Standard Average European: A Linguistic Area #

[Has01]

Formalization of [Has01]'s argument that the core European languages — the Romance, Germanic, Balkan, and Balto-Slavic families plus the westernmost Finno-Ugric languages — form a Sprachbund (linguistic area) called Standard Average European (SAE), defined by a dozen shared structural features that are absent in the geographically and genealogically adjacent languages.

The file first sets out a framework-neutral areal schema (§0: Isogloss, ArealReference, ArealProfile, IsArealAt, LinguisticArea) and then instantiates it: SAE is a LinguisticArea SAELanguage SAEFeature whose feature set is the twelve major Europeanisms of §2 of the paper, whose reference frame is the four samples §1 demands (area, cofamilial, adjacent, world), and whose areality is verified feature-by-feature against [Has01]'s Maps 107.1–107.12.

The cluster-map gradience of §4 (most notably the Charlemagne nucleus formed by French and German) is recovered automatically from the discrete feature data via LinguisticArea.clusterScore and nucleus.

Architectural notes #

Languages and features are local inductives: the paper surveys a specific sample and committing to its boundaries is appropriate here. A SAELanguage.toWALS bridge maps each language to its WALS code where one exists.
The paper's Maps 107.1–107.12 are the primary source for every isogloss. Each paper-based isogloss (e.g. articleLgs, vplusNILgs) is the explicit set Haspelmath plots on the corresponding map. For the six features with a directly comparable WALS chapter (37A, 38A, 47A, 101A, 107A, 115A, 121A), a sibling wals*Lgs set is derived from Data.WALS.F*.allData via walsClassifies. The two are intentionally not unioned: where they disagree, that disagreement is a fact about Haspelmath's classification vs. WALS's encoding and should remain visible to readers and to bridge theorems.
Isoglosses are Finset SAELanguage (the schema's Isogloss is a transparent abbreviation for Finset L). All filter predicates are decidable, so the four areality criteria reduce to decide against the computable rational Isogloss.density.

A schema for typological linguistic areas (Sprachbünde) — groups of geographically contiguous, often genealogically heterogeneous languages that share structural features attributable to contact rather than inheritance. The schema follows the methodology of [Has01] but is gradient-first, in the spirit of his §4 cluster maps: the primary data per feature is an ArealProfile (a 4-tuple of densities — in-area, in-cofamilial, in-adjacent, in-world). Binary judgments are derived from profiles by comparison against an ArealThresholds parameter, defaulting to the qualitative 1/2 cut.

Design notes. Isogloss is Finset L, not Set L: typology works over finite samples, and Finset gives computable density without Classical. Reference samples carry Finset.Nonempty proofs: [Has01] evaluates his criteria against actual evidence — there is no "vacuous absence" in the paper — so requiring non-emptiness eliminates the degenerate 0/0 = 0 density and lets the rational and Nat formulations of the criteria agree on all in-use cases (cf. Isogloss.density_gt_iff / density_lt_iff). The threshold layer is parameterized rather than hard-coded so that stricter or paper-specific cutoffs (Heine-Kuteva, Grambank conventions) can be instantiated without rewriting the predicates; a future gradient treatment can replace Isogloss = Finset L with L → ℚ strength functions and reuse ArealProfile and ArealThresholds unchanged.