(24)a.
A: When does the train leave?
B: (At) 6:30.
b.A: How do you keep fro mgetting nervous?
B: By drinking.
Speaker B's utterancesinisolationare ill-formed syntacticallyand semantically. Inearlygenerativegrammar, onewould
have claimed that these phrases are derived by deleting the rest of a full answer. Such a solution, however, has been
unacceptable since the early 1970s, when such wholesale deletion was rejected. The alternative is to let the grammar
generate the mas is.^207 Speaker A must interpret them by incorporating conceptual information borrowed from the
preceding question. Thus in a strict sense this too is enriched composition: the reply has conceptual structure that
comes from neither its words nor its syntactic structure. We discuss some related cases in section 12.5.
I mention this case only to open the door. It remains to be seen to what extent other discourse phenomena might be
integrated into the framework of enriched composition, including important phenomena such as conversational
implicature.
12.3 The referential tier
A majorconcern offormal semanticsis absent fromthesemantictheory workedoutso far inthischapterand thelast:
the existential or referential claims a sentence makes about the entities named in it. I am going to take a somewhat
novel approach to this aspect of meaning without (I hope) discarding the traditional insights. This approach, based
most directlyon theworkof Piroska Csuri (1996), might betaken tobe a versionof Discourse Representation Theory
(Kamp and Reyle 1993) or of Dynamic Semantics (Groenendijk et al. 1996); it also bears some resemblance to
Fauconnier's (1985) theory of“mental spaces,”within cognitive grammar.
Csuri's basic idea is that, just as phonology breaks naturally into semi-independent tiers, so does conceptual structure.
The semantic structures presented so far in this chapter and the last constitute thedescriptive tier, the organization of
conceptual functions, arguments, and modifiers. But conceptual structurealso containsareferential tier, whichorganizes
the referential claims about the
394 SEMANTIC AND CONCEPTUAL FOUNDATIONS
(^207) An intermediate possibility would be that they have the full syntactic structure of a sentence, but almost every node is phonologically empty. This solution, proposed in
Jackendoff (1972) , is in a way forced by the syntactocentric model, where interpretation must be a consequence of syntactic structure. However, there is no evidence for
such syntactic structure other than the interpretation. Therefore the parallel model should do without it.