views updated

# EVENTS IN SEMANTIC THEORY

It is an ancient idea that many verbs are used to describe eventsthings that happen, in places and at times. Frank Plumpton Ramsey introduced an important twist, in the context of distinguishing events from facts. Suppose that Aggie hit Pat. Then on Ramsey's view, the fact reported with (1)

• (1) Aggie hit Pat

is the general proposition that there was a hitting of Pat by Aggie. This existential generalization, unlike any event, has no specific spatiotemporal properties. But any event of Aggie hitting Pat verifies (1). So the action report seems to mean that an event of a certain sort occurred. Though at least initially, it is not clear how to square this with the compositionality of linguistic meaning. Let the invented monadic predicate Aghipatish1 be satisfied by z if and only if (iff) z was a hitting of Pat by Aggie. Then plausibly, (1) is true iff z[Aghipatish1(z)]. But this biconditional reveals nothing about how the meaning of (1) is determined by the constituent words. And prima facie, the logical form of (1) is Hit2(a, p ); where a and p are names for Aggie and Pat, respectively, and Hit2 is satisfied by an ordered pair x, y iff x hit y. Donald Davidson (1967, 1985) shows how to represent the meaning of (1) compositionally and "eventishly," and he offers an argument for doing so. Others develop his proposal and provide independent support for it.

Let Hit3 be satisfied by an ordered triple x, y, z iff z was a hitting of y by x, so that xy{Hit2(x, y) z[Hit3(x, y, z)]}. Then (1a),

• (1a) z[Hit3(a, p , z)]

which is true iff z[Aghipatish1(z)], has parts corresponding to the words in (1). If we represent the meaning of (1) with (1a), we can explain the apparent synonymy of (1) with (2):

• (2) There was a hitting of Pat by Aggie

But one wants to see further evidence of the alleged covert variable.

An action report can be extended as in (38):

• (3) Aggie hit Pat softly
• (4) Aggie hit Pat with a red stick
• (5) Aggie hit Pat in March
• (6) Aggie hit Pat with a red stick in March
• (7) Aggie hit Pat in March with a red stick
• (8) Aggie hit Pat softly with a red stick in March

One might be inclined to represent the meaning of (3) with SoftlyHit2(a, p ); where the invented binary predicate is satisfied by x, y iff x hit y softly. But if an inference from (3) to (1) is of the form Φ2(α, β), so ψ2(α, β), we need some other explanation for why the truth of (3) guarantees the truth of (1). One can stipulate that xy[SoftlyHit2(x, y) Hit2(x, y)]. However, one wants to know why the corresponding English sentences are related in this fashion. Furthermore, (3) seems to be synonymous with "There was a soft hitting of Pat by Aggie," which implies (2). Such implications are reminiscent of conjunction-reduction, as in inferences like the following: x[Red(x) & Stick(x)], so x[Stick(x)].

This invites Davidson's (1967) hypothesis that the logical form of (3) is (3a);

• (3a) z[Hit3(Aggie, Pat, z) & Soft1(z)]

where Soft1(z) means that z was done softly. Furthermore, sentences like (18) exhibit a network of entailments. The truth of (4) or (5) also guarantees the truth of (1); (6) implies (7), which implies each of (46); and (8) implies each of (17). These facts, which illustrate that adverb-reduction is often a valid form of inference in natural language, go unexplained if we represent the meanings of (38) with binary predicates like HitInMarchWithARedStick2. But we can explain the entailments by analyzing the adverbial modifiers as predicates conjoined with others, as in (7a);

• (7a) z[Hit3(Aggie, Pat, z) & In2(z, March) & With2(z, a red stick)]

where In2(z, March) and With2(z, a red stick) mean that z occurred in March, and that z was done with a red stick. By analogy, "There was a red stick on the table touching the chalk" implies that there was a red stick on the table, a stick touching the chalk, a red stick, and so on.

There are also nonimplications to account for. Each of (35) could be true, even if (8) is false. Aggie may have hit Pat more than once, but never softly with a red stick in March. Let's suppose, though, that Aggie hit Pat exactly twice: once in March with a red stick, and once in April with a blue stick. Then (911) are true, like (46), but (1213) are false.

