|Title:||Direct Optimality Theory: Representation as Constraint Violation|
|Abstract:|| Direct Optimality Theory: Representation as Constraint Violation
This paper argues for a model of morphology and phonology, Direct
Optimality Theory, in which the phonological form of words is
represented directly by constraint violations rather than by strings
of segments. I propose a restructuring of Optimality Theory (Prince &
Smolensky 1993) in which two of the five elements of the theory ('UR'
and 'SR') are eliminated and the role of a third ('GEN') is reduced;
this is done without adding any machinery to the model or reducing its
ability to capture significant generalizations. The argument runs as
follows: (i) underlying representations are syllabified, (ii) they are
relatively ill-formed, (iii) so they can be represented directly in
terms of constraint violations. Step (i) brings to its logical
conclusion the incorporation of prosody into generative phonology
begun in the 1970s and 1980s: prosody is relevant at all phonological
levels of description. As will become clear, representation as
constraint violation is possible only in a theory of grammar that
allows violable constraints. Hence Direct OT.
The paper is organized in the following way. Section 1 provides
the necessary background to the discussion and an overview of the model and
the assumptions it relies on. Section 2 argues that underlying
representation includes prosodic information including syllable structure.
Section 3 shows in detail how underlying representation can be reduced to
constraint violation and section 4 does the same for surface representation
and for alternation. Section 5 offers novel solutions for subtractive
morphology, affixation and templatic morphology in terms of the theory.
Section 6 presents some striking evidence for the theory from speech error
patterns. Section 7 compares Direct Optimality theory to some recently
proposed alternatives to lexical representation (Hammond 1995, Neef 1995,
Russell 1995). The final section offers a short conclusion.