Corbett (1991): Gender — typology of noun-class systems #
[Cor91] [Cor13] [DH13a] [Dix72]
Greville Corbett. Gender. Cambridge University Press, 1991. Plus [Cor13]'s WALS Chs 30, 31, 32.
This study file holds the 22-language exemplar sample and Corbett's
cross-linguistic generalisations over it. The per-language records
(Profile) live here — per-language paper-specific data belongs in the
study file that consumes it, not in Fragments. The WALS chapter values
are stored explicitly per language and grounded against
Data.WALS.F30A/F31A/F32A by the *_wals_grounded theorems; the
Ch 30 bin is derived from the controller-gender count
(Profile.genderCount), so bin/count consistency holds by construction.
Deviations from earlier revisions #
- The former
GenderProfilesubstrate (fromWALSsmart constructor with silent fallbacks, stored count + bin + consistency predicates) was dissolved into this file 2026-06-10. - The former
IsCanonicalGenderpredicate was renamedIsPrototypicalSexBased: "canonical gender" is [CF16], whose Canonical Gender Principle makes purely semantic assignment canonical — the prototypical Indo-European mixed-assignment cell this predicate picks out is non-canonical in their sense. - The former Dyirbal basis-override ("Corbett 1991 treats Dyirbal as non-sex-based") was dropped as unverifiable against the monograph: ch. 2 presents Dyirbal classes I/II as organised around male/female humans, which is WALS's sex-based criterion. The non-sex-based reading is Dixon/Lakoff framing and belongs to a study anchored on them.
Sample composition #
22 languages spanning all five WALS Ch 30 values: no gender (Mandarin, Japanese, Turkish, Finnish, Korean, Quechua), 2 genders (French, Spanish, Hindi-Urdu, Irish, Hebrew, Hausa), 3 genders (German, Russian, Latin, Romanian, and — as a pronominal system — English), 4 genders (Dyirbal, Archi), 5+ noun classes (Swahili, Zulu, Fula).
Substrate: semantic bases and per-language profiles #
Semantic dimensions that can underlie gender / noun-class assignment. Editorial five-way cut over [Cor91]'s assignment discussion; consumed by [Kra20]'s Semantic Core Generalization.
- sex : SemanticBasis
- animacy : SemanticBasis
- humanness : SemanticBasis
- shape : SemanticBasis
- rationality : SemanticBasis
Instances For
Equations
- Corbett1991.instDecidableEqSemanticBasis x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
Equations
- Corbett1991.instBEqSemanticBasis.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Corbett1991.instReprSemanticBasis = { reprPrec := Corbett1991.instReprSemanticBasis.repr }
A language's gender record for this study: the controller-gender
count ([Cor91]'s ch. 6 counting — Swahili has 7 genders over
~15 traditional classes; Fula ~20), the WALS Ch 31/32 codings stored
explicitly (grounded against Data.WALS below), agreement-target
inventory, semantic bases, and the comparative-label bridge for
sex-based systems.
- name : String
- iso639 : String
- rawCount : ℕ
Controller genders ([Cor91] ch. 6): sets of nouns taking the same agreements, with singular/plural pairings counted once.
- basis : Data.WALS.F31A.GenderBasis
WALS Ch 31 value ([Cor13]).
- assignment : Data.WALS.F32A.SystemsOfGenderAssignment
WALS Ch 32 value ([Cor13]).
- agreementTargets : List Agreement.AgreementTarget
Targets where gender agreement surfaces. An inventory, not the Agreement Hierarchy itself (which is a monotonicity claim about hybrid nouns; not yet formalized).
- semanticBases : List SemanticBasis
Semantic dimensions organising the system.
- attestedGenders : List Gender
Comparative-label bridge for systems the
masculine/feminine/neuter/commonvocabulary covers;[]for noun-class systems (Bantu, Dyirbal).
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Corbett1991.instReprProfile = { reprPrec := Corbett1991.instReprProfile.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
WALS Ch 30 bin, derived from the controller-gender count. A count of
1 maps to .none: one agreement pattern for all nouns is the absence
of a system, not a one-gender system ([Kra15]'s two-class
minimum; cf. Gender.System's Nontrivial convention).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Whether the language has any gender agreement.
Equations
- p.HasAgreement = (p.agreementTargets ≠ [])
Instances For
Equations
- p.instDecidablePredHasAgreement = inferInstance
"Noun class" system: 5+ controller genders per [Cor91]'s conventional cut.
Equations
- p.IsNounClassSystem = (p.rawCount ≥ 5)
Instances For
Equations
- p.instDecidablePredIsNounClassSystem = inferInstance
Internal coherence: a no-gender language has no-gender WALS codings and no agreement targets; a gendered one has gendered codings.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Prototypical European-style gender system: sex-based, 2 or 3 genders, semantic + formal assignment.
Not [CF16]'s canonical gender: their Canonical Gender Principle makes purely semantic assignment canonical, so this mixed-assignment cell is non-canonical in their sense. The predicate picks out the prototypical Indo-European configuration.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- p.instDecidablePredIsPrototypicalSexBased = inferInstance
WALS grounding predicates #
The stored Ch 30/31/32 values agree with Data.WALS wherever WALS covers
the language; for WALS-silent languages (Fula, Irish, Latin, Romanian)
the predicates hold vacuously and the stored value is this file's
commitment.
