relativization

[Matthew Dryer]

Paul Kroeger posted a query to the Austronesian list which I found
interesting and which I offered to post on his behalf to LINGTYP.  The
following includes his original query as well as some further questions
that arose in email between us.  Since other members of this list (such as
myself) may find answers interesting, I would encourage you to send
responses to the list.  If you wish to respond directly to Paul, his email
is pkroeger at stanford.edu.  I will also forward any messages to LINGTYP
about this to Paul.

Matthew Dryer

One of the most frequently cited tests for subjecthood in AN languages is
relativization, based on the Keenan and Comrie accessibility hierarchy.
Can anyone give me examples of languages which are neither Austronesian nor
ergative in which only subjects can be relativized?  (The non-ergative
restriction is to avoid getting entangled in the issue of what counts as
the subject in languages like Dyirbal and Mayan.)

Does anyone know of cases where only one argument can be relativized and it
is not the subject?  Are there languages in which only objects, or only
possessors, etc. can be relativized?  As far as I know, the only such
examples are languages which might be described as being syntactically
ergative (on other grounds besides relativization), in which transitive
objects but not transitive subjects can be relativized.

Another interesting issue is the generalization about gapping being
preferred at the high end of the hierarchy, resumptive pronouns at the
lower end (including "islands" etc.).  Tongan is often cited (including I
think by K&C 1977) as a counter-example, but some people claim it is
syntactically ergative (absolutive = subject), in which case it is
consistent with the generalization.  (But Mike Dukes does not accept this
analysis.)  A true counter-example seems to be Vata, a Kru language (W.
Africa) described by Hilda Koopman, in which resumptive pronouns are
obligatory when the subject is relativized and impossible in all other
positions.  Does anyone know of any other similar cases?

Many thanks,

Paul Kroeger