Multi-grammar Serial OT Approach to Opacity

Abstract

Multi-grammar Serial OT Approach to Opacity Jongho Jun It is well-known in the literature that phonological opacity poses a serious problem for classic Optimality Theory (OT) although it can be easily captured in derivational rule-based theories. Arguing that only derivational theories can provide a proper account for opacity, McCarthy (2007) recently proposes a single-grammar serial OT approach, i.e., OT-CC, where derivational steps (formalized as candidate chains) are evaluated. In this paper, I have shown that although OT-CC, unlike classic OT, can explain opacity effects well, OT-CC assumes a lot more complicated mechanism than classic OT, mainly due to the candidate structure with multiple elements and well-formedness conditions on candidate chains such as gradualness and harmonic improvement. A candidate in OT-CC includes derivational steps, the record of faithfulness constraint violations and their relative ordering whereas a candidate in classic OT is always a single output form. For the purpose of explaining opacity while maintaining a simplified candidate structure of classic OT, this paper proposes a multi-grammar serial OT approach, building on the idea of OT-CC in which the violation of a single basic faithfulness constraint defines a valid derivational step. In the present proposal, a language is characterized by a set of ordered grammars, called strata, the final of which is the same as the one that classic OT would assume for the same language. In the initial stratum, all basic faithfulness constraints are ranked at the top of constraint hierarchy. In the second stratum with the optimal output of the initial stratum as the input, only a single basic faithfulness constraint is demoted to a position in the ranking that it would take in the final stratum. In the following strata, each of remaining faithfulness constraints is analogously demoted to its own position in the final ranking. Which faithfulness constraint is demoted first depends on the language. Thus, even if languages have exactly the same grammar in the final stratum, they may differ in the derivational steps. This difference determines whether the resulting patterns are opaque or transparent. In summary, the present study proposes an OT model which is much simpler than OT-CC although there is no significant difference in the ability of explaining opacity effects.