Three tablets copied through the mouth

Claim (verbatim)

Three rongorongo tablets carry long stretches of one and the same text: the Great Santiago tablet (H, Museo Nacional de Historia Natural, Santiago) and the Great and Small St Petersburg tablets (P and Q, Kunstkamera) — the parallel Boris Kudrjavtsev noticed as a student in wartime Leningrad (published posthumously, 1949), worked through by Butinov and Knorozov and by Barthel, and traced glyph by glyph in Paul Horley's comparative studies (Rapa Nui Journal). The triple is the closest thing to a stemma the tradition left, and it should betray its copying channel. A visual copyist confuses what looks alike and skips at arbitrary points; a scribe re-inscribing a memorized chant holds the spoken sequence fixed while the graphic packaging floats — the same base signs turn up differently fused, appended and spaced — because on Rapa Nui the tablet was a score for recitation and the exemplar was the mouth, not the wood. The claim: H, P and Q are recitation-anchored re-inscriptions, not eye-copies of a wooden exemplar. Prediction: in a sign-level alignment of the shared H/P/Q text, at least 70% of non-identical aligned positions will differ only allographically under Barthel's coding — same base-sign number with divergent ligature, appendage or hand variant — while substitutions of graphically unrelated signs stay under 10%, and insertions or deletions fall at boundaries of repeated formulas rather than mid-formula (primary clause: the 70% allographic share of mismatches; the verdict follows it). Kill: Barthel's 1958 transcriptions of texts H, P and Q in the CEIPP digitization, checked against Paul Horley's published comparative tracings of the three tablets (Rapa Nui Journal); the alignment and the mismatch classification are a mechanical computation over the coded lines.

Prediction clause (verbatim)

Prediction: in a sign-level alignment of the shared H/P/Q text, at least 70% of non-identical aligned positions will differ only allographically under Barthel's coding — same base-sign number with divergent ligature, appendage or hand variant — while substitutions of graphically unrelated signs stay under 10%, and insertions or deletions fall at boundaries of repeated formulas rather than mid-formula (primary clause: the 70% allographic share of mismatches; the verdict follows it).

Kill-dataset (verbatim)

Kill: Barthel's 1958 transcriptions of texts H, P and Q in the CEIPP digitization, checked against Paul Horley's published comparative tracings of the three tablets (Rapa Nui Journal); the alignment and the mismatch classification are a mechanical computation over the coded lines.

Novelty / leakage triage

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

The H/P/Q parallel is the classic tool for identifying allographs, ligatures and graphic variants (Kudrjavtsev's discovery; Horley's tracings; Davletshin's substitution-method study), so the allographic character of much inter-tablet variation is qualitatively established. But the copying-channel question — recitation-anchored re-inscription versus eye-copy — has not been posed as a quantitative mismatch-classification test with an allographic-share threshold.

• B.G. Kudrjavtsev, 'Pis'mennost' ostrova Paskhi', Sbornik Muzeya Antropologii i Etnografii 11 (1949), 175-221
• A. Davletshin, 'Allographs, Graphic Variants and Iconic Formulae in the Kohau Rongorongo Script of Rapa Nui (Easter Island)', Journal of the Polynesian Society 126.1 (2017), 61-92
• P. Horley, 'Structural analysis of rongorongo inscriptions', Rapa Nui Journal 21 (2007)

Predictions

No prediction registered yet.