And:
>I don't see that at all. Indeed, one of the best candidates for nullhood,
>namely VP ellipsis, looks far more like phonological deletion.
>
>Moreover I personally find it much much much more plausible that
>phonology is deleted in certain syntactic environments than that there
>are certain lexical items that happen, by remarkable coincidence, to have
>null phonology and, unlike other nouns, are restricted to a very very
>specific syntactic environment (namely, above the line subject of gerunds
>and to-infintives). When you consider that no other item can apparently
>be subject of to-infinitives it seems even more suspicious that there
>is exactly one lexical item that can be subject of these and that it
>happens to be null. If, on the other hand, you treat this as a case
>of obligatory deletion, then there are no suspicious coincidences or
>surprising distributional anomalies.
## Yes, I can see the point of these arguments. But I suspect this is an
area where the best hope for progress is via mentalism - psycholinguistic
experimentation. You're predicting that the individual words will be
activated but not pronounced; I'm not. I suspect you will be proved right,
especially when we extend the discussion to VP ellipsis and pronominal
anaphora. E.g. when "these" = trousers, the word "trousers" is activated
but not pronounced (and similarly for pronouns in languages that have
arbitrary gender). And of course VP ellipsis is normally in a context where
the VP concerned has just been said, so it's already active and just needs
a bit of a boost. 
	One small worry about assuming specific words rather than PRO is that it
may be hard for the analyst to know which (unpronounced) words are there;
e.g. in (1):
(1)	Discussing linguistics is fun.
What's the understood subject of "discussing" here? "You"? "One"? And what
about "laying" in (2)?
(2)	Laying eggs takes time.
"A bird"? "A dinosaur"? "An egg-layer"? 
Maybe we do need an all-purpose abstract word (pro or PRO or Pro) for
non-anaphoric patterns as well as specific words?

>
>> As you say, PRO has to inherit the gender and other characteristics of
>> its antecedent, so there must be some place to put this information.
>> The obvious analysis is that in Italian PRO is just like the
>> other pronouns and agrees in gender with its antecedent.
>
>My objection to this is that there needn't be any actual antecedent, so
>I don't think a gender agreement analysis will work.
>
>> Similarly for person, gender and number in your English examples. I
>> think your examples (which I've left below) are compatible with this
>> analysis.
>
>Again, lexically idiosyncratic number in English ("getting tangled up
>with themselves", said of trousers) indicates to me that the antecedent
>+ agreement analysis is not viable.
## I'm puzzled here because I thought my analysis was what you were
suggesting! I.e. an unpronounced specific word. Whether it's recovered
anaphorically or exophorically doesn't matter. But now see below:


>In my view, each dependency corresponds to a separate nonterminal node of
>the regent, and such nodes can be coindexed even when there is no
>dependent, so I don't need to posit null items in such cases. But WG
>does.
## Aha, I see. So your nonterminal nodes don't identify specific words, but
just the possibility of a word? I don't understand how you propose to
handle the anaphoric/exophoric cases like VP ellipsis and arbitrary
gender/number on a pronoun.



Richard (= Dick) Hudson

Phonetics and Linguistics, University College London, 
Gower Street, London WC1E  6BT.
+44(0)171 419 3152; fax +44(0)171 383 4108;
http://www.phon.ucl.ac.uk/home/dick/home.htm


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