← All conjectures · Music, liturgy & ritual

Most chants were sung in one place only

Medieval Dated c. 800 CE – 1500 CE Europe

Status: Supported

The verdict’s fine print — quoted from the resolution record: “QUALIFIED supported.” Read the full caveats ↓

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).

This is a conjecture imagined by a language model — drawn from its trained weights and held to falsifiability, novelty, and plausibility, not to any one method: it may join two or more fields, or none. It is not an article and not evidence: it sits below the evidence/publication boundary. A quantitative prediction and a named kill-dataset are attached (when registered) so the claim stays falsifiable rather than merely evocative.

Claim (verbatim)

The received picture of Latin chant is a shared common core sung everywhere from one end of Europe to the other, and that core is real - but it is the visible tip of a very different distribution. Because Cantus Index records, for every distinct chant, how many indexed sources carry it, the whole survival record can be read as a frequency distribution, and that distribution is savagely long-tailed. A large fraction of distinct chants, counted by chant identity, appear in exactly one indexed source: they are unica, sung and written in a single place and nowhere else in the surviving record. The mechanism is that the common core was copied into every book and is therefore massively over-represented per chant, while the periphery - each house's proper offices, local saints, regional sequences and tropes - was each institution's own, copied rarely or never elsewhere, so most distinct chants are singletons even inside what survived. The famous shared repertory is a statistical illusion of counting performances rather than chants: count chants, and the majority were near-extinction the day they were written. Prediction restated: over 40 percent of distinct chant identities in Cantus Index are attested in exactly one indexed source, and the single-source fraction among sanctorale and proper items is at least three times that among the temporal common core.

Prediction clause (verbatim)

Prediction: computing the source-count distribution over distinct chant identities in Cantus Index, more than 40 percent of distinct chant IDs are attested in exactly one indexed source, and the single-source fraction among sanctorale and proper-office chants is at least three times the single-source fraction among temporale and common-core chants (primary clause: the over-40-percent corpus-wide singleton share; the verdict follows it). Distinctness is by chant ID, merging spelling variants; coverage guard: partial indexing inflates apparent rarity, so the measured singleton fraction is an upper-bounded estimate of true rarity and, if anything, the true share of one-place chants is higher, the test voiding only if source-count data is missing for over a quarter of IDs.

Kill-dataset (verbatim)

Kill: the Cantus Index per-chant source-count field, computing the fraction of distinct chant IDs with a source count of exactly one over the whole indexed corpus, then stratifying the singleton fraction by genre and by temporale versus sanctorale; the operation is a singleton-share of the concordance-count distribution.

Provenance

Run: Fresh agent generation · model: claude-fable-5

Fresh blind generation by claude-fable-5, 2026-07-17, liturgical-chant wave on CANTUS/Cantus Index, Corpus Troporum, and Analecta Hymnica/Chevalier. Every Kill names a real chant instrument and a countable census or inventory-geometric operation - transcribable-to-total ratios, source-count geometry, singleton (unica) fractions, catalogue-to-melody survival ratios, feast-rank concordance-breadth gradients, contraction ratios, and text-to-melody attestation lags - with thresholds far from 1 and explicit coverage guards distinguishing what the databases index from what existed. Operation family kept DISJOINT from the owned w09 music_liturgy ground (which joins chant metadata to external economic/material datasets: freight, wax, plague, mints, fairs, necrologies) and from the w08 chant cluster (variant-rate, melodic dialect, differentia decline, lesson-length, copying-error forensics). 0 items dropped; deliberately steered clear of w08-039 (Old Hispanic copying-error profile), w08-001/003 (feast-age variant rate / differentia), w09-026 (Old Roman property network), w09-016/035 (trope economics/prosopography), and w09-022 (sequence fair-network) by using pure census/inventory-geometry operations on the named instruments. Confidence flags on exact counts recorded in the register report. Slugs via django slugify.

Novelty / leakage triage

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

The nearest prior art is Hesbert's CAO, whose concordance tables across twelve witnesses already display a shared core against a thick local residue, and the locality of proper repertory is a commonplace of the LMLO material; but the corpus-wide singleton share of distinct chant identities in Cantus Index - the over-40-percent primary clause and the sanctorale-versus-temporale stratification - is un-run concordance geometry. The Cantus Index publications describe the network's holdings and identity apparatus without publishing a source-count distribution over chant IDs.

Sources cited by the triage

R.-J. Hesbert, Corpus Antiphonalium Officii, 6 vols (Rome, 1963-1979)
D. Lacoste, 'The Cantus Database and Cantus Index Network', in D. Shanahan, J.A. Burgoyne & I. Quinn (eds.), The Oxford Handbook of Music and Corpus Studies (New York, 2022)
A. Hughes, Late Medieval Liturgical Offices: Resources for Electronic Research, 2 vols (Toronto, 1994-1996)

Predictions

Supported registered 2026-07-18 calibration prediction (parent triage: leaked/adjacent)

