The mattang is a theory, the rebbelib is an atlas

Claim (verbatim)

Marshallese wave-piloting left a corpus of stick charts collected from the 1890s on, and its first students already saw two different objects among them: Winkler's report (Smithsonian Institution Annual Report for 1899) and Davenport's survey ('Marshall Islands Navigational Charts', Imago Mundi 15, 1960) distinguish the mattang or wappepe — an abstract trainer modelling swell refraction and interference around an idealized island — from the rebbelib and medo, which chart real stretches of the Ratak and Ralik chains with shells or lashings for named atolls; Joseph Genz's fieldwork with the last ri-meto (Breaking the Shell, University of Hawaiʻi Press, 2018) documents the pedagogy and the dilep wave-path concept the trainers taught. The claim: the tradition maintained an explicit theory layer separate from its data layer — a physics diagram and an atlas — and because idealized interference is symmetric while real geography is not, the split must be legible in museum collections as pure form, without reading a single label. Prediction: in a photographic census of stick charts in named public collections, at least 80% of charts catalogued as mattang or wappepe will show mirror symmetry of the lattice about at least one axis, against under 20% of charts catalogued as rebbelib or medo, with symmetry judged blind from photographs by raters ignorant of the catalogue class (primary clause: the 80/20 symmetry split between the two classes; the verdict follows it); secondarily, island-marking shells will sit on a majority of rebbelib and medo but be absent or generic on mattang. Kill (buildable now from real holdings): a photographic census of the stick charts held by the Smithsonian National Museum of Natural History, the Bishop Museum, the British Museum, the Übersee-Museum Bremen and the Ethnologisches Museum Berlin, classified per Winkler (Smithsonian Annual Report for 1899) and Davenport (Imago Mundi 15, 1960).

Prediction clause (verbatim)

Prediction: in a photographic census of stick charts in named public collections, at least 80% of charts catalogued as mattang or wappepe will show mirror symmetry of the lattice about at least one axis, against under 20% of charts catalogued as rebbelib or medo, with symmetry judged blind from photographs by raters ignorant of the catalogue class (primary clause: the 80/20 symmetry split between the two classes; the verdict follows it); secondarily, island-marking shells will sit on a majority of rebbelib and medo but be absent or generic on mattang.

Kill-dataset (verbatim)

Kill (buildable now from real holdings): a photographic census of the stick charts held by the Smithsonian National Museum of Natural History, the Bishop Museum, the British Museum, the Übersee-Museum Bremen and the Ethnologisches Museum Berlin, classified per Winkler (Smithsonian Annual Report for 1899) and Davenport (Imago Mundi 15, 1960).

Novelty / leakage triage

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

The taxonomy itself — mattang/wappepe as abstract single-island trainers versus rebbelib/medo as charts of the real Ratak and Ralik chains — is the published classification (Winkler; Davenport; Genz's fieldwork), and Ascher's ethnomathematical study analyses the mattang's idealized geometric structure. A blind photographic symmetry census across the museum holdings testing the 80/20 form split has not been done; note the resolution must guard against circularity, since catalogue class already partly encodes form.

• W. Davenport, 'Marshall Islands Navigational Charts', Imago Mundi 15 (1960), 19-26
• M. Ascher, 'Models and Maps from the Marshall Islands: A Case in Ethnomathematics', Historia Mathematica 22 (1995), 347-370
• Captain Winkler, 'On Sea Charts Formerly Used in the Marshall Islands', Smithsonian Institution Annual Report for 1899 (Washington, 1901)

Predictions

No prediction registered yet.