AI-generated conjecture · a wager, not a finding
← All conjectures · Literature & poetics
The grammars taught from a short shelf
Status is derived only from the shepherd-authored triage/prediction data above -- community submissions and claims are a separate overlay and can never change it (see the participation panel below).
Claim (verbatim)
The tradition wrote its own textbooks: Raimon Vidal's Razos de trobar - the first grammar of a Romance vernacular, written for would-be composers - Uc Faidit's Donatz proensals made in Italy in the 1240s, Terramagnino da Pisa's verse Doctrina d'acort, Jofre de Foixa's Regles de trobar from Sicily. The prose ones teach by quotation, citing lines of real songs as specimens of right and wrong usage, and that quotation layer is an independent sample of what circulated where the teaching happened - drawn outside the anthology channel, by different hands, for a different purpose. This conjecture claims the sample carries a double signal. Overwhelmingly, the grammarians should quote the same narrow canon the chansonniers canonized: teaching ran on the export repertoire, so the quotation census will look like the anthology census - the filter seen from a second instrument, which is itself evidence that one filter, not many, made the tradition's memory. But the residue is loss made visible: a non-trivial share of quoted lines that match no surviving song, each one a piece that existed in a teacher's working memory around 1200-1240 and exists nowhere now. Marshall's editions identify the quotations line by line; the join to the survival record is mechanical, and both signals - the canon-echo and the unmatched residue - come out of the same count.
Prediction clause (verbatim)
Prediction: over the identified quotations in the Razos de trobar tradition and associated texts (Marshall's 1972 edition, with the Donatz proensals in Marshall 1969), the quoted songs' pooled median BEdT witness count will be at least twice the corpus-wide median (primary clause: the two-fold quoted-over-corpus median ratio; the verdict follows it). Secondary: at least one in twelve distinct quotations matches no PC/BEdT item, counted on Marshall's own non-identifications. Distinct quotations are counted once per quoted song per treatise. The test voids if the distinct-quotation count runs under 40.
Kill-dataset (verbatim)
Kill: Marshall's quotation apparatus (1972 and 1969 editions) joined to BEdT witness counts - the transmission profile of quoted songs, and the unmatched-quotation share.
Nobody has run this test. The kill-data is named above. If you can run it — or you know the paper that already settles it — claim the kill or submit the prior scholarship. Kills and prior scholarship are credited here, by name, as they come in.
Provenance
Run: Fresh agent generation · model: claude-fable-5
Fresh blind generation by claude-fable-5, 2026-07-17, troubadour/Old Occitan wave instrument-anchored on the BEdT (Bibliografia Elettronica dei Trovatori, bedt.it - per-song PC identities, per-chansonnier attestations, per-witness attributions, genres, dates, music fields), with the printed censuses as controls: Pillet-Carstens 1933, Frank's Repertoire metrique 1953-57, the chansonnier sigla corpus (Brunel 1935; Zufferey 1987), Boutiere-Schutz for vidas/razos, Marshall 1969/1972 for the grammars, van der Werf 1984 for melodies. Every Kill names one instrument and one countable/positional/attestation-geometric operation; thresholds far from 1; disambiguation rules and coverage guards pinned inside each prediction; one item (no. 8, the mention-census diff) carries an honest 'Kill (not yet built)'. DISJOINTNESS from owned European lyric ground, checked by grep across all packets: w01 no.30 'Biography replaces the stage' owns vida/razo geography-of-production (this wave never tests vida presence by region); w07 no.2 'The life replaces the tune' owns the vida-melody substitution and the A/B/I/K-vs-G/R/W/X contrast (this wave's melody items run different arithmetic - per-poet text-mass concentration in no.1 and R's internal empty-stave audit in no.16, both flagged adjacent); w07 no.38 'Anthologies are funerals' owns the composition-to-witness time-lag operation, so that steer was DROPPED here and replaced by the home-vs-export archiving order (no.10, a different geometry, flagged adjacent); w07 no.49 'Hymns steal from love songs' owns melody-concordance contrafacture direction (no.6 here is metrical, within-Occitan, a Frank-scheme ghost-model census, flagged adjacent); w07 no.41 owns Dante-vs-Franco-Italian epic (no.12 here is the De vulgari eloquentia citation channel, a different claim). Honesty flags: melody totals (roughly 250) and the trobairitz roster vary by edition and are pinned to named instruments inside each prediction; gallery-roster identifications follow the critical editions; no in-house BEdT extract exists yet, so 'immediately resolvable' means the instrument is live and public while the harvest itself is a small build.
Novelty / leakage triage
anticipated in the literature — this exact test has never been run
Marshall's editions identify the grammarians' quotations line by line, note the canonical tilt of the quoted poets, and leave the non-identifications on record; but the two instruments have never been joined - no published count sets the quoted songs' witness profile against the corpus-wide median, so the two-fold ratio of the primary clause and the one-in-twelve unmatched share exist only as scattered apparatus, not as a run measurement.
- J.H. Marshall (ed.), The Razos de trobar of Raimon Vidal and Associated Texts (London, 1972)
- J.H. Marshall (ed.), The Donatz proensals of Uc Faidit (London, 1969)
Predictions
No prediction registered yet.
Weigh in
No community feedback yet.
New here? Create an account first
Create an account or sign in and your feedback is tied to you — you can track it, get replies, and claim this conjecture so others know you’re working on it. Prefer not to? Just leave your take below as a guest — only the name you type is shown.
Add your take
Posted immediately (spam is removed). Community feedback is never an adjudicated verdict and never changes this conjecture's triage label or status above.