Root Typology: States and Changes of State ([BEJ+21], B&[BKG20]) [BEJ+21] [BKG20] [Coo19] #
[Ara05] [Dix82] [Emb04] [Dow91] [Emb09] [RSN21] [Lev93]
Beavers, Everdell, Jerro, Kauhanen, [BEJ+21] "States and changes of state: A crosslinguistic study of the roots of verbal meaning." Language 97(3), 439–484.
Core contribution #
Change-of-state verb roots split into two types:
- Property concept (PC) roots: flat, red, long — underlie deadjectival CoS verbs
- Result roots: crack, break, shatter — underlie non-deadjectival CoS verbs
The key semantic distinction: result roots lexically entail change, while PC roots do not. This refutes the Bifurcation Thesis: contra bifurcation, some roots introduce templatic meaning (change = BECOME).
The deepest theorem #
A root's entailment of change determines ALL of its morphosyntactic behavior in a single four-way biconditional (§§3–8):
entailsChange R ↔ ¬hasSimpleStative R ∧ ¬verbalFormIsMarked R ∧ ¬allowsRestitutiveAgain R
This holds crosslinguistically (88-language typological study), ruling out bifurcation as an explanation. The deeper explanation is the Markedness Generalization (eq. 44): morphological markedness reflects semantic mismatch between functional head and root.
Bridges #
EntailmentProfile.changeOfState= result root entailment (ProtoRoles §8)TemplateHead.vBecome= the templatic operator that result roots lexicalizeCoSType.inception= BECOME as ¬P→P transition (ChangeOfState/Theory)Template.achievement/.accomplishment= templates containing BECOME (EventStructure)- Crosslinguistic data validates the correlations (
Studies/, e.g.Coon2019,BeaversEtAl2021)
Unified Root (§§15–17) #
Extends the file with the RootClassification structure (§16) bundling [Coo19]'s arity
dimension (does the root select an internal argument?) with Beavers et al.'s
change-entailment dimension. The two axes cross-classify orthogonally
(arity_changeType_orthogonal): knowing whether a root selects a theme tells
you nothing about whether it entails change, and vice versa.
- Coon, J. (2019). Building verbs in Chuj: Consequences for the nature of roots. Journal of Linguistics 55(1): 35–81.
- Hale, K. & Keyser, S.J. (2002). Prolegomenon to a Theory of Argument Structure. MIT Press.
- Harley, H. (2014). On the identity of roots. Theoretical Linguistics 40, 225–276.
Two types of change-of-state verb roots ([BEJ+21] §3.1).
Property concept (PC) roots: underlie deadjectival CoS verbs. The root describes a gradable property (dimension, color, value, etc.). Examples: flat, red, long, warm.
Result roots: underlie non-deadjectival CoS verbs. The root describes a specific result state that arises from a particular kind of event (breaking, cooking, killing, etc.). Examples: crack, break, shatter.
Instances For
Equations
- instDecidableEqRootType x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprRootType.repr RootType.propertyConcept prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "RootType.propertyConcept")).group prec✝
- instReprRootType.repr RootType.result prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "RootType.result")).group prec✝
Instances For
Equations
- instReprRootType = { reprPrec := instReprRootType.repr }
Whether a root lexically entails prior change ([BEJ+21] §3.6).
PC roots denote simple states that can hold without any prior change event. Result roots denote states that entail a prior change event.
Equations
- RootType.propertyConcept.entailsChange = false
- RootType.result.entailsChange = true
Instances For
The semantic denotation domain of a root ([Coo19], (3); extended by [HKG25]).
- indivStatePred ⟨e, ⟨v,t⟩⟩: individual/state relation (√TV, √ITV; also PC Class 1/3 roots per [HKG25]: λx_e λs_v[P(x)(s)])
- statePred ⟨v,t⟩: predicate of states, no individual argument ([HKG25] Class 2 PC roots: λs_v[P(s)], e.g., √I:YEL 'big' in Wá·šiw)
- measureFn ⟨e, ⟨s,d⟩⟩: entity → event → degree (√POS; [Hen19])
- entityPred ⟨e,t⟩: entity → truth-value, no event (√NOM)
- indivStatePred : RootDenotationType
- statePred : RootDenotationType
- measureFn : RootDenotationType
- entityPred : RootDenotationType
Instances For
Equations
- instDecidableEqRootDenotationType 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
- instReprRootDenotationType = { reprPrec := instReprRootDenotationType.repr }
Whether a root denotation type includes an individual argument.
Types with an individual argument (⟨e, ...⟩) can compose directly with v_become (which requires ⟨e, ⟨v,t⟩⟩). Types without one (⟨v,t⟩) cause a type mismatch and require type-shifting (e.g., ∇) or possessive semantics (v_have) to predicate of individuals. ([HKG25] §5.1)
Equations
Instances For
Unified root characterization bundling all classification dimensions.
A root is characterized along five independent axes:
- Arity: does it select an internal argument?
- Change entailment: does it lexically entail a prior change event?
- Denotation type ([Coo19], (3)): event predicate, measure function, or entity predicate.
- Quality dimensions ([SMcN26]): within-class root content
- Class membership: verb class taxonomy
Axes 1, 2, and 3 cross-classify: Coon's four Chuj root classes are recovered as (arity × denotationType) pairs: √TV = selectsTheme + indivStatePred, √ITV = noTheme + indivStatePred, √POS = noTheme + measureFn, √NOM = noTheme + entityPred.
