Event Structure Templates #
Verbs decompose into templates (structural meaning that determines argument realization) filled by roots (idiosyncratic content). Templates compose via CAUSE; which sub-predicate determines argument realization yields different syntactic frames.
Key insight from @cite{rappaport-hovav-levin-2024}: verbs of motion + sustained contact (sweep, rub, scrape) have two grammatically relevant predicates — MOVE and CONTACT. Which predicate determines argument realization yields variable agentivity and distinct syntactic frames.
Bridges #
Template.toAspectualProfile→AspectualProfile(aspect)motionContact_variable_agentivity→sweepBasicSubjectProfile(proto-roles)contact_determines_implies_effector_subject→isEffector(proto-roles)lexicalize_increases_agentivity→pAgentScoreordering (proto-roles)HasResultState→ bieventive sub-event boundary (@cite{krejci-2012}; @cite{dowty-1979})cause_implies_resultState→ CAUSE entails result stateintransitiveVariant→ causative/inchoative alternation (@cite{krejci-2012}; @cite{rappaport-hovav-levin-1998})
Canonical event structure templates.
The first four are from @cite{rappaport-hovav-levin-1998}. motionContact is from @cite{rappaport-hovav-levin-2024}
for the sweep/rub/scrape class: [x MOVE y] WHILE [x CONTACT y].
- state : Template
- activity : Template
- achievement : Template
- accomplishment : Template
- motionContact : Template
Instances For
Equations
- Features.EventStructure.instDecidableEqTemplate 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
Does the template involve CAUSE?
Equations
- Features.EventStructure.Template.accomplishment.HasCause = True
- x✝.HasCause = False
Instances For
Equations
- One or more equations did not get rendered due to their size.
Does the template have an external causer position?
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
How many grammatically relevant predicates?
Equations
- Features.EventStructure.Template.state.predicateCount = 1
- Features.EventStructure.Template.activity.predicateCount = 1
- Features.EventStructure.Template.achievement.predicateCount = 1
- Features.EventStructure.Template.accomplishment.predicateCount = 2
- Features.EventStructure.Template.motionContact.predicateCount = 2
Instances For
Predicted aspectual profile for each template.
Equations
- Features.EventStructure.Template.state.toAspectualProfile = Features.stateProfile
- Features.EventStructure.Template.activity.toAspectualProfile = Features.activityProfile
- Features.EventStructure.Template.achievement.toAspectualProfile = Features.achievementProfile
- Features.EventStructure.Template.accomplishment.toAspectualProfile = Features.accomplishmentProfile
- Features.EventStructure.Template.motionContact.toAspectualProfile = Features.activityProfile
Instances For
Predicted Vendler class for each template (derived from profile).
Equations
Instances For
Templates with complex internal structure — multiple sub-events connected by CAUSE or embedding BECOME — license scopal ambiguities that mono-eventive templates do not.
At the template level, three diagnostics from @cite{dowty-1979} reduce
to two structural properties already defined above:
1. ***again*/*re-* ambiguity** tracks `HasResultState`: templates
embedding [BECOME [STATE]] allow restitutive readings where a
scopal modifier targets just the result sub-event.
2. **Negation over CAUSE** (@cite{koontz-garboden-2009}) tracks
`HasCause`: negation can scope narrowly over CAUSE, denying
the causal link while maintaining the result.
3. **"By itself" licensing** (@cite{koontz-garboden-2009}) also tracks
`HasCause`: "without outside help" requires CAUSE in the meaning.
@cite{krejci-2012}'s insight is that some verbs assigned simpler templates
(eat, wash, dress, learn) nonetheless pass all three diagnostics — evidence
that they have bieventive, causative event structures in their simple forms.
This verb-level property is captured in `RootTypology` and `ArgDerivation`,
not at the template level here.
Does the template embed a result state under BECOME? Templates with [BECOME [STATE]] have a sub-event boundary that scopal modifiers (again, re-, almost) can target independently.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
CAUSE implies a result state (accomplishment embeds BECOME).
Causative/Inchoative Alternation #
The accomplishment template [[x ACT] CAUSE [BECOME [y STATE]]]
has an intransitive variant. On the **deletion** analysis
(@cite{krejci-2012}; @cite{rappaport-hovav-levin-1998}), this is
the achievement [BECOME [x STATE]], obtained by stripping the
external cause — yielding a monoeventive representation.
