Re: Comments on Tokana Reference Grammar

On Fri, 11 Dec 1998, John Cowan wrote:

> ellipsis vs. gaps:  I am troubled by the blanket statement that
> noun-phrase ellipsis in Tokana is generally allowed, coupled with
> the use of gaps to indicate coreferencing pronouns in embedded
> clauses.  That could lead to ambiguities like
>
>         Han believes that (gap) saw (ellipsis)
>         Han believes that (ellipsis) saw (gap)
>
> I think that allowing a resumptive pronoun in embedded clauses
> that have ellipsis would be very useful.  In Lojban, the resumptive
> is required for precisely this reason, so as to allow ellipsis
> to do its work.  Lojban uses a unique resumptive, but in Tokana,
> the existing determiners could do the work --- except that they
> have been preempted for precisely the opposite use!
> Or is the rule about ellipsis applicable only to main clauses?
To return to this question, here are the rules, as I see them,
that govern null arguments:

(1) In main clauses, any argument may be null if its reference is
arbitrary.  So "The dog bites" (with "the dog" in topic position)
means "The dog bites people".

(2) A topic may be null if it is coreferential with the topic of
an immediately preceding clause.  (Recall that topics are only found
in main clauses, so trivially this rule applies only to main clauses.
So if this is ellipsis, then I guess the answer to your question above
is "yes".)

(3) An argument of an embedded verb is null if and only if it is
coreferential with the topic of the main clause (or, in the case of
relative clauses, the relativised noun).

Rules (1) and (3) are to some degree violable, but that's the *basic*
pattern, as I see it.

So to return to your example, "Han thinks that saw": Only topics can
be elided, so both of the null arguments in the embedded clause must
be gaps, which corefer with the topic "Han".  But in that case the
sentence is bad, for the same reason that "Han saw him" is bad in English
with "him" = "Han".  Just as in English, having an object pronoun corefer
with the subject of the same clause is ungrammatical because it violates
the (anti-)locality conditions on pronominal reference.  (In Chomskyan
terms, it violates Condition B of the binding theory.)

I suppose that "Han thinks that saw" would be *marginally* acceptable
on the reading "Han thinks that he (Han) saw someone" or "Han thinks
that he was seen", where one of the null arguments is a gap coeferring
with the topic and the other one has arbitrary reference.  But such an
utterance would have to have sufficient context to be unambiguous.
Otherwise you'd need to stick in an overt pronoun:

        Han thinks that saw him
                = Han thinks that he (Han) saw him (someone else)

        Han thinks that he saw
                = Han thinks that he (someone else) saw him (Han)

Does that address your concern?

Any other questions and comments on the Tokana webpages would be most
welcome.  However, as I said before, I'll be leaving town on Monday for
a two-week holiday, so any messages sent after then won't be answered
until the end of the month.

Matt.