EXERCISES 167 corresponding queries fail Since this is not the case, it seems more appro-

priate to assign an indeterminate truth values in these cases. The upshot of this discussion is that in the situation described, (28) is true, and (29) indeterminate.

5. Exercises

EXERCISE13. Prove corollary 4. (Hint: apply definition 40.)

EXERCISE14. Discuss the difference between the sentences

(30) a. #The coin is coming up heads.

b. It is important for the catcher to know what the ball is going to do when it hits the fence.

in view of the fact that both describe largely unpredictable physical phe- nomena.

EXERCISE15. Formalize Bonomi’s ‘Multiple Choice Paradox’ and the

CHAPTER 11

Coercion

Aspectual properties orAktionsartof verbs cannot belexicalproperties of these verbs alone. For instance, John drank is certainly an activity and John drank beer is an activity as well. However, John drank a glass of beer is not an activity but an accomplishment. Therefore, the choice of the object NP (a mass term versus an indefinite NP) is a determining factor for assigning aspectual class. The next example from [109] shows that in general Aktionsartis not fully determined prior to the sentence level. The verb ‘arrive’ in

(1) Chapman arrived.

is an achievement, as can be seen from the ungrammaticality of (2) *Chapman arrived all night.

But if we choose a bare plural subject instead of a proper name as subject NP, we get the grammatical sentence

(3) Visitors arrived all night.

In the last sentence, ‘arrive’ denotes an activity. Therefore the choice of the object and subject NPs can force verbs to be interpreted in different aspec- tual classes. However, in several quarters of the semantic world, Vendler’s classification of verbs into states, achievements, activities, accomplishments and points is often conceived of as pertaining to inherent properties of a verb (or, rather verb phrase). By contrast, various forms of cognitive grammar treat aspect as a way to impose temporal structure on events, or a way to conceptualize the temporal constitution of events. In the words of Croft [21, p. 70]:

The aspectual grammatical constructions determine in part the tem- poral structure of the event it describes via conceptualization.

The difference between these ways of treating aspect is subtle but important, as can be made clear by means of an example due to Croft (op. cit.). If the verbloveis considered to be inherently stative, then

(4) *I am loving her

is ungrammatical, and cognitive linguists would agree; here the intended content has to be expressed by

170 11. COERCION (5) I love her.

But if (4) is provided with a context, it may become grammatical, as in (6) I am loving her more and more every day, the more I get to know her. This phenomenon is hard to explain if stativity is taken to be a property of the verb love, fixed in the lexicon. It seems more profitable to explain the contrast by saying that the events referred to in (5) and (6) are conceptu- alised differently, so that the event taken as an atom in (5) is considered to have stages and phases in (6). Thus, the phenomenon ofaspectual coercion comes to the fore: the potential of grammatical constructions, such as the progressive, to ‘move’ a verb or verb phrase from one aspectual category to another.

The term coercionwas introduced in [81]. Unfortunately, in the liter- ature the term ‘coercion’ does not refer to a coherent class of phenomena. Consider example (7):

(7) Pollini played the sonata for two days.

Pollini played the sonata is an accomplishment and accomplishments are generally bad with for–adverbials; so one might think that (7) is not felic- itous. Nevertheless sentence (7) can mean that Pollini played this sonata repeatedly within a timespan of two days. Thus the accomplishmentPollini played the sonata is coerced to an iterative reading. But here we cannot identify a linguistic constituent of the sentence which is responsible for this reinterpretation, as we did above in example (6). The expression for two daysdoes not force an iterative reading in John slept for two daysand nei- ther does the accomplishmentPollini played the sonata.

Worse, even cases of metonymy are often considered as instances of coercion. Thus a sentence likeThe pianists are on the top shelf interpreted with respect to a scene within a CD–shop means that the CDs of pianists are on the top shelf. The NPThe pianistsis in this context reinterpreted as CDs of pianistson the basis of a threatening type conflict with the VPbe on the top shelf.

In this chapter we will certainly not do justice to all phenomena which are grouped together under the linguistic term coercion, but we will con- centrate on those which pertain to Vendler’s classification outlined above. Moreover, we will not comment on the vast literature on reinterpretation but refer the interested reader to a selection of more recent work1.

Note that in general we cannot expect to arrive at the result of a coercive move in a compositional way2_{. To see this, consider again example (7). As} 1_{See especially Moens [81], Steedman [109] and the unpublished [108], Pulman [92],} Pustejovsky [93], Dolling [29], De Swart [26], and Egg [34].

2_{The reader will have noticed that we have paid scant attention to compositionality in} this book. One reason is that is impossible to do justice here to the recent wave of interest in compositionality inaugurated by Hodges [51]. Another reason is that the approach to

1. ADDITIVE COERCION 171