- arity : Verb.Root.Arity
Does this root select an internal argument?
- changeType : RootType
Does this root lexically entail prior change?
- denotationType : Option RootDenotationType
Semantic denotation domain ([Coo19], (3)). Optional — not all roots have been annotated.
- profile : Verb.Root.Profile
Within-class quality dimensions ([SMcN26])
- levinClass : Option Semantics.Lexical.LevinClass
Verb class membership
Instances For
Equations
- One or more equations did not get rendered due to their size.
- instBEqRootClassification.beq x✝¹ x✝ = false
Instances For
Equations
- instBEqRootClassification = { beq := instBEqRootClassification.beq }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- instReprRootClassification = { reprPrec := instReprRootClassification.repr }
Does this root lexically entail prior change?
Equations
Instances For
Root change type, and its blindness to the outcome axis #
RootType is a projection of a root's entailment signature, derived not
stored — Root.changeType is the derived analog of the stored
RootClassification.changeType. Crucially it is blind to the [Bha24]
outcome axis (Root.outcomes, the orthogonal dimension Root now carries): two
roots with the same entailments share a changeType whatever their outcomes —
which is precisely why outcome cardinality is a genuinely independent dimension
of root content, the one that drives reversative un- where the manner/result
typology cannot.
The change-entailment type of a root, derived from its kind signature:
a root entails change iff its signature carries result ([BEJ+21]).
The derived analog of RootClassification.changeType.
Equations
- r.changeType = if Verb.LexKind.result ∈ r.kinds then RootType.result else RootType.propertyConcept
Instances For
changeType is blind to outcomes: same entailments ⇒ same changeType,
whatever the outcomes. The formal statement of why [Bha24]'s outcome
cardinality is an independent axis the manner/result signature cannot see.
Property concept root subclasses ([Dix82]; [BEJ+21] ex. 5).
[Dix82] identifies seven semantic categories. [BEJ+21] omits HUMAN PROPENSITY from their Table 2 sample but the category is attested crosslinguistically ([HKG25] Table A1: hungry, afraid, sick, brave, generous, etc. in Wá·šiw).
- dimension : PCClass
- age : PCClass
- value : PCClass
- color : PCClass
- physicalProperty : PCClass
- humanPropensity : PCClass
- speed : PCClass
Instances For
Equations
- instDecidableEqPCClass x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprPCClass = { reprPrec := instReprPCClass.repr }
Equations
- instReprPCClass.repr PCClass.dimension prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.dimension")).group prec✝
- instReprPCClass.repr PCClass.age prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.age")).group prec✝
- instReprPCClass.repr PCClass.value prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.value")).group prec✝
- instReprPCClass.repr PCClass.color prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.color")).group prec✝
- instReprPCClass.repr PCClass.physicalProperty prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.physicalProperty")).group prec✝
- instReprPCClass.repr PCClass.humanPropensity prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.humanPropensity")).group prec✝
- instReprPCClass.repr PCClass.speed prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PCClass.speed")).group prec✝
Instances For
Result root subclasses ([Lev93]; [BEJ+21] ex. 6).
- entitySpecificCoS : ResultClass
- cooking : ResultClass
- breaking : ResultClass
- bending : ResultClass
- killing : ResultClass
- destroying : ResultClass
- calibratableCoS : ResultClass
- inherentlyDirectedMotion : ResultClass
Instances For
Equations
- instDecidableEqResultClass x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprResultClass = { reprPrec := instReprResultClass.repr }
Equations
- One or more equations did not get rendered due to their size.
- instReprResultClass.repr ResultClass.entitySpecificCoS prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.entitySpecificCoS")).group prec✝
- instReprResultClass.repr ResultClass.cooking prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.cooking")).group prec✝
- instReprResultClass.repr ResultClass.breaking prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.breaking")).group prec✝
- instReprResultClass.repr ResultClass.bending prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.bending")).group prec✝
- instReprResultClass.repr ResultClass.killing prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.killing")).group prec✝
- instReprResultClass.repr ResultClass.destroying prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.destroying")).group prec✝
- instReprResultClass.repr ResultClass.calibratableCoS prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ResultClass.calibratableCoS")).group prec✝
Instances For
How a result root's change entailment is restricted (B&[BKG20] §2.4).
Break-type result roots (√CRACK, √SHATTER) entail change of ANY kind — spatial or temporal. A crack can "run from the tree to the house" without a temporal becoming event; the state itself extends spatially.
Cook/kill-type result roots (√COOK, √KILL) entail only TEMPORAL change. Cooking, killing, and melting are necessarily temporal processes. "✱The meat cooked from the oven to the table" is ruled out.
This three-way refinement (PC / break-type / cook-type) is invisible to
the binary entailsChange flag but has consequences for spatial predication
and the interpretation of directional PPs (B&[BKG20] §2.4).
- anyChange : ChangeRestriction
- temporalOnly : ChangeRestriction
Instances For
Equations
- instDecidableEqChangeRestriction x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprChangeRestriction = { reprPrec := instReprChangeRestriction.repr }
Equations
- One or more equations did not get rendered due to their size.
