4.4 Event semantics

Event Semantics

Event semantics treats sentences not just as descriptions of the world, but as descriptions of events that happen in the world. Instead of saying a sentence is simply true or false, you introduce an event variable into the logical representation. This gives you a way to talk about actions, states, and processes as things in their own right, which turns out to be powerful for handling modifiers, temporal structure, and thematic roles compositionally.

Basics of Event Semantics

The core idea comes from philosopher Donald Davidson: a sentence like "John kicked the ball" doesn't just relate John and the ball. It also asserts that there exists a kicking event. In a Davidsonian representation, you'd write something like:

$\exists e[\text{kick}(e) \wedge \text{agent}(e, \text{John}) \wedge \text{patient}(e, \text{ball})]$

That variable $e$ is the event. Once events are treated as entities in your semantics, you can attach extra information to them (location, time, manner) just by adding more conjuncts. This is why event semantics is so useful for compositionality: modifiers don't require special rules, they just predicate over the event variable.

Event semantics distinguishes three main ontological categories:

• Events are occurrences at a specific time and place. "John kicked the ball" and "Mary ate an apple" each introduce a bounded event.
• States are conditions that hold over a stretch of time without implying change. "The book is on the table" and "John loves Mary" describe states.
• Processes are ongoing activities that unfold over time. "John is running" and "The water is boiling" are processes.

Temporal and Causal Representation

Because events are entities, you can represent temporal and causal relationships between them directly.

Temporal structure:

• Events can be punctual (happening at essentially a single point, like a flash of lightning) or durative (extending over a period, like a meeting).
• Temporal ordering uses relations like before, after, and during to sequence events relative to one another. For example, "John left before Mary arrived" links two events with a before relation.

Causal structure:

• Thematic roles (agent, patient, instrument) attach participants to events, specifying how each entity is involved.
• Causal relations like cause, enable, and prevent link one event to another. "The storm caused the power outage" connects a storm event to an outage event through a causal relation.

Application to Predicate Types

Different verb classes map onto different event types, and event semantics gives you the tools to represent those differences precisely.

• Action verbs like "kick" and "eat" denote events with an agent performing an action. "John kicked the ball" introduces an event with John as agent and the ball as patient.
• Stative verbs like "love" and "believe" denote states rather than events. "John loves Mary" is represented as a state holding over some interval, with no implication of change or culmination.
• Aspectual modifiers specify temporal properties of the event. "John ran for an hour" describes a process with a one-hour duration, while "John ran a mile in an hour" describes an event with a natural endpoint reached within that time.

That last contrast connects to an important distinction: telic events have a natural endpoint (running a mile), while atelic events lack one (just running). Event semantics captures this by representing the internal temporal structure of the event differently for each type. The classic test: you can say "John ran for an hour" (atelic, fine with for-phrases) but "John ran a mile in an hour" (telic, natural with in-phrases). Swapping the modifiers sounds odd, which reveals the underlying event structure.

Advantages for Semantic Analysis

Event semantics earns its place in the toolkit for several reasons:

• Compositionality of modifiers. Adding "in the park" or "with a stick" to "John hit the ball" just means adding more predicates over the same event variable. You don't need a separate rule for each modifier type.
• Telic vs. atelic distinctions. The framework naturally represents whether an event has an inherent endpoint, capturing contrasts that matter for aspect and temporal reasoning.
• Causal reasoning. By treating events as entities linked by causal relations, you can represent complex sentences about why things happened.
• Broad applicability. Event semantics has been successfully applied to tense, aspect, modality, and discourse structure, making it one of the more versatile tools in formal semantics.

The key takeaway: by reifying events (treating them as things that exist in your logical representations), you gain a flexible, compositional way to handle a wide range of semantic phenomena that would otherwise require ad hoc solutions.