next up previous contents
Next: Logical Metonymy and Pragmatics Up: An Alternative Previous: Conclusions about Logical Metonymy

The Representation

  To sum up, the behaviour of the eventive verbs in logical metonymy constructions can be described as follows:
I.
Aspectual verbs are control verbs in the logical metonymy constructions.
II.
The complement NP in the logical metonymy construction must refer to a bounded entity (from Godard & Jayez).
III.
The complement NP of an aspectual verb in the logical metonymy construction must refer to an object.
IV.
Default interpretations for the logical metonymy construction stem from either the agentive role of the complement NP's qualia structure, or, if present, the telic role of that qualia structure. These default interpretations thereby derive only from lexical specification of information in qualia structure for particular NPs. For some eventive verbs (e.g. begin on), no default interpretation is derived from lexical information. This difference must be lexically captured.

   figure24044
Figure 5.2: Part of the lexical entry for book, specifying default telic and agentive events

On the basis of the above observations, and building on the representation established in Chapter 2, I will propose two lexical entries for coercing uses of begin in this section. Other aspectual verbs will be represented analogously. A representation for nouns as in Figure 5.2 is assumed, in which qualia structure is distinct from other nominal semantics (following Lascarides & Copestake 1995). Values are specified as default, necessary in an inheritance framework when subtypes of a lexical type require distinct qualia values, via the slash notation introduced in Section 2.6: information to the left of the slash (if any) is indefeasible and that to the right is defeasible. I use these typed default feature structures (TDFSs) to allow for the treatment of the influence of pragmatics, to be discussed further in Section 5.5.4, because I need the lexical defaults to persist into the logical form of the interpretation of a sentence so they are available for interaction with pragmatic reasoning and can be overridden through that reasoning. I further assume that the roles of qualia structure representations for different NP forms are ``computed'' appropriately using the qualia structure of the head noun.

I also assume a type hierarchy for begin which corresponds to that proposed in Copestake & Briscoe (1995) for enjoy: a general begin type has subtypes corresponding to its different complement forms, each of which has a specific semantic definition. The uses of begin in various constructions must be distinguished from one another, as has been discussed previously in this chapter, because each incorporates the semantic contribution of its complement in a different way. Thus a polymorphic treatment of metonymic verbs has been adopted, to allow for different constraints to be implemented for each use of begin.

The missing event in a begin logical metonymy can either come from the agentive or telic roles, as specified in Figure 5.3. The two lexical entries for the coercing forms of begin define the behaviour of this verb on metonymic uses.

   figure24283
Figure 5.3: Coercing forms of begin: One entry picks out agentive role, the other telic role

I argued in the previous section that the interpretation for the missing event in a {begin on + NP} construction is filled in by context. As a result of the context-dependent properties of this form, no defaults/conventions specified in the lexicon are accessed via unification for this phrase; pragmatics must provide all the clues to an interpretation. The lexical entry in Figure 5.4 captures the proposal for the behaviour of begin in this construction.

   figure25043
Figure 5.4: Lexical entry for begin in a {begin on + NP} construction

   figure25260
Figure 5.5: Lexical entry for enjoy on a coercing use

The lexical entry for the coercing form of enjoy is shown in Figure 5.5. When it is combined with the representation of an NP argument which has a telic role specified, the default unification framework ensures that the default information specified in the qualia structure for the NP remains as default in the representation of the metonymic event. Thus the metonymic event predicted for enjoy the book would be act_on_pred/read. This allows for the possibility that read can be overridden pragmatically.

If no telic role is specified for an NP, as will be the case for many potential arguments of enjoy, we must account for the fact that when enjoy is combined with such an NP a default interpretation is still arrived at. I propose that the information which comes from the lexicon into the pragmatic component will be as represented in Figure 5.5 except that the metonymic event will simply correspond to act_on_pred, without any default event coming from the NP. Any default interpretation for an {enjoy + NP} construction with these NPs is then determined in the pragmatic component, by world knowledge. Clearly world knowledge includes information about what kinds of events are associated with particular objects. Some of these events will be more strongly associated with the objects than others. It is these strongly associated events which will provide the default interpretation for the act_on_pred metonymic event.


next up previous contents
Next: Logical Metonymy and Pragmatics Up: An Alternative Previous: Conclusions about Logical Metonymy