Event Semantics

Origin. Donald Davidson, "The Logical Form of Action Sentences" (in Rescher, ed., The Logic of Decision and Action, University of Pittsburgh Press, 1967), argued that action sentences quantify over events, so that adverbial modification is conjunction and adverb-drop inferences become logical rather than lexical. Castañeda, commenting on that paper in the same volume (1967), proposed the neo-Davidsonian variant in which every participant is severed into its own role conjunct. Terence Parsons, Events in the Semantics of English: A Study in Subatomic Semantics (MIT Press, 1990), developed it; Angelika Kratzer, "Severing the External Argument from its Verb" (1996), gave it a syntactic basis. Vendler (1957) and Dowty (1979) supplied the aspectual classes; Bach, "The Algebra of Events" (Linguistics and Philosophy 9, 1986), the mereology.

Models. The domain of individuals is enlarged with events, a verb denotes a property of events rather than a relation among participants, and a clause asserts that some event with the stated properties exists. Modifiers are further predicates of the same variable, conjoined, so the entailment from a modified sentence to an unmodified one is conjunction elimination — Davidson's original argument and still the main one for the whole apparatus. Thematic roles are relations between an event and a participant, taken as partial functions under Thematic Uniqueness. Events themselves are structured: they form a join semilattice, which is what makes plurality and telicity expressible in the same algebra as plural individuals.

Formalism.

Davidsonian events: "John kissed Mary" → ∃e[kiss(e) ∧ Agent(e,j) ∧ Patient(e,m)]. Event variable e. Quantified over. Modification target.

Adverbial modification: "John kissed Mary slowly" → ∃e[kiss(e) ∧ Agent(e,j) ∧ Patient(e,m) ∧ slow(e)]. Slow modifies event. Conjunction. Explains entailments.

Adverb drop (Davidson's argument): "John buttered the toast slowly in the bathroom with a knife" ⊨ "John buttered the toast in the bathroom" ⊨ "John buttered the toast". Each modifier is one conjunct on e; dropping it is conjunction elimination. No meaning postulate per adverb, and no variable-arity verb.

Thematic roles: Agent, Patient, Theme, Goal, etc. Link participants to event. Decomposed predication. Role-based.

Thematic uniqueness: ∀e∀x∀y[(Agent(e,x) ∧ Agent(e,y)) → x = y]. Each role is a partial function from events to participants. One Agent per event; this is what licenses role-based inference.

Neo-Davidsonian: Verb contributes only event type. Roles added separately. kiss: λe.kiss(e). Agent: λxλe.Agent(e,x).

Voice decomposition (Kratzer): vP introduces Agent. VP introduces event. Severed external argument. Syntactic decomposition.

Existential closure: The verb contributes λe.kiss(e) ∧ Agent(e,j) ∧ Patient(e,m) — a property of events. ∃e is supplied at the clausal level, not by the verb. Leaving the variable open is what lets adverbs and aspect apply first.

Aspect: Events vs states. Telic vs atelic. Event structure. Temporal properties.

Aspectual classes (Vendler 1957): States: know, resemble — no internal stages. Activities: run, push — atelic, cumulative. Accomplishments: build a house — telic, with a culmination. Achievements: notice, reach the summit — telic, punctual. Dowty (1979) derives the classes from a decomposition with BECOME and CAUSE.

Telicity from mereology (Krifka): Cumulative: P(e) ∧ P(e') → P(e ⊔ e'). Quantized: no proper part of an event satisfying P also satisfies P. Telicity of the VP tracks quantization of the incremental theme. "eat an apple" quantized, hence telic; "eat apples" cumulative, hence atelic. The object's reference type is transferred to the event by a homomorphism.

Plurality: Collective vs distributive. Event summing. Mereology of events. Plural predication.

Event algebra (Bach 1986): ⊑ parthood on events; ⊔ event sum; a join semilattice, as for plural individuals. Collective reading: one event with a plural Agent. Distributive reading: a sum of events, each with a singular Agent. Link's (1983) lattice treatment of plurals applies unchanged to eventualities.

Symbols.

Metatheory. The framework adds no expressive power: Davidsonian logical forms are ordinary first-order formulas over a domain enlarged with events, so what is gained is an inference pattern, not a class of truth conditions unavailable before. That pattern is the whole original argument — Davidson (1967) observed that modifiers analysed as conjuncts on a shared event variable make adverb-drop entailments instances of conjunction elimination, requiring neither a meaning postulate per adverb nor a variable-arity verb. Castañeda (1967) showed the same effect follows if subject and object are severed into role conjuncts too, the neo-Davidsonian form that Parsons (1990) developed and Kratzer (1996) grounded syntactically by placing the Agent on a Voice head while leaving the internal argument with the verb. Bach (1986) put events in a join semilattice, so Link's (1983) plural mereology applies to them unchanged; Krifka (1989, 1998) used the same algebra to derive telicity from quantized reference of the incremental theme. Dowty (1991) replaced the discrete role inventory with Proto-Agent and Proto-Patient entailment clusters.

Applies to. Semantics. Syntax-semantics interface. Aspect. Modification.

Limitations. The ontology is a commitment, not a notation: events must be in the domain, and the theory supplies no identity criterion for them — Davidson's causal criterion (1969) was later abandoned in favour of Quine's spatiotemporal one (1985), and neither settles when two descriptions name one event, which leaves event counting ("knocked three times") underdetermined. The conjunctive analysis covers only intersective modifiers; "almost", "allegedly", and "intentionally" license no adverb drop, so the framework's central argument does not reach them and they need separate machinery. The role inventory is fixed by nothing internal to the theory — Dowty (1991) argued that no discrete list works and replaced it with prototype clusters. Whether states carry event arguments is disputed: Maienborn (2005) argued that stative verbs fail the locative-modification and perception-report tests, which would confine the analysis to eventive predicates.

© 2026 Lingenic LLC