Blurb:

Extended exponence (Matthews (1972; 1974)) poses a challenge for many theories of inflectional morphology, including optimality theory: If one exponent α satisfies two faithfulness constraints C1, C2, and another exponent β only satisfies one of them (say, C1), then, other things being equal, β should be blocked due to harmonic bounding. I argue that this problem can be solved by integrating a constraint MinimizeSatisfaction (MinSat), which requires outputs to gradually improve the constraint profile of their inputs (by requiring a minimization of new satisfactions of constraints in outputs), and which is independently motivated both for phonology (it derives otherwise unexplained counter-bleeding effects in harmonic serialism; see McCarthy (2007)) and syntax (it derives Merge over Move effcts; see Chomsky (2000; 2015)). On this view, optimization is slowed down: Extended exponence can arise because β is locally optimal (minimizing constraint satisfaction), and α is ultimately also optimal because it can later satisfy an additional constraint (again, in accordance with MinSat). This result cannot be achieved in standard parallel OT.