The sign froze and the songs kept moving

Claim (verbatim)

The Masoretes of Tiberias finished, by the tenth century, one of history's most exact notation systems: the te'amim, cantillation accents fixed sign by sign in the great codices (Aleppo, c. 930; Leningrad, 1008/9) and copied since with checksum discipline. But the te'amim notate syntax and hierarchy - which words bind, where clauses cut - not pitch. The melodic realization of each sign stayed oral, and once the sign system froze, the melodies were free to drift apart regionally: a Yemenite, a Baghdadi, and a Moroccan reader chant the identical sign sequence of the identical verse to different musical substance. The drift is measurable because it was transcribed: Idelsohn's Thesaurus of Hebrew Oriental Melodies (ten volumes, 1914-1932) notates Pentateuch cantillation separately for the Yemenite, Babylonian, Persian, Eastern Sephardi, and Moroccan communities - parallel realizations of one frozen text. This is the exact inverse of the Greek case: there the notation died and took the music with it; here the notation lived precisely because it never carried the music, a syntax-script with an oral melodic plugin supplied per community. The mechanism: what is written is what gets audited - communities checked the signs against codices for a millennium, while melody, unwritten, was checked only against local memory, and memory diverges at the rate of community separation. The signs are the same age everywhere; the songs are only as old as each diaspora's isolation, and the contour scatter across communities should record exactly that.

Prediction clause (verbatim)

Prediction: coding, from Idelsohn's Thesaurus transcriptions, the Torah-reading realization of the twelve most frequent prose accents in five communities (Yemenite, Babylonian, Persian, Eastern Sephardi, Moroccan) into pinned contour classes (rise, fall, rise-fall, fall-rise, level, compound melisma, assigned from the transcription's note sequence), at least half the accents will show three or more distinct contour classes across the five communities, and no more than two accents will show all five communities agreeing in one class (primary clause: the >=6-of-12 accents at >=3 contour classes; the verdict follows it), while the underlying sign inventory is identical across all five; any accent not transcribed for at least four of the five communities is excluded, and the test voids entirely if fewer than eight accents clear that bar.

Kill-dataset (verbatim)

Kill (not yet built): a contour-coding table from Idelsohn's Thesaurus of Hebrew Oriental Melodies, vols. I-V (1914-1932), twelve accents by five communities, coded on the pinned contour classes.

Novelty / leakage triage

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

The drift itself is published: Idelsohn's Thesaurus transcribes Pentateuch cantillation separately for exactly these five communities (vols. I-V), his prefaces and his synthetic handbook compare the traditions in parallel motive tables, and the frozen-sign/moving-melody decoupling is his stated theme. But the pinned instrument - the twelve most frequent prose accents by five communities coded into six contour classes with agreement arithmetic - is an un-run coding: Idelsohn juxtaposes melodies without coding contours, and no later publication was located that tabulates per-accent contour-class counts across his five corpora.

• A.Z. Idelsohn, Thesaurus of Hebrew Oriental Melodies, vols. I (Yemenite, 1914), II (Babylonian, 1922), III (Persian, 1922), IV (Oriental Sephardi, 1923), V (Moroccan, 1929), prefaces and transcriptions
• A.Z. Idelsohn, Jewish Music in its Historical Development (New York, 1929), comparative cantillation tables

Predictions

No prediction registered yet.