- instReprChangeRestriction.repr ChangeRestriction.anyChange prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "ChangeRestriction.anyChange")).group prec✝
Instances For
Change restriction for each result root subclass. Breaking/bending roots and directed motion allow spatial change descriptions; cooking/killing/destroying/scalar-change roots are temporally restricted.
Equations
- ResultClass.breaking.changeRestriction = ChangeRestriction.anyChange
- ResultClass.bending.changeRestriction = ChangeRestriction.anyChange
- ResultClass.inherentlyDirectedMotion.changeRestriction = ChangeRestriction.anyChange
- ResultClass.entitySpecificCoS.changeRestriction = ChangeRestriction.temporalOnly
- ResultClass.cooking.changeRestriction = ChangeRestriction.temporalOnly
- ResultClass.killing.changeRestriction = ChangeRestriction.temporalOnly
- ResultClass.destroying.changeRestriction = ChangeRestriction.temporalOnly
- ResultClass.calibratableCoS.changeRestriction = ChangeRestriction.temporalOnly
Instances For
Breaking roots allow spatial change: "The road cracked from the tree to the house."
Cooking roots restrict to temporal change: "✱The meat cooked from A to B."
PC roots have simple (unmarked) stative forms; result roots lack them.
Crosslinguistic evidence (§6, Fig. 1): PC median = 95.67% of languages have simple statives; result median = 1.59% (Mann-Whitney U = 1266.5, p < 0.001, n₁ = n₂ = 36).
English: "bright" (PC, simple adj) vs *"shattered" requires prior change.
Equations
Instances For
PC root verbs TEND to be morphologically marked (deadjectival: wid-en, flat-ten); result root verbs tend to be unmarked (break, crack, shatter).
This captures the cross-linguistic DEFAULT pattern, not a universal. [HKG25] §4: Wá·šiw Class 1 PC roots predicate as bare verbs without verbal morphology (zero categorization), deviating from the tendency captured here.
Crosslinguistic evidence (§7, Fig. 5): PC median = 56.01% marked; result median = 15.20% (U = 1291, p < 0.001).
Equations
Instances For
PC roots allow restitutive 'again' (scope over root only); result roots allow only repetitive 'again' (scope over BECOME).
§3.4: "John sharpened the knife again" allows restitutive reading (could be just one sharpening), but "#Chris thawed the meat again" in a restitutive context is unacceptable (necessarily two defrostings).
Under the analysis in §3.6: 'again' can target √ROOT. For PC roots, this yields a restitutive reading (return to prior state without prior change). For result roots, since the root itself entails change, 'again' over the root still entails a prior change event.
Equations
Instances For
The main theorem of [BEJ+21].
A root's entailment of change determines ALL of its morphosyntactic behavior in a single four-way biconditional. This is the paper's deepest result: four independently testable properties form a perfect correlation package.
For result roots (entailsChange = true):
- No simple stative forms (§6)
- Unmarked verbal forms (§7)
- No restitutive 'again' — only repetitive (§3.4)
For PC roots (entailsChange = false):
- Simple stative forms exist (§6)
- Marked verbal forms (§7)
- Restitutive 'again' available (§3.4)
This correlation holds crosslinguistically (88 languages, §§4–7) and refutes the Bifurcation Thesis: if roots couldn't introduce templatic meaning (change), there would be NO semantic basis for the morphological and syntactic correlations.
The converse: NOT entailing change determines the opposite package.
The Bifurcation Thesis for Roots ([Emb09]:1, [Ara05]:79; [BEJ+21] eq. 2):
"If a component of meaning is introduced by a semantic rule that applies to elements in combination [i.e. by templatic operators], then that component of meaning cannot be part of the meaning of a [lexical semantic] root."
Under bifurcation, change (= BECOME) is introduced only by templates, never by roots. Therefore NO root should entail change.
Equations
- bifurcationThesis rootEntailsChange = ∀ (rt : RootType), rootEntailsChange rt = false
Instances For
[BEJ+21] main result: bifurcation does not hold. Result roots entail change, violating the thesis (§§3.3, 3.6, 9).
Corollary: result roots are a witness to bifurcation failure.
PC roots are consistent with bifurcation (they don't entail change).
B&[BKG20] strengthened bifurcation failure via Root.Kinds.
[BEJ+21] show roots can entail CHANGE (one templatic notion). B&[BKG20] show roots can entail CHANGE, CAUSATION, and MANNER — ALL notions traditionally reserved for templates. This is a strictly stronger refutation: even if one accepted that change is "special", roots encoding manner+cause (√GUILLOTINE, √HAND) violate bifurcation on three separate dimensions simultaneously.
Witness: Root.Kinds.fullSpec carries all four kinds.
Multiple Levin classes witness the stronger bifurcation failure.
Whether a form is morphologically marked (=derived/complex) or unmarked (=basic/simple).
- unmarked : Markedness
- marked : Markedness
Instances For
Equations
- instDecidableEqMarkedness x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprMarkedness = { reprPrec := instReprMarkedness.repr }
Equations
- instReprMarkedness.repr Markedness.unmarked prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Markedness.unmarked")).group prec✝
- instReprMarkedness.repr Markedness.marked prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "Markedness.marked")).group prec✝
Instances For
The Markedness Generalization ([BEJ+21] eq. 44).
