Recursive Constraint Evaluation in Optimality Theory: Evidence from Cyclic Compounds in Shanghai

Abstract

An important assumption in Optimality Theory is parallelism, and a proper analysis of cyclic effects is crucial. I examine a typical case of cyclicity, namely, stress in Shanghai compounds, where the layers of embedding are in principle unlimited. I show that alignment constraints are inadequate. Instead, identity constraints are needed, in particular Stress-ID which requires that stress locations in the immediate constituents of a compound be the same as when the constituents occur alone. In addition, Stress-ID (and other constraints) must be checked recursively, namely, at every layer of syntactic bracketing. This analysis incorporates the essential properties of the cycle and can therefore handle all cyclic cases. Finally, I discuss the compatibility of recursive constraint evaluation with parallelism, and the remaining differences between a cyclic analysis and recursive constraint evaluation.