Request for references/ideas

[Emily Bender]

Dear Funknetters,

I am looking for ideas, pointers to the literature,
etc. regarding the following problem:

Production studies of African American Vernacular English
(since Labov 1969) have turned up a remarkably stable set of
constraints on the appearance of overt vs. "empty" or
"deleted" forms of the verb 'be', as in (1):

(1) a. Kim a doctor.
    b. Kim is a doctor.

Both (1a) and (1b) are grammatical in AAVE.  However, one
finds that the distribution of the two forms is conditioned
in part by the part of speech of the predicate, as follows:

(2) NP < PP(loc)/AP < V+ing < gonna

That is, the ratio of forms like (1a) to forms like (1b) is
much lower than the ratio forms like (3a) to forms like (3b):

(3) a. Kim gonna leave.
    b. Kim is gonna leave.

The PP(loc) and AP categories are usually tabulated
separately, but their ordering is not very consistent across
different studies.

What I am looking for are some possible functional (pragmatic,
processing, etc.) constraints that could explain this
pattern.  One consideration in formulating such constraints
is that copulaless sentences with NP predicates do not seem
to be particularly dispreferred crosslinguistically, at
least from the typological evidence that I have to hand.

Any ideas and/or references would be appreciated, and I'll
post a summary to this list.

Many thanks,
Emily Bender