Heard on the Today Show

[Arnold M. Zwicky]

On Mar 26, 2008, at 7:19 AM, Wilson Gray wrote:

> "She can't wait to vote, but _she refuses to tell us for whom she's
> going to_." [1]
>
> I can't find anything wrong with this sentence, but it's different,
> somehow, even ignoring the stricture that ending a sentence with a
> preposition is something up with which one should not put.

the "to" here is not a preposition, but the infinitival "to", with its
complement VP omitted by Verb Phrase Ellipsis (VPE); this is the
configuration i've called "stranded _to_" (see "Stranded _to_ and
phonological phrasing in English, Linguistics 20.3-57 (1982)), and
it's now entirely standard.

> You kinda
> expect:
>
> "... she refuses to tell us for who(m)." [2]

an instance of what's called Sluicing in the literature -- an ellipsis
construction in which an initial WH-containing phrase is left as a
remnant of an embedded interrogative clause.  [2] is Sluicing with a
fronted P.

> "... she refuses to tell us who(m) she's going to vote for." [3]

this has the full (non-elliptical) embedded clause, with stranded P in
it.  the fronted P variant is

"... she refuses to tell us for whom she's going to vote." [3']

(grammatical but, to my ear, awfully stiff, though people who hate
stranded Ps will opt for it).

> "... she refuses to tell us who for." [4]

another Sluicing example, but with a stranded P.

the original, [1], is, in my judgment, flagrantly ungrammatical.  (it
might have arisen as a blending of VPE (with stranded P) with Sluicing
(with fronted P).)  the problem would at first seem to be that what's
omitted is just a V, not a whole VP (as VPE would require); the "for"
is still around.  so [1] is roughly as bad as

"... she refuses to tell us who(m) she's going to for."  [1']

(with stranded rather than fronted P).

but that can't be the whole story, since a version without the "for"
is just as bad:

"... she refuses to tell us who(m) she's going to."  [1"]

well, maybe [1"] is bad because there is no "for" in the antecedent
material "She can't wait to vote", so the P can't be omitted in
ellipsis.

maybe the problem with [1] is that the ellipsis (the omitted material)
itself contains a gap:

... she refuses to tell us _for whom_ she's going to [vote ___]

(ellipsis in brackets, with the underline indicating the gap and the
material "extracted" from the gap site in italics).  now, the ellipsis
in VPE can contain a gap, as in this example from my VPE files:

   Doesn't DOMA say the Feds don't have to pay attention to any state
marriage laws they don't want to [pay attention to ___]?  [5]

(again, ellipsis in brackets).

but [5] has a gap of relativization, while [1] has a gap of
interrogative-WH-extraction, and these two types of gaps are different
in a number of ways; in particular, gaps in relative clauses function
like anaphors -- in effect, like zero pronouns -- while gaps in
interrogative WH clauses work rather differently.

i'm not sure what the limits are on relativization gaps in VPE, but
the constraint on WH extraction gaps seems general, as in these
unfortunate examples:

   I know you have to finish some work; explain to me _what_ you need
to [finish ___].
   (ok, of course, on the reading: explain to me what you need to
[explain to me].

   I have to finish some work now; _what_ I need to [finish ___] is an
article for Language.

there's probably some literature on this stuff; i'll ask around.

arnold

------------------------------------------------------------
The American Dialect Society - http://www.americandialect.org