@cite{krejci-2012} — Lexical Reflexivity and the Ingestive/Middle Class

Krejci, Bonnie. 2012. The Event Structural Properties of the Transitive Alternation: A Cross-Linguistic Study. Master's Report, University of Texas at Austin.

Core claims

1. Lexically reflexive verbs (eat, wash, dress, learn) have bieventive, causative event structure in their simple forms: [[ACT(x)] CAUSE [BECOME ⟨STATE⟩ (x, y)]] with causer = causee. All four bieventivity diagnostics confirm this (§4.3–4.4).

2. Two subtypes: middles (wash, dress) where the agent acts on their own body, and ingestives (eat, drink) where the agent acts on a consumed substance directed at themselves. Both are lexically reflexive. learn is proposed as a "metaphorical ingestive" (§4.2.4).

3. Four diagnostics — again ambiguity, re- prefixation, almost ambiguity, and negation over CAUSE — all detect bieventive structure in the simple forms of these verbs. All diagnostic data is for English.

4. Antireflexivization: the lexical causatives of these verbs (eat→feed, learn→teach) are derived by splitting the coidentified causer-causee argument into two distinct participants (§4.2).

5. The causativizability hierarchy — unaccusatives > middles/ingestives

   unergatives > simple transitives — is validated across 12 languages (Table 2.8). This data is formalized in MorphologicalCausation.krejciLanguages.

Bridges

• reflexive_is_bieventive → IntransitivizationType (central claim)
• eat_is_causativeResult → LevinClass.rootEntailments (root typology)
• eat_licenses_accomplishment → RootLicensesTemplate (ArgDerivation)
• eat_argTemplate_is_consumption → LevinClass.argTemplate (LevinClassProfiles)
• accomplishment_has_variant → Template.intransitiveVariant
• anticausative_no_theta / middle_no_theta → VoiceHead (Minimalist syntax)

inductive Krejci2012.LexReflexiveSubtype : Type

Subtype of lexically reflexive verb (§4.1–4.2).

• middle : LexReflexiveSubtype

  Middles: agent acts on own body (wash, dress). @cite{kemmer-1994}: one internally complex participant.

• ingestive : LexReflexiveSubtype

  Ingestives: agent causes substance to enter self (eat, drink). Includes "metaphorical ingestives" (learn).

@[implicit_reducible]

instance Krejci2012.instDecidableEqLexReflexiveSubtype : DecidableEq LexReflexiveSubtype

Equations

• Krejci2012.instDecidableEqLexReflexiveSubtype x✝ y✝ = if h : x✝.ctorIdx = y✝.ctorIdx then isTrue ⋯ else isFalse ⋯

@[implicit_reducible]

instance Krejci2012.instReprLexReflexiveSubtype : Repr LexReflexiveSubtype

Equations

• Krejci2012.instReprLexReflexiveSubtype = { reprPrec := Krejci2012.instReprLexReflexiveSubtype.repr }

def Krejci2012.instReprLexReflexiveSubtype.repr : LexReflexiveSubtype → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

structure Krejci2012.LexReflexiveVerb : Type

A lexically reflexive verb: a verb whose simple form has bieventive causative event structure with coidentified causer and causee.

The four bieventivity diagnostics (§4.3–4.4) test for complex internal structure — the presence of distinct sub-events connected by CAUSE that scopal modifiers can target independently:

• again: restitutive (result state restored) + repetitive (whole event repeated)
• re-: restitutive reading with re- prefix
• almost: scope over action sub-event vs. result sub-event
• negation over CAUSE: deny simple form while asserting causative variant

• gloss : String
• subtype : LexReflexiveSubtype
• lexicalCausative : Option String

  Lexical causative counterpart (antireflexivized form). none when the verb uses the same form transitively (wash, dress).

• againAmbiguity : Bool

  Does postverbal again produce restitutive + repetitive readings?

• rePrefixation : Bool

  Does re- prefixation produce a restitutive reading?

• almostAmbiguity : Bool

  Does almost produce scope ambiguity over sub-events?

