@cite{konnelly-cowper-2020} #
Konnelly, Lex and Elizabeth Cowper. 2020. Gender diversity and morphosyntax: An account of singular they. Glossa: a journal of general linguistics 5(1): 40. 1--19. DOI: https://doi.org/10.5334/gjgl.1000
Core Contribution #
Variation in acceptance of singular they reflects speakers' position in a grammatical change in progress. The change has three stages, differing only in the contrastivity status of [MASC] and [FEM] on the gender features of the pronoun system. The pronoun Vocabulary Items themselves are constant across all three stages; what changes is which feature bundles nouns and names project.
| Stage | [MASC]/[FEM] status | Singular they distribution |
|---|---|---|
| 1 | Contrastive | Quantified/generic epicene only |
| 2 | Contrastive, fewer nouns | Extends to ungendered names |
| 3 | Non-contrastive modifiers | Default for all singular animate |
The key theorem: the VIs are constant across stages. At Stage 3, [MASC]/[FEM] are optional modifiers (non-contrastive in the sense of @cite{wiltschko-2008}), so their absence is the norm, and the Elsewhere Condition yields they as the default singular animate pronoun.
Connection to @cite{arnold-2026} #
Arnold's pragmatic account (underspecified vs. personal singular they) complements K&C's grammatical account. At Stage 1, underspecified they is licensed only when discourse elaboration is thin (quantified/generic antecedents). At Stage 3, they is the grammatical default for any singular animate referent regardless of discourse elaboration, because the grammar itself no longer requires a contrastive gender feature.
Formalization #
We define the pronoun VI rules using FeatureVI from
Morphology.DM.VocabularyInsertion, parameterize the three
stages via Contrastivity from Morphology.DM.Categorizer,
and prove bridge theorems connecting to Arnold's discourse conditions.
The morphosyntactic features relevant to English 3rd-person pronoun spellout (K&C §4, (11a--d)).
K&C place [MASC] and [FEM] on the φ head above Number (following @cite{kramer-2009}; @cite{kramer-2015}), [SG] on Num, and [INANIM] on n. For VI purposes, only the feature bundle at the pronoun terminal matters, not where in the tree it originates.
- sg : PronFeature
- masc : PronFeature
- fem : PronFeature
- inanim : PronFeature
Instances For
Equations
- KonnellyCowper2020.instDecidableEqPronFeature x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Equations
- KonnellyCowper2020.instBEqPronFeature.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
The four English 3rd-person pronoun VIs.
These are constant across all three stages — the change from Stage 1 to Stage 3 is in which feature bundles nouns project, not in the VI rules themselves.
(11) a. [SG] [FEM] ↔ she b. [SG] [MASC] ↔ he c. [SG, INANIM] ↔ it d. Elsewhere ↔ they
Equations
- KonnellyCowper2020.vi_she = { features := [KonnellyCowper2020.PronFeature.sg, KonnellyCowper2020.PronFeature.fem], exponent := "she" }
Instances For
Equations
- KonnellyCowper2020.vi_he = { features := [KonnellyCowper2020.PronFeature.sg, KonnellyCowper2020.PronFeature.masc], exponent := "he" }
Instances For
Equations
- KonnellyCowper2020.vi_it = { features := [KonnellyCowper2020.PronFeature.sg, KonnellyCowper2020.PronFeature.inanim], exponent := "it" }
Instances For
Equations
- KonnellyCowper2020.vi_they = { features := [], exponent := "they" }
Instances For
The complete VI rule set for English 3rd-person pronouns. Order does not matter — the Subset Principle selects by specificity (feature-list length).
Equations
Instances For
A plural bundle (no [SG]) → they (elsewhere).
A singular bundle with both [MASC] and [FEM] → the VI system is well-defined even for this impossible configuration (no nominal bears both features). FEM wins by being listed first among equal-length matches.
The three stages of grammatical change for singular they (K&C §4).
The stages differ only in the contrastivity status of [MASC] and [FEM]. The pronoun VIs (§2) are constant across stages.
- stage1 : Stage
Stage 1 (pre-modern / conservative): [MASC] and [FEM] are fully contrastive. Absence of a gender feature on a singular animate nominal is possible only when the referent's gender is unknown (quantified/generic contexts).
- stage2 : Stage
Stage 2 (innovative / current): [MASC] and [FEM] remain contrastive, but the inventory of nouns that obligatorily bear them shrinks. Ungendered proper names (Kelly, Morgan) may lack a gender feature even for referents of known gender.
- stage3 : Stage
Stage 3 (most innovative): [MASC] and [FEM] become optional, non-contrastive modifiers. Their absence carries no semantic content — they is the default for any singular animate.
