Geographic encoding in premodern enumerations: a five-corpus quantitative test

Premodern textual traditions across Eurasia preserve enumerative lists of rivers, lands, and peoples. The qualitative observation that such lists sometimes "trace a geographic sweep" has been made informally since the nineteenth century, but has rarely been tested with quantitative method. We test the hypothesis that five canonical enumerations encode monotonic geographic order in the sequence of items they name: the ten-river enumeration of Ṛgveda 10.75.5 (Nadīstuti-sūkta), the sixteen Mahājanapadas of the Anguttara Nikāya, the sixteen lands of Vendīdād Fargard 1, the twenty-three subject lands (dahyāva) of Darius I's Bīsutūn inscription, and the Table of Nations in Genesis 10. For each corpus we compute Kendall's τ between sequence position and the modern longitude (and latitude) of the identified items, using a permutation null distribution, a Procrustes superimposition test, a Mantel distance-matrix test, and a likelihood-ratio comparison, under a Mallows model, of geographic against uniform-random generative models. The two Indic corpora (RV 10.75.5 and the Mahājanapadas) show strong monotonic ordering along the region's east-to-west axis (τ = -0.911 and -0.783, both p < 10⁻⁴). The Bīsutūn dahyāva is moderately ordered, with marked pairwise spatial clustering (τ = +0.47, p = 0.0014; Mantel r = +0.50, p < 10⁻⁴). Vendīdād Fargard 1 shows no robust geographic structure under the main tests. Genesis 10 shows only a weak monotonic signal but significant pairwise spatial coherence (Mantel r = 0.17, p = 0.0004): two of its three Noachic sub-sections are significantly latitudinally ordered, in opposite directions, so that they cancel at the surface. A suggestive within-tradition contrast is Bīsutūn against Vendīdād: both Iranian and naming much the same regions, though not matched in date, yet the administrative inscription is geographically ordered and the cosmological text is not. Across the five corpora, genre appears to track geographic ordering more closely than language family does; with n = 5 this is a suggested pattern, not an established finding. We make no claims about the date, composition, or cultural identity of any of the five traditions; the tests bear only on the texts as transmitted.