[SynSem] Heidi Harley (Arizona)

retour Séminaire Syntaxe et Sémantique

Heidi Harley (U. Arizona)

(Joint work with Raffaella Folli, University of Ulster)

Object 'drop', noun incorporation and intentionality: How do we individuate events?

We introduce a pattern of interaction between object drop and the animacy of the external argument in English, particularly with verbs of contact. Object drop produces an agentivity/animacy constraint on the newly intransitive verb, which doesn't exist in the transitive equivalent. Compare John pushed the car to the curb and The glacier pushed the boulder to the sea. The former, but not the latter, allows for object drop: At the signal, John pushed! vs #In winter, the glacier pushed.

Our analysis follows Martí 2015, according to which object drop instantiates noun incorporation of a null indefinite N. As with all noun incorporation, this tends to yield a 'conventionalized community activity' denotation for the VP—compare Michael lifted the glass with Michael lifted. 

In both object drop and (overt) noun incorporation contexts, the denotation of the object nominal is nonreferential—no object participant in the verbal event is introduced. Instead, the object nominal is integrated into the predicate denotation via some flavor of predicate restriction or predicate modification (Chung & Ladusaw 2004, e.g.) We contend that the absence of an event participant creates a challenge for the application of the predicate by the speaker. Philosophers have long supposed that events are individuated by virtue of their participant configurations (Strawson,  Carlson). Linguists in turn have reified this intuition in a number of proposed principles or conditions on argument structure, such as Levin & Rappaport (20xx)'s 'Argument-per-Subevent Condition' (see also Ramchand 2008, a.o.). A predicate which has been subject to object drop or noun incorporation, then, lacking an event participant, does not pick out the same class of events that the transitive verbal predicate does. 

      Instead, we contend, intentionality is recruited to derive the sortal content of the verbal predicate. Martin 2015 showed that an agent's intention by itself can constitute a so-called "indicative property", providing enough individuating properties to license the application of a change-of-state predicate in nonculminating contexts even when the change of state encoded by the verb has not yet begun to occur. Without an intentional agent, she shows, such 'Zero-CoS' readings of telic predicates are unavailable, since no indicative properties which can identify the ongoing event as an instance of the predicate exist until the CoS actually begins to occur. Intentions, then, can identify events even when 'normal' event identification criteria, such as the configuration of participants or the existence of a CoS, do not apply. This, we propose, is why the subjects of such predicates have to be (intentional) Agents—it's because without intention, the indicative properties necessary for application of the predicate cannot be identified. This also lets us understand why object drop contexts so often produce such specialized 'conventionalized community activity' readings—it's a subcase of the wider case where simply intending to execute an action of a particular type yields enough indicative properties to license predication. We term this effect the 'Goal Oriented Condition' on object drop. This condition on event individuation in turn constrains the types of subject argument that are compatible with object drop contexts, since external arguments must be 'teleologically capable' of executing the event type denoted by the predicate. In the case of intentionally individuated event types, the only kind of external argument that has the relevant teleological properties are animate Agents.