• (9) Aggie hit Pat with a blue stick
• (10) Aggie hit Pat in April
• (11) Aggie hit Pat with a blue stick in April
• (12) Aggie hit Pat with a red stick in April
• (13) Aggie hit Pat with a blue stick in March

The truth of (9) and (5) does not guarantee the truth of (13). Nor does the truth of (4) and (10) guarantee the truth of (12). This is what Davidson's (1967) account predicts.

If (4) and (10) are true, there were events z1 and z2, such that Hit3(Aggie, Pat, z1) & With2(z1, a red stick) & Hit3(Aggie, Pat, z2) & In2(z2, April). But it does not follow that there was an event z such that With2(z, a red stick) & In2(z, April). So (12) can be false, and likewise for (13). If we represent the meanings of (113) with predicates like HitWithARedStickInApril2, we must add stipulations corresponding to the network of implications, and then explain why the English sentences exhibit these implications and not the others. (Note that appealing to times, instead of events, will not account for the facts. Aggie may have hit Pat simultaneously with a red stick softly and with a blue stick sharply.)

## Other Evidence

We can specify the meaning of (14) with (14a):

• (14) Aggie fled after Pat fell
• (14a) z{Fled2(a , z) & w[(After2(z, w) & Fell2(p, w )]}

which is true iff an event of Aggie's fleeing occurred after an event Pat's falling. Words like after can thus be analyzed as devices for expressing relations between events. Perceptual reports are also relevant. In (15) as opposed to (16):

• (15) Pat heard Aggie shout
• (16) Pat heard that Aggie shouted

shout is untensed, and replacing Aggie with another name for the same individual is sure to preserve truth. Thus, one might render (15) as (15a),

• (15a) zw[Heard3(p , w, z) & Shout2(a , w)]

which is true iff there was a hearing by Pat of a shouting by Aggie. This lets us account for the ambiguity of (16), which can imply that the hearing took place in the hall or that the shout did:

• (16) Pat heard Aggie shout in the hall
• (16a) zw[Heard3(p , w, z) & Shout2(a , w) & In2(the hall, z)]
• (16b) zw[Heard3(p , w, z) & Shout2(a , w) & In2(the hall, w)]

Another kind of evidence concerns the intuitive unacceptability of certain adverbial modifications. While (18) sounds somehow wrong, (17) and (19) do not:

• (17) Aggie ran for an hour
• (18) Aggie ran in an hour
• (19) Aggie ran to the store in an hour

If we represent the verb-phrase meanings in (17) and (19) with RanForAnHour2 and RanToTheStoreInAnHour2, not only do we fail to capture implications, we are left wondering why "ran in an hour" is a defective complex monadic predicate. By contrast, (17a) and (19a)

• (17a) z[Ran2(a , z) & For2(z, an hour)]
• (19a) z[Ran2(a , z) & To2(z, the store) & In2(z, an hour)]

suggest that "for an hour," unlike "in an hour," can be part of an event description that does not provide an independent way of saying when the events described are finished. This hypothesis is confirmed by examples like (2021):

• (20) Aggie painted the walls for/in an hour
• (21) Aggie painted walls for/in an hour

While (20) is fine with either modifier, (21) is not. Furthermore, with "in an hour" (20) implies an event that ended when the walls in question were covered with paint. With "for an hour" neither sentence implies that Aggie finished painting any wall.

A striking generalization about action reports of the form "Subject Verb Object," with the verb in active voice, is that such reports invariably imply that the subject of the sentence was the actor. We can invent a predicate Tih2 satisfied by x, y iff y hit x; and we can imagine a language in which a homophone of (1), with Aggie as the subject, means that Tih2(a, p )or equivalently, Hit2(p, a ). However, there are no such expressions in natural human languages. And while the source of this fact is a matter of debate, a great deal of evidence suggests a constraint on how grammatical relations are related to thematic relations that hold between events and their participants. In which case, event variables (and thematic relations) are introduced somehow. However, there is more than one way to introduce them.

