ππ·ππΎπ· β βhe delivered the tinβ β the Uluburun cargo, reanalyzed in public
Around 1320 BC a ship sank off Uluburun, on the Lycian coast of Turkey, carrying about a ton of tin in 105 ingots β the largest tin assemblage of the Bronze Age Mediterranean. Where that tin was mined is the sharpest open question in the archaeology of the period's metal trade. Since 2022 two research groups have argued incompatible answers from the same isotope measurements: Central Asia and the Taurus mountains on one side; south-west Britain and exclusion-based reasoning on the other.
This project asks the question underneath that impasse: is deposit-level attribution statistically decidable from these data at all? Every model, parser, and number is public and re-runnable; every claim traces to a script.
Identifiable at the family level, not at the deposit level. On the tin-isotope channel, the 86-ingot cargo bulk excludes every negative-Ξ΄Sn source β Mushiston (Tajikistan) at ~7Ο, robustly across every stress test we could construct β but the five positive-Ξ΄Sn candidates (Cornwall, Erzgebirge, Taurus, Serbia, Portugal) are mutually unidentifiable, and informative priors pass through the model nearly unchanged: the positive apportionments published on both sides are, on this channel, restatements of their authors' priors. The exclusions are what the data themselves insist on.
Part II walks through the ceiling. Joining the ingots' trace-element chemistry to the isotope model β with censored likelihoods, measured smelting-partition priors, and the same adversarial review protocol β every candidate source anyone has chemically measured is excluded except Cornwall, and the isotope channel's deadlocked pair resolves to Cornwall over the Erzgebirge at roughly 10:1 per batch. What survives beside Cornwall is precisely what no one has measured: the posterior is now a map of the field's measurement gap. The part-II draft, figures, and review record →
Three structural gaps cap what any analysis can decide: the Taurus reference values exist only as points in published figures, no Karnab ore data exists at all, and roughly 570 of the ~857 cassiterite analyses cited in the dispute have never been deposited as data. The data page documents all three.
working draft, three adversarial review rounds absorbed β last updated 2026-08-01