Morphological markedness reflects semantic mismatch between a functional head and its root complement. A form is unmarked when the head's semantic contribution is REDUNDANT with the root's meaning:
(a) Default realization of v_become with complement √ROOT: - If √ROOT entails change → v_become is redundant → UNMARKED verb - If √ROOT does not entail change → v_become adds content → MARKED verb
(b) Default realization of Asp_{S/R} with complement X: - If X does not entail change → already stative → UNMARKED stative - If X entails change → change must be stripped → MARKED stative
This explains three attested language types:
- English-type: markedness asymmetry realized overtly (-en, -ed)
- Equipollent: both marked (high-marking languages like Hebrew)
- Labile: neither marked (low-marking languages like Kinyarwanda)
And rules out the unattested fourth type (where the markedness pattern is the inverse of what the generalization predicts).
Equations
- verbalMarkedness rt = if rt.entailsChange = true then Markedness.unmarked else Markedness.marked
Instances For
Stative markedness is the mirror image of verbal markedness.
Equations
- stativeMarkedness rt = if rt.entailsChange = true then Markedness.marked else Markedness.unmarked
Instances For
Verbal and stative markedness are always complementary.
Result roots produce unmarked verbs.
PC roots produce marked verbs.
Result roots produce marked statives.
PC roots produce unmarked statives.
The markedness generalization is equivalent to the semantic distinction.
What a ditransitive root entails about possession (B&[BKG20] §3.3). Templates always introduce PROSPECTIVE possession (via ◇ modality). Whether the ROOT adds actual or prospective possession on top determines cancellability: "gave X the ball, #but X never had it" vs. "sent X the ball, but it never arrived" (OK).
- none : PossessionEntailment
- prospective : PossessionEntailment
- actual : PossessionEntailment
Instances For
Equations
- instDecidableEqPossessionEntailment x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprPossessionEntailment = { reprPrec := instReprPossessionEntailment.repr }
Equations
- One or more equations did not get rendered due to their size.
- instReprPossessionEntailment.repr PossessionEntailment.none prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "PossessionEntailment.none")).group prec✝
Instances For
Six classes of ditransitive verb roots (B&[BKG20] §3.6). The ditransitive parallel to the PC/result distinction for CoS roots: templates contribute only PROSPECTIVE states; roots can contribute ACTUAL states.
- causedPossession : DitransitiveRootClass
- futureHaving : DitransitiveRootClass
- ballisticMotion : DitransitiveRootClass
- sending : DitransitiveRootClass
- accompaniedMotion : DitransitiveRootClass
- carrying : DitransitiveRootClass
Instances For
Equations
- instDecidableEqDitransitiveRootClass 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
- instReprDitransitiveRootClass = { reprPrec := instReprDitransitiveRootClass.repr }
What a ditransitive root entails about the transfer event. Each field is a distinct semantic entailment from the ROOT, independent of what the template provides.
- possession : PossessionEntailment
- causedMotion : Bool
- manner : Bool
- accompaniment : Bool
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Equations
- instReprDitransitiveEntailments = { reprPrec := instReprDitransitiveEntailments.repr }
Equations
- One or more equations did not get rendered due to their size.
Instances For
Entailment profile for each ditransitive root class (B&[BKG20] §3.6).
Equations
- DitransitiveRootClass.causedPossession.entailments = { possession := PossessionEntailment.actual, causedMotion := false, manner := false, accompaniment := false }
- DitransitiveRootClass.futureHaving.entailments = { possession := PossessionEntailment.prospective, causedMotion := false, manner := false, accompaniment := false }
- DitransitiveRootClass.ballisticMotion.entailments = { possession := PossessionEntailment.none, causedMotion := true, manner := true, accompaniment := false }
- DitransitiveRootClass.sending.entailments = { possession := PossessionEntailment.none, causedMotion := true, manner := false, accompaniment := false }
- DitransitiveRootClass.accompaniedMotion.entailments = { possession := PossessionEntailment.none, causedMotion := true, manner := false, accompaniment := true }
- DitransitiveRootClass.carrying.entailments = { possession := PossessionEntailment.none, causedMotion := true, manner := true, accompaniment := true }
Instances For
√GIVE entails actual possession — "Kim gave Sandy the ball, #but Sandy never had it" is contradictory (§3.3, ex. 17).
√SEND does NOT entail possession — "Kim sent Sandy the ball, but it never arrived" is felicitous (§3.3, ex. 18).
√PROMISE entails only prospective possession.
√THROW entails manner + caused motion but not possession.
√BRING entails accompaniment — agent travels with theme.
√CARRY entails manner + accompaniment but not possession.
Bridge to LevinClass: ditransitive Levin classes → root classes.
Equations
- Semantics.Lexical.LevinClass.give.ditransitiveRootClass = some DitransitiveRootClass.causedPossession
- Semantics.Lexical.LevinClass.contribute.ditransitiveRootClass = some DitransitiveRootClass.causedPossession
- Semantics.Lexical.LevinClass.exchange.ditransitiveRootClass = some DitransitiveRootClass.causedPossession
- Semantics.Lexical.LevinClass.throw.ditransitiveRootClass = some DitransitiveRootClass.ballisticMotion
- Semantics.Lexical.LevinClass.send.ditransitiveRootClass = some DitransitiveRootClass.sending
- Semantics.Lexical.LevinClass.carry.ditransitiveRootClass = some DitransitiveRootClass.carrying
- x✝.ditransitiveRootClass = none
Instances For
The six MRC diagnostics developed in B&KG (2020 §§4.2–4.3). Three test for result entailment in the root; three test for manner. An MRC violation is a verb that passes diagnostics from BOTH sets.
