Missing PREP - redux

[Arnold M. Zwicky]

On Apr 10, 2008, at 6:00 AM, Damien Hall wrote:

> Randy posted:
>
>> CGEL has some similar examples, but using gaps:
>>
>> p1047 ex [40]ia. I want a car(i) [that I know [___(i) is safe]].
>> p1083 ex [16]ii. *He's the man(i) [they think [that ___(i) attacked
>> her]].
>> p1090 ex [36]iii Here's a book(i) [I think [___(i) might help us]].
>>
>> Damien -- when you wrote: "(1') Please forward to anyone(i) you think
>> t(i) who(i) might be interested.", does "t(i)" signify a gap?
>
> Yes, it does, or should.  That is, my intended postulated analysis
> is exactly
> parallel to the CGEL ones.  It's just too long since my syntax
> classes, so I
> used the wrong terminology.

but it's not parallel.  the CGEL analyses co-index the noun that the
relative clause modifies and the gap inside the relative clause that
represents the NP that is "relativized on"; this is just the pattern
of english relative clauses in general.  in this particular case, the
NP that's relativized on is the subject of a complement clause.  so
the first example is different only in detail from examples in which
an object is relativized on:
   I want a car(i) [that I know [I can drive ___(i)]]

but your analysis, which I re-state here to make the representation of
missing elements conform to the scheme that CGEL uses, has *three* co-
indexed elements -- the modified noun, plus two elements within the
relative clause:
   ... to anyone(i) [you think [___(i) who(i) might be interested]]
or possibly:
   ... to anyone(i) [you think [who(i) ___(i) might be interested]]
one of these is presumably for the (omitted) NP that's relativized
on.  where does the other one come from?  it's not licensed by any
principles of standard english.

so it seems you're just stipulating that some varieties of english can
have an anaphoric relative pronoun within the relative clause, in the
position of the NP that's relativized on, *and* in combination with NP
omission in this position.  that just looks like unmotivated complexity.

consider the parallel in gapless relatives like
   ... to anyone(i) [you think [they(i) might be interested]]
there's no reason to assume a gap or trace here *in addition to* the
anaphoric personal pronoun.  so the analysis for
   ... to anyone you think who might be interested
is presumably
   ... to anyone(i) [you think [who(i) might be interested]]
which is an innovation, but one that could at least be related to the
standard english system of relativization (and to the non-standard
gapless systems).  (the question of how the innovation arose is a
separate issue.)

on this analysis, the examples we're looking at are just new kinds of
gapless relatives, in which the NP that is relativized on is marked by
a WH relative pronoun, rather than a personal pronoun.  most of the
rest of the data will follow from this.  as in ordinary gapless
relatives, all three types of relativizers are possible at the
beginning of the relative clause:
   ... to anyone(i) [who(i)/that/0 you think [they(i) might be
interested]]
   ... to anyone(i) [who(i)/that/0 you think [who(i) might be
interested]]

in addition, complement clauses can be marked (with the complementizer
"that") or unmarked (as in the examples above), and for gapless
relatives, there ought to be no problem with that+trace:
   ... to anyone(i) [you think [that/0 they(i) might be interested]]
   ... to anyone(i) [you think [that/0 who(i) might be interested]]
(and of course the possibilities for relative-marking and complement-
marking can combine).  but i *think* that these predictions aren't
quite right.  in particular,
    ... to anyone(i) [you think [that who(i) might be interested]]
sounds *much* worse to me than any of the other examples (which i find
are growing on me).  one way of looking at this oddity is that "who"
in the innovative examples in some sense absorbs the complementizer
"that".

i've tried to check if the innovative construction is possible for
object relatives as well as subject relatives (and, if so, whether
"that who(i)" is also bad there), but the judgments are beyond me; we
need some informants.  here's the story:  parallel to a subject
relative like
   ... to anyone(i) [you think [___(i) might be interested]]
there are object relatives like
   ... to anyone(i) [you think [I like ___(i)]]
for the subject relatives, we get innovative
   ... to anyone(i) [you think [who(i) might be interested]]
the question is whether there are innovative object relatives like
   ... to anyone(i) [you think [who(i) I like ___(i)]]  (X)
my first impression is that these are considerably worse than the
subject relatives (if they're acceptable for some speakers, then i'll
have to re-think my criticism of damien's analysis that started this
posting, since it would have both relative "who" *and* a gap/trace/
omission site).  but let it stand for now.  then the question is: can
you get complementizer "that" plus WH anaphor "who", as in
   ... to anyone(i) [you think [that who(i) I like ___(i)]]  (Y)
this strikes me as phenomenally bad.

going with an absorption analysis, then, if (Y) is out but (X) is
acceptable (for the relevant speakers), an anaphoric "who" (in any
syntactic function) absorbs a preceding complementizer "that".  if (X)
is out, then the WH anaphor "who" is possible only when it represents
a subject, and it absorbs a preceding complementizer "that" then.

that's enough for today of thinking out loud and making this stuff up
as i go along.

arnold

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