Long distance dependencies and the unaccusative/unergative distinction

[Mary Dalrymple]

This is really interesting.  I am not sure the -r approach would work,
since I don't see how it would rule out extraction from transitive
subjects.

Could it be that the "object" of a transitive verb is really a subject,
as is the subject of an unergative, so that the generalization is about
extraction from subjects?  This would make Copala Trique a syntactically
ergative language.  Or could it be that the generalization is about
thematic roles rather than grammatical functions?  It would be
interesting to see more examples.

 - Mary

George Aaron Broadwell wrote:

> Dear LFG list readers,
>
>
>
> I'm working with a language (Copala Trique) where extraction in
> wh-questions shows the following schematic pattern. For a transitive,
> extracting the possessor of the object is good, but extracting the
> possessor of the subject is bad:
>
>
>
> OK Whose did John see [__ dog]?
>
> *Whose [__ dog] bit John?
>
> For intransitives, extracting the possessor of an unaccusative subject
> is good, but extraction from an unergative is bad:
>
>
>
> OKWhose [__ sister] died/fell?
>
> *Whose [__ sister] ran/sang?
>
>
>
>
>
> The language is VSO, and the interrogative is to the left of the verb,
> with the remaining arguments to the right, so it's pretty clear where
> the extraction has taken place and where it has not.
>
>
>
> My question is how to account for this pattern in LFG. I can see that
> one possibility is that the unaccusatives actually have their sole
> argument bearing the OBJ role, while unergatives have their argument
> in the SUBJ role. In that case, the analysis is relatively simple --
> the path for the interrogative may not include a SUBJ.
>
>
>
> But suppose the sole argument in an unaccusative is really the SUBJ.
> Is there a good way to express the generalization that extraction is
> good from transitive objects and unaccusative subjects? Could we use
> the [+/- r], [+/-o] features from the mapping theory and say that the
> extracted element is GF [-r]?
>
>
>
> Thoughts and comments welcome,
>
>
>
>         Aaron Broadwell
>
>
>
>
>