optimality and lexical rules

[Suzanne E Kemmer]

Here's the crux of the lexicon-based objection to OT, for those who
don't want to wade through the correspondence.

[The author says that evidence has been mounting since SPE days that
lexical rules are emergent generalizations that do not exist beyond
the forms they describe -- e.g. they generalize to novel forms on an
ad hoc "analogical" basis; at intermediate stages the "rules" are
scattered messily across individual lexical items; they can interact
with "analogical," non-rule-like patterns of irregular inflection in a
way requiring reference to individual items or sets of items...]


"       The insight I THOUGHT we were coming to, then, was that
lexical rules are neither rules nor constraints but something
like emergent properties of lexical items.  You can get this with
connectionism or you can get it some other way, but it's got to
be gotten sooner or later, don't you think?  OT is thus especially
irritating because it sidetracks and tames the more heretical
students with the misleading claim that OT is sort of like
connectionism, but then it just goes ahead and replaces general
rules with even more general constraints, neither of which
makes sense given what we already know about the lexicon.
OT constraints may or may not be a notational variants of rules,
but you can bet OT will be there with rules, up against the wall
when the real revolution comes.   "