Should Output-Output Correspondence be Invoked to Account for Analogy?

A guiding principle of all scientific inquiry is that the power of a theory should be extended only if such an extension leads to greater insight than a more constrained theory. This principle, which goes by the name of Occam’s Razor, should be applied to the problem of constraining the set of universal constraints in an Optimality Theoretic (OT) model of grammar. Recent work in OT has seen a burgeoning of proposals for new kinds of constraints, including a family of constraints known as Output-Output Correspondence (OOC) constraints. ‘Normal’ OT Correspondence constraints value similarity between underlying form and surface form, but OOC constraints are posited to value similarity between morphologically related surface forms.1 OOC constraints have been proposed to capture a variety of phenomena, including those known as ‘cyclic effects’ in earlier generative literature. One of the first and most influential proposals for OOC was McCarthy’s (1995) use of OOC to account for the complete/incomplete phase alternations of Rotuman. McCarthy adopted the suggestion of Churchward (1940) that the phase alternations were conditioned by syntacticosemantic factors including definiteness in noun phrases and perfectiveness in verbs. In addition to pointing out some problems with these assumptions (such as the fact that I would like to thank Mark Hale, Morris Halle, James Harris, Madelyn Kissock and Ida Toivonen for discussion and comments. 1It has been generally acknowledged, e.g. by Kenstowicz (1995:433) that the notion ‘morphologically related’ needs to be made more specific, but it is not clear that any progress has yet been achieved on this important problem.