I
will show, however, that a change of state is not necessarily
causative and the distinction between directed manner of motion
constructions and other resultative constructions is precisely that
the former express non-causative changes of state while the latter
express causative changes of state.
Let us consider two similar examples. res23 is a directed manner of motion construction and res24 is a change-of-state resultative involving the same verb.
John walked to the store. John walked his feet sore.
The account of the interpretation of the resultative
construction is based on compositionality, as in most semantic
theories, with the additional assumptions that an eventuality which
expresses a process followed by a state is interpreted as an
accomplishment (cf. Moens and Steedman
moens_steedman:88), and that an accomplishment always
describes causative changes of state (:54). The meaning of
resultative constructions consists of the basic meaning expressed by
the main verb, which I will represent in the case of res23 as
walk(John) in line with predicate logic representations, and
the meaning expressed by the resultative predicate, approximately
at_the_store(John). The
process expressed in the main verb is composed with the state
expressed by the resultative phrase to generate an interpretation in
which the main event causes a change to the state conveyed by the
resultative phrase. Sentence res23, then, is interpreted as
meaning essentially John's walking causes the state of John
being at the store on the account (cf. Dowty
dowty:79). Similarly, sentence res24 expresses
John's walking causes the state of John's feet being sore.
One identifiable weakness of this account is the failure to explain the distinction in meaning between motion sentences containing directional phrases which explicitly include the endpoint of the path, and those containing directional phrases which don't.
John walked to the school. John walked towards the school/along the river/around the park.
In res41, in contrast to res40, John does not necessarily reach the school. This sentence can therefore not be considered a true resultative: there is no change of location to a specified point. Rather, it expresses that the activity of walking took place along a particular path. do not make any distinction between these two types of directional phrases. A uniform treatment of these directed manner of motion constructions as resultative constructions ignores the possibility of non-resultative interpretations for them. fail to explicitly address the issue of how and why certain directional phrases are interpreted as result predicates, while others are not.
There is in addition a distinction in the causative nature of the unergative resultative construction, as in res24, in comparison with the unaccusative resultative construction. The former requires a causative interpretation, while the latter explicitly does not. This additional required component of meaning of the unergative resultative construction becomes evident upon examination of cases in which the causation component is incompatible with the meaning of the verb, and which are as a result infelicitous. The problem can be seen clearly in comparison with a set of sentences for a non-agentive verb, as in res25-res26.
The bottle floated to the bridge. ??The bottle floated itself broken.
The interpretation of res25 as would be predicted by is
The bottle's floating caused it to be at the bridge, and that
of res26 is The bottle's floating caused it to be broken.
These two sentences seemingly should be equally felicitous on the
given interpretations, but are not. I suggest that this is due to the
association of different interpretations with these two constructions:
res25 actually means The bottle is at the bridge as a
result of it floating while res26 should mean The
bottle caused itself to be broken by floating. The unergative
resultative construction expresses a relation of causation
(cause-eff-rel) between the event in the main clause (headed by
the verb) and the state expressed by the resultative phrase, in which
the subject of the main verb must be interpreted as the
causer (i.e. the actor, the initiator of the
causation). This is in contrast to the unaccusative resultative
construction which does not express a causative
relation. Hence
res25 is felicitous and sentences like res42-res43 are
interpreted, respectively, as The river was solid as a result of
it freezing and The vase was in pieces as a result of it
breaking.
The river froze solid. The vase broke into pieces.
A verb which instantiates the unergative resultative construction, then, must be compatible with a causative interpretation. As the interpretation of res26 as dictated by the construction is semantically incompatible with the specifics of bottles and floating (how can a bottle floating cause breakage?), the sentence as a whole is infelicitous. On the other hand, the interpretation of res24 as John caused his feet to be sore by (John) walking is perfectly felicitous. This causation is a necessary component of the meaning of the unergative resultative -- if the semantics of a particular verb or verb/argument phrase are incompatible with that meaning, the construction is infelicitous.
It could be argued that what is reflected by the contrast in the data for these two constructions is not a causativity relation but rather a volitionality constraint which applies to the unergative resultative construction but not the unaccusative resultative construction. The non-volitionality of bottles with respect to floating res26, etc. could then explain the infelicity of these events in the unergative resultative construction. However, consider the data in res45-res48 below.
John cried himself to sleep.
John laughed tomato soup up his
nose.
The ball squashed the tin flat.
In res45 and res46, John is not volitional with respect to the crying or the dreaming (if he were volitional, it should make sense to say #John accidentally cried/laughed but it does not) yet he does instigate a change of state to the state expressed in the resultative phrase by being the agent of the event which leads to the change of state. Similarly in res48 the agent of the main events which cause the change of state are clearly non-volitional as they are inanimate.
The difference in interpretation between these two kinds of
constructions becomes even more clear if we consider a verb which
seems to be possible in both the unaccusative resultative construction
and the unergative resultative construction, as in
res28. On the account, such a verb is impossible since
the lexical rule forces the manner of motion verb swing to be unaccusative in the
presence of the directional phrase apart, ruling out its use
in the unergative resultative construction. Both the unaccusative and
unergative resultative sentences, however, seem to be equally
felicitous in this example. Note that my judgement on the unergative
resultative construction for these verbs differs from that which I
assume Levin and Rappaport Hovav would assign, given the * they assign
to the sentences in res17 I cited in above.
The clowns swung apart. The clowns swung themselves apart.
The interpretations of these sentences are closely related, and in fact the meaning of res28a is entailed by the meaning of res28b. The latter conveys the meaning of the former, with the additional suggestion that the clowns are actively swinging in such a way that they end up apart. That is, res28b means The clowns cause themselves to be apart by swinging, and res28a simply means The clowns are apart as a result of their swinging. Intuitively the unergative resultative construction conveys a certain meaning which the unaccusative resultative construction does not. The former could be used in a context, such as res49, in which the clowns are not agentive with respect to the swinging, and can therefore not be construed as causers, while the latter cannot.
The director pulled the puppet strings and
the clowns swung apart. *the clowns swung themselves apart.
Consider also the contrast in res51.
The clowns swung over the net. The clowns swung themselves over the net.
res51a is highly ambiguous due to the nature of over: the PP could be expressing either the location of the swinging event, a point on or the direction of the path along which the swinging occurs, or a result phrase. On the other hand, res51b is not ambiguous. The PP in that case can only be interpreted as a result phrase.
The treatment of these two constructions as identical semantically
fails to capture the distinctions between them. In addition, the
meaning of the unergative resultative construction does not seem to
lend itself to a solely compositional treatment, since the element of
causation conveyed by these contructions cannot be derived solely from
the constituents of the sentence (the constituents parallel that in
unaccusative resultative constructions, yet the meanings of the two
constructions differ). An adequate treatment of the interpretation of these
constructions is missing in the account of the resultative
constructions. I will introduce a possible solution to this problem
in Section 4.6, on the basis of Jackendoff (1990)'s
adjunct rules and Goldberg (1995)'s analysis of
these PPs in Construction Grammar, in which certain syntactic
configurations have a particular semantic content not fully
predictable from the semantics of their components.