I begin by precisely defining the data of interest. Consider the following sentences:
John began reading/to read the book. John began the book.
John enjoyed reading the book. John enjoyed the book.
In both of these sets of sentences, the two sentences express the same
meaning although in the (b) sentences no reading event is explicitly
mentioned.
The (b) sentences exemplify logical metonymy, for
which more meaning arises than is directly attributable to the
sentential components. That is, the interpretation attributed to
these sentences includes an event which is not explicitly introduced,
in contrast to the (a) sentences in which the event is specified in
the VP complement.
This phenomenon involves the use of a noun phrase to suggest an event
associated with that noun phrase -- it is metonymy (e.g.
Nunberg 1978) in the sense that one phrase is used
in place of another (the noun phrase in place of the full eventive
verb phrase) and, under the analysis I will adopt, logical in
that it is triggered by type requirements which a verb places onto its
arguments (in the examples above, the verb requires an eventive
complement). It occurs for certain verbs which have alternate
syntactic complement forms with only a single (default) semantic
interpretation.
We will see in this chapter that the characteristics of logical metonymy can only be captured through an interaction of syntax, semantics, and contextual influence. This interactivity will be shown to have implications for the representation of lexical information, and the balance between the lexicon and pragmatic reasoning in interpretation. I will present extensive corpus evidence for a conventional, lexicalised basis for the interpretation of logical metonymies, and will argue that a purely pragmatic treatment of this phenomenon (e.g. Hobbs 1993) cannot accurately predict grammaticality judgements on the data.
The verbs which will be examined in this chapter are of interest because they can appear in various syntactic constructions which ultimately may be assigned the same interpretation, while each construction is subject to specific constraints. These constraints are either syntagmatic in nature, or reflect idiosyncrasies in the interaction of word meaning with pragmatic reasoning. These facts highlight the importance of the lexicon as a repository of knowledge about verb use and their interpretation in distinct linguistic contexts. In this chapter I will identify the different constructions and introduce data which exemplifies a range of syntactic, lexical semantic, and discourse-level constraints on the constructions. I will argue for an analysis of the phenomena which hinges on lexical semantic information. I will formalise my proposals through the introduction of lexical entries which build on the representation developed in Chapter 2, and the account of the felicitous and infelicitous data will exploit the interaction between discourse reasoning and the information in such lexical entries.
The so-called eventive verbs (aspectual verbs like begin and finish and verbs like enjoy, which all require eventive complements) can take on a huge range of meanings in the logical metonymy construction. In each of the sentences in beg91, began takes on different possible meanings which vary with the particular noun phrase which is the complement of the main verb. Similar patterns hold for the other eventive verbs.
John began the book. (began = began reading; began writing) John began the sandwich. (began = began eating; began making) John began the beer. (began = began drinking; began brewing) John began the cigarette. (began = began smoking)