"Syntatic Theory" Query

[Tom Wasow]

Here is a partial answer from 2/3 of the authors of the textbook.  Ivan is
on vacation just now, so his response will have to wait.  We are confident
he will have something to add.

The mechanisms we introduce in Chapter 14 would in fact block the analysis
in b).  The lexical entry for 'smiled' is [STOP-GAP < >], because this is
a lexical default that is obligatorily preserved by lexical rules (p.
441).  Hence, the Head-Filler Rule, which requires a nonempty STOP-GAP
value on the head, will not be able to apply.

We hasten to add, however, that a slight modification of the example
raises essentially the same question.  'John smiled knowingly' does permit
an analysis in which 'John' is the filler for a subject gap.  Nothing in
our grammar blocks 'smiled knowingly' from having the requisite STOP-GAP
value.

Our book makes no pretense of giving more than a very incomplete sketch of
a theory of long-distance dependencies.  A fully worked out treatment
should clearly eliminate this difference between one-word and multi-word
VPs:  either both should allow an adjacent subject NP to be analyzed as a
filler, or neither should.  Frankly, we're not sure which of these
alternatives is preferable.  If root NP-VP strings can get the prosody and
discourse interpretations normally associated with topicalization in
English, then it would make sense to give them dual analyses.  There are
empirical issues here, which we haven't looked into.

Tom & Emily

On Tue, 22 Jul 2003, Rui Pedro Chaves wrote:

> Greetings,
>
> can anyone please explain me how the 'Subject Extraction
> Lexical Rule' as stated in Sag & Wasow (& Bender)
> "Syntactic Theory" textbook is blocked from predicting b)
> for a sentence such as 1):
>
>
> 1) John    smiled.
>
>     subj-dtr   head-dtr
> a) John       smiled.
>     [1]NP     [SPR <[1]NP>
>                COMPS <>
>                GAP <> ]
>
>     filler-dtr head-dtr
> b) John       smiled.
>     [1]NP      [SPR <>
>                 COMPS <>
>                 GAP <[1]NP> ]
>
> ----------------------
> subj-extraction-lir effect:
>   INPUT ( 'smiled'
>           SPR <[1]>
>           COMPS <>
>           GAP <> )
>
>   OUTPUT( 'smiled'
>           SPR <>
>           COMPS <>
>           GAP <[1]>)
> ----------------------
>
> Note that would not be possible in Pollard & Sag (1994:350).
> Thank you very much in advance.
> Cheers,
> Rp
>