Tupaia kept the bearings and let the distances go

Claim (verbatim)

In 1769 Tupaia, the Raʻiātean navigator-priest aboard Cook's Endeavour, produced a chart of the Pacific naming some 74 islands, surviving in the Banks copy (British Library Add MS 21593.C); the associated island lists run longer still. Two modern reconstructions — Di Piazza and Pearthree (Journal of the Polynesian Society 116, 2007) and the full study by Eckstein and Schwarz ('The Making of Tupaia's Map', Journal of Pacific History 54, 2019), which supplies an island-by-island concordance and reads the chart's bearing frame from its avatea (noon) mark — agree it is no scaled map but a flattened sailing-directory: star-compass courses put to paper in a one-off translation between the Pacific's two navigational cultures. If so, the artifact's information is radically anisotropic: bearing is data; radial distance is page layout. The claim turns that asymmetry into a joint quantitative test on the reconstructions' own concordance, where either half failing kills it. Prediction: for the islands securely identified in the Eckstein–Schwarz concordance, chart azimuth measured from that reading's reference island will fall within one 32-point compass house (±22.5°) of the true initial great-circle bearing for at least 60% of islands, while the rank correlation between chart radial distance and true distance over the same set stays below 0.3 (primary clause: the joint 60-percent-within-one-house and sub-0.3 distance-correlation asymmetry; the verdict follows it). Exact computation: digitize island positions from the British Library scan, take identifications and reference islands from the 2019 concordance, and compare against bearings and distances from modern coordinates. Kill: the British Library digitization of Add MS 21593.C with the identification and reference-island tables in Eckstein and Schwarz (Journal of Pacific History 54, 2019), against Di Piazza and Pearthree (Journal of the Polynesian Society 116, 2007) as the rival plotting model.

Prediction clause (verbatim)

Prediction: for the islands securely identified in the Eckstein–Schwarz concordance, chart azimuth measured from that reading's reference island will fall within one 32-point compass house (±22.5°) of the true initial great-circle bearing for at least 60% of islands, while the rank correlation between chart radial distance and true distance over the same set stays below 0.3 (primary clause: the joint 60-percent-within-one-house and sub-0.3 distance-correlation asymmetry; the verdict follows it). Exact computation: digitize island positions from the British Library scan, take identifications and reference islands from the 2019 concordance, and compare against bearings and distances from modern coordinates.

Kill-dataset (verbatim)

Kill: the British Library digitization of Add MS 21593.C with the identification and reference-island tables in Eckstein and Schwarz (Journal of Pacific History 54, 2019), against Di Piazza and Pearthree (Journal of the Polynesian Society 116, 2007) as the rival plotting model.

Novelty / leakage triage

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

Both kill-source reconstructions argue precisely this asymmetry: Eckstein and Schwarz read the chart as bearing-encoding via the avatea frame (with bearing accuracy quantified in places and contested in the subsequent Journal of Pacific History exchange), and Di Piazza and Pearthree treat it as plotted sailing directions in which radial distance is layout. The joint statistical test as specified — 60% of identified islands within one compass house plus a sub-0.3 distance rank-correlation over the full 2019 concordance — has not been run.

• L. Eckstein and A. Schwarz, 'The Making of Tupaia's Map', Journal of Pacific History 54.1 (2019)
• A. Di Piazza and E. Pearthree, 'A New Reading of Tupaia's Chart', Journal of the Polynesian Society 116.3 (2007)

Predictions

No prediction registered yet.