hypercorrection/grammatical evolution (was: murky days)

[Laurence Horn]

At 4:54 PM +0000 3/28/02, Lynne Murphy wrote:
>Yes, but might hypercorrection lead to syntactic reanalysis?  Perhaps this
>will evolve so that all relative 'who's become 'whom's, for example.  (Just
>like the plural form seems to be evolving to apostrophe-s, but we've been
>down that road enough times...)
>
>Then there's the old favorite "...and I", never "...and me" in my mother's
>variety (which seems like a decent argument for Construction Grammar).  It
>sort of brings into question the idea of 'impossible lexical items'.  Was
>it Jim McCawley who used the example of "my uncle and" as an impossible
>lexical item?  So "X and" is an impossible lexical item, but we've got
>definite examples of (phrasal) lexical items of the form "and X" ("and I",
>"and so forth", "et cetera")--which I suppose has to do with English being
>a head-initial language?
>
cf. the currrent hit Mexican movie (now imported into the U.S.), "Y
tu mama tambien"  ('And so's your old lady').
But there are also restaurants that have a menu header "coffee and"

Notice that again we're talking about single words here either.  I
thought McCawley's claim had more to do with blocking lexical items
whose derivation would violate syntactic constraints (in particular
Ross constraints), e.g. a word that would mean 'to kiss a girl who is
allergic to' or 'to drink coffee and'.  As others pointed out, this
is hard to test because of different paraphrase possibilities.  One
example I remember being brought up was 'to cuckold', which
could--although it doesn't have to--be regarded as derived from 'to
have sex with the wife of', thereby violating the complex NP
constraint.

larry