CROSS–COERCION 175 (14) a The light flashed.

b. The light was flashing all night.

‘Flash’ is a point; for this example we need therefore consider only the event flash. Sentence (14-a) can then be formalised as the integrity constraint

?Happens(flash, t), t <now succeeds.

In sentence (14-b) ‘flash’ occurs in the progressive, hence it must for- mally be represented by a fluent, or rather a pair of fluents – one an activity f1, the other a (possibly parametrized) state f2, related by a dynamics. At

first f1 andf2 are parameters, which then have to be unified with material

given by sentence (14-b). However, at first sight there seems to be no flu- ent with which the parameters can be unified, since the lexicalmaterial of (14-b) provides only an event type: the representation of the point ‘flash’. The coercion process requires therefore as a preliminary stage coercing an event type to a fluent. In [47] (see also Section 3.5 of Chapter 12 below) we have indicated (for the case of coerced nominals) a procedure which achieves just this. Namely, this form of coercion is represented by a map-

pingflash _5→ Happens[flash,ˆt], which maps an event type to a fluent5. By

construction, in any given model the fluent has the same temporal profile as the event type, but it belongs to a different syntactic category and is now available for unification. We still need one to fill argument place however, because we have made an absolute syntactic distinction between activities and states in terms of the argument places that they can fill (see Chapter 7). A fluent may change syntactic category due to a change in the scenario, but we do not allow that a fluent represents simultaneously a state and an activity. Therefore Happens[flash,ˆt] unifies either withf1 or with f2, but

not with both. We then imagine that semantic memory is searched for an activityf, or rather a mechanism, which drives the flashing via dynamics (it is clear what this would be in the case of a lighthouse, say); and in the next stepf is unified withf1 andHappens[flash,ˆt]withf2. Sentence (14-b) is

then represented by the integrity constraint6

?HoldsAt(f, s), s <now succeeds.

3.4. Activity/accomplishment ! state. To conclude the Section on cross–coercion we briefly discuss two examples about which much more could be said, in particular with regard to the syntax-semantics interface.

3.4.1. Negation. It has often been held (see for example Verkuijl [128]) that negating an activity results in a stative predicate. This intuition can to a certain extent be reproduced in the present framework. An activity (in the wide sense) consists of a fluent representing the activity proper, and a parametrized state. In a minimal model, the state changes only when the activity is ‘on’; outside of those intervals, the parametrized state is really

5_{At this point the material on coding in Chapter 6 is used essentially.}

6_{Disregarding for the moment the temporal adverbial ‘all night’; these adverbials will} be treated in Section 4.

176 11. COERCION

static. It follows that the set of intervals complementary to that representing the activity indeed has a state-like character. It is a somewhat atypical state in that it is initiated and terminated by events, since of course a terminating event for an activity is an initiating event for its negation.

3.4.2. Passive. In German and Dutch there is a form of the passive, the ‘Zustandspassiv’ (indicated by a special auxiliary), which transforms an activity or an accomplishment into a (consequent) state.

(15) a. Johann baut ein Haus. Es ist weiss.

b. Das Haus ist von Johann gebaut. Es ist weiss. c. Das Haus wird von Johann gebaut. *Es ist weiss.

In (15-b), the auxiliary sein indicates that the house is finished; by con- trast, in (15-c) the auxiliary werdenrefers to the process of building. This explains the * in (15-c).

In English the passive is apparently ambiguous between the two read- ings. Compare

(16) a. John is building a shed in his garden. This causes his neighbours much distress [because of the tremendous noise].

b. The shed that is built in John’s garden causes his neighbours much distress [because it spoils their view].

c. A shed is built in John’s garden. This causes his neighbours much distress.

‘This’ in sentence (16-c) can refer both to an activity, as in (16-a), and the result of that activity, as in (16-b). If one conceives of the passive syntac- tically as movement of the object NP into subject position, an interesting tension between syntax and semantics comes to the fore. The NPa house in build a house is an incremental theme, and need not denote an object in the ontological sense. In our setup, the denotation of a house is as it were distributed over the changing partial object, the canonical terminating event, and the consequent state whose relations are governed by the sce- nario. Similarly, the verb build is not a two–place predicate, but a fluent with one (subject) parameter. So what happens semantically when the NP is moved into subject position?

One possibility is that nothing happens, in which case a passive sentence retains the process reading of the corresponding active sentence. Another possibility is that the NP is re-interpreted as a real object, to which an ad- jective can be applied. Indeed, in English grammar this form of the passive is known as ‘adjectival passive’. For example,builtwould now be an adjec- tive, obtained from the verb build by existentially quantifying the subject position. But the upshot of this is that ‘a house is built’ now corresponds to the consequent state of the accomplishment, as indicated in Section 2.1.