Stassen (2013): WALS chapter 121 on comparative constructions #
WALS Chapter 121 by Leon Stassen covering comparative-construction typology
across 167 languages: the five-way classification
(locational | exceed | conjoined | particle | mixed) and its areal
distribution.
This study file holds cross-linguistic generalisations that consume the
Fragment-side def comparison : ComparativeProfile data with non-trivial
semantic content (particle_implies_svo_in_sample,
conjoined_no_degree_marking, morph_comp_implies_morph_super, etc.).
Per-language Fragment-vs-WALS data-equality theorems are deliberately absent
— see feedback_no_per_lang_wals_grounding_in_studies for the rationale.
WALS aggregate distribution theorems live in the substrate
(Linglib/Typology/Comparison.lean). Stassen's 1985 fine-grained 6-way
typology and the chaining-based universals live in
Studies/Stassen1985.lean.
The 17-language sample drawn from per-language Fragment Comparison files.
Sample shrunk from the dissolved file's 18 (dropped Martuthunira: no
Fragments/Martuthunira/ directory exists).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Sample comparative-type distribution.
All five comparative types are represented in the sample.
Sample degree-word distribution.
Degree-word counts sum to sample size.
Sample superlative-strategy distribution.
All particle languages in the sample (English, German, Russian, French) are SVO (or V2). This reflects [Has01]'s identification of the comparative particle as a Standard Average European feature.
Equations
Instances For
Exceed-comparative languages in the sample include Yoruba (W Africa), Mandarin (E Asia), Swahili (E Africa), Thai (SE Asia), and Tagalog (Austronesian).
Equations
Instances For
Equations
Instances For
Conjoined-comparative languages universally lack overt degree marking in the sample: comparison without any morphological apparatus.
Conjoined-comparative languages also lack dedicated superlative strategies — if you can't grammaticalize "more than", you typically can't grammaticalize "most" either.
SOV languages in the sample (Japanese, Korean, Turkish, Hindi-Urdu, Navajo, Latin) tend toward locational comparatives.
Equations
- Stassen2013Comparison.sovLanguages = List.filter (fun (x : Comparative.ComparativeProfile) => x.isSOV) Stassen2013Comparison.allLanguages
Instances For
Among SOV languages, locational is the dominant single type.
Equations
- Stassen2013Comparison.svoLanguages = List.filter (fun (x : Comparative.ComparativeProfile) => x.isSVO) Stassen2013Comparison.allLanguages
Instances For
Among SVO languages, exceed and particle types are both attested.
Exceed-comparative languages in the sample mostly lack bound comparative morphology on adjectives. Comparison is expressed via the verb.
Languages with morphological comparative degree marking ('-er'/'-ee') also have morphological superlatives in the sample.
Every locational comparative in the sample uses a standard marker that also has spatial/ablative meaning ('from', ablative case, partitive case). This is definitional but worth verifying: the standard marker is never semantically empty.
Equations
Instances For
Particle comparatives in the sample all have overt degree marking (free word or comparative morphology).
Exceed comparatives show a split on degree words (Mandarin/Tagalog have them, Yoruba/Swahili/Thai do not).
Conjoined-comparative languages (which lack a dedicated comparative construction in Stassen's terms) also lack dedicated superlative strategies. Implicational universal: SUPERLATIVE → COMPARATIVE (contrapositive).