Instances For
Equations
- KonnellyCowper2020.instDecidableEqStage x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- KonnellyCowper2020.instReprStage = { reprPrec := KonnellyCowper2020.instReprStage.repr }
Equations
- KonnellyCowper2020.instBEqStage.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
The contrastivity status of [MASC]/[FEM] at each stage.
Equations
Instances For
Stages 1 and 2 share contrastive gender features — they differ only in the inventory of nouns carrying them, not in the feature system itself.
The Stage 1→3 transition is a change in the feature system itself (contrastive → non-contrastive), not just in lexical inventory.
Whether a contrastive gender feature must be present on a singular animate nominal of known binary gender, at a given stage.
At Stage 1: always (contrastive features are obligatory for known gender). At Stage 2: depends on the noun (not modeled here — lexical variation). At Stage 3: never obligatory (non-contrastive = optional).
Equations
Instances For
Gender obligatoriness is DERIVED from the contrastivity status of [MASC]/[FEM], not stipulated independently. Contrastive features are obligatory (absence = ¬F); non-contrastive features are optional (absence = vacuous). This is the formal content of K&C's claim that the three stages differ only in feature system properties.
The feature bundle projected by a singular animate nominal with no gender specification — the critical case for singular they.
At all stages, this bundle is just [SG], with no [MASC], [FEM], or [INANIM]. The Elsewhere Condition yields they.
Instances For
The feature bundle projected by a singular animate feminine nominal.
Equations
Instances For
The feature bundle projected by a singular animate masculine nominal.
Equations
Instances For
The feature bundle projected by a singular inanimate nominal.
Equations
Instances For
Classification of pronoun–antecedent configurations that differ in whether [MASC]/[FEM] is obligatorily projected at a given stage.
K&C §4.1: the three stages are distinguished by which referent contexts obligatorily project gender features.
- (12): "Every student brought {his/their} lunch" — quantifier-bound, gender optional at all stages.
- (13a): "Jess₁ said he₁ couldn't make it" — referential with known gendered name, gender obligatory at Stages 1--2.
- (13b): "Jess₁ said they₁ couldn't make it" — rejected at Stage 1 for a known-gender referent.
- Stage 2 innovation: ungendered proper names (Kelly, Morgan, Pat) may lack gender features even though the referent's gender is known.
- referentialKnownGender : ReferentContext
Referential pronoun with a known-gender antecedent (e.g., "Jess said he couldn't make it").
- referentialUnknownGender : ReferentContext
Referential pronoun with a gender-unknown/irrelevant antecedent (e.g., "the student said they would be late").
- quantifierBound : ReferentContext
Quantifier-bound / generic pronoun (e.g., "every student brought their lunch").
- ungenderedProperName : ReferentContext
Referential pronoun with an ungendered proper name antecedent (e.g., "Kelly said they would be late").
Instances For
Equations
- KonnellyCowper2020.instDecidableEqReferentContext x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Equations
- KonnellyCowper2020.instBEqReferentContext.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
Whether [MASC]/[FEM] is obligatorily projected for a given referent context at a given stage.
This captures the full empirical landscape of K&C §4:
- Gender-unknown referents never project [MASC]/[FEM] (at any stage).
- Quantifier-bound pronouns optionally lack [MASC]/[FEM] at all stages.
- Known-gender referential pronouns require [MASC]/[FEM] at Stages 1--2.
- Ungendered names require [MASC]/[FEM] at Stage 1 but not Stage 2.
- At Stage 3, nothing is obligatory (non-contrastive).
Equations
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.referentialUnknownGender x✝ = false
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.quantifierBound x✝ = false
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.referentialKnownGender KonnellyCowper2020.Stage.stage1 = true
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.referentialKnownGender KonnellyCowper2020.Stage.stage2 = true
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.referentialKnownGender KonnellyCowper2020.Stage.stage3 = false
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.ungenderedProperName KonnellyCowper2020.Stage.stage1 = true
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.ungenderedProperName KonnellyCowper2020.Stage.stage2 = false
- KonnellyCowper2020.genderObligatoryFor KonnellyCowper2020.ReferentContext.ungenderedProperName KonnellyCowper2020.Stage.stage3 = false
Instances For
K&C (12): At Stage 1, quantifier-bound pronouns can be they because [MASC]/[FEM] is optional in bound variable contexts. "Every student brought their lunch" is acceptable at Stage 1.
K&C (13): At Stage 1, referential pronouns with known-gender antecedents MUST project [MASC]/[FEM] — singular they is not available for a referent of known gender. "Jess₁ said they₁ couldn't make it" is rejected at Stage 1 (for a known-gender Jess).
