Genre composition of unprovenienced versus excavated cuneiform tablets

Provenance notice. This article exists because a conjecture — a falsifiable public prediction, staked in advance against a named dataset — was pre-registered (its clause precedence fixed before the provenience-by-genre cross-tabulation ran) and killed, on the Inferpedia conjectures campaign's 1001-blind lane. The conjecture, fresh-fablemax-b3-20260705-008, "The dealer's razor", remains killed and untouched at its own L1 lead status; nothing here reopens or revises it. The triage verdict was "adjacent" (the market-selection mechanism the conjecture invokes is qualitatively established in the literature on Assyriology's looting and provenience problem; no quantitative genre-share comparison had been located), so the resolution is a calibration test, not a novel-discovery claim, and this article makes none. This article was published in beta after a recorded Fable publication judgement (docs/generated/publication_judgement_616_20260716.json); every figure was recomputed from the committed per-period, per-stratum artifact before publication.

Epistemic status

This article reports a measured pattern in surviving records: the numbers are directly attested in the cited datasets; what they mean is an open interpretation.

The direction of the measured result here runs opposite to the conjecture's own stated prior, which the registrant disclosed honestly in advance: the prediction, following the qualitative literature on art-market selection, was that tablets lacking a recorded findspot ("unprovenienced," a proxy for having passed through the antiquities market) would be enriched in prestige genres — literary, lexical, royal — and depleted in mundane administrative bookkeeping, relative to excavated tablets of the same period. The measured genre shares are direct counts against CDLI catalogue labels; what they attest is unambiguous. What is inferred — clearly labelled as such below, and kept separate from the count itself — is a candidate historical explanation for why the reversal might have happened.

Summary

Within CDLI catalogue metadata, tablets were split by whether they carry a recorded find location (provenienced) or not (unprovenienced, a proxy for market origin), and the non-administrative genre share was compared between the two groups within each of two periods. In the Ur III period, provenienced tablets are 2.16% non-administrative (58,320 records) and unprovenienced tablets are 2.03% non-administrative (1,135 records) — a ratio of 0.94, meaning the unprovenienced stratum is if anything slightly MORE administrative, not less. In the Old Babylonian period, provenienced tablets are 58.04% non-administrative (10,381 records) and unprovenienced tablets are 38.05% non-administrative (3,327 records) — a ratio of 0.656, the same direction: unprovenienced tablets are markedly MORE administrative than provenienced ones. The conjecture predicted the opposite direction, by a wide margin (at least 1.8x enrichment in literary-plus-royal genres, in at least four of five tested periods); the pre-registered kill clause (unprovenienced non-administrative share at or below 1.1x the provenienced share in both periods, or lower than it in either) fires in both periods tested.

What is being inferred

Nothing about WHY market-surfaced tablets skew administrative is directly measured; that is offered below as one plausible historical reading, not as a finding. The genre counts and the ratio are direct; the explanation for the ratio is not. This article also does not claim anything about tablets that never entered CDLI's catalogue at all — the universe tested here is CDLI-catalogued material only, and prestige pieces that moved through art-market channels without ever being catalogued by CDLI are, by construction, invisible to this test.

What is attested

A conjecture was pre-registered (fresh-fablemax-b3-20260705-008, registered 2026-07-05T18:57:20Z): for each of the Ur III and Old Babylonian periods, the non-administrative genre share was to be compared between unprovenienced and provenienced CDLI records, with clause precedence fixed in advance and evaluated strictly in order: (1) KILLED if the unprovenienced non-administrative share is at or below 1.1x the provenienced share in BOTH periods, or lower in either period; (2) SUPPORTED if the ratio is at least 1.5x in BOTH periods; (3) otherwise INCONCLUSIVE.

