Categorizing Heads (Distributed Morphology) @cite{harley-2014} #
@cite{embick-2004} @cite{marantz-1997} @cite{kramer-2015}
@cite{harley-2014} "On the identity of roots" addresses three questions about roots in DM:
What are roots? (§2) Root terminal nodes are individuated by arbitrary indices, not by phonological or semantic content. The Categorization Assumption holds: roots must merge with a categorizing head (n, v, a) to enter the syntax.
Can roots take complements? (§3) Yes — roots can Merge directly with internal arguments without mediation by a functional head. Evidence: one-replacement in argument structure nominals, verb-object idioms, Hiaki suppletive verbs conditioned by the root's complement.
What delimits the domain of special interpretation? (§4) VoiceP, not the first categorizing head. Idiosyncratic interpretation can extend past the first categorizer (evidence: multiply derived words like editorial, classifieds, nationalize). Voice is the phase head.
DM Three-Lists Architecture (@cite{marantz-1997}, @cite{harley-2014} §5) #
- List 1: Root terminal nodes — syntactic atoms with opaque indices
- List 2: Vocabulary Items — phonological realizations competing for insertion
- List 3: Encyclopedia entries — interpretations conditioned by context
Phi-Features on n (@cite{kramer-2015} Ch 3) #
@cite{kramer-2015} argues that grammatical gender is a phi-feature located on the nominalizing head n, not on roots. The feature system is parameterized across languages by dimension (what binary feature is used):
| Language | Dimension | Four types of n |
|---|---|---|
| Amharic | [±FEM] | n i[+FEM], n i[−FEM], n, n u[+FEM] |
| Spanish | [±FEM] | n i[+FEM], n i[−FEM], n, n u[+FEM] |
| Maa | [±FEM] | n i[+FEM], n i[−FEM], n, n u[−FEM] |
| Algonquian | [±ANIM] | n i[+ANIM], n i[−ANIM], n, n u[+ANIM] |
(@cite{kramer-2015} Chs 3, 5-7; @cite{adamson-2024} extends this to Teop [±ANIM] and Jarawara [±MASC])
This module formalizes the categorization layer, its phi-feature content,
and its relationship to Voice. List 2 (Vocabulary Insertion) is formalized
in VocabularyInsertion.lean.
A categorizing head that merges with an acategorial root to project syntactic structure. The three options correspond to the functional heads n, v, a in Distributed Morphology (@cite{marantz-1997}, @cite{harley-2014} §2).
- n : Categorizer
- v : Categorizer
- a : Categorizer
Instances For
Equations
Equations
- Morphology.DM.instBEqCategorizer.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
Map a categorizer to its syntactic category.
Equations
Instances For
Gender feature dimension. Different languages locate different binary features on n (@cite{kramer-2015} Chs 3, 5-7):
- FEM: [±FEM] dimension (Amharic, Spanish, Maa, Dieri, Wari', Lavukaleve)
- MASC: [±MASC] dimension (Jarawara; @cite{adamson-2024})
- ANIM: [±ANIM] dimension (Algonquian, Teop, Lealao Chinantec)
- fem : GenderDimension
- masc : GenderDimension
- anim : GenderDimension
Instances For
Equations
- Morphology.DM.instBEqGenderDimension.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
Polarity of a gender feature value. The binary [±VAL] system from @cite{kramer-2015} Ch 3.
Note: polarity is about the feature value (+/−), not about markedness. In Set 1 languages, u[+FEM] is the arbitrary gender; in Set 2, u[−FEM] is. Neither polarity is inherently "marked."
Instances For
Equations
- Morphology.DM.instBEqPolarity.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
A gender feature value: a dimension (what kind of feature) combined with a polarity (positive or negative).
Examples:
⟨.fem, .pos⟩= [+FEM] (female, as in Amharic innat 'mother')⟨.fem, .neg⟩= [−FEM] (male, as in Amharic abbat 'father')⟨.anim, .pos⟩= [+ANIM] (animate, as in Teop body-part nouns)⟨.masc, .pos⟩= [+MASC] (masculine, as in Jarawara)
- dim : GenderDimension
- pol : Polarity
Instances For
Equations
Instances For
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Feature interpretability (@cite{kramer-2015} §3.4.2).
- Interpretable (natural gender): legible at LF, restricts the denotation to male/female referents. Licensed by Encyclopedia (List 3).
- Uninterpretable (arbitrary gender): invisible at LF, visible only at PF. Licensed by Vocabulary Insertion (List 2).
- i : Interpretability
- u : Interpretability
Instances For
Equations
- Morphology.DM.instBEqInterpretability.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
A gender feature annotated for interpretability.
