And:
>> (1)     Down with the government.
>> (2)     How about a cup of tea?
>> (3)     He sneezed the foam off his cappucino.

>I think you correctly characterize the dilemma. Certainly I think that
>the more we learn about valency patterns, the more they seem to be
>generalizable above the level of idiosyncrasy, and as you implicitly
>note, that finding is, if anything, supported by constructionism.
>
>So as I said to Dylan, the main issue seems to be whether we handle
>constructions and nonidiosyncratic valency patterns in terms of
>relations that derive sublexemes, or whether we take a prefabricationist
>view.
>
>As for examples like (1-2), these can simply be seen as constructions
>that are more restricted/specific than (3); they're not a real threat
>to prefabricationism.
## This is the kind of situation where I tend to look around for the Third
Way. There are clearly *some* idiosyncratic valency facts - after all, four
of us went public with a list of synonyms with different valencies (e.g.
LIKELY/PROBABLE, TRY/ATTEMPT). But as you say, *most* valency facts seem to
be generalisable. So maybe we need both lexicalism and prefabricationism:
a. Lexicalism: my version of WG where lexical items fully define their own
valents (except subject, which is handled at the level of Verb + various
inflections);
b. Prefabricationism: your (= And's) version of WG in which each valency
pattern defines a word class, e.g. Transitive, Ditransitive, ...  I used to
argue hard against that, but I now think you're probably right because each
such verb class actually combines its valency with other properties, mostly
semantic.
        Why not mix and match? Indeed, WG theory predicts that Prefabricationism
should arise naturally out of Lexicalism - the child starts (a la
Tomasello) learning lots of patterns as facts about individual lexical
items, then spots generalisations and raises them to a higher level whence
they can be inherited to further examples.
        Re my earlier examples:
(1,2)   WITH and HOW ABOUT remain idiosyncratic and ungeneralisable so their
valency stays at the 'lexical level' of the hierarchy.
(3)     Causative is recognised as a verb class that has causative semantics,
an object and a sharer (e.g. "blow it off"). Some verbs are linked
explicitly (in the mental grammar) to Causative, and BLOW (Dylan's example)
may be one of these; but other verbs can be added as the need arises - i.e.
you want to describe a scene in which I sneeze while drinking a coffee etc
etc. and the best fit with available resources is by combining SNEEZE with
causative syntax and semantics.
        But I feel all this is very different from the 'prefabricationism' that we
find in GB, where the whole of sentence structure is rigidly prefabricated
(by machinery that nobody bothers to spell out). What I'm suggesting is
still strictly lexicalist in that everything is projected from single
words. What's at stake is simply how general the rules are. A far cry from
prefabricating a whole sentence structure and then inserting lexical items.

Richard (= Dick) Hudson

Phonetics and Linguistics, University College London,
Gower Street, London WC1E  6BT.
+44(0)20 7679 3152; fax +44(0)20 7383 4108;
http://www.phon.ucl.ac.uk/home/dick/home.htm