• negationOverCause : Bool

  Can the simple form be negated while asserting the causative variant?

def Krejci2012.instReprLexReflexiveVerb.repr : LexReflexiveVerb → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance Krejci2012.instReprLexReflexiveVerb : Repr LexReflexiveVerb

Equations

• Krejci2012.instReprLexReflexiveVerb = { reprPrec := Krejci2012.instReprLexReflexiveVerb.repr }

@[implicit_reducible]

instance Krejci2012.instBEqLexReflexiveVerb : BEq LexReflexiveVerb

Equations

• Krejci2012.instBEqLexReflexiveVerb = { beq := Krejci2012.instBEqLexReflexiveVerb.beq }

def Krejci2012.instBEqLexReflexiveVerb.beq : LexReflexiveVerb → LexReflexiveVerb → Bool

Equations

• One or more equations did not get rendered due to their size.
• Krejci2012.instBEqLexReflexiveVerb.beq x✝¹ x✝ = false

All diagnostic data is from English. @cite{krejci-2012} tests four verbs — eat, wash, dress, learn — each representing a subtype of the lexically reflexive class. All four pass all four diagnostics.

def Krejci2012.eat : LexReflexiveVerb

eat: ingestive. [[ACT⟨manipulate food⟩(x)] CAUSE [BECOME ⟨potentially digest⟩ (x, y)]]. Lexical causative: feed (§4.2.1).

Equations

• One or more equations did not get rendered due to their size.

def Krejci2012.wash : LexReflexiveVerb

wash: middle. [[ACT⟨manipulate water⟩(x)] CAUSE [BECOME ⟨washed⟩ (x)]]. Transitive wash (someone) is the antireflexivized form (§4.2.2).

Equations

• One or more equations did not get rendered due to their size.

def Krejci2012.dress : LexReflexiveVerb

dress: middle. [[ACT⟨manipulate clothes⟩(x)] CAUSE [BECOME ⟨dressed⟩ (x)]]. Transitive dress (someone) is the antireflexivized form (§4.2.3).

Equations

• One or more equations did not get rendered due to their size.

def Krejci2012.learn : LexReflexiveVerb

learn: proposed metaphorical ingestive. [[ACT(x)] CAUSE [BECOME ⟨know⟩ (x, y)]]. Lexical causative: teach (§4.2.4).

Equations

• One or more equations did not get rendered due to their size.

def Krejci2012.allVerbs : List LexReflexiveVerb

Equations

• Krejci2012.allVerbs = [Krejci2012.eat, Krejci2012.wash, Krejci2012.dress, Krejci2012.learn]

theorem Krejci2012.all_again : (allVerbs.all fun (x : LexReflexiveVerb) => x.againAmbiguity) = true

All four verbs pass the again ambiguity diagnostic.

theorem Krejci2012.all_re : (allVerbs.all fun (x : LexReflexiveVerb) => x.rePrefixation) = true

All four verbs pass the re- prefixation diagnostic.

theorem Krejci2012.all_almost : (allVerbs.all fun (x : LexReflexiveVerb) => x.almostAmbiguity) = true

All four verbs pass the almost ambiguity diagnostic.

theorem Krejci2012.all_negation : (allVerbs.all fun (x : LexReflexiveVerb) => x.negationOverCause) = true

All four verbs pass the negation-over-CAUSE diagnostic.

theorem Krejci2012.all_bieventive : (allVerbs.all fun (v : LexReflexiveVerb) => v.againAmbiguity && v.rePrefixation && v.almostAmbiguity && v.negationOverCause) = true

All four diagnostics pass for every verb in the dataset: the simple forms of eat, wash, dress, and learn are bieventive.

theorem Krejci2012.both_subtypes : (allVerbs.any fun (x : LexReflexiveVerb) => x.subtype == LexReflexiveSubtype.middle) = true ∧ (allVerbs.any fun (x : LexReflexiveVerb) => x.subtype == LexReflexiveSubtype.ingestive) = true

Middles and ingestives are both represented.

@cite{krejci-2012}'s central operation: the lexical causative of a lexically reflexive verb is derived by antireflexivization — splitting the coidentified causer-causee into two distinct participants.

Reflexive form (eat):
  `[[ACT⟨manipulate food⟩(x)] CAUSE [BECOME ⟨potentially digest⟩ (x, y)]]`
  — one participant (x) is both causer and causee.

Antireflexive form (feed):
  `[[ACT⟨manipulate food⟩(x)] CAUSE [BECOME ⟨potentially digest⟩ (y, z)]]`
  — causer (x) and causee (y) are distinct participants.