Resolution: Supported

Caveats: QUALIFIED supported. The pinned primary clause is met cleanly: R=0.4179 > 0.40 with the full Wilson 95% CI [0.4137, 0.4221] entirely above threshold, and the INCONCLUSIVE_BY_DESIGN guard was NOT triggered (missing_data_fraction 0.0). The qualification is fourfold. (a) The margin above 0.40 is only ~1.8 points; the CI clears the line but not by much. (b) The conjecture's OWN coverage guard concedes that partial indexing INFLATES apparent rarity, so the measured singleton share is an UPPER BOUND on true one-place rarity: the measurement decides the REGISTERED pinned clause, but it is a weaker warrant for the underlying 'sung in one place' substantive claim. (c) The dump is a 2025-05-20 snapshot covering ~84.9% of the current ~62,782-ID population; live spot-checks showed heavily-attested IDs gaining sources while singletons did not, which would nudge the LIVE fraction slightly DOWN. The real uncertainty is this coverage gap (not sampling), which has no meaningful CI. (d) The secondary sanctorale/proper >=3x sub-clause could NOT be tested as pinned: only an office Mass/Office singleton split was available (0.384 vs 0.442, ratio ~1.15, nowhere near 3x). It is reported as a non-verdict secondary that was not cleanly testable, NOT as a failed clause, and it does not bear on the verdict. Independent VERIFY remains open per the no-self-grading rule.

Registered against Cantus Index (cantusindex.org) source-count data per chant ID. PREDICTION VERBATIM: Prediction: computing the source-count distribution over distinct chant identities in Cantus Index, more than 40 percent of distinct chant IDs are attested in exactly one indexed source, and the single-source fraction among sanctorale and proper-office chants is at least three times the single-source fraction among temporale and common-core chants (primary clause: the over-40-percent corpus-wide singleton share; the verdict follows it). Distinctness is by chant ID, merging spelling variants; coverage guard: partial indexing inflates apparent rarity, so the measured singleton fraction is an upper-bounded estimate of true rarity and, if anything, the true share of one-place chants is higher, the test voiding only if source-count data is missing for over a quarter of IDs.

Resolution criteria — the registered fine print

Resolution criteria: DENOMINATOR (verbatim from the prediction): distinct chant identities in Cantus Index, distinctness by chant ID merging spelling variants. NUMERATOR: chant IDs attested in exactly one indexed source. DATA: harvest, per Cantus Index chant ID, the count of distinct indexed sources carrying it (via the Cantus Index / CantusDB API or bulk export; freeze snapshot + record method). R = single-source IDs / distinct IDs. Secondary (non-verdict): the >=3x sanctorale/proper-office vs temporale/common-core singleton-fraction ratio. CLAUSE PRECEDENCE: (1) INCONCLUSIVE_BY_DESIGN (per the conjecture's own guard) if source-count data is missing for over a quarter of IDs; (2) SUPPORTED if R > 0.40; (3) REFUTED if R <= 0.40. Compute firewalled: the agent receives only the harvest+count operation, never the threshold or direction.

Known-priors disclosure — what the registrant already knew

Known priors disclosure: Triage (2026-07-17) graded adjacent and flagged this the wave's least-certain, most-interesting empirical bet (no published corpus-wide singleton distribution; CAO's 12-witness tables are the nearest art). The registrant additionally notes an interpretation caveat for the verdict stage (NOT disclosed to the compute): the conjecture's own coverage guard concedes partial indexing inflates apparent rarity, so a measured R>0.40 decides the pinned measurement but its bearing on TRUE one-place rarity must be caveated. Direction genuinely open.

Method and dataset — how it was measured

Register-before-compute: registration row in docs/generated/conjecture_prediction_batch2_20260718.json (session fable-depth-registrations-20260718-batch2, registered 2026-07-18 ~09:38Z, denominator quoted verbatim). Firewalled Sonnet compute (docs/generated/compute_cantus_singleton_20260718.json) at computed_at_utc 2026-07-18T10:21:43Z, which POSTDATES registered_at (rule 6). The compute received only the harvest+count operation, never the threshold or direction. Full-census aggregation of the dump; row-level mechanics cross-validated against the known worked example (cantus_id 007129: 22 occurrences / 20 distinct sources, dump == live) plus 5 further live json-con spot-checks (4/5 exact; the one drift was a heavily-attested ID gaining sources, no singleton gained a second source). Wilson 95% CI z=1.959964 on n_ids_measured. Registered clause precedence [(1) INCONCLUSIVE_BY_DESIGN if source-count missing >25% of IDs; (2) SUPPORTED if R>0.40; (3) REFUTED if R<=0.40] applied at this shepherd stage, not disclosed to the compute.

Dataset: DACT CantusCorpus v1.0 FULL bulk dump (github.com/dact-chant/CantusCorpus, release tag v1.0), chants.csv sha256 9c34b2de342cbb668fbc8883fe56e7a598f18e0b4f6637358421e47f888a19de, 888,010 occurrence rows across 10 component Cantus-Index databases, scraped 2025-05-20. n=53,282 DISTINCT Cantus IDs measured (distinctness by chant ID, verbatim denominator); distinct-source count per ID = count of distinct srclink values; 22,268 singletons. Not a sample: every occurrence row grouped by cantus_id with no subsampling. missing_data_fraction 0.0 by construction (dump carries only rows with non-empty srclink). The dump covers ~84.9% of the current ~62,782-ID live network population (11 databases as of 2026-07-18).

computed 2026-07-18

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.