Wechsler () suggests an analysis of resultatives which does not rely on the syntactic encoding of unaccusativity/unergativity. In fact, he explicitly rejects this encoding as an adequate basis for explaining the distribution of data in the resultative construction. Instead, he draws a distinction between types of resultative constructions which parallels the control/raising distinction for complement-taking verbs (see e.g. Pollard and Sag 1994, ch. 3). Wechsler argues that differences between the semantic restrictions that control resultatives impose on their complement resultative phrase and those that are imposed by raising resultatives can be used to account for the resultative data. He provides a formalisation of the distinction between these resultative types in HPSG, and shows how his analysis can be used to account for a range of resultative data.
Briefly, control resultatives are resultatives in which the
resultative phrase is predicated of an argument of the main verb,
w1, and raising
resultatives are resultatives in which the resultative phrase is
predicated of something which is not an argument of the main verb,
w2.
John hammered the metal flat. Sally jumped free of the car. The joggers ran their Nikes threadbare. The audience laughed the speaker off the stage.
The main distinction between these two types of resultatives according to Wechsler's analysis is that control resultatives place semantic restrictions on their complement resultative phrases, while raising resultatives do not. Specifically, the resultative predicate in a control resultative must express a ``canonical or generic result (or intended result) of the action denoted by the verb'' (Wechsler , p. 2). The nature of this canonical result is encoded in the lexical semantics of the verb. The resultative phrase is required to unify with the encoded result through the mechanisms of the formalisation. In contrast, raising resultatives do not specify any canonical result or other specification in their lexical semantics and therefore do not constrain the kinds of resultative phrases they can appear with. This distinction is supported on the basis of data such as that in w3, which shows control resultatives to be acceptable only with result phrases that are compatible with the canonical result of the action expressed by the main verb, and w4, which shows that the result phrases appearing with raising resultatives do not need to have any connection to the meaning of the main verb.
Robert ran clear of the car/*exhausted. [Wechsler , p. 2, (6a)] John hammered the metal flat/*safe. [Wechsler , p. 2, (6c)]
Olof painted himself into a corner. [Wechsler , p. 2, (7b)] We laughed ourselves silly. [Wechsler , p. 2, (7d)]
Wechsler captures his insights in HPSG by utilising the background feature in verbal synsems which encodes pragmatic information relevant to the proper use of the word. This feature has a value which may contain the feature telos, which in turn has a value representing the result state or endpoint of the event conveyed by the verb. Control resultatives have the telos value specified while raising resultatives do not. Resultatives are licensed via a lexical rule which adds resultative phrases (APs or PPs) to a verb's subcategorisation list and requires the unification of the content value of the resultative phrase with the telos value of the verb. If that telos value is unspecified, then no restrictions are placed on the semantic content of the resultative phrase.
Wechsler's use of a unification-based grammar formalism such as HPSG allows him to easily specify interactions between the semantics of the main verb and the semantics of the result phrase in the resultative. In particular, control resultatives can indicate in their lexical semantics precisely what role their arguments must play in the canonical result state they are associated with. This is achieved via structure-sharing between the elements in the core semantics of the verb and elements in the value of the telos feature. It ensures that resultative phrases are predicated of the appropriate verbal argument in control resultatives, and enables Wechsler to account for the data in w5 (Wechsler , p. 11, (29a-b)), in which the result phrase is predicated of the subject of a transitive verb. These sentences would violate L&RH's Direct Object Restriction in DOR, as the resultative is not predicated of the direct object as required by the DOR. would therefore be unable to account for these sentences.
The wise men followed the star out of Bethlehem. The sailors rode a breeze clear of the rocks.
Syntactic issues associated with the resultative construction are mainly handled by the mechanisms of HPSG of which Wechsler takes advantage, with the addition of a lexical rule which adds the non-argument NPs which serve as the subjects of resultative predicates in raising resultatives (the Raising Rule). This addition, in conjunction with the control/raising resultatives distinction, leads to an account of the appearance of fake reflexives for resultative constructions with many unergative intransitives, w6, in contrast with the lack of reflexives for other unergative intransitives, specifically the manner of motion verbs w7, which behave as control resultatives.
*The dog barked hoarse. [Wechsler , p. 4, (11a)] The dog barked itself hoarse. [Wechsler , p. 4, (11b)]
In w6a, Wechsler argues, hoarseness cannot be a canonical result of barking and thus this result state is compatible only with a raising use of bark. In this case, the Raising Rule is triggered and the requirement for a post-verbal NP is added. Semantic considerations will determine that the NP is token-identical with the subject, and syntactic constraints will ensure that it appears as a reflexive. In contrast, all manner of motion verbs optionally encode a location in their telos value. Thus the result state in w7 is compatible with the canonical result of running, and so this is a control resultative which does not trigger the addition of any non-argument noun phrases. Instead the single argument of run is structure-shared with the subject of the resultative predicate due to unification between the verb's telos value and the content of the result phrase.
This treatment does not give unaccusativity a role in the modeling of the resultative construction, thereby avoiding many of the pitfalls of the Levin and Rappaport Hovav approach. In particular, the treatment avoids any notion of a lexical shift in the underlying syntactic classification of certain verbs and is therefore not subject to many of the criticisms given in this chapter. There are, however, other difficulties, to which I now turn.