The proposal

  I propose a model of the data introduced in this chapter which has the following characteristics:

• A non-uniform treatment of the range of data.
  Constructions with the syntactic configuration identified in resalt1a are treated differently than those in the syntactic configuration in resalt1b. Only the latter fall under the heading Resultatives.
• Syntactic constraints and semantics: Resultatives derive from fixed constructions.
  Following Goldberg (1995), Resultatives will be licensed by a lexicalised construction with specific syntactic and semantic characteristics.
• Conventionalisation: certain instantiations of the Resultative construction will be assumed to be lexicalised.
  Again following Goldberg (1995), this property seems to be required to account for certain idiosyncrasies in the range of data allowed.
• Pragmatics: plays an important role in determining the felicity of an instantation of the Resultative construction.
  As suggested in Section 4.4.3, and indeed by Goldberg's (1995) direct-causation restriction (see Section 4.5), some of the apparent idiosyncrasy of Resultatives can be accounted for by requiring there to be a natural causal relation between the event expressed by the main verb and the result state provided by the resultative phrase.
• Compositionality: non-Resultatives acquire their meaning through composition.
  The constructions in the syntactic configuration identified in resalt1a will be assumed to be licensed via the normal process of adjunction and shown to be interpreted compositionally, given a certain perspective on the possible behaviour of adjunct phrases (to be outlined below).
• Sense shifts: verbs of sound emission receive their motion sense from a lexical rule.

 
• Resultatives
  • Formalisation in HPSG
  • Ruling out ungrammatical instances
• The other constructions
  • Manner of Motion verbs
  • Unaccusatives
  • Verbs of Sound Emission
  • Conclusions
• Causativisation revisited
• Idiosyncrasy and Pragmatics