dick:

> And:
>
> At 23:40 23/12/1999 -0000, you wrote:
> >Supposing that I am right in thinking that a that-clause
> >or for-to subject must also be extractee of a finite
> >verb it's subject of (thereby explaining why such
> >subjects don't invert, & why they block all extraction
> >out of the clause, and why they don't occur in raising
> >structures),
> ## Good idea. A standard transformational analysis takes them as
> topics but not subjects, doesn't it?

I didn't know that. I believe they at least used to be just
subjects, but that's not in modern TG.

> Yours is better because it preserves the need for
> a tensed verb to have a subject. So you're excluding examples like (1)?
> (1)	That it would rain being certain, we cancelled the picnic.
> If this is ok your analysis fails, doesn't it, because non-finite verbs
> (being) can't initiate extraction.

Do you think (1) is okay? Or "For him to win being unlikely,..."?
The rule as I stated it excludes these, but I think it could be
adjusted to allow them, since I think nonfinite verbs do have a
topic position, but they are subject to a special constraint that
their topic must also be their subject. However, I'd have to add
other stipulations to rule out THAT and FOR-TO as raised
subjects.

> what is the best way to formulate this?
> >
> >A. If X is subject of Y then X is a noun or [X is
> >   subject of Z and X is extractee of Z (and Z is
> >   finite) and X is a THAT or a FOR-TO.
> ## Don't like this - it involves two words X and Y whose dependency
> relation isn't defined. I don't think such rules are possible.

You mean Y and Z?

On what basis do you take such rules to be impossible?

> >B. (i)  If X is subject of Y then X is a noun or a
> >        THAT or a FOR-TO
> >   (ii) If X is subject of Y and X is a THAT or a
> >        FOR-TO then X is subject of Z and X is
> >        extractee of Z (and Z is finite)
> ## Same again. Also i and ii have different logical structures so it's not
> clear that ii will override i.

(ii) isn't actually overriding (i); it's just adding an extra
condition.

> >C. (i)  If X is subject of Y then X is a noun
> >   (ii) If X is subject of Y and X is subject of Z
> >        and X is extractee of Z (and Y is finite)
> >        then X is a noun or a THAT or a FOR-TO
> ## Same again, though the logical structures of i and ii are more similar.

Here (ii) is overriding (i).

> >D. something simpler than A, B or C.
> >
> >What I haven't been able to put my finger on is a
> >simple way to capture the idea that a subject must
> >be a noun, but that this requirement is overridden
> >when the subject is subject and extractee of a finite
> >verb. A FOR-TO or THAT can be subject of something
> >other than a finite verb, so long as it is also
> >subject and extractee of a finite verb.
> ## Yes. Not easy. How about recognising a special sub-type of 'subject'?
>
> a. A word's subject isa noun.
> b. C-subject isa subject and extractee.  ('c' = 'clausal')
> c. A word's c-subject isa THAT or FOR-TO.

The problem with this is that "That it will rain is unlikely", "that"
is subject but not extractee of "unlikely". Or so I think.

> This guarantees that clauses will be extractees as well as subjects, but
> I'm not quite sure how that prevents inversion.

Because extractees must precede the finite verb.

BTW, I'd like to rename "extractee" "topic", I think, because while
extractees must be topics (in the syntactic sense of that term), not
all topics are extractees, as indeed in the constructions under
discussion.

> I think I once suggested a special sub-type of subject for this as
> well, 'i-subject' (i = inverted).

I'm not sure why you need this. Also, I think inverted subjects aren't
subjects; they're something else, which has failed to raise to
subject. Uninverted subjects are somethingelses that have raised
to subject.

> If this was on the same level of specificity as c-subject (I-subject isa
> subject.) there would be an irresoluble clash between the two word orders:
> c-subject before the verb, i-subject after it. That would be a
> good outcome because there is a gap - (2) has no interrogative or form for
> use after e.g. "not only":
> (2)	That it rained was a nuisance.
> (3)	*Was that it rained a nuisance?
> (4)	*Not only was that it rained a nuisance, but ...
> (5)	*Not only that it rained was a nuisance, but ...
> What we have to avoid is an analysis in which the rules for c-subjects
> *override* those for ordinary subjects, and (2) itself ends up as
> being its own interrogative and (5) ends up as its not-only version.

These seem straightforward to me. In (3-4) "that" is not topic (or, for
that matter, subject) (and, as it happens, something else is topic &
subject, namely QUE in (3) and "not only" in (4)). In (5), "that" is
subject but not topic -- which is not allowed, since it must also
be topic -- and "not only" is topic but not subject -- which is not
allowed, because it must also be subject.

> I don't know whether this analysis solves all the problems, but it's at
> least worth exploring.

And.



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%