This chapter has reviewed various proposals for the modeling of the resultative construction and suggested criteria which any model must meet in order to adequately account for the relevant data. I offered a hybrid model, combining insights from the different proposals and from related work, to account for the full range of data.
I have argued, in direct contrast to the proposal put forth by Levin and Rappaport Hovav (1995), that there is no underlying syntactic difference between the verbs that appear in the different kinds of ``resultative'' constructions. I showed that the evidence presented by in favor of the syntactic difference is weak. Furthermore, their account requires postulation of a lexical rule which applies only to two verb classes and does not follow from any general principles. The syntactic shift and corresponding semantic shift induced by the lexical rule requires addition of an explicit directional phrase, but I have shown that discourse context can trigger at least the semantic shift and that therefore this rule is overrestrictive. The assumed change in underlying argument structure also potentially creates problems for other components of grammar such as linking rules.
This evidence therefore calls into question the analysis of manner of motion verbs on a directed motion use. More strongly, the idiosyncrasy of the resultative data and the differing interpretations associated with the various forms of the resultative construction suggests that a purely syntactic explanation is impossible. I showed that semantic differences exist between the different forms and that pragmatic coherence is a critical factor in determining their felicity. A single syntactic constraint is too restrictive, treating as ungrammatical examples which are merely infelicitous independent of a context which establishes the appropriate semantic relations, and grouping together constructions which actually display entirely distinct syntactic and semantic behaviour.
Instead, I argued that the differing behaviours which verbs display with respect to these constructions follow from the specific semantic properties of the various construction types and the lexical semantics of the verbs themselves. I motivated this through an overview of the interpretations associated with various forms of the construction, a discussion of the syntactic flexibility of those forms, and an analysis of the lexical semantics of both the verbs and the ``resultative'' phrases appearing in the different forms.
The proposals in this chapter draw on an insight identified by Wechsler () that certain verbs encode an inherent endpoint in their lexical semantics. Although I do not support Wechsler's contention that there are verbs for which this endpoint is optional background information, preferring a more general explanation for these cases in terms of lexicalisation and pragmatic constraints on the construction, I showed that the behaviour of certain classes of verbs, specifically manner of motion verbs and unaccusative verbs, does stem from the semantics they express.
The work by Goldberg (1995) provides a fundamental contribution to the proposals here: her introduction of independent form-meaning pairs to a grammar allows phenomena which have no clear generative basis to be captured. This approach seems to derive from an acknowledgement of the highly conventional nature of language use and the largely fixed nature of the conventions. Acceptance of this insight from Construction Grammar led directly to a view of resultative constructions in terms of a form-meaning pair. This view is supported by the fixed syntactic and semantic behaviour of the construction, as discussed in Section 4.7.2.
In Section 4.6 I identified criteria which any model of the resultative construction must satisfy. I review them here, pointing out how the proposals in this chapter address each one.
The analysis proposed, in sum, relies on generative mechanisms which have been shown to have wider applicability within grammar, as constrained by conventionalisation of language use and general pragmatic principles. I have taken advantage, where possible, of the compositional nature of much of language yet allowed for the apparent existence of non-compositional units within language. In doing so, I have been able to provide a model of the resultative construction and similar constructions which accounts for the syntactic and semantic differences between them and some of their idiosyncrasy. I have not had to involve the unaccusative/unergative distinction or assume semantic constraints specific to the constructions, but have shown how their behaviour follows from principles more generally applicable to the lexicon and discourse construction.