest surviving corpora belong to late poets who bypassed the anthology market: Cerveri de Girona's hundred-plus pieces ride almost entirely on one Catalan collection (Sg, the Cansoner Gil), Guiraut Riquier's 89 on his own book as copied into R with its dated rubrics, Bertran Carbonel's dozens of coblas on scarcely more - while the mid-size canonical corpora of the classic age, a Bernart de Ventadorn or a Folquet or a Gaucelm Faidit, are spread across dozens of witnesses each. The mechanism: the classic canon was replicated for a century by an international market, but the late giants were archived once, by self-collection or a single compiler's initiative, after that market had died - big because self-collected, fragile because self-collected. So the best-attested poets by count are the worst-attested by redundancy, and one lost binding would have deleted the biggest single bodies of troubadour verse outright. This inverts the intuition that large survival means robust survival, and it recalibrates the loss estimate: if corpora of a hundred pieces could reach us only by single-thread luck, the expected number of comparable corpora that had no such luck - and so are invisible - is not small.

Prediction clause (verbatim)

Prediction: in BEdT, among the three largest attributed corpora by piece count, at least two will have over two-thirds of their pieces on a single witness, while no poet in the fifth-through-fifteenth ranks by corpus size shows a single-witness rate above one-third (primary clause: the two-of-the-top-three over-two-thirds unica rates; the verdict follows it). Coblas transmitted in series count as pieces per BEdT's own itemization; razo-embedded copies do not count as witnesses. The test voids if per-song witness lists are incomplete for any of the top three corpora.

Kill-dataset (verbatim)

Kill: BEdT per-poet witness matrices - single-witness rates down the corpus-size ranking.

Novelty / leakage triage

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

The single-collection dependence of the late giants is established case by case - Cerveri's corpus rides on the Cansoner Gil, and Guiraut Riquier's self-collected libre with its dated rubrics is a classic study - but the comparative single-witness profile down the corpus-size ranking has never been computed. Note against the claim's framing that Riquier's libre is transmitted in C as well as R, so his pieces are typically double-witness and the two-of-top-three clause must carry on Cerveri plus the third-ranked corpus.

• M. de Riquer (ed.), Obras completas del trovador Cerveri de Girona (Barcelona, 1947)
• V. Bertolucci Pizzorusso, 'Il canzoniere di un trovatore: il libro di Guiraut Riquier', Medioevo romanzo 5 (1978)

Predictions

No prediction registered yet.