|Title:||Multiple Input-Output Faithfulness Relations in Japanese|
|Abstract:||Multiple Input-Output Faithfulness Relations in Japanese
University of Maryland, College Park
Japanese lexical stratification has been considered problematic in
Optimality Theory (OT) (Prince and Smolensky 1993). Since some
phonological phenomena are observed only in a certain stratum (or
strata), multiple kinds of constraint rankings are necessary to
account for the entire Japanese grammar. This does not respect a
fundamental principle of OT: a single invariant constraint ranking
represents a grammar of each language.
Ito & Mester (1995a) try to solve this problem by introducing that
lexical stratification can be explained in OT by re-ranking of
faithfulness constraints among the strata. However, their approach
still compromises the invariant ranking hypothesis in OT.
Building on Correspondence Theory (McCarthy & Prince 1995), I claim
that there is a system which can account for the whole grammar in a
language with more than one sub-lexicon. Correspondence Theory
recognizes the different types of faithfulness relations, and all
types, including Input-Output (IO), Base-Reduplicant (BR) or Output-
Output (OO) faithfulness, are evaluated in the same way, on the
basis of correspondence relations, regulated by ranked and violable
faithfulness constraints. I propose that Japanese instantiates
multiple sets of IO faithfulness constraints interacting in the same
grammar. With multiple types of IO faithfulness, all the stratum
specific phonological phenomena can be explained within the same
grammar. By proposing four kinds of IO faithfulness relations in
the same language, I contend that a language consists of a single
grammar; or a total ordering of constraints.
Moreover, I introduce the recent study on lexical stratification by
Fukazawa, Kitahara, and Ota (1998) to indicate that some Japanese
hybrids are crucial evidence for differentiating my claim from Ito
& Mester's re-ranking system.