another Hocank/Helmbrecht article question

[Catherine Rudin/HU/AC/WSC]

Ok, here's another question sparked off by the recent IJAL article.  This
one is really totally irrelevant to Johannes' point, just something he
mentioned in passing -- I almost hesitate to bring it up for fear it'll be
seen as an unfair criticism of the article.   So Johannes, if you're
reading this, it's really just a question!   I was interested in the brief
discussion of relative clauses starting on p. 11, and especially the
structure (22) which shows the Hocank relative clause as having an external
head:  N-head [null relativizer Predicate-Determiner] with the part in
[...] being  the relative clause.  This is pretty surprising for a Siouan
language -- relative clauses in Lakota, Crow, Hidatsa, Omaha are
internal-headed.  (Though of course it's possible to have both internal and
external headed relatives in the same family, or even in the same
language... as far as I know no Siouan language has been shown to have
clearly external-headed relatives.)   So it would be really interesting if
Hocank does have this structure.

What I'm wondering is -- did Helmbrecht just assume the external-head
structure, or is there actually evidence for it in Hocank?  The few
examples given are inconclusive; none of them have more than one
constituent besides the predicate, so it's not possible to distinguish N [
predicate] from [N predicate].    What happens if instead of just "the meat
I cooked" (23b) we have "the meat my mother cooked" or "the meat I cooked
yesterday" -- does "meat" necessarily come at the beginning, or can you
have orders like [my mother meat cooked determiner] or [yesterday meat
I-cooked determiner] where "meat" is a clearly internal head?   If "meat"
has to be first, it would argue for the external-head structure.  One
indication in (23b) that the head is actually external might be the
definite determiner on "meat", given the apparently universal fact that
internal heads of RCs must be indefinite (Williamson's indefiniteness
restriction)....   The indefiniteness restriction is robust enough and has
enough raison d'etre -- required to allow operator binding to work right,
etc. -- that I'd take it seriously as an argument.

Any thoughts?
Catherine