## Thematic Elaboration

Let Hitting1 be satisfied by z iff z was an event of hitting, ignoring tense for simplicity. Let Agent2 and Patient2 signify thematic relations, without worrying here about how to get beyond intuitive specifications of these relations, so that xy{z[Hit3(x, y, z)] iff z[Agent2(z, x) & Hitting1 (z) & Patient2(z, y)]}.

This makes it easy to explain why it follows from (1) that Aggie did something, there was a hitting, and something happened to Pat. This view also preserves a sense in which the transitive verb hit is a binary predicate. For while the verb itself is associated with a monadic predicates of events, hit is also associated with two thematic relations. Correlatively, one can capture the distinctionindependently motivated in many languages that mark nominative and accusative casebetween intransitive verbs like fled that implicate action, and those that do not. Intuitively, events of falling (like deaths) are things that happen to individuals (not things done), even if such events are intended effects of actions. Besides, one can supplement Davidson's (1967) original proposalas Davidson (1985) didwith hypotheses like the following:
x{z[Fled2(x, z)] {z[Agent2(z, x)] & Fleeing1(z)]};
x{z[Fell2(x, z)] {z[Patient2(z, x)] & Falling1(z)]}.
But there are at least two construals of such hypotheses.

One might view them as analyses of multiply unsaturated verb-meanings. From this perspective the verb hit is satisfied by ordered triples x, y, z such that z was a hitting whose Patient was y and whose Agent was x. In which case, given standard assumptions about semantic compositionality, hit Pat is satisfied by ordered pairs x, z such that z was a hitting whose Patient was Pat and whose Agent was x. Less standardly, one might say that hit is satisfied by events of hitting (period) and that hit Pat is satisfied by each event of hitting whose Patient is Pat. This is, in effect, to adopt the following hypothesis: combining hit with a direct object corresponds to predicate-conjunction, not predicate-saturation; and the thematic relation "being the Patient of" is expressed by a certain grammatical relation, between the verb and its object, not simply by the lexical meaning of hit.

Barry Schein (1993, 2002) and others argue that considerations involving plurality, along with the need for second-order quantification, favor the second perspective. On this kind of view, "Five boys ate two pizzas" has a (collective) reading according to which there some events of eating whose Agents were five boys, and whose Patients were two pizzas; where this does not imply that any one event had all five boys and both pizzas as participants. The first view fits more naturally with the following formulation of the collective reading: There was an event whose (plural) Agent was a collection of five boys, and whose Patient was a collection of two pizzas. Still, however one thinks of thematic elaboration, it both extends the scope of event analyses and highlights difficulties.

As many authors have discussed, verbs like boil can appear in transitive and intransitive forms, with a characteristic entailment illustrated in (22):

• (22) Aggie boiled the soup; so the soup boiled

Treating the two forms as independent predicates, Boiled2(x, y) and Boiled1(x), makes the implication mysterious. So one might analyze (22) as in (23) or (24):

• (23) z{Agent2(z, a ) & w[C2(z, w) & Boiling1(w) & Patient2(w, s )] }; so z[Patient2(z, s ) & Boiling1(z)]
• (24) z{Agent2(z, a ) & w[M2(z, w) & Boiling1(w)] & Patient2(z, s )}; so z[Patient2(z, s ) & Boiling1(z)]

Here, s stands for the soup, C2 indicates a causal relation holding between an action and some of its effects, and M2 indicates a merelogical relation holding between processes that start with actions and end with effects of those actions. Many linguists argue that some such analysis is required, especially given the constraints on how grammatical relations are mapped to thematic relations. But specifying the requisite causal/mereological relation has proven difficult. Moreover, (23) fails to represent Aggie and the soup as coparticipants in some event describable with an adverbial phrase; yet, if Aggie boiled the soup on Monday, both Aggie's action and the resultant boiling occurred on Monday. And given (24), a background premise is required to reveal the inference as valid: yzw[M2(z, w) & Patient2(z, y) Patient2(w, y)].

This raises hard questions about the individuation of actions and their relation to thematic relations. More generally, it is unclear what event variables range over, given the kinds of considerations that motivate such variables. Suppose that Aggie drank a pint of beer (and nothing else) in ten minutes. Then for those ten minutes Aggie drank beer. Let z1 be this event of Aggie drinking beer, and let z2 be the event of Aggie drinking the pint in question. Intuitively, z1 is z2; Aggie's beer drinking was none other than the drinking of that pint. In which case, z1 satisfies "in ten minutes" iff z2 does; and z2 does. However, if z1 satisfies "in ten minutes," why is "Aggie drank beer in ten minutes" anomalous? This kind of question arises often.

