[Author Login]
Title:Morphemes and Candidates in Optimality Theory
Authors:Kevin Russell
Abstract: Morphemes and Candidates in Optimality Theory




Kevin Russell


This paper, like Mike Hammond's "There is no lexicon!", argues that

underlying representations are unnecessary in OT and that the

phonological information related with morphemes should be coded

in/enforced by constraints in Eval. I argue that doing things this

way is not just conceptually simpler, but might actually give better

empirical results.

I look at two cases of coalescence, where it seems as though a single

piece of a representation belongs to more than one morpheme at the

same time. The first is coronal coalescence in Nisgha, where, for

example, "underlying" /naks-t-s-t/ is realized as [naks] for reasons

that have nothing to do with syllable structure or phonotactics. The

second is verbal ablaut in Hua, where the stem final vowel undergoes

changes based on a complicated interaction of factors such as

person/number agreement, verb class, and the identity of the following

morpheme. Standard OT would have problems with these cases, since much

of its machinery is designed to forbid multiple exponence. But if

morphemes are constraints, Nisgha and Hua are simply unexceptional

cases where the same piece of a representation satisfies more than one


There's also some discussion about how many of the complications in

the OT architecture are caused by the need to deal with URs, and about

how devices like faithfulness constraints and multiple grammars might

be done away with in a morphemic constraints framework.

Type:Paper/tech report
Article:Version 1