Universality of GR's

[Stephen M. Wechsler]

What about an economy-driven approach to GR's, as in OT?  I'm imagining
something along the following lines.  A predicator's input contains an
argument structure (or perhaps lexical entailments from which the a-str is
derived) but little or no gf specification.  Constraints include a-to-f str
mapping rules.  The (infinite) candidate set would include some
f-structures containing gf's like SUBJ and OBJ; some f-structures which
include reentrant a-structure items (i.e. token identical with a-structure
representations); and various combinations of the two.  So at one extreme
would be a language with a non-a-structure GF label for each argument (i.e.
the way all languages are represented in current LFG); at the other would
be a language where the entire a-structure is copied into the f-structure
(an active language of the sort Rachel described).

Which candidate is selected would depend on the ranking of the mapping
constraints.  Economy constraints would generally disfavor unneeded
structure.  For example, mapping A-subject (most prominent arg in
a-structure) to SUBJ would be less economical than just copying A-subject
into the f-structure.  So the ranking of economy constraints relative to
mapping constraints would be crucial.

This seems to me to be analogous to the approach to deriving c-structure
taken in some of Joan's OT/LFG papers (and in Grimshaw's OT/GB).  Instead I
am deriving f-structure.

Turning to the typological issue raised by Chris:  a language which, due to
the constraint ranking, lacks SUBJ altogether could not have rules
referring to SUBJ, like functional control.

Another consequence for many (all?) languages might be that oblique roles
would no longer be duplicated by distinct GF values.  That is, instead of
mapping Goal to OBLgo, "Goal" itself would become a GF.

Of course all this needs to be worked out in detail...

--Steve

p.s. for some reason my posts to this list sometimes take several days to
appear, so this msg may seem like a non sequitor.

Prof. Stephen Wechsler, Linguistics Dept.
U of Texas, Austin, TX 78712-1196
http://uts.cc.utexas.edu/~wechsler/
Calhoun Hall 403; (512)471-1701