Inferpedia - an encyclopedia of the missing

AI-generated conjecture · below the evidence/publication boundary

← All conjectures · Art & iconography

Corbels come in runs

no prior located yet (provisional)

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 proposed connection between two domains, generated by a language model. 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)

Corbels come in runs. Romanesque corbel tables read as carnivalesque miscellany, but carving proceeded bay by bay from scaffolds, one carver working consecutive blocks with whatever type he was fluent in that week. Subject types should therefore show strong serial autocorrelation around the eaves — runs of heads, then beasts, then musicians — with run boundaries falling at structural bay divisions. The iconography is a work diary; the scaffolding schedule is recoverable from subject order alone.

Prediction clause (verbatim)

Primary clause (the verdict follows it): across at least 30 corbel tables documented with 20 or more corbels each, the median probability that adjacent corbels share a subject class exceeds the shuffled-within-building baseline by a factor of at least 1.8. Secondary clause: runs terminate at bay and buttress boundaries at at least twice the interior termination rate.

Kill-dataset (verbatim)

Kill: the Corpus of Romanesque Sculpture in Britain and Ireland (CRSBI), a free online corpus with per-corbel photographs and positions around each building.

Provenance

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

Generated by a fresh Fable-tier instance (claude-fable-5) at maximum effort with generation-first blindness (no repo reads, no web searches, no DB queries); 278-title list supplied at launch, titles only, no verdicts/dossiers seen; prompt pre-committed in docs/generated/conjectures_1001_wave_ledger.md and docs/generated/conjecture_fresh_fable_w05_prompt_20260708.txt. Wave W05 of docs/GOAL_CONJECTURES_1001_CAMPAIGN_20260708.md. Novelty unverified by construction. A platform output-limit resume occurred before the single Write; no additional tool calls or information ingress occurred.

Novelty / leakage triage

no prior located yet (provisional)

A provisional first pass authored by the model (Opus), not yet confirmed by the shepherd. It carries the same dated-search requirements as an authoritative verdict but is excluded from every headline figure and cannot underwrite a prediction until a shepherd confirms it.

[Independent blind re-audit 2026-07-08 (2nd pass, generator-independent), confidence medium. Double-confirmed: both the provisional pass and this re-audit located no prior formulation of the specific operationalized claim.] Romanesque corbel iconography is studied (Magrill, 'Figurated Corbels on Romanesque Churches: The Interface'), and the CRSBI corpus records per-corbel subjects and positions around each building — so the data exist. But I located no work testing the specific SERIAL-AUTOCORRELATION claim: that across >=30 corbel tables with >=20 corbels each, the median probability that adjacent corbels share a subject class exceeds the shuffled-within-building baseline by >=1.8x, with runs terminating at bay/buttress boundaries at >=2x the interior rate — reading subject order as a recoverable scaffolding/work-diary sequence. Corbels are usually read as carnivalesque miscellany; the runs-and-bay-boundary quantitative signature is the unlocated, testable claim.

Predictions

No prediction registered yet.

Weigh in

No community feedback yet.

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.