Key evidence (§4.2.1): with *eat*, the eater must agentively
manipulate food; with *feed*, this entailment shifts to the feeder.
With *make eat* (syntactic causative), the eater retains the
manipulation entailment. This shows *feed* ≠ *make eat*: the lexical
and syntactic causatives have different event structures. 

structure Krejci2012.AntireflexivePair : Type

An antireflexive pair: a lexically reflexive verb and its suppletive lexical causative. Only verbs with a distinct lexical causative form have such pairs (eat→feed, learn→teach). Middles (wash, dress) use the same form transitively — the antireflexivization is structural, not morphological.

• reflexive : String
• causative : String

def Krejci2012.instReprAntireflexivePair.repr : AntireflexivePair → ℕ → Std.Format

Equations

• One or more equations did not get rendered due to their size.

@[implicit_reducible]

instance Krejci2012.instReprAntireflexivePair : Repr AntireflexivePair

Equations

• Krejci2012.instReprAntireflexivePair = { reprPrec := Krejci2012.instReprAntireflexivePair.repr }

@[implicit_reducible]

instance Krejci2012.instBEqAntireflexivePair : BEq AntireflexivePair

Equations

• Krejci2012.instBEqAntireflexivePair = { beq := Krejci2012.instBEqAntireflexivePair.beq }

def Krejci2012.instBEqAntireflexivePair.beq : AntireflexivePair → AntireflexivePair → Bool

Equations

• Krejci2012.instBEqAntireflexivePair.beq { reflexive := a, causative := a_1 } { reflexive := b, causative := b_1 } = (a == b && a_1 == b_1)
• Krejci2012.instBEqAntireflexivePair.beq x✝¹ x✝ = false

def Krejci2012.eatFeed : AntireflexivePair

Equations

• Krejci2012.eatFeed = { reflexive := "eat", causative := "feed" }

def Krejci2012.learnTeach : AntireflexivePair

Equations

• Krejci2012.learnTeach = { reflexive := "learn", causative := "teach" }

theorem Krejci2012.eat_causative : eat.lexicalCausative = some "feed"

The suppletive pairs are derivable from the verb data.

theorem Krejci2012.learn_causative : learn.lexicalCausative = some "teach"

theorem Krejci2012.wash_same_form : wash.lexicalCausative = none

Middles use the same form transitively (no suppletive pair).

theorem Krejci2012.dress_same_form : dress.lexicalCausative = none

@cite{krejci-2012}'s claim that eat and dress have causative event structure aligns with their RootEntailments classification: both are causativeResult (state + result + cause), meaning the root itself entails external causation.

theorem Krejci2012.eat_is_causativeResult : Semantics.Lexical.LevinClass.eat.rootEntailments = Semantics.Lexical.Roots.RootEntailments.causativeResult

eat roots are causativeResult: the root entails caused consumption.

theorem Krejci2012.dress_is_causativeResult : Semantics.Lexical.LevinClass.dress.rootEntailments = Semantics.Lexical.Roots.RootEntailments.causativeResult

dress roots are causativeResult: the root entails caused dressed state.

theorem Krejci2012.causativeResult_entails_cause : Semantics.Lexical.Roots.RootEntailments.causativeResult.cause = true

causativeResult roots entail cause — consistent with @cite{krejci-2012}'s analysis that these verbs have CAUSE in their simple forms.

The root→template pipeline predicts that eat and dress roots license the accomplishment template (since they entail causation). The accomplishment template in turn has an intransitive variant (achievement). But eat/dress do NOT undergo the standard causative/inchoative alternation (*"The food ate", *"The clothes dressed") — the root's obligatory agentive entailments block it.

Instead, the causativization operation is antireflexivization:
splitting a coidentified participant, not adding a new external
cause. 

theorem Krejci2012.eat_licenses_accomplishment : Semantics.ArgumentStructure.ArgDerivation.RootLicensesTemplate Semantics.Lexical.LevinClass.eat.rootEntailments Features.EventStructure.Template.accomplishment

eat roots license the accomplishment template.

theorem Krejci2012.eat_primary_accomplishment : Semantics.ArgumentStructure.ArgDerivation.primaryTemplate Semantics.Lexical.LevinClass.eat.rootEntailments = some Features.EventStructure.Template.accomplishment

eat's primary template is accomplishment.

theorem Krejci2012.eat_argTemplate_is_consumption : Semantics.Lexical.LevinClass.eat.argTemplate = some Features.LevinClassProfiles.consumption

eat's ArgTemplate is consumption (agent + incremental theme).

