Gender Resolution in Coordination #
@cite{adamson-anagnostopoulou-2025} @cite{carstens-2026}
When singular DPs are conjoined, the resulting &P must bear phi-features for agreement. For number, sg + sg → pl by summation (@cite{corbett-2000} §6.3). For gender, the mechanism is more complex: percolation of interpretable features to &P, followed by intersection and selection.
This file formalizes the percolation-and-intersection mechanism proposed by @cite{adamson-anagnostopoulou-2025} and adopted by @cite{carstens-2026}, parameterized over the feature type so it applies across language families.
The mechanism #
Percolation: each conjunct's i(nterpretable) gender features percolate to &P; u(ninterpretable) features are excluded.
Intersection: percolated features common to all conjuncts form a single set on &P.
Result: non-empty intersection →
some features(gender-matching agreement); empty →none(language-specific default).
Cross-linguistic instantiation #
- Bantu (@cite{carstens-2026}): features are i[entity] flavors ([human], [animal], [inanimate]). u-genders (Xhosa 3/4, 5/6) have no i-features → empty intersection → default cl 2 or cl 8.
- Greek (@cite{adamson-anagnostopoulou-2025}): features are organized in a hierarchy (CLASS > MASC > FEM). FEM entails MASC, so mismatched humans yield {CLASS,MASC} → masculine (via Subset Principle). Inanimates bear only iCLASS, so mismatch yields {CLASS} → neuter (least specified exponent). No default insertion — all outcomes from intersection.
- Icelandic: MASC and FEM are independent under CLASS. Mismatched human intersection yields {CLASS} → neuter. Same vocabulary as Greek; different geometry → different result.
- BCS: MASC is under INDIV (not CLASS). All coordinatable nominals have INDIV → intersection always includes {INDIV} → masculine.
Selection grammars #
When the intersection contains multiple i-features (from stacked nPs), the language must select which one determines the agreement class. @cite{carstens-2026} §5.1 identifies two grammars available to Xhosa speakers: Highest Wins (outermost nP layer wins) and Best Semantic Match (most specific semantic core wins).
A gender feature annotated with interpretability. @cite{kramer-2015}: gender features on categorizing heads n are either interpretable (i) — natural gender — or uninterpretable (u) — arbitrary. Only i-features enter the resolution calculus.
Uses the unified Interpretability type from Features.lean.
- value : F
- interp : Interpretability
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
A feature bundle on a conjunct DP. Features are ordered from outermost (highest nP layer) to innermost (lowest/core). Single-layer DPs have singletons or empty bundles; stacked nPs have multiple entries.
Equations
Instances For
Percolation: extract only interpretable feature values. u-features are excluded from the resolution calculus (@cite{adamson-anagnostopoulou-2025}).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Intersection of percolated feature lists from two conjuncts. Features present in both lists survive; others are eliminated. Result retains the ordering from the first list.
Equations
- Minimalist.Agreement.GenderResolution.intersectFeatures xs ys = List.filter (fun (x : F) => ys.contains x) xs
Instances For
Resolve gender agreement for two conjoined DPs via the percolation-and-intersection mechanism.
- Percolate: extract i-features from each conjunct
- Intersect: find shared i-features
- Return:
some featuresif non-empty,noneif empty
This is the single compositional endpoint for gender resolution. All language-specific resolution functions are projections of this.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A singleton interpretable feature bundle always self-matches under
resolution: resolving ⟨f, i⟩ & ⟨f, i⟩ yields some [f].
This is the foundation of the A&A mechanism: each geometry node,
coordinated with itself, yields matching agreement. The proof
requires LawfulBEq to ensure f == f = true.
Uninterpretable features are excluded from resolution:
a singleton u-feature bundle always yields none.
When multiple i-features survive intersection (from stacked nPs), the language selects which one determines the agreement class. @cite{carstens-2026} §5.1 proposes two grammars available to speakers.
- highestWins : SelectionGrammar
The feature from the outermost (highest) nP layer wins. Since features retain their percolation ordering, this is the first element of the intersection.
- bestSemanticMatch : SelectionGrammar
The most specific semantic match wins. A core gender (i[human], i[inanimate], i[animal]) beats a plain i[entity] from an arbitrary gender member.
Instances For
Equations
- Minimalist.Agreement.GenderResolution.instDecidableEqSelectionGrammar 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
Select the determining feature from a non-empty intersection.
specificity maps features to a ranking (higher = more specific).
Equations
- One or more equations did not get rendered due to their size.
- Minimalist.Agreement.GenderResolution.selectFeature Minimalist.Agreement.GenderResolution.SelectionGrammar.highestWins specificity features = features.head?
Instances For
Intersecting a feature list with itself is the identity.
Every element is trivially contained in itself (via LawfulBEq).
TODO: induction on xs; each element satisfies xs.contains x
by LawfulBEq reflexivity.
General idempotency: resolving a bundle with itself always yields its
percolated i-features (when non-empty). This generalizes
singleton_self_matching to bundles of any size.
Linguistically: uniform conjuncts (same gender) always produce gender-matching agreement, regardless of bundle complexity.
Resolve gender agreement across n conjuncts by iterated intersection.
- Percolate i-features from each conjunct
- Intersect all percolated lists pairwise (left fold)
- Return
some featuresif non-empty,noneif empty
Binary resolve is the special case resolveN [fs1, fs2].
Equations
- One or more equations did not get rendered due to their size.
Instances For
N-ary resolution subsumes binary: resolveN [fs1, fs2] = resolve fs1 fs2.
A set of feature bundles satisfies MRH (Mismatch Resolution Hypothesis) if resolution succeeds for every pair of bundles — i.e., no pair produces an empty intersection, and no default insertion is ever needed.
@cite{adamson-anagnostopoulou-2025}: Greek satisfies MRH because all resolution outcomes emerge from intersection alone. Bantu does NOT satisfy MRH because uninterpretable genders produce empty intersections.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A feature geometry that defines entailment between feature nodes. @cite{adamson-anagnostopoulou-2025}: cross-linguistic variation in resolution follows from differences in feature geometry — the same labels with different entailment relations yield different outcomes.
nodes: the feature nodes in this geometrybundle: maps each node to its full feature bundle (itself + everything it entails, as interpretable features)
- nodes : List F
- bundle : F → FeatureBundle F
Instances For
Feature entailment: f₁ entails f₂ iff every i-feature in f₂'s
bundle is also present in f₁'s bundle. In Greek: FEM entails MASC
(because FEM's bundle {CLASS,MASC,FEM} ⊇ MASC's bundle {CLASS,MASC}).
Equations
- order.entails f₁ f₂ = (Minimalist.Agreement.GenderResolution.percolateI (order.bundle f₂)).all fun (x : F) => (Minimalist.Agreement.GenderResolution.percolateI (order.bundle f₁)).contains x
Instances For
A feature order satisfies MRH if all bundles from its geometry produce non-empty intersections under resolution.
Equations
- order.satisfiesMRH' = Minimalist.Agreement.GenderResolution.satisfiesMRH (List.map order.bundle order.nodes)