failures of parallelism

[Laurence Horn]

At 12:37 PM -0700 7/5/04, Arnold M. Zwicky wrote:
>i've been collecting examples where parentheticals, postnominal
>modifiers, and coordination set things up for what are, on reflection,
>errors.  these include cases of "determination by the nearest" (in
>agreement and government) and several kinds of failures of parallelism
>in coordination.  here's a new type of example, involving the
>triggering of negative-polarity "any":
>
>(NYT Week in Review, 7/4/04, p. 6, "Strange Bedfellows: 'Imperial
>America' Retreats from Iraq" by Roger Cohen)
>-------
>It was a low-key exit, reflecting problems that Mr. Bremer, and perhaps
>any American, could not resolve.
>-------

I actually noticed this one.  But for me, it's not terrible, since
there's a free-choice reading for the "any", so the offending clause
below ("Any American could not resolve these problems").  A lot worse
is the variant I've not infrequently noticed of the form

...that Bremer, and perhaps no American, could resolve.

Clearly, in Jerry Cohen's terms, a syntactic blend of "Bremer could
not resolve..." and "No American could resolve".  But on the same
footing with, say,

"*one of the best, if not THE best, {resolution/resolutions}..."

>
>this has a parenthetical conjunct NP "and perhaps any American", with a
>negative-polarity "any"

not necessarily, as noted.

Larry

>  in it that is apparently triggered by the
>negative VP "could not resolve".  the problem is that when we unpack
>the coordination into two relative clauses, we get:
>   problems that Mr. Bremer could not resolve
>       and
>   problems that perhaps any American could not resolve
>the second of which is ungrammatical, or only very marginally
>acceptable, because the "any" precedes the trigger "not":  it's the
>same problem as
>   *Any American could not resolve these problems.
>
>fixing the second to
>   problems that perhaps no American could resolve
>no longer allows it to combine with the first in such a way that two
>subject NPs are coordinated:
>   *problems that Mr. Bremer, and perhaps no American, could (not)
>resolve
>
>there *are* solutions, with "conjunct raising" (nonconstituent
>coordination) --
>   problems that Mr. Bremer could not, and perhaps no American could,
>resolve
>or with a postposed negative tag --
>   problems that Mr. Bremer could not solve, nor perhaps could any
>American
>   problems that Mr. Bremer could not solve, and perhaps no American
>could
>or with a postposed noncoordinate parenthetical  --
>   problems that Mr. Bremer could not solve (perhaps no American could).
>but the first is awkward (conjunct raising in general seems to present
>perceptual difficulties that constituent coordination does not) and the
>other two don't put Mr. Bremer and Americans in general into a direct
>comparison.  all three are also longer than the version that appeared
>in the article.
>
>so: go for a punchy direct comparison, and you fall afoul of NPI
>triggering.
>
>arnold (zwicky at csli.stanford.edu)