theorem Krejci2012.accomplishment_has_variant : Features.EventStructure.Template.accomplishment.intransitiveVariant = some Features.EventStructure.Template.achievement

The accomplishment template has an intransitive variant (achievement). This is the template-level possibility that eat/dress could alternate — but root semantics blocks it.

theorem Krejci2012.eat_root_alternation_possible : (Semantics.ArgumentStructure.ArgDerivation.licensedTemplates Semantics.Lexical.LevinClass.eat.rootEntailments).length = 3

causativeResult roots derive 3 ArgTemplates (state, achievement, accomplishment). The template infrastructure predicts alternation; the blocking is root-level, not template-level.

theorem Krejci2012.alternation_preserves_result : Features.EventStructure.Template.achievement.HasResultState

The intransitive variant retains the result state (the BECOME sub-event persists).

theorem Krejci2012.alternation_loses_cause : ¬Features.EventStructure.Template.achievement.HasCause

The intransitive variant loses CAUSE.

@cite{krejci-2012}'s central theoretical claim: lexically reflexive verbs have reflexive intransitivization (coidentification of causer and causee), not anticausative intransitivization (removal of the external cause). This is what makes their simple forms bieventive — the causer position is retained, filled by the same participant as the causee.

The connection: all four verb-level diagnostics (§4.3–4.5) detect
bieventive structure, and `IntransitivizationType.reflexive` is the
type-level characterization of exactly that structure. 

theorem Krejci2012.reflexive_is_bieventive : Semantics.Causation.Morphological.IntransitivizationType.reflexive.isBieventive = true

Reflexive intransitivization is bieventive — matching what the four verb-level diagnostics detect.

theorem Krejci2012.anticausative_is_monoeventive : Semantics.Causation.Morphological.IntransitivizationType.anticausative.isBieventive = false

Anticausative intransitivization is monoeventive — the diagnostics would NOT detect bieventive structure for true anticausatives.

theorem Krejci2012.reflexive_has_coidentification : Semantics.Causation.Morphological.IntransitivizationType.reflexive.hasCoidentification = true

Reflexive intransitivization involves coidentification of causer and causee — the structural basis of lexical reflexivity.

theorem Krejci2012.reflexive_licenses_bySelf : Semantics.Causation.Morphological.IntransitivizationType.reflexive.licensesBySelf = true

Reflexive intransitives license "by itself" (§4.5, (114a–d)), because a causer position exists (even if coidentified).

theorem Krejci2012.anticausative_blocks_bySelf : Semantics.Causation.Morphological.IntransitivizationType.anticausative.licensesBySelf = false

True anticausatives do NOT license "by itself" — no causer position to negate with "without outside help".

@cite{krejci-2012}'s reflexive/anticausative distinction maps onto the Voice typology in Minimalism: reflexive intransitives correspond to middle Voice (bieventive, coidentification), and true anticausatives correspond to anticausative Voice (monoeventive, cause removed). Both lack an external argument in Spec,VoiceP.

theorem Krejci2012.anticausative_no_theta : ¬Minimalist.voiceAnticausative.AssignsTheta

Anticausative Voice does not assign a θ-role (no external argument).

theorem Krejci2012.middle_no_theta : ¬Minimalist.voiceMiddle.AssignsTheta

Middle Voice does not assign a θ-role.

@cite{krejci-2012} Table 2.8 validates the causativizability hierarchy across 12 languages. This data is formalized in MorphologicalCausation.krejciLanguages and verified by MorphologicalCausation.krejci_hierarchy_holds.

The hierarchy is implicational:
unaccusatives > middles/ingestives > unergatives > simple transitives

The key contribution is establishing middles/ingestives as a
distinct tier: Type 1 languages (Slave, Mapudungun, Classical
Nahuatl) causativize only unaccusatives; Type 2 (Cora, Marathi,
Amharic) add middles/ingestives; Type 3 (Ahtna, Tariana, Malayalam)
add unergatives; Type 4 (Basque, Dulong/Rawang, Koyukon) add simple
transitives. No language skips tiers. 

theorem Krejci2012.hierarchy_holds : (Semantics.Causation.Morphological.krejciLanguages.all fun (x : Semantics.Causation.Morphological.CausativizabilityData) => x.respectsHierarchy) = true

The causativizability hierarchy holds for all 12 languages.