Consider two billiard balls, b and c , that came into contact exactly once. At that moment, b touched c , and c touched b . Perhaps touched, used in this way, does not mark its subject as an Agent, but letting Sub2 and Ob2 signify the relevant thematic relations, whatever they are: z[Sub2(z, b ) & Touching1(z) & Ob2(z, c )], and z[Sub2(c , z) & Touching1(z) & Ob2(b , z)]. One might have expected the touching of c by b to be identical with the touching of b by c . But how can any one event of touching, z, be such that: Sub2(z, b ) & Ob2(z, c ) & Sub2(z, c ) & Ob2(z, b )? Presumably, Sub2(z, b ) implies that b is the unique individual that bears the relevant thematic relation to zand likewise for Sub2(z, c )since Aggie touched/lifted Pat does not mean merely that there was a touching/lifting with Aggie as one of potentially many touchers/lifters, and Pat as one of potentially many things touched/lifted.

One can avoid the false implication that b = c by denying that event variables in semantic theories range over language-independent occurrences. However, this has implications for the relations among meaning, truth, and ontology. Another option is to elaborate further, treating notation like Sub2(z, b ) as shorthand for claims of the form: e[R2(e, z) & P2(e, b )]; where e ranges over language-independent spatiotemporal particulars individuated (at least roughly) in accordance with intuitions about events, R2 signifies a relation that holds between such particulars and their grammatical presentations, and P2 signifies a suitable participation relation. But if z ranges over things individuated partly in terms of the grammatical relations that verbs bear to their arguments, we still face questions about the individuation of events and their relation to any thematic relations appealed to in theories of meaning.

## Bibliography

Davidson, Donald. "Adverbs of Action." In Essays on Davidson: Actions and Events, edited by Bruce Vermazen and Merrill B. Hintikka. Oxford, U.K.: Clarendon Press, 1985.

Davidson, Donald. "The Logical Form of Action Sentences." In The Logic of Decision and Action, edited by Nicholas Rescher. Pittsburgh: University of Pittsburgh Press, 1966.

Fodor, Jerry. "Three Reasons for Not Deriving 'Kill' from 'Cause to Die.'" Linguistic Inquiry 1 (1970): 429438.

Higginbotham, James. "The Logical form of Perceptual Reports." Journal of Philosophy 80 (1983): 100127.

Higginbotham, James, Fabio Pianesi, and Achille C. Varzi, eds. Speaking of Events. New York: Oxford University Press, 2000.

Landman, Fred. "Plurality." In The Handbook of Contemporary Semantic Theory, edited by Shalom. Lappin. Oxford, U.K.: Blackwell, 1996.

LePore, Ernest, and Brian P. McLaughlin, eds. Actions and Events: Perspectives on the Philosophy of Donald Davidson. Oxford, U.K.: Blackwell, 1985.

Parsons, Terry. Events in the Semantics of English. Cambridge, MA: MIT Press, 1990.

Pietroski, Paul. Events and Semantic Architecture. New York: Oxford University Press, 2005.

Ramsey, Frank. "Facts and Propositions." Aristotelian Society (Supplement) 6 (1927): 153170.

Schein, Barry. "Events and the Semantic Content of Thematic Relations." In Logical Form and Language, edited by Gerhard Preyer and Georg Peter. New York: Oxford University Press, 2002.

Schein, Barry. Plurals. Cambridge, MA: MIT Press, 1993.

Taylor, Barry. Modes of Occurrence. Oxford, U.K.: Blackwell, 1985.

Tenny, Carol. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht, Netherlands: Kluwer Academic, 1994.

Thomson, Judith. Acts and Other Events. Ithaca, NY: Cornell University Press, 1977.

Vendler, Zeno. Linguistics in Philosophy. Ithaca, NY: Cornell University Press, 1967.

Paul M. Pietroski (2005)