Resolution: Killed
Caveats: (a) The qualitative direction survives: Pseudo-Jonathan IS more expansive than Onkelos -- R=1.2762 with a tight per-book range (1.2157-1.2947 across all five books of the Torah), consistent with the standard characterization of the expansive Palestinian recension vs the terse Babylonian-authorized one (Klein; Flesher-Chilton). What fails is the specific magnitude the conjecture wagered: a per-verse expansion ratio > 2.0. Measured R=1.28 <= 2.0 falls under the registered REFUTED clause. (b) The secondary fragment-coverage clause HELD: the Fragment Targum ('Targum Jerusalem' on Sefaria) covers 14.85% of Torah verses, under the predicted 0.5 -- but that clause was registered as non-verdict and cannot rescue the primary. (c) Coverage is clean (Onkelos 100.0%, PsJ 99.86%, both well above the 90% void floor), so this is a genuine refutation, not a void. (d) CALIBRATION SIGNAL -- this is the third consecutive resolution of the identical shape: direction correct, bold quantitative threshold overshooting the measured value. PELLA wagered posthumous:lifetime >= 3:1, measured 1.69; Yerushalmi wagered Bavot/rest segment-length ratio < 0.67, measured 0.87; Targum wagered PsJ/Onkelos > 2.0, measured 1.28. In each case the scholarly qualitative claim holds while the Fable-generated numeric threshold was too aggressive: the noetome tends to over-pin magnitudes on directionally-correct claims. This is an emerging finding about the conjecture instrument itself, worth its own essay or calibration note. It is NOT a flaw in the resolution machinery, which is working exactly as designed by catching the overshoot.
Registered against Sefaria's open Targum corpus (verified live open). PREDICTION VERBATIM: Prediction: comparing targum word-count per Torah verse in Sefaria, Targum Pseudo-Jonathan (the expansive Palestinian recension) will exceed Targum Onkelos by more than a factor of two (a per-verse expansion ratio > 2.0), while the strictly fragmentary Palestinian witnesses (the Fragment Targums) survive on under half of the Torah's verses (< 0.5 coverage) where Onkelos survives on essentially all - primary clause: the > 2.0 Pseudo-Jonathan-to-Onkelos per-verse word-count ratio; the verdict follows it. The fragment-coverage clause resolves in Sefaria only if those witnesses are digitized, else it defers to Klein's critical editions; the test voids for coverage if Pseudo-Jonathan or Onkelos is under 90% digitized on the Torah.
Resolution criteria — the registered fine print
Resolution criteria: DENOMINATOR (verbatim): targum word-count PER TORAH VERSE in Sefaria, for Targum Pseudo-Jonathan and Targum Onkelos. DATA: fetch both targumim from the Sefaria API/export at the verse (Torah chapter:verse) level; word-count per verse = whitespace-token count of the Aramaic text; freeze snapshot+sha256; report the digitization coverage (fraction of the Torah's ~5,845 verses each targum carries non-empty text for). R = mean(PsJ words/verse) / mean(Onkelos words/verse) over verses where BOTH are present. SECONDARY (non-verdict, verbatim): the Fragment Targums survive on <0.5 of Torah verses where Onkelos survives on essentially all — resolves in Sefaria only if the Fragment Targums are digitized, else defers to Klein's critical editions (report which route). CLAUSE PRECEDENCE: (1) INCONCLUSIVE_BY_DESIGN (the conjecture's void rule) if Pseudo-Jonathan OR Onkelos is under 90% digitized in Sefaria; (2) SUPPORTED if R > 2.0; (3) REFUTED if R <= 2.0. Compute firewalled from threshold and direction.
Known-priors disclosure — what the registrant already knew
Known priors disclosure: Triage (2026-07-17) graded ADJACENT: the expansive-vs-literal character of Pseudo-Jonathan vs Onkelos is textbook (Klein 1980/1986; Flesher-Chilton), but the specific >2.0 per-verse word-count ratio over Sefaria is un-run and the exact multiple is uncertain (PsJ's expansion is uneven across the Torah). Direction genuinely open — the >2.0 threshold is a real test, not a foregone conclusion.
Method and dataset — how it was measured
Register-before-compute: the prediction and its clause precedence (void if PsJ or Onkelos <90% digitized on the Torah; SUPPORTED if R>2.0; REFUTED if R<=2.0) were registered verbatim in conjecture_prediction_batch4_20260718.json at 2026-07-18T12:15:13Z, before any measurement. Compute was firewalled (claude-sonnet, threshold and direction withheld) at 2026-07-18T12:25:48Z, postdating registration. Identity verification against Sefaria's own index metadata: 'Targum Jonathan on {book}' = Targum Pseudo-Jonathan (per Sefaria's enDesc), 'Onkelos {book}' = Targum Onkelos, 'Targum Jerusalem' = the Fragment Targum tradition. Tokenization: Sefaria 'he' Aramaic text, HTML tags stripped to whitespace, strict whitespace word-count per verse; canonical denominator 5,846 Torah verses from the Sefaria shape API. Coverage gate applied first: Onkelos 100.0%, PsJ 99.86% -- both above the registered 90% void floor, so the void clause does not trigger. R = mean(PsJ words/verse)/mean(Onkelos words/verse) over the 5,838-verse intersection where both are present.
Dataset: Sefaria Targumim; PsJ 'Targum Jonathan' vs 'Onkelos', 5838-verse intersection, per-verse word counts
computed 2026-07-18