rote, analogy and OT

Brian MacWhinney macw at CMU.EDU
Sun Dec 12 20:57:31 UTC 1999


Dear Wally,
  Many thanks for your thoughtful response to my queries and concerns about
OT.  I believe I share your basic perspective.  In my 1978 account of the
acquisition of morphophonology in Hungarian, I relied largely on the
mechanisms of rote and analogy which you (and Dedre Gentner) suggest.  Of
course, one of the big challenges facing connectionist theory is trying to
cash out the meaning of an account based on analogy.  It seems to me that OT
might have some promise in this regard.  However, there seem to be two core
problems that have to be ironed out.
  I would guess that you and I would agree that rote and analogy play their
largest role in the area of morphophonology.  But, when we turn to issues in
phonotactics, assimilatory processes, and metrical phonology, then basic
OT-type constraints seem to move to the fore.  So, how do we link up the
mechanisms of rote and analogy with constraints in a single language system?
We certainly want to avoid modularizing two systems that, in fact, far from
separable.  After all, many morphophonological processes are echoes of
phonotactic and assimilatory processes.  So, a better solution is to try to
figure out how physiologically-grounded constraints impact complex word
combinations.  To do this, we need to have a system of graded (or scalar)
constraints and a method for talking about their interaction.  Ideally, OT
would have provided this, but the jury is still out on whether strengths and
interactions will be a part of OT.
   The second problem is the ongoing reliance on underlying form in OT.  In
my still not properly informed opinion (see my forthcoming email reply to
Joan Bresnan), the basic problem in OT is its commitment to the doctrine of
underlying form.  If we build a system that is designed to work off of forms
that have little relation to anything stored in the brain, we will be lacing
our synchronic theories with diachronic commitments.
  I will read your paper from the Historical Linguistics volume as soon as I
can.  The question will be how to reshape OT in a way that solves these
problems.
  Thanks again for the helpful comments.  I will continue to think more about
these issues.

--Brian MacWhinney



More information about the Funknet mailing list