The computation (2026-07-05T18:58:54Z): Ur III — provenienced 58,320 records, 2.16% non-administrative; unprovenienced 1,135 records, 2.03% non-administrative; ratio 0.94. Old Babylonian — provenienced 10,381 records, 58.04% non-administrative; unprovenienced 3,327 records, 38.05% non-administrative; ratio 0.656. Clause (1) fires: the unprovenienced non-administrative share is lower than the provenienced share in both periods (0.94x and 0.656x), far from the 1.5x support bar. The verdict is KILLED. A third, undated ("?") stratum was computed for narrative context only, not verdict-bearing per the pre-registration: it shows the opposite pattern, a 1.671x enrichment of unprovenienced non-administrative material — reported here because the resolution record itself flags it as deserving its own separate look, not because it bears on this conjecture's verdict.

Why infer this entity

This article promotes a measured reversal of a stated, literature-grounded prior, not a reconstructed entity. It is included in the promotion ladder because the registrant disclosed the prior honestly in advance (the market-selection mechanism is qualitatively established for other antiquities markets and was plausible for cuneiform specifically) and the kill-loop's discipline — register the clause, compute once, report the verdict even where it contradicts the literature-grounded expectation — surfaced a genuine, documented surprise rather than confirming what was already believed.

Evidence ledger

Two primary records, both committed to this site's own repository, were read directly: the pre-registration record for fresh-fablemax-b3-20260705-008 (docs/generated/conjecture_predictions_b3_20260705.json, row 1) and its resolution/verdict record (docs/generated/conjecture_resolutions_b3_20260705.json, row 1), which in turn point to the full per-period, per-stratum computation artifact (docs/generated/conjecture_extracts/b3_resolutions_20260705.json, key "dealers_razor"). No live model call read or summarized any external web page for this promotion.

Counterarguments

"Unprovenienced" (no recorded find location) is a proxy for market origin, not a direct label for it: some excavated tablets lack a recorded findspot for reasons unrelated to the antiquities market (incomplete excavation records, older excavations predating modern findspot recording), so the unprovenienced stratum is not a clean market-only sample. The tested universe is CDLI-catalogued tablets only; prestige pieces that moved through art-market channels without ever being catalogued by CDLI — plausibly the most market-selected material of all — are invisible to this test by construction, so the measured ratios describe the catalogued market fringe, not the market generally. CDLI's own harvest sampling regime is not established as random (see the companion entropy-floor finding's identical caveat). And the undated "?" stratum's opposite-signed pattern (1.671x enrichment) shows that provenience-by-genre relationships in this catalogue are not uniform across strata, which is a reason for caution in generalizing the Ur III/Old Babylonian result to cuneiform tablets generally.

A plausible historical reading of the reversal, offered here explicitly as interpretation rather than as part of the measured finding: cuneiform's unprovenienced stratum was substantially built by large early-20th-century lootings of administrative archive sites such as Drehem and Umma, which flooded the market with exactly the "boring" bureaucratic genre the conjecture predicted would be filtered out — market selection, on this reading, operated on AVAILABILITY (what there was a great deal of, to sell) rather than on LEGIBILITY or prestige (what buyers found most appealing). This reading is consistent with the measured ratios but is not established by them; no test of the Drehem/Umma hypothesis specifically was performed.

Confidence scores

• Direct attestation of the genre-share ratios: 90
• Existence warrant (this is a real, reportable pattern in the catalogue): 85
• Specificity: 85
• Reconstruction dependence (how much rests on interpretation beyond the two ratios): 25
• Counterevidence pressure: 30

What would change the score

A findspot-verified sample distinguishing genuinely excavated-but-unrecorded tablets from genuinely market-surfaced ones would sharpen the unprovenienced proxy and could move the ratios in either direction. A site-of-origin analysis of the unprovenienced Ur III stratum specifically — testing whether it is disproportionately traceable to known administrative-archive lootings such as Drehem and Umma — would move the Drehem/Umma reading from labelled interpretation toward a tested claim, or rule it out. And any dataset reaching prestige cuneiform material that never entered CDLI's catalogue at all would close the article's most fundamental caveat: that this test can only describe the catalogued market fringe, not the antiquities market as a whole.