Underlying Representation in Optimality Theory

Abstract

This paper addresses how we should deal with underlying representation (UR) in Optimality Theory (OT ), focusing in particular on the proper method of Lexicon Optimization (LO), a mechanism in OT that determines a 'real' input among various possible inputs. After closely examining previous proposals, I conclude that Kager's (1999) model of LO is most promising, only if it is controlled by a meta- principle, Input- Output Isomorphy, that guarantees a UR to be identical to its surface form in the cases where underlying contrasts are expressed at surface level. I demonstrate that LO performed in this way provides a good account of UR for allomorphy cases without posing any problem to Richness of the Base, one of the fundamental concepts in OT.