Haspelmath (1997): Polarity-Side Indefinite Typology #

@cite{haspelmath-1997} @cite{haspelmath-2013} @cite{kadmon-landman-1993} @cite{ladusaw-1979} @cite{wals-2013}

Haspelmath, Martin (1997). Indefinite Pronouns. Oxford Studies in Typology and Linguistic Theory. Oxford University Press.

Polarity-side projection of @cite{haspelmath-1997}'s 9-function implicational map for indefinite pronouns. Where the sibling file Phenomena/Indefinites/Studies/Haspelmath1997.lean formalises the indefinite-typology angle (Fragment-derived IndefiniteParadigms for a 6-language sample, with WALS-bridge theorems checking the F46A classification), this file owns the polarity-side claims:

17-language sample with hand-stipulated IndefiniteParadigm instances
NPI-cluster + FC-region + neg-concord theorems
WALS Ch 46 distribution and per-language wals46A lookups
FragmentBridges decide-checks against Fragments/{Lang}/PolarityItems.lean entries (verifies that NPI items the polarity Fragments declare are licensed in the contexts the Haspelmath polarity-sensitive forms cover)

The substrate (HaspelmathFunction, IndefiniteEntry, IndefiniteParadigm, MorphologicalBasis, contiguity / coverage / disjointness predicates, wals46A and converters) lives in Typology/Indefinite.lean.

Sample #

17 typologically diverse languages:

Indo-European: English, Russian, German, Italian, Hindi-Urdu
Uralic: Finnish, Hungarian
Turkic: Turkish
Sino-Tibetan: Mandarin
Japonic / Koreanic: Japanese, Korean
Kartvelian: Georgian
Quechuan: Quechua (Imbabura)
Niger-Congo: Yoruba, Swahili
Kra-Dai: Thai
Austronesian: Tagalog

The 17 paradigms are hand-stipulated here rather than derived from Fragments/{Lang}/Indefinites.lean because the per-form IndefiniteEntry.functions field commits to a particular analysis of how forms partition the 9-function map, and the polarity-side analysis (Haspelmath 1997's contiguity-driven encoding) genuinely differs from the existing Fragment-side analysis (Degano-Aloni 2025 / Bubnov 2026's competition-driven encoding) on three of the 17 languages where Fragments already exist (English, German, Russian).

Concrete disagreement: Haspelmath polarity-view English some- covers {SK, SU} only, with any- (NPI) owning {irrealis, question, conditional, indirectNeg}; the D&A-shape Fragment's someEntry covers {SK, SU, irrealis} with no any- form. This is a real analytical disagreement, not a missing-data gap.

Audit history (see project_indefinite_substrate_contested.md memory note):

A first-pass audit framed this as a 5-framework conflict and recommended substrate evolution (split functions → attestedFunctions). That framing was wrong: re-audit verified actual writers are 2 (Fragments + this file); other consumers (D&A, Bubnov, Dekier, Chierchia, Modal Indefinites) are READ-only, parallel substrate, or no contact.
A second-pass check against the most recent canonical paper (@cite{degano-aloni-2025}, How to be (non-)specific, L&P 2025) verified that D&A 2025 explicitly works within Haspelmath's 9-function map unchanged — they do NOT split irrealis into specific/nonspecific sub-functions. So substrate-level function-inventory refinement would put linglib ahead of the field.
D&A 2025 Table 2 also explicitly allows two forms to cover the same function (Russian -то and -нибудь both NS). This invalidates universal FormsDisjoint as a constraint. The Russian paradigm here now follows D&A 2025's classification (-то = Epistemic {SU, NS}); see disjoint_languages_count + russian_not_disjoint.
Phenomena/Polarity/Studies/Chierchia2006.lean already consumes this file's italian/english/german/mandarin paradigms as substrate. Hand-stipulation here is therefore a working pattern with multiple consumers, not an isolated stipulation.

The Fragment-vs-Studies disagreement is two published analyses, lifted to theorem level in Phenomena/Indefinites/Studies/Bubnov2026.lean §11: fragment_polarity_disagree_on_kto_to proves the Russian case; fragment_polarity_disagree_on_some proves the English case. Both are decide-checked extensional inequalities on the Haspelmath function sets. The disagreement source is documented there: D&A read profiles theoretically (semantic permission); Bubnov reads them distributionally (actual coverage net of paradigmatic competition with sibling forms). Promotion of the 14 missing-language paradigms to Fragments is deferred on the same grounds: each promoted paradigm would have to pick a classification, replicating the disagreement at more sites.

Relation to Indefinites/Studies/Haspelmath1997.lean #

CLAUDE.md permits placing the same paper's formalisation under multiple phenomena when the contributions split cleanly. The split here:

Indefinites/Studies/Haspelmath1997.lean — typological coverage of indefinites, advancing claims about IndefiniteParadigm's F46A bridge
Phenomena/Polarity/Studies/Haspelmath1997.lean (this file) — polarity- side projection, advancing claims about NPI/FC clustering, neg-concord, and Fragment-PolarityItem consistency

WALSCount is imported from Linglib/Data/WALS/Aggregation.lean.