Result diagnostics (§4.2):
denialOfResult: "X cut Y, #but Y wasn't separated" — contradictoryobjectDeletion: result verbs resist unspecified object deletionrestrictedResultatives: result verbs block adding new result XPs
Manner diagnostics (§4.3):
selectionalRestriction: subject restricted to agents capable of the mannerdenialOfAction: "X cut Y, #but X didn't do anything" — contradictoryactorParaphrase: paraphrasable as "X did manner to Y" (§4.3)
- denialOfResult : MRCDiagnostic
- objectDeletion : MRCDiagnostic
- restrictedResultatives : MRCDiagnostic
- selectionalRestriction : MRCDiagnostic
- denialOfAction : MRCDiagnostic
- actorParaphrase : MRCDiagnostic
Instances For
Equations
- instDecidableEqMRCDiagnostic 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.
- instReprMRCDiagnostic.repr MRCDiagnostic.denialOfResult prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "MRCDiagnostic.denialOfResult")).group prec✝
- instReprMRCDiagnostic.repr MRCDiagnostic.objectDeletion prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "MRCDiagnostic.objectDeletion")).group prec✝
- instReprMRCDiagnostic.repr MRCDiagnostic.denialOfAction prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "MRCDiagnostic.denialOfAction")).group prec✝
- instReprMRCDiagnostic.repr MRCDiagnostic.actorParaphrase prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "MRCDiagnostic.actorParaphrase")).group prec✝
Instances For
Equations
- instReprMRCDiagnostic = { reprPrec := instReprMRCDiagnostic.repr }
Whether a diagnostic tests for result entailment.
Equations
- MRCDiagnostic.denialOfResult.TestsResult = True
- MRCDiagnostic.objectDeletion.TestsResult = True
- MRCDiagnostic.restrictedResultatives.TestsResult = True
- x✝.TestsResult = False
Instances For
Equations
- One or more equations did not get rendered due to their size.
Whether a diagnostic tests for manner entailment.
Equations
- MRCDiagnostic.selectionalRestriction.TestsManner = True
- MRCDiagnostic.denialOfAction.TestsManner = True
- MRCDiagnostic.actorParaphrase.TestsManner = True
- x✝.TestsManner = False
Instances For
Equations
- One or more equations did not get rendered due to their size.
Each diagnostic tests for exactly one of manner or result.
A verb's diagnostic profile: which of the six diagnostics it passes. B&KG (2020 §§4.2–4.3) survey these for each verb class.
- denialOfResult : Bool
- objectDeletion : Bool
- restrictedResultatives : Bool
- selectionalRestriction : Bool
- denialOfAction : Bool
- actorParaphrase : Bool
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
- instReprMRCDiagnosticProfile = { reprPrec := instReprMRCDiagnosticProfile.repr }
Whether the profile shows result entailment (passes ≥1 result diagnostic).
Equations
- p.showsResult = (p.denialOfResult || p.objectDeletion || p.restrictedResultatives)
Instances For
Whether the profile shows manner entailment (passes ≥1 manner diagnostic).
Equations
- p.showsManner = (p.selectionalRestriction || p.denialOfAction || p.actorParaphrase)
Instances For
An MRC violation is a verb whose diagnostics show BOTH manner and result.
Equations
- p.showsMRCViolation = (p.showsManner && p.showsResult)
Instances For
Cut passes all 6 diagnostics (B&KG §4.4): manner of cutting + separation. "Kim cut the bread, #but the bread wasn't separated" (result) "Kim cut the bread, #but Kim didn't do anything" (manner)
Equations
- cutDiagnostics = { denialOfResult := true, objectDeletion := true, restrictedResultatives := true, selectionalRestriction := true, denialOfAction := true, actorParaphrase := true }
Instances For
Break passes result diagnostics only: pure result (√CRACK), no manner.
Equations
- breakDiagnostics = { denialOfResult := true, objectDeletion := true, restrictedResultatives := true, selectionalRestriction := false, denialOfAction := false, actorParaphrase := false }
Instances For
Hit passes manner diagnostics only: pure manner (impact), no result.
Equations
- hitDiagnostics = { denialOfResult := false, objectDeletion := false, restrictedResultatives := false, selectionalRestriction := true, denialOfAction := true, actorParaphrase := true }
Instances For
Drown passes all 6 (B&KG §4.5): manner of drowning + death result.
Equations
- drownDiagnostics = { denialOfResult := true, objectDeletion := true, restrictedResultatives := true, selectionalRestriction := true, denialOfAction := true, actorParaphrase := true }
Instances For
Cut is MRC-violating by BOTH diagnostics AND kind signature.
Break is MRC-respecting by BOTH diagnostics AND kind signature.
Hit is MRC-respecting by BOTH diagnostics AND kind signature.
Dowty's P-Patient entailment (a) "undergoes change of state" is
precisely the result root entailment. An object bearing a result
root's state predicate has changeOfState = true.
