optimality theory

[Martin Haspelmath]

Hi Funknetters,

I would be interested in hearing some views on Optimality Theory. Is this
a development within formalist linguistics that has the potential of
bringing it closer to a realistic, cognitive-functional view of language?
Or is it just another formalism that attracts attention mainly because it
exploits recent advances in word processing technology?
  I remember hearing Smolensky and Prince talk about connectionism and a
"soft reduction" of generative grammar to connectionism, back in 1991,
and I was curious where they were heading. If Optimality Theory ("OT")
incorporates aspects of connectionism, isn't it of great interest to
cognitive-functional linguistics?
  Some of us have been talking about preferences and their hierarchies in
a way very similar to OT parlance, for instance Natural Morphologists
(Dressler, Mayerthaler, Wurzel). The terms used in this research
tradition are parameters of naturalness and markedness, but basically
these are identical to OT's ranked constraints. When Theo Vennemann
listened to the first OT talks here in Germany, he kept saying: Isn't
this what I've been saying for the past 15 years?
  "Optimality" suggests "good" and "bad" linguistic structures--this is
not the way formalists have been talking, but functionalists have not
been afraid to do so (even if informally). Theo Vennemann even once
suggested that language change is ALWAYS language improvement, i.e. local
optimization. In a recent paper forthcoming in "Linguistics", my FU
Berlin colleague Renate Raffelsiefen presents an analysis of the
phonology of German schwa within OT, which is based on the idea that
language change is always the satisfaction of some OT constraint. True,
this is not standard OT, but works like this one suggest to me that OT
and functionalism may perhaps be quite compatible.
  But I have also heard the view that OT is just an empty formalism that
can be filled with any kind of content, functionalist or formalist or
whatever.
  So what do you all out there think?

Martin Haspelmath