@cite{kramer-2015} Ch 3 identifies four attested combinations on n (per dimension):
- i[+VAL]: natural gender, positive polarity (e.g. female)
- i[−VAL]: natural gender, negative polarity (e.g. male)
- u[+VAL]: arbitrary gender, positive polarity (Set 1: Amharic, Spanish)
- u[−VAL]: arbitrary gender, negative polarity (Set 2: Maa, Wari')
A fifth option is plain n with no gender feature at all (the default).
- interp : Interpretability
- val : GenderVal
Instances For
Equations
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Whether a gender feature is interpretable (natural).
Equations
Instances For
Whether a gender feature is uninterpretable (arbitrary).
Equations
- { interp := Morphology.DM.Interpretability.i, val := val }.isArbitrary = false
- { interp := Morphology.DM.Interpretability.u, val := val }.isArbitrary = true
Instances For
Number feature on the n head (@cite{kramer-2015} §3.5).
Split plurality: irregular plurals are marked on n (within the categorization domain), while regular plurals are marked on Num (outside nP). Only irregular number lives on the categorizer.
Instances For
Equations
Equations
- Morphology.DM.instBEqNumberOnN.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
Phi-features hosted on a categorizing head.
Following @cite{kramer-2015} Ch 3, the n head is the locus of gender features and (for irregular nouns) number features. The v and a heads do not host phi-features in the standard analysis.
- gender : Option GenderFeature
Instances For
Equations
Instances For
Equations
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
A categorizing head enriched with phi-features and selectional properties.
This extends the basic three-way Categorizer distinction with the
feature content that @cite{kramer-2015} argues sits on the categorizer
head. For n heads, this includes gender and (for irregular nouns) number.
For v and a heads, the phi-bundle is typically empty.
The selectsD field captures the selectional feature {D} from
@cite{adamson-2024} (following Myler 2016): when true, the n head
creates a specifier position for an iPossessor DP in Spec,nP.
- cat : Categorizer
- phi : PhiBundle
- selectsD : Bool
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
The syntactic category of a phi-enriched categorizer head.
Equations
- ch.toCategory = ch.cat.toCategory
Instances For
An iPossessable n head: has {D} (selectsD = true) by construction. Use this for any n that licenses an iPossessor in Spec,nP. The phi-bundle determines gender; selectsD is not a free parameter.
Examples:
- Teop body-part n:
.iPoss { gender := some ⟨.u, ⟨.anim, .pos⟩⟩ } - Jarawara iPossessable n:
.iPoss(no gender feature → feminine) - Inherited-gender n:
.iPoss(gender comes from iPossessor via Agree)
Equations
- Morphology.DM.CatHead.iPoss phi = { cat := Morphology.DM.Categorizer.n, phi := phi, selectsD := true }
Instances For
FEM dimension (Amharic, Spanish, Romance; @cite{kramer-2015} Chs 3, 6) #
n with interpretable [+FEM]: female natural gender. Examples: Amharic -it suffix on animate female nouns.
Equations
- One or more equations did not get rendered due to their size.
Instances For
n with interpretable [−FEM]: male natural gender. Examples: Amharic animate male nouns.
Note: iMasc is a mnemonic for the gender this n yields (masculine),
not the feature dimension. The feature is i[−FEM] — negative polarity
in the FEM dimension. For the separate MASC dimension used in
Jarawara (@cite{adamson-2024}), see n_uMasc.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Plain n: no gender feature. Default nominal categorizer. Examples: inanimate nouns with no gender marking.
Equations
Instances For
n with uninterpretable [+FEM]: feminine arbitrary gender. Examples: Amharic nouns arbitrarily assigned to feminine class (door, lip, sun, ear, eye). In Set 1 languages (@cite{kramer-2015} Chs 5-6), the u-feature has positive polarity, making feminine the arbitrary gender and masculine the default. Languages: Amharic, Spanish.
Equations
- One or more equations did not get rendered due to their size.
Instances For
n with uninterpretable [−FEM]: masculine arbitrary gender in the FEM dimension. In Set 2 languages (@cite{kramer-2015} Ch 6), the u-feature has negative polarity, making masculine the arbitrary gender and feminine the default. Languages: Maa, Wari' (@cite{kramer-2015} Chs 6-7).
Equations
- One or more equations did not get rendered due to their size.
Instances For
u[+FEM] and u[−FEM] are distinct n heads: Set 1 vs Set 2.
ANIM dimension (Teop, Algonquian, Lealao Chinantec; #
@cite{kramer-2015} Chs 5-6; @cite{adamson-2024} §3.1)
n with interpretable [+ANIM]: animate natural gender. Examples: Teop gender I nouns (article a).
Equations
- One or more equations did not get rendered due to their size.
Instances For
n with interpretable [−ANIM]: inanimate natural gender. Examples: Teop gender II nouns (article o).
Equations
- One or more equations did not get rendered due to their size.
