What does it mean to say there is such a conventionalized unit in the lexicon? It means that speakers can“abbreviate”
their utterances by leaving this unit unexpressed and can trust hearers to “reconstruct” it in the course of
interpretation.In some cases of reference transfer such as theham sandwichexample, the hearer is cued by the anomaly
of the simple (or literal) reading. In other cases such asTosca wasn't in good voice tonight, the hearer has to rely on
contextual cues outside the sentence itself.
This accountseems tometoreflectintuitionprecisely. Itsays thatreferencetransferis“pragmatics,”inthesensethatit
is part of contextualized interpretation but not part of the utterance. On the other hand, it is also part of
language—part of grammar—in the sense that (a) it is conventionalized and (b) it is integrated into conceptual
structure just as if it were a word orfixed phrase such asperson who is contextually associated with.... Harking back to the
discussionofsection 9.7: only ifweinsistonan ideologicallyfixed boundary betweensemanticsand pragmatics is such
an account problematic; taken on its own terms it seems to me perfectly satisfactory.
This is a typical,ifunusuallyclear, case ofenrichedcomposition.I wanttobrieflymention oneother characteristiccase
(see Jackendoff 1997a: ch. 3 for several more): the widely discussed phenomenon ofaspectual coercion(Talmy 1978;
Verkuyl 1993; Jackendoff 1991; Pustejovsky 1995; and many others). Compare the following sentences.
(20) a.Sa mslept until the bell rang.
b. Sa mju mped until the bell rang.(20a) implies that Sam's sleeping was a single continuous process, whereas (20b) implies that Sam jumped repeatedly.
What makes the difference?
The insight behind all approaches to this proble mis thatuntil(along with various other expressions likefor an hour)
expresses a temporal bound at the