When word meanings vary under presure from the contsructions that they find
themselves in (a proces called coercion) you can attribute the variation to
either the construction (in which case the notion 'construction' has to
have a real status in your theory of grammar) or to discourse.

One kind of coercion is mass-count coercion as in (1).

(1) we're having rabbit for dinner (from Jackendoff 1997)

This usually comes with extra bits of meaning (ie rabbit without fur,
intestinal tract, etc -- ie edible bits of rabbit). You can do this with
lexical polysemy (there's a count and a mass entry for rabbit in the
lexicon; or with some kind of count-to-mass coercion function. The function
could be claimed to work under discourse effects (convert a count animal
noun to a mass one if the thing referred to is a food item) although
Goldberg (1995 book) claims that coercion only works when licensed by a
construction. (So on a Goldberg a/c, you'd have to say that HAVE X FOR
DINNER was a construction, and whatever you replaced the x variable with
could be mass even if it's normally a count N.

I'm toying with the idea that the count/mass distinction is really a
referential one, not identified as part of sense in a lexical entry
(except, perhaps, as a preference). Any comments?

Nik.

Nikolas Gisborne
http://www.hku.hk/english/staff/gisborne.htm




%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%