Instances For
n with uninterpretable [+ANIM]: animate arbitrary gender. Examples: Teop body-part n when iPossessed (@cite{adamson-2024} §3.1).
Equations
- One or more equations did not get rendered due to their size.
Instances For
MASC dimension (Jarawara; @cite{adamson-2024} §3.2) #
Note: Maa uses the FEM dimension (Set 2: u[−FEM]), not the MASC
dimension. The MASC dimension is used only by Jarawara in our
current coverage (@cite{adamson-2024} §3.2).
n with uninterpretable [+MASC]: masculine arbitrary gender. In Jarawara, masculine is the marked gender; feminine is unmarked (plain n).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Verbal categorizer (no phi-features).
Equations
Instances For
Adjectival categorizer (no phi-features).
Equations
Instances For
Two types of root–n licensing condition (@cite{kramer-2015} §3.4.1).
- Semantic licensing (Encyclopedia / List 3): restricts interpretation. A root with a female natural gender referent must combine with n i[+FEM] because the Encyclopedia entry is only defined in that context.
- Arbitrary licensing (PF / List 2): restricts exponence. A root is listed in a VI rule's context as requiring [+FEM] on n, even though there is no semantic motivation.
- semantic : LicensingType
- arbitrary : LicensingType
Instances For
Equations
- Morphology.DM.instBEqLicensingType.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
A root–n licensing condition: specifies that a particular root (identified by index) is licensed to combine with an n head bearing specific features, and the type of licensing (semantic or arbitrary).
- rootIdx : RootIdx
- requiredGender : Option GenderFeature
- licensingType : LicensingType
Instances For
Whether a CatHead satisfies a licensing condition's gender requirement.
Equations
Instances For
All four Amharic n types are nominal categorizers.
The four Amharic n types are pairwise distinct.
Natural gender features are interpretable.
Arbitrary gender features are uninterpretable.
Plain n has no gender feature — it is the default/unmarked case.
Natural and arbitrary gender are mutually exclusive on any feature.
Interpretable gender is semantically licensed; uninterpretable gender is arbitrarily licensed (@cite{kramer-2015} §3.4.1).
Equations
- { interp := Morphology.DM.Interpretability.i, val := val }.licensingType = Morphology.DM.LicensingType.semantic
- { interp := Morphology.DM.Interpretability.u, val := val }.licensingType = Morphology.DM.LicensingType.arbitrary
Instances For
Natural gender → semantic licensing.
Arbitrary gender → arbitrary licensing.
Canonical encoding of gender values as natural numbers for the
Minimalism PhiFeature.gender constructor. Each dimension × polarity
pair maps to a unique Nat.
Equations
- { dim := Morphology.DM.GenderDimension.fem, pol := Morphology.DM.Polarity.pos }.toNat = 0
- { dim := Morphology.DM.GenderDimension.fem, pol := Morphology.DM.Polarity.neg }.toNat = 1
- { dim := Morphology.DM.GenderDimension.masc, pol := Morphology.DM.Polarity.pos }.toNat = 2
- { dim := Morphology.DM.GenderDimension.masc, pol := Morphology.DM.Polarity.neg }.toNat = 3
- { dim := Morphology.DM.GenderDimension.anim, pol := Morphology.DM.Polarity.pos }.toNat = 4
- { dim := Morphology.DM.GenderDimension.anim, pol := Morphology.DM.Polarity.neg }.toNat = 5
Instances For
Map a DM gender feature to a Minimalist phi-feature.
Equations
Instances For
Map a DM gender feature to a valued or unvalued grammatical feature, determined by interpretability: interpretable gender is valued (legible at LF), uninterpretable gender is unvalued (probe).
Equations
- One or more equations did not get rendered due to their size.
Instances For
Interpretable gender maps to a valued feature.
Uninterpretable gender maps to an unvalued feature.
Amharic n i[+FEM] produces a valued gender feature.
Amharic n u[+FEM] produces an unvalued gender feature (probe).
Cross-dimensional verification #
Animacy-dimension n types are distinct from FEM-dimension types.
Animacy-dimension n types are distinct from plain n.
A morphosyntactic context that can trigger impoverishment.
@cite{adamson-2024} ex. 63: [MASC] → ∅ in context of [PL] or [PARTICIPANT]. Each context is a separate impoverishment rule.
- plural : ImpoverishmentContext
- participant : ImpoverishmentContext
Instances For
Equations
- Morphology.DM.instBEqImpoverishmentContext.beq x✝ y✝ = (x✝.ctorIdx == y✝.ctorIdx)
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Impoverishment: postsyntactic deletion of morphosyntactic features.
In DM, impoverishment rules apply after syntax but before Vocabulary Insertion, deleting features from terminal nodes. This can neutralize gender distinctions in certain contexts.