On the competing **reflexivization** analysis (@cite{koontz-garboden-2009};
@cite{chierchia-2004}), anticausativization does NOT delete CAUSE.
Instead, the reflexive clitic (*se*, *sich*) identifies the EFFECTOR
with the THEME: the derived inchoative retains the full causative
structure [∃v[CAUSE(v,e) ∧ EFFECTOR(v,x) ∧ BECOME(e,s) ∧ THEME(s,x)]].
This preserves the Monotonicity Hypothesis and explains why
anticausative morphology cross-linguistically coincides with reflexive
morphology (Haspelmath 1990: 9/13 languages).
`Template.intransitiveVariant` below implements the deletion view at the
template level. The reflexivization analysis is formalized in
`KoontzGarboden2009`.
The intransitive variant of a template on the deletion analysis, stripping the external cause. Only accomplishments have an alternation partner.
NOTE: this implements one specific analysis. On the reflexivization analysis (@cite{koontz-garboden-2009}), the intransitive variant retains CAUSE with reflexivized arguments.
Equations
Instances For
The intransitive variant retains the result state (BECOME STATE survives stripping of ACT CAUSE).
The intransitive variant loses CAUSE (on the deletion analysis).
@cite{koontz-garboden-2009} disputes this on Monotonicity-Hypothesis
grounds; see Phenomena/Causation/Studies/KoontzGarboden2009.lean.
Only accomplishments have an intransitive variant (only templates with CAUSE can undergo the alternation).
Predicted subject entailment profile for each template.
Equations
- One or more equations did not get rendered due to their size.
- Features.EventStructure.Template.motionContact.subjectProfile = Semantics.ArgumentStructure.EntailmentProfile.sweepBasicSubjectProfile
Instances For
Predicted object entailment profile (if any).
Equations
- One or more equations did not get rendered due to their size.
- Features.EventStructure.Template.state.objectProfile = none
- Features.EventStructure.Template.activity.objectProfile = none
- Features.EventStructure.Template.achievement.objectProfile = none
Instances For
Accomplishment subject is a full agent (5 P-Agent entailments).
motionContact subject has the same profile as basic sweep (movement + IE, underspecified for agentivity).
For motionContact verbs, which sub-predicate determines argument realization (@cite{rappaport-hovav-levin-2024} §3.3–3.4).
- motion : DeterminingPredicate
- contact : DeterminingPredicate
Instances For
Equations
- Features.EventStructure.instDecidableEqDeterminingPredicate 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
When CONTACT determines argument realization, the subject is an effector (movement + IE → external argument). This yields the simple transitive frame: "The wind swept the deck".
The motionContact object profile exists and is a force recipient.
Instrument lexicalization adds agentivity to a template by restricting the moving entity to a specific instrument, forcing volition + sentience
- causation. The result is the broom-sweep subject profile.
Equations
Instances For
Instrument lexicalization strictly increases agentivity for motionContact templates (@cite{rappaport-hovav-levin-2024} §3.5: broom-sweep is more agentive than basic sweep).
Lexicalized motionContact has full proto-agent profile (5 entailments).
Levin classes map to event structure templates via meaning components. The core correspondence (@cite{rappaport-hovav-levin-1998} §3; @cite{rappaport-hovav-levin-2010} §2):
| Meaning component pattern | Template | Example class |
|---|---|---|
| CoS + causation | accomplishment | break (45.1), destroy (44) |
| CoS, no causation | achievement | appear (48.1), calve (28) |
| No CoS, no motion | state | exist (47.1), admire (31.2) |
| Otherwise | activity | hit (18.1), run (51.3) |
The `motionContact` template is specific to the sweep/rub/scrape
class and requires a class-specific override.
The fundamental binary distinction in event types: whether a predicate encodes a process (PROC only) or a state change (involves CHANGE).
This crosscuts Vendler's four-way classification: degree achievements are Vendler activities or accomplishments depending on scale boundedness but are event-structurally state-change predicates (@cite{bohnemeyer-2004} §5).
@cite{bohnemeyer-2004} argues this is the primary semantic distinction governing verb classification in Yukatek Maya — more fundamental than Vendler classes for predicting argument linking and transitivization.