This bridges the root typology to the entailment profile via the shared property.
Equations
- rootTypeFromChangeEntailment p = if p.changeOfState = true then RootType.result else RootType.propertyConcept
Instances For
A result verb's object (accomplishment template) has changeOfState = true, so it patterns with result roots. Contact-verb objects (kick: CA+St, no entailed change per [Bea11] eq. (60c)) fall on the other side of the bridge.
Die subject undergoes change → result-type pattern.
Result roots MUST combine with a template containing BECOME (achievement or accomplishment), because the root's change entailment is semantically redundant with BECOME.
PC roots CAN combine with any template — with BECOME (achievement/ accomplishment) they get change compositionally; without it (state) they denote simple states.
Equations
- RootType.result.requiresBECOME = true
- RootType.propertyConcept.requiresBECOME = false
Instances For
Result roots always get templates with BECOME.
Achievement and accomplishment templates contain BECOME.
Equations
Instances For
The templates result roots combine with always have BECOME.
State template lacks BECOME — only available to PC roots.
Bridge: templates with BECOME map to achievement/accomplishment Vendler classes, both of which are telic (bounded by the result state). This connects the template operator to the existing aspectual profile.
Aspectual profile for root types in their stative use.
Equations
Instances For
Result root verbs are typically achievements or accomplishments (they have BECOME in their template). PC root verbs in their inchoative (change-of-state) use are also achievements/accomplishments, but their stative use is a state (derived from profile).
Equations
Instances For
PC roots in stative use are states; result roots pattern as achievements even in adjectival use (the state entails prior change).
[Emb04] posits two adjectival structures:
(8a) BASIC STATIVES: [AspP AspS √ROOT] Simple adjective directly combining with stativizer. Only available to PC roots.
(8b) RESULT STATIVES: [AspP AspR [vP DP v_become √ROOT]] Deverbal adjective containing v_become. Available to result roots; superficially also to PC roots.
Under the non-bifurcated analysis, result root adjectives are ALWAYS (8b) because the root requires v_become. PC root adjectives are (8a) by default but can also be (8b).
- basicStative : AdjectivalStructure
- resultStative : AdjectivalStructure
Instances For
Equations
- instDecidableEqAdjectivalStructure 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
- instReprAdjectivalStructure = { reprPrec := instReprAdjectivalStructure.repr }
PC roots admit both structures; result roots only admit resultStative.
Equations
Instances For
This is equivalent to NOT entailing change.
Reading-count tabulation of sublexical again by root type
([BKG20] §1.3.2 (25), §2.4 (46)–(47)). again attaches
either low (to the root, restitutive) or high (over vbecome, repetitive):
- RESTITUTIVE: presupposes only a prior state — available iff the root is change-free (PC roots, BKG (46));
- REPETITIVE: presupposes a prior change — always available.
A result root's state itself entails change, so its low/restitutive attachment
still presupposes change and collapses into the repetitive reading (BKG (47),
"result roots never admit truly restitutive readings"). The mechanism — the
presuppositional operator and the reading hierarchy — is the canonical
Verb.CosModel.again (see result_restitution_entails_change); this enum just
records how many distinct readings each RootType admits.
- restitutive : AgainReading
- repetitive : AgainReading
Instances For
Equations
- instDecidableEqAgainReading x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprAgainReading.repr AgainReading.restitutive prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "AgainReading.restitutive")).group prec✝
- instReprAgainReading.repr AgainReading.repetitive prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "AgainReading.repetitive")).group prec✝
Instances For
Equations
- instReprAgainReading = { reprPrec := instReprAgainReading.repr }
Which readings of 'again' are available for each root type.
Equations
Instances For
PC roots have strictly more 'again' readings than result roots.
Result roots lack the restitutive reading.
PC roots have the restitutive reading.
Whether a ditransitive verb is obligatorily telic in the IO frame (B&[BKG20] §3.7).
Telicity correlates with whether the root spells out a state in the template. If the root entails actual possession, the template's prospective possession is discharged → definite endpoint → telic.
"Kim gave Sandy balls for an hour" — # (bounded by possession transfer) "Kim sent Sandy balls for an hour" — OK (endpoint only prospective)
Equations
- DitransitiveRootClass.causedPossession.obligatorilyTelic = true
- x✝.obligatorilyTelic = false
Instances For
Telicity aligns exactly with actual possession entailment.
Predicted again readings for each ditransitive root class.
Equations
Instances For
Give has one reading (collapse); send has two (distinct).
Beavers et al.'s conclusion (§9): accepting that result roots entail change does NOT blunt the predictive power of event structures. It simply means the theory of possible root meanings is richer than bifurcation allows. A verb that entails change should have v_become-type grammatical behavior (argument structure, morphology) even if the change comes from the ROOT rather than the template.
Formally: if a root entails change, then the verb should be associated with a template containing BECOME.
And conversely: if a root does NOT require BECOME, it doesn't entail change (it can stand alone as a simple stative).
The full correlation package.
Starting from the single Boolean entailsChange, we can derive
ALL of the paper's morphosyntactic predictions. This is the formal
content of the paper's main contribution: one semantic property
(whether the root lexically entails change) is the sole determinant
of six independently observable properties.
Predicted verbal markedness from change entailment.