Derived Ch 30 bin agrees with WALS F30A where covered.
Equations
- p.CountGrounded = ∀ d ∈ Data.WALS.F30A.lookupISO p.iso639, d.value = p.genderCount
Instances For
Stored Ch 31 basis agrees with WALS F31A where covered.
Equations
- p.BasisGrounded = ∀ d ∈ Data.WALS.F31A.lookupISO p.iso639, d.value = p.basis
Instances For
Stored Ch 32 assignment agrees with WALS F32A where covered.
Equations
- p.AssignmentGrounded = ∀ d ∈ Data.WALS.F32A.lookupISO p.iso639, d.value = p.assignment
Instances For
Equations
- p.instDecidablePredCountGrounded = inferInstance
Equations
- p.instDecidablePredBasisGrounded = inferInstance
Equations
- p.instDecidablePredAssignmentGrounded = inferInstance
The 22-language exemplar sample #
Public (not private): [Kra20] and [Car26] consume
individual profiles.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- Corbett1991.japanese = { name := "Japanese", iso639 := "jpn", rawCount := 0, basis := Data.WALS.F31A.GenderBasis.noGender, assignment := Data.WALS.F32A.SystemsOfGenderAssignment.noGender }
Instances For
Equations
- Corbett1991.turkish = { name := "Turkish", iso639 := "tur", rawCount := 0, basis := Data.WALS.F31A.GenderBasis.noGender, assignment := Data.WALS.F32A.SystemsOfGenderAssignment.noGender }
Instances For
Equations
- Corbett1991.finnish = { name := "Finnish", iso639 := "fin", rawCount := 0, basis := Data.WALS.F31A.GenderBasis.noGender, assignment := Data.WALS.F32A.SystemsOfGenderAssignment.noGender }
Instances For
Equations
- Corbett1991.korean = { name := "Korean", iso639 := "kor", rawCount := 0, basis := Data.WALS.F31A.GenderBasis.noGender, assignment := Data.WALS.F32A.SystemsOfGenderAssignment.noGender }
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Archi. The stored assignment is this file's [Cor91] commitment
(semantic and formal: the monograph documents phonological and
morphological correlates for the gender III/IV split
-- UNVERIFIED: locate the passage), diverging from WALS F32A's
semantic-only coding — see archi_corbett1991_vs_corbett2013.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
All 22 language profiles in the Corbett 1991 sample.
Equations
- One or more equations did not get rendered due to their size.
Instances For
WALS grounding #
Every derived Ch 30 bin agrees with WALS F30A where covered.
Every stored Ch 31 basis agrees with WALS F31A where covered.
Every stored Ch 32 assignment agrees with WALS F32A where covered — except Archi, whose 1991-vs-2013 divergence is first-class below.
Corbett 1991 vs 2013: Archi. The 1991 monograph treats assignment as semantic and formal; the 2013 WALS chapter (F32A) codes Archi as semantic-only.
Sample-level structural sanity #
All profiles are internally coherent.
All five WALS Ch 30 values are attested.
All three Ch 31 values are attested.
All three Ch 32 values are attested.
Cross-linguistic generalisations #
All 2- and 3-gender systems in the sample are sex-based. Reflects the cross-linguistic pattern that small gender systems organise around sex.
[Cor91]'s key finding: no language assigns gender on a purely formal basis without any semantic core. In the sample, every gendered language has semantic assignment (alone or combined with formal).
A gender system without any agreement is not a gender system — genders are precisely the categories that trigger agreement.
Agreement-target inventories #
[Cor79] / [Cor91] ch. 8's Agreement Hierarchy has four
positions: attributive > predicate > relative pronoun > personal pronoun.
Verbal agreement (the substrate's .verb target, tested below) is
subsumed under the predicate position in Corbett's scheme.
The theorems below are inventory observations on the sample — which
targets mark gender at all. The hierarchy's actual content (monotone
availability of semantic agreement along the positions) is formalized
for number agreement in Studies/Corbett2000.lean
(AgreementProfile.RespectsHierarchy, with British English committee
as witness); the gender-side analogue for hybrid nouns like vrač
needs per-target assignment on Gender.System.Assigned and is still
open.
Verb agreement implies higher-target agreement in the sample (none of the languages agree only on verbs).
No language in the sample agrees only on verbs.
English is the sample's pronominal gender system: it has gender (3 genders, he/she/it) but gender surfaces only on personal pronouns.
Noun-class systems (5+) show agreement on more targets than smaller systems. In the sample, all noun-class systems agree on ≥4 of the 5 target types.
Basis × count interactions #
Non-sex-based systems in the sample have ≥4 genders. When gender is not organised around sex, the system tends to proliferate.
Prototypical sex-based systems #
The European languages in the sample (and Hindi-Urdu) all instantiate
the prototypical sex-based configuration. (See
Profile.IsPrototypicalSexBased for why this is not
[CF16]'s canonical gender.)
Gender bridge #
Every 2- or 3-gender sex-based language in the sample exposes the
appropriate Gender comparative labels via attestedGenders.