The Stage 2 innovation: ungendered proper names (Kelly, Morgan, Pat) may lack [MASC]/[FEM] even though the referent's gender is known. This extends singular they beyond quantifier-bound contexts.
At Stage 3, no referent context obligatorily projects [MASC]/[FEM]. Singular they is available for all singular animate referents.
At all stages, ungendered singular animates get they. The stages differ only in how many nominals project ungendered bundles, not in what VI does with such a bundle.
At all stages, feminine bundles still get she. Stage 3 speakers still have he/she available when the non-contrastive gender modifier is present.
At all stages, masculine bundles still get he.
At all stages, inanimate bundles still get it. The inanimate feature [INANIM] is not affected by the contrastivity change — it remains contrastive at all stages.
At Stage 3, non-contrastive [MASC]/[FEM] means gender features are optional. For any singular animate referent — regardless of known gender — a speaker may omit the gender modifier, yielding they. This is the core K&C prediction: at Stage 3, singular they is unrestricted for animate referents.
At Stage 1, gender features are obligatory for known-gender referents. Singular they arises only when gender is genuinely unknown.
At Stage 1, singular they is restricted to contexts where discourse elaboration is underspecified — the grammatical obligation to project [MASC]/[FEM] for known-gender referents means they arises only when gender is unknown, which correlates with thin discourse models.
At Stage 3, singular they is the grammatical default — it does not require underspecified discourse elaboration. Arnold's personal they (elaborated discourse, known they/them pronouns) is naturally accommodated: the grammar produces they regardless of discourse state.
Project a DM n-head's phi-features into K&C's flat pronoun VI feature system.
This bridges @cite{kramer-2015}'s structured features on n (GenderFeature with dimension, polarity, interpretability) to K&C's flat feature system for VI competition.
The mapping for English (FEM dimension):
- i[+FEM] on n → [FEM] (K&C's feature that yields "she")
- i[−FEM] on n → [MASC] (K&C's feature that yields "he")
- i[−ANIM] on n → [INANIM] (yields "it")
- no gender on n → ∅ (elsewhere, yields "they")
Note: K&C adopt @cite{harley-ritter-2002}'s feature geometry with independent [MASC] and [FEM], while @cite{kramer-2015} uses [±FEM] with polarity. For English pronoun VI, these are equivalent: K&C's [FEM] = Kramer's [+FEM], K&C's [MASC] = Kramer's [−FEM].
Equations
- One or more equations did not get rendered due to their size.
Instances For
The end-to-end pipeline: DM n-head → projected features → VI → "she". A singular nominal categorized by n i[+FEM] gets "she" via the Elsewhere Condition.
n i[−FEM] (masculine natural gender) → "he".
n i[−ANIM] (inanimate) → "it".
Plain n (no gender feature) → "they" (elsewhere). This is the structural correlate of K&C's core claim: they is what VI produces when the n-head lacks a gender feature.
K&C's (14): the nominal structure relevant to pronoun feature projection.
DP
├── D (determiner/quantifier)
└── NumP
├── Num [SG]
└── nP
├── n [gender, animacy]
└── NP (root)
Features are distributed across heads: [SG] on Num, [MASC]/[FEM] on the φ head (which K&C identify with the n domain following @cite{kramer-2009}; @cite{kramer-2015}), and [INANIM] on n.
The DP is the maximal projection of the nominal extended projection n (F1) → Num (F3) → D (F4), a well-formed subsequence of @cite{borer-2005}'s full nominal spine N → n → Q → Num → D.
- nHead : Morphology.DM.CatHead
The categorizing head n, bearing gender and animacy features.
- singular : Bool
Whether the Num head bears [SG] (singular number).
- isQuantifier : Bool
Whether D hosts a quantifier (relevant for bound-variable pronouns at Stage 1, where [MASC]/[FEM] are optional for quantifier-bound DPs).
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
- KonnellyCowper2020.instReprPronDP = { reprPrec := KonnellyCowper2020.instReprPronDP.repr }
Project the flat feature bundle for VI competition from the DP structure.
This composes features from their structural positions in K&C's (14):
- [SG] comes from the Num head
- Gender features ([MASC], [FEM]) come from n (via
catHeadToPronFeatures) - [INANIM] comes from n (also via
catHeadToPronFeatures)
The resulting list is what the Subset Principle operates over.
Equations
- ns.toPronBundle = (if ns.singular = true then [KonnellyCowper2020.PronFeature.sg] else []) ++ KonnellyCowper2020.catHeadToPronFeatures ns.nHead
Instances For
Spell out the pronoun for a nominal structure via VI competition.
This is the complete pipeline: DP structure → feature projection → Subset Principle → phonological exponent.
