She danced/swam free of her captors. They slowly swam apart.
*She danced/swam herself free of her captors. *They slowly swam themselves apart.
*She danced/swam sore. *Don't [you] expect to swim sober.
She danced her feet sore. Don't [you] expect to swim yourself sober!
From the treatment of resultatives in terms of the Direct Object Restriction in DOR it follows that the verbs in res16 must be given an unaccusative analysis. In these sentences there is no postverbal NP of which the change of location resultative phrase can be predicated. Instead, the surface subjects are interpreted as entities undergoing a change of location. Under the DOR, these surface subjects must then be underlying (deep structure) objects. That is, the verbs in these sentences are unaccusative. In contrast, the required reflexive postverbal elements in res19 (cf. res18) suggest that the relevant verbs in these sentences have an unergative analysis and therefore do not provide the resultative phrase with a subject in the underlying argument structure.
The main verbs of the sentences in res16-res17 are identical to those in res18-res19, yet the former set of examples are felicitous only when they appear without a (reflexive) postverbal element, while the reverse holds for the latter. Comparison of these two sets shows that, in addition to this syntactic difference, the resultative phrases differ: in the first set the resultative phrases express a change of location, while in the second set, the resultative phrases express a change of state. identify this distinction in resultative phrase type as the explanation for the syntactic differences: manner of motion verbs behave like all other unergative verbs in the resultative construction, except when the resultative phrase expresses a change of location, in which case they behave like unaccusative verbs. therefore propose that the change of location resultative phrase is added to the subcategorisation frame of manner of motion and sound emission verbs by a special lexical rule which shifts the sense of the verb to a directed motion interpretation and simultaneously changes the argument structure of the verb to be unaccusative. This is in contrast to their treatment of all other resultative constructions, for which the resultative phrases (regardless of their type) are not syntactically licensed by a lexical rule, but instead on the basis of a more general process resulting from the identificaiton of event positions in the argument structures of the verb and the resultative phrase (see Levin and Rappaport Hovav 1995:53).
This analysis accounts for the data above. The sentences in res17 are ruled out because the verb is in the context of a directional phrase and so must have unaccusative structure due to the lexical rule which licenses the directional phrase. The single underlying argument position cannot be filled twice. The sentences in res18 in contrast do not appear with a directional phrase. The verb in these cases is unergative, and hence the sentences are ruled out because there is neither an underlying object nor an overt postverbal element of which the resultative phrase can be predicated.
Nonetheless, the account has a flaw which can immediately be identified. The extended sense of manner of motion verbs is linked to an explicit directional phrase through the lexical rule, causing the sense shift to be dependent on the presence of this phrase. It is, however, possible for a manner of motion verb to acquire a directed motion sense through context, without an explicit directional phrase appearing as one of its arguments. In res30, for example, the sentence He walked means He walked to the store.
John had to go to the store to buy some flowers for Mary. He walked, arriving just before it closed.
This example implies that the directional phrase requirement imposed by the lexical rule must be overridable by the discourse context, or that the unergative-unaccusative shift can be induced by discourse context. This fact calls into question the dependency of the sense of the verb on a lexical rule and, as a result, the association of the verb's underlying argument structure with a particular syntactic frame. I will return to this issue in Section 4.3.1.
In addition, there is a problem with using resultative data to support an analysis of the directed manner of motion verbs as unaccusative in that the analysis presupposes a particular treatment of the resultative construction. The effect is that the argumentation concerning resultatives is rather circular. The arguments in favor of the proposal for the syntax of the resultative construction depend in part on the distinct behaviour of unaccusative and unergative verbs with respect to this construction. In particular, the Direct Object Restriction is justified in large part on the basis of unaccusative data. Then in order to bring the directed manner of motion resultative data in line with the analysis required under this restriction, must assume that these verbs are unaccusative. So in fact the data do not provide independent evidence for the unaccusativity of these verbs; rather, their unaccusativity follows from the analysis of resultatives assume. A different treatment of resultatives may not require these verbs to be unaccusative.