Equations
Instances For
Predicted stative markedness from change entailment.
Equations
Instances For
√BREAK: selects theme + entails change (result root, Levin 45.1). "Break X" — the root obligatorily takes a patient that undergoes breaking, and the root lexically entails a prior change event.
Equations
- One or more equations did not get rendered due to their size.
Instances For
√HIT: selects theme + does not entail change (Levin 18.1).
"Hit X" — the root takes a contactee, but hitting does not entail
that the patient undergoes a change of state ([Lev93] pp. 5–8).
.propertyConcept is used broadly here: the formal content
(entailsChange = false) is what matters, not the label.
Equations
- One or more equations did not get rendered due to their size.
Instances For
√DIE: no theme + entails change. "Die" — intransitive; the dying entity is introduced by functional structure (unaccusative vGO/vBE), not selected by the root. Dying lexically entails a prior change event (becoming dead).
Equations
- RootClassification.die = { arity := Verb.Root.Arity.noTheme, changeType := RootType.result, denotationType := some RootDenotationType.indivStatePred }
Instances For
√SIT: no theme + does not entail change (positional root). "Sit" — Coon's √POS class: denotes a spatial configuration state. No internal argument, no entailed change.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Orthogonality of arity and change entailment.
All four cells of the 2×2 cross-classification are inhabited. This proves the two dimensions are genuinely independent: knowing that a root selects a theme tells you nothing about whether it entails change, and vice versa.
Change entailment does not determine arity (and vice versa).
Beavers et al.'s grand unification (§14) shows that entailsChange
determines ALL morphosyntactic correlates (markedness, simple stative,
again readings, template requirements). But it determines NOTHING
about internal argument selection. Coon's arity adds an independent
dimension of prediction: whether the root will surface with an
internal argument across voice alternations.
Theme persistence ([Coo19] main empirical claim).
If a root selects a theme, the internal argument persists regardless of what v/Voice⁰ head combines with it. In Chuj, √TV roots surface with an internal argument in transitive (Ø), passive (-ch, -j), and antipassive (-w) constructions alike.
This is expressed by design: arity is a field of RootClassification, not of
the derived verb. No functional head modifies it.
Change entailment determines markedness in the unified Root.
Roots with the same change type have identical morphosyntactic behavior regardless of arity — markedness, stative forms, and again readings are orthogonal to internal argument selection.
Full root specification: entailment features + structural position. This is [BKG20]'s ch. 5 root typology (display (12)) in full — the 4 binary entailment features × 2 positions give 32 theoretical cells, of which 7 are attested and the rest are principled gaps.
- entailments : Verb.Root.Kinds
- position : Verb.Root.Position
Instances For
Equations
- One or more equations did not get rendered due to their size.
Instances For
Adjoined position requires +manner: a root in adjunct position must specify a manner of action. Without manner, there is nothing to adjoin — the adjunct slot expects an action modifier.
Equations
Instances For
+manner +state −result −cause is semantically incoherent (B&KG §5.4.1): the root would specify both a manner of action and a state, but with no result or cause linking them. What would such a verb mean? "Perform manner M while in state S" — with no causal connection.
Equations
- s.SemanticallyCoherent = ¬(Verb.LexKind.manner ∈ s.entailments ∧ Verb.LexKind.state ∈ s.entailments ∧ Verb.LexKind.result ∉ s.entailments ∧ Verb.LexKind.cause ∉ s.entailments)
Instances For
Full well-formedness: entailment constraints + position licensing + semantic coherence.
Equations
- s.WellFormed = (s.entailments.WellFormed ∧ s.PositionLicensed ∧ s.SemanticallyCoherent)
Instances For
Equations
Attested cells of Table 12 #
√FLAT: +S −M −R −C, complement. Property concept root.
Equations
- FullRootSpec.flat = { entailments := Verb.Root.Kinds.propertyConcept, position := Verb.Root.Position.complement }
Instances For
√BLOSSOM: +S −M +R −C, complement. Pure result root.
Equations
- FullRootSpec.blossom = { entailments := Verb.Root.Kinds.pureResult, position := Verb.Root.Position.complement }
Instances For
√CRACK: +S −M +R +C, complement. Causative result root.
Equations
- FullRootSpec.crack = { entailments := Verb.Root.Kinds.causativeResult, position := Verb.Root.Position.complement }
Instances For
√JOG: −S +M −R −C, adjoined. Pure manner root.
Equations
- FullRootSpec.jog = { entailments := Verb.Root.Kinds.pureManner, position := Verb.Root.Position.adjoined }
Instances For
√DROWN: +S +M +R +C, complement. Manner+result in complement position — the manner restricts HOW the state is caused.
Equations
- FullRootSpec.drown = { entailments := Verb.Root.Kinds.fullSpec, position := Verb.Root.Position.complement }
Instances For
√TOSS: +S +M +R +C, adjoined. Manner+result in adjunct position — the manner is the primary event that happens to cause a state change.
Equations
- FullRootSpec.toss = { entailments := Verb.Root.Kinds.fullSpec, position := Verb.Root.Position.adjoined }
Instances For
√HAND: same entailments + position as √TOSS. The difference is in the ditransitive layer (DitransitiveRootClass.causedPossession vs .ballisticMotion), not in FullRootSpec's 4+1 features.
