The prosodic structure of Serbo - Croatian function words : An argument for tied constraints * Carson

The question of the proper treatment of clitics has received considerable attention in recent literature on the syntax-morphology and morphology-phonology interfaces (e.g., Marantz 1988, 1989, Zec and Inkelas 1992, Schütze 1994, Selkirk 1996, and references cited there). Selkirk (1996) proposes an elegant theory of the prosodification of clitic function words crosslinguistically, demonstrating that variation in the behavior of function words both within a language (English) and across dialects of a language (Serbo-Croatian) follows straightforwardly from re-rankings of universal constraints in an Optimality Theory (OT) framework (McCarthy and Prince 1993, in press; Prince and Smolensky in press). In this paper I argue that, in addition to strict re-rankings of constraints, tied constraints are also needed within such a system, in order to capture the Serbo-Croatian facts.1 I discuss three empirical shortcomings of her analysis, all involving optionality, and show how they can be remedied by appealing to a particular notion of what it means for constraints to be tied in rank. To the extent that Selkirk’s basic insights are correct, this supports the conclusion that tied constraints play an important role in OT accounts of the ways in which dependent and independent morphemes are combined into larger prosodic units. It adds to the growing evidence (cf. Anttila 1995, Reynolds 1994, and sources cited there) that a necessary part of OT theories of morphophonology is a particular notion of tied constraints or “crucial nonranking” (Prince and Smolensky in press), whereby separate tableaux are computed for each ordering of the relevant constraints and the output of each is a valid possibility in the language.