Equations
Instances For
A singular feminine nominal (n i[+FEM], Num [SG]) → "she".
A singular masculine nominal (n i[−FEM], Num [SG]) → "he".
A singular inanimate nominal (n i[−ANIM], Num [SG]) → "it".
A singular nominal with plain n (no gender, Num [SG]) → "they". The DP structure DETERMINES the elsewhere outcome — plain n projects no gender feature, so the Subset Principle selects the elsewhere VI.
A plural nominal (no [SG]) → "they" (elsewhere).
A quantifier-bound singular nominal with plain n → "they".
The isQuantifier flag records the D-head status but does not
affect VI: feature projection reads n and Num, not D content.
K&C's (14) spine [n, Num, D] is a well-formed nominal extended projection: all heads share [-V, +N] features (category consistency).
K&C's (14) spine [n, Num, D] has monotonically increasing F-values: n (F1) ≤ Num (F3) ≤ D (F4). The gap between n (F1) and Num (F3) reflects K&C's omission of the Q (classifier) layer, which sits at F2 in @cite{borer-2005}'s full nominal spine.
The F-values of K&C's (14) are strictly increasing: n (F1 = 1) < Num (F3 = 3) < D (F4 = 4).
K&C's (14) spine belongs to the nominal EP family.
At Stage 3, every referent context permits an ungendered nominal structure (plain n), which produces "they". This composes two independent results:
stage3_nothing_obligatory: no context requires gender featuressingular_ungendered_spellout: plain n → "they" via VI
The DP structure at all three stages is identical — K&C's (14). What changes across stages is which n-heads are available in which referent contexts, not the syntactic architecture itself.
The VI elsewhere exponent they corresponds to the Fragment's epicene gender paradigm. This makes explicit the connection that K&C's analysis reveals: epicene is not a positive gender class but the absence of a contrastive gender feature — the elsewhere case in VI competition.
The VI competition and the Fragment's gender paradigm are consistent across all four pronoun forms: each VI exponent maps to its expected gender paradigm class.
This is the bridge between the morphosyntactic mechanism (VI) and the descriptive grammar (GenderParadigm): VI derives which form is used, and the Fragment's paradigm classification records the result.
@cite{bjorkman-2017}'s dynamic discourse condition: a pronoun's gender features must be a subset of its antecedent's gender features. A pronoun that is "less specified" than its antecedent is acceptable; one that adds features the antecedent lacks is not.
K&C §5 argue this condition presupposes contrastive features and cannot extend to Stage 3, where features are non-contrastive.
Equations
- KonnellyCowper2020.bjorkmanSubset pronounFeatures antecedentFeatures = pronounFeatures.all fun (x : KonnellyCowper2020.PronFeature) => antecedentFeatures.contains x
Instances For
Under Bjorkman's condition, they (no gender features) can always corefer with any antecedent, because ∅ ⊆ S for any set S.
Under Bjorkman's condition, he ([MASC]) cannot corefer with an antecedent that lacks [MASC] — the pronoun adds a feature the antecedent doesn't have. Example: "Every student₁ brought his₁ lunch" — his has [MASC] but every student has no gender feature → Bjorkman predicts this is blocked, matching Stage 1 speakers' judgments.
At Stage 3, K&C predict that antecedents' gender feature bundles are typically empty (non-contrastive features are optional modifiers whose absence is the norm). Bjorkman's condition is then vacuously satisfied for ALL pronouns, providing no constraint at all.
K&C argue this is a problem: Bjorkman's system gives no mechanism for why Stage 3 speakers would still optionally use "he"/"she" (which K&C handle via non-contrastive modifiers that are present but not obligatory).
K&C §5.2: Bjorkman's condition also cannot explain the GRADUAL nature of the Stage 1→2 transition. If the change were simply learning that some names lack gender features (Bjorkman's account), it should be abrupt — once you learn "Kelly" has no [MASC], you accept they for Kelly. But K&C argue the change is gradual because it involves restructuring the feature system, not just updating lexical entries.
Formally: at Stage 1, ungendered names obligatorily project gender (so "Kelly₁ said they₁ couldn't make it" is rejected). The Stage 2 innovation is that certain names no longer project gender. This is NOT modeled by Bjorkman's condition, which only constrains pronoun selection given a fixed feature bundle — it doesn't address which bundles nominals project.
At Stage 3, animate non-human referents (pets) can also receive singular they — the ungendered bundle is available for all animate referents, not just humans.
Inanimate singulars cannot receive they even at Stage 3: the [INANIM] feature remains contrastive, and it is the most specific match for [SG, INANIM].
K&C (10): "I put my favourite watch down somewhere, and now I can't find it / *them."