Jasp:
> And:
> 
> > I understand your & Nik's & others' disquiet at the existence of
> > null words, since they definitely relax restrictions on the set
> > of grammars that the theory can express. However, I gave an 
> > argument in an earlier message that even before you consider
> > any actual data, if you think about how language must work, 
> > you're forced to the conclusion that in principle some null 
> > words must be able to exist, unless you can find some 
> > nonmetatheoretical reason for excluding them.
> 
> I agre entirely. We have no trouble in allowing _agre_ to bea 
> AGREE, even though it doesn't have one of the _e_s. Then we 
> should be very surprised if there weren't sometimes cases where 
> the whole form was missing.

But this is a different kind of example. The grammar says the form is
/agree/, but when that gets interpreted at the phonetic interface we
will under certain circumstances accept [agr] or even [] as a valid
interpretation of /agree/. This is different from the grammar itself
saying that // is a valid form of AGREE.

I'm delighted to be agreed with, though. 

And btw, I didn't even notice the <agre> in the first sentence on first
reading.

> It is for this reason that I favour your position (&), that null words 
> are whatever word they have to be less the form, rather than PRO. I 
> should point out, though, that I'm not convinced by your argument 
> that it's too unlikely that a lexeme with the form // should have the 
> proeprties Dick wants to give to PRO: it's no more unlikely than 
> that a lexeme with the form /kat/ should mean 'cat'.

So would you claim that there is an equal probability of CAT having the
form // and PRO having the form /kat/?

No, my dear Watson, the phonologylessness of PRO is far far too convenient
and neat an alibi...

[I'm afraid I momentarily usurped your Holmeshood there.]


--And.


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