@cite{adamson-2024} ex. 63: Jarawara [MASC] → ∅ in the context of [PL] or [PARTICIPANT].
- targetGender : GenderVal
The feature to be deleted.
- context : ImpoverishmentContext
The conditioning context (feature that triggers deletion).
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Apply impoverishment: if the rule matches, delete the gender feature.
Equations
- One or more equations did not get rendered due to their size.
Instances For
A root that has been merged with a categorizing head, yielding a syntactically projectable unit (@cite{harley-2014} §2).
- root : Root
The acategorial root (arity, change-type, etc.)
- categorizer : Categorizer
The categorizing head that gives it syntactic category
Instances For
Equations
Instances For
Equations
Equations
- One or more equations did not get rendered due to their size.
Instances For
The syntactic category of a categorized root, derived from its categorizer.
Equations
- cr.category = cr.categorizer.toCategory
Instances For
Same root + different categorizer → different syntactic category. This is the formal content of the claim that √HAMMER can surface as either a noun (hammer) or a verb (to hammer) — same root, different category, determined entirely by the categorizer (@cite{harley-2014} §2).
Complement selection is a root-level property, not contributed by the categorizer (@cite{harley-2014} §3). Evidence:
- one-replacement in argument structure nominals: "the proud owner of a large dog" → "the proud one" — one replaces nP including √OWN + complement, showing the root took its complement directly.
- Verb-object idioms: kick the bucket — √KICK selects the bucket directly under vP, not via mediation by v.
- Hiaki suppletive verbs: suppletive forms are conditioned by the root's complement (singular vs. plural object), showing locality between root and argument below the categorizer.
In our formalization, RootArity.selectsTheme captures this: the
root obligatorily selects an internal argument at the root level,
and this persists regardless of which categorizer it merges with.
A theme-selecting root maintains its complement requirement regardless of whether it surfaces as a noun, verb, or adjective (@cite{harley-2014} §3).
Layered derivational morphology: a root categorized by one head can be further categorized by another, yielding derived forms. For example, √SHELF + n → shelf, then + v → to shelve (denominal verb).
Harley (2014 §4) uses multiply derived words (editor-ial, class-ifi-eds, national-ize) to argue that idiosyncratic interpretation can extend past the first categorizer — the phase boundary is at Voice, not at the inner categorizer.
- denominal : Recategorization
- deadjectival : Recategorization
- deverbal_n : Recategorization
- deverbal_a : Recategorization
Instances For
Equations
- Morphology.DM.instBEqRecategorization.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
The source categorizer of a re-categorization.
Equations
- Morphology.DM.Recategorization.denominal.source = Morphology.DM.Categorizer.n
- Morphology.DM.Recategorization.deadjectival.source = Morphology.DM.Categorizer.a
- Morphology.DM.Recategorization.deverbal_n.source = Morphology.DM.Categorizer.v
- Morphology.DM.Recategorization.deverbal_a.source = Morphology.DM.Categorizer.v
Instances For
The target categorizer of a re-categorization.
Equations
- Morphology.DM.Recategorization.denominal.target = Morphology.DM.Categorizer.v
- Morphology.DM.Recategorization.deadjectival.target = Morphology.DM.Categorizer.v
- Morphology.DM.Recategorization.deverbal_n.target = Morphology.DM.Categorizer.n
- Morphology.DM.Recategorization.deverbal_a.target = Morphology.DM.Categorizer.a
Instances For
Apply a re-categorization to a categorized root. Returns none if the
root's current categorizer doesn't match the expected source.
Equations
Instances For
Denominal verbs start from n-categorized roots.
Re-categorization yields the target categorizer.
A denominal verb and a directly verbal root yield the same syntactic category (V), but have different internal structure. √HAMMER + v gives V directly; √HAMMER + n + v also gives V but via layered derivation. This structural ambiguity is invisible at the category level (@cite{harley-2014} §2).
Deadjectival derivation (a → v) connects to @cite{embick-2004}'s resultStative structure: what RootTypology calls
AdjectivalStructure.resultStative is, in DM terms, a root
first categorized by a, then further categorized by v.
Categorizers are never phase heads (@cite{harley-2014} §4).
Equations
- x✝.isPhaseHead = false
Instances For
No categorizer is a phase head (@cite{harley-2014} §4).
Agentive Voice IS a phase head — it demarcates the boundary above which interpretation must be compositional (@cite{harley-2014} §4).
The phase-boundary asymmetry: Voice can be a phase head while categorizers never are. This is why idiosyncratic interpretation extends past categorizers but not past Voice (@cite{harley-2014} §4).
Voice introduces the external argument (@cite{harley-2014} §4, following @cite{kratzer-1996}). The categorizer does NOT introduce arguments — complement selection is a root property (§3).