Seven scholars, seven different songs

Claim (verbatim)

The one nearly complete piece of Bronze Age music, the Hurrian hymn h.6 from Ugarit, has been 'performed' many times - and the performances refute each other. Since publication, at least seven independent scholarly realizations have appeared: Wulstan (1971), Kilmer (1974, recorded in 1976 as Sounds from Silence), Duchesne-Guillemin (1975/1984), Vitale (1982), West (1994), Krispijn, and Dumbrill (1998/2005), with later entrants still arriving. They disagree not at the level of ornament but at the level of what kind of object the notation describes: whether the Akkadian interval terms denote simultaneous dyads or melodic motion; what the numerals following each term count; how, and whether, the Hurrian syllables align under the notes. The mechanism of the deadlock is that the Mesopotamian tuning texts fix the terms - each names a string-pair on a nine-stringed instrument - but the performance semantics died with the tradition; we inherited a vocabulary without its grammar, so every realization must import a grammar from outside, and the imports do not converge. That makes decipherment instability itself a measurable index of loss: a recovered system shows convergence over time (as Ventris's Linear B did within a decade), while an unrecoverable one shows a stable scatter. Fifty years of h.6 scholarship is the experiment already run; the conjecture is that its scatter has not contracted and can be shown not to have contracted by coding the realizations against one another.

Prediction clause (verbatim)

Prediction: coding every published scholarly realization of h.6 on three pinned binary axes - (a) interval terms read as simultaneous dyads versus melodic steps, (b) numerals read as repetition or note counts versus anything else, (c) syllable-to-note underlay treated as strict versus free - at least seven realizations will occupy at least five of the eight possible cells with no cell holding a majority (primary clause: the >=5-of-8-cells-occupied, no-majority-cell result; the verdict follows it), and no two realizations will yield an identical pitch sequence for the first ten instruction units; a realization qualifies if published in a peer-reviewed venue or research monograph; the test voids if fewer than six qualifying realizations can be located.

Kill-dataset (verbatim)

Kill (not yet built): a comparative alignment table of the published h.6 realizations - Wulstan (Music & Letters 1971), Kilmer (RA 1974; Sounds from Silence 1976), Duchesne-Guillemin (1975/1984), Vitale (Ugarit-Forschungen 1982), West (Music & Letters 1994), Krispijn, Dumbrill (1998/2005) - coded on the three pinned axes and note by note for the opening units.

Novelty / leakage triage

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

The mutual incompatibility of the h.6 realizations is itself a published topos: West's 1994 survey critiques his predecessors precisely on whether the interval terms are dyads or melodic motion and what the numerals count, Duchesne-Guillemin argued against the dyadic reading from 1975 on, and Dumbrill's monograph reviews the rival systems one by one. What was not located is the pinned comparison instrument - a three-axis cell coding of the seven-plus realizations with occupancy and majority arithmetic, or any note-by-note alignment table of the opening units across realizations; the disagreement is notorious in print, but nobody has tabulated it.

Predictions

No prediction registered yet.