Equations
Instances For
√EXIST: −S −M −R −C, complement. Minimal stative root.
Equations
- FullRootSpec.exist = { entailments := Verb.Root.Kinds.minimal, position := Verb.Root.Position.complement }
Instances For
√DROWN and √TOSS have identical entailments but different positions.
Templatic functional heads in event structure (B&[BKG20] Table 13).
Each root type PREDICTS which templatic heads its verb will entail. If the root's own meaning already includes what a template head provides, that head is "entailed by the root" — its semantic contribution is redundant (though structurally still present).
v_act, v_cause, v_become are verbal heads. P_loc and P_have are prepositional heads specific to ditransitive structures.
- vAct : TemplateHead
- vCause : TemplateHead
- vBecome : TemplateHead
- pLoc : TemplateHead
- pHave : TemplateHead
Instances For
Equations
- instDecidableEqTemplateHead x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯
Equations
- instReprTemplateHead = { reprPrec := instReprTemplateHead.repr }
Equations
- instReprTemplateHead.repr TemplateHead.vAct prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "TemplateHead.vAct")).group prec✝
- instReprTemplateHead.repr TemplateHead.vCause prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "TemplateHead.vCause")).group prec✝
- instReprTemplateHead.repr TemplateHead.vBecome prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "TemplateHead.vBecome")).group prec✝
- instReprTemplateHead.repr TemplateHead.pLoc prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "TemplateHead.pLoc")).group prec✝
- instReprTemplateHead.repr TemplateHead.pHave prec✝ = Repr.addAppParen (Std.Format.nest (if prec✝ ≥ 1024 then 1 else 2) (Std.Format.text "TemplateHead.pHave")).group prec✝
Instances For
Which template heads a root's entailments make redundant (Table 13).
The mapping is monotone: more root entailments → more heads entailed.
- +result → v_become (root entails the change v_become would provide)
- +cause → v_cause (root entails the causation v_cause would provide)
- +manner ∧ +cause → v_act (manner that causes = activity)
- +manner alone → no v_act (manner without causation doesn't entail activity — √JOG specifies jogging manner but v_act still provides the activity frame)
Equations
- One or more equations did not get rendered due to their size.
Instances For
For ditransitive roots, additional prepositional heads beyond
the verbal heads predicted by entailedHeads.
Equations
- DitransitiveRootClass.causedPossession.additionalHeads = [TemplateHead.pLoc, TemplateHead.pHave]
- DitransitiveRootClass.ballisticMotion.additionalHeads = [TemplateHead.pLoc]
- DitransitiveRootClass.sending.additionalHeads = [TemplateHead.pLoc]
- DitransitiveRootClass.accompaniedMotion.additionalHeads = [TemplateHead.pLoc]
- DitransitiveRootClass.carrying.additionalHeads = [TemplateHead.pLoc]
- DitransitiveRootClass.futureHaving.additionalHeads = []
Instances For
Table 13 verification #
√JOG (pureManner): no template heads entailed. The root specifies jogging manner, but v_act still provides the activity frame — the root doesn't make it redundant.
√FLAT (propertyConcept): no template heads entailed. The root names a state, but doesn't entail change or cause.
√BLOSSOM (pureResult): v_become entailed. The root entails change — v_become's contribution is redundant.
√CRACK (causativeResult): v_become + v_cause entailed. The root entails change AND causation.
√DROWN (fullSpec): v_become + v_cause + v_act entailed. The root entails change, causation, AND activity (manner that causes).
√TOSS (fullSpec + ballistic): v_become + v_cause + v_act + P_loc. Verbal heads from entailments + P_loc from ditransitive class.
√HAND (fullSpec + causedPossession): all 5 heads. Verbal heads from entailments + P_loc + P_have from ditransitive class.
Monotonicity: more root entailments → weakly more heads entailed. Pure result ⊂ causative result ⊂ full spec (by inclusion).
The attested cells of B&KG's Table 12: five complement types (√FLAT, √BLOSSOM, √CRACK, √DROWN, √EXIST) and two adjoined types (√JOG, √TOSS/√HAND).
Equations
Instances For
Whether a FullRootSpec cell is attested in B&KG's Table 12.
Equations
- s.IsAttestedCell = (s.WellFormed ∧ s ∈ FullRootSpec.attestedCells)
Instances For
Equations
Gap explanations #
B&KG (§5.4.1) identify three principled gap types:
Adjoined without manner: the adjunct position in event structure hosts manner modifiers. Without +manner, there's nothing to adjoin. This rules out all −manner roots in adjoined position.
+manner +state −result −cause: the root would need to encode both a manner of action and a state, with no causal/change connection between them. No known verb has this pattern.
Well-formedness violations: +result −state and +cause −result are ruled out by the entailment constraints (result→state, cause→result).
Gap type 1: adjoined position requires manner.
Gap type 2: +manner +state −result −cause is incoherent.
Gap type 3: well-formedness violations.
Attested cells are well-formed and recognized #
The open question: +S +M +R −C (mannerResult without cause) in complement position. B&KG note this cell may be inhabited by verbs like slide — manner of motion + change of location, without external causation. Left as NOT attested per Table 12, pending further research.
The complement/adjoined split for fullSpec roots is the only case where position differentiates otherwise identical entailments.