Muqabala has ears, not eyes

Claim (verbatim)

Muqabala has ears, not eyes. This connects collation practice with the error spectrum of the resulting copies. Muqabala was typically performed aloud: one party reads the exemplar while the other follows the new copy. An acoustic channel catches what the ear can hear, namely omitted words, skipped lines, and transposed clauses, and it passes what only the eye could catch, namely pointing errors, orthographic variants, and substitutions that sound alike. Collation should therefore not shrink the error distribution uniformly; it should truncate it asymmetrically, slashing large omissions while leaving small substitutions almost untouched. The collation note is a filter with a knowable frequency response.

Prediction clause (verbatim)

Comparing witness pairs of the same work where one witness carries a collation statement (balaghat muqabalatan, qubila) and another does not, collated witnesses show at least a 3x lower rate of omissions of three or more consecutive words relative to the critical text, while their rate of single-word substitutions differs from uncollated witnesses by less than 25%. Primary clause: the 3x-or-greater omission-rate ratio; the verdict follows it.

Kill-dataset (verbatim)

Kill: KITAB passim pairwise alignments over OpenITI texts with multiple transcribed witnesses, filtered by collation statements recorded in FIHRIST and the source catalogues.

Novelty / leakage triage

provisional — model-triaged, shepherd review pending

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. Provisional reading: Adjacent (closely related prior work exists).

That collation (muqabala) was standardly an ACOUSTIC operation -- one party reading the exemplar aloud while another follows the new copy -- is documented in the handbook literature (Gacek's Vademecum and the standard collation-note glossaries), which is exactly the item's premise that the channel is the ear, not the eye. Not located: the item's frequency-response claim -- that collated witnesses show a >=3x lower rate of >=3-word omissions than uncollated ones while single-word (often homograph/near-homophone) substitutions differ by <25% -- run over KITAB passim alignments of multi-witness OpenITI texts filtered by collation statements. The oral character of muqabala is established; the asymmetric error-filter measurement is unrun.

• Adam Gacek, Arabic Manuscripts: A Vademecum for Readers (Brill, 2009) -- entries on muqabala/collation — Collation performed by reading aloud from the exemplar; the acoustic channel the item's asymmetry depends on
• KITAB project passim text-reuse alignments over OpenITI (multi-witness works) — The kill dataset for the omission-vs-substitution error-spectrum comparison

Predictions

No prediction registered yet.