Instances For
Equations
- Features.EventStructure.instDecidableEqEventType 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
Derive event type from template. Activities and motionContact are processes; states, achievements, and accomplishments involve state change.
Equations
Instances For
Whether a process is internally caused — the event is instigated by a participant — or externally caused — occurring "spontaneously" without an instigator.
This is a per-verb property of the ROOT, not of the template. Two activity verbs can differ: sing (internal) vs roll (external).
@cite{levin-hovav-1995} §4; @cite{bohnemeyer-2004} §2,6.
@cite{koontz-garboden-2009} §4.1: externally caused COS verbs have CAUSE+EFFECTOR in their LSR and license por sí solo 'by itself'. Internally caused COS verbs (empeorar, hervir, crecer) lack CAUSE in their LSR and reject por sí solo.
- internal : CausationType
- external : CausationType
Instances For
Equations
- Features.EventStructure.instDecidableEqCausationType 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
Externally caused COS verbs have CAUSE in their LSR; internally caused COS verbs do not (@cite{koontz-garboden-2009} §4.1; @cite{levin-hovav-1995} §4).
This determines whether derived inchoatives (on the reflexivization analysis) retain a CAUSE operator.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
por sí solo / by itself is licensed iff CAUSE is present in the LSR (@cite{koontz-garboden-2009} §4.1). Stative predicates and internally caused COS verbs reject it.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Predicted event structure template from meaning components.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Predicted template for a Levin class, with class-specific overrides.
Equations
Instances For
Verification: canonical quadruple #
Break → accomplishment (CoS + causation → [ACT CAUSE BECOME]).
Hit → activity (contact + motion, no CoS → [ACT]).
Touch → activity (contact only, no CoS).
Cut → accomplishment (CoS + causation).
Change-of-state classes → accomplishment #
All §45 CoS classes map to accomplishment.
Motion classes → activity #
Motion verbs are activities (no CoS, have motion).
Stative classes → state #
Perception/psych statives map to state template.
Achievement classes #
Appear (CoS without causation) → achievement.
Calve (CoS without causation) → achievement.
Special: motionContact #
Wipe class → motionContact (class-specific override).
Template → aspectual class consistency #
Accomplishment classes (break, cut) are predicted telic.
Activity classes (hit, run) are predicted atelic.
State classes (exist, admire) are predicted stative.
§ 9. Bridge: Event Structure ↔ Diathesis Alternation #
predictedTemplate and predictedAlternation are two predictions computed from the
same MeaningComponents feature vector. This section proves their agreement and shows
that MeaningComponents.fuse simultaneously derives both template shift and new
alternation predictions from a single componentwise OR operation.
The central theorem — ci_alternation_iff_template_alternates — says the
causative/inchoative alternation is exactly the syntactic reflex of having an
accomplishment event template (which has an intransitive variant), modulo
instrumentSpec. This connects @cite{levin-1993}'s diathesis alternation diagnostics
to @cite{rappaport-hovav-levin-1998}'s event structure decomposition.
The causative/inchoative alternation is available iff the verb's event template has an intransitive variant (i.e., is an accomplishment), given no instrumentSpec.
Both conditions reduce to changeOfState ∧ causation, making the alternation
prediction and the event structure prediction two views of a single semantic fact.
instrumentSpec breaks the template↔alternation correspondence: cut verbs have accomplishment template (they cause state change) but cannot undergo causative/inchoative alternation (instrument specification requires an agent).
This is why ci_alternation_iff_template_alternates requires
instrumentSpec = false — the hypothesis is necessary, not just sufficient.
Fusion with CoS + causation yields accomplishment template regardless of the verb's original template. The resultative construction adds [CAUSE [BECOME [STATE]]], upgrading any verb to accomplishment.
One fusion, three consequences: accomplishment template, causative/inchoative alternation, and intransitive variant — all from a single componentwise OR.
Fusion-induced Vendler class shift: fusion with CoS + causation yields accomplishment Vendler class (telic, bounded).
Fusion with CoS + causation yields result state, enabling again/re- restitutive readings (@cite{dowty-1979}).
Fusion with CoS + causation yields CAUSE structure, enabling negation-over-CAUSE readings and by itself modification (@cite{koontz-garboden-2009}).