Correspondence and Identity Constraints in Twolevel Optimality Theory

Abstract

One goal of post-SPE generative phonology has been to constrain possible input-output pairs. While SPE (Chomsky and Halle 1968) permitted any input-output pair to be stipulated, later work in Autosegmental Phonology and Feature Geometry (see Goldsmith 1990 for an overview) restricted rules and representations so as to constrain input-output relations. In particular, a given rule could target only a single node in a featuregeometric representation and was restricted to one of the following operations: spreading, delinking, insertion, or deletion. While it is not clear that these restrictions had any effect on the formal power of the theory, they resulted in a reflection of markedness that many linguists I thank the following people for discussing this paper with me: Abigail