1 Why the phonological component must be serial and rule-based ♣ Bert Vaux, University of Cambridge 18 January 2007 1. Introduction This chapter provides general arguments for replacing Optimality Theory with a theory that employs ordered rules and derivations. Between 1968 and 1993 the majority of phonologists worked within a theoretical framework of Derivational Phonology (DP), whose central proposition is that the surface representation of a sequence of morphemes derives from their abstract underlying representations by the application of a series of ordered rules. The introduction of Optimality Theory (OT) in the early 1990s by McCarthy, Prince, and Smolensky has resulted in a drastic realignment of the field of phonology, in terms both of the questions that are being asked and of the ways in which these questions are being addressed. In canonical OT the underlying and surface representations are related by means of universal violable constraints, and the differences among languages are claimed to be due exclusively to differences in the rankings of these constraints. The rapid acceptance of OT in North America, Europe, and East Asia was due in part to a dissatisfaction among phonologists with various aspects of DP such as its perceived lack of universality and the stipulative nature of its extrinsic rule orderings. I have summarized in (1) the claimed advantages of OT that I have been able to find in the literature. (1) Arguments adduced in favor of OT over DP a. New directions, new empirical results (McCarthy and Prince 1993, McCarthy 2002a) b. Generality of scope (OT framework can be used for all components of the grammar, not just phonology and morphology; McCarthy 2002a) c. Parsimony (McCarthy 2002a:243: “if a constraints-only theory is workable, then it is preferable [to a theory combining rules and constraints], all else being equal”; cf. Kager 1999:187: OT is “conceptually superior” in that “we find that a rule-based analysis uses excessive machinery to achieve effects that an OT analysis attributes to a single interaction”.) d. Direct incorporation of markedness (Constraints actually produce cross-linguistic distributions and markedness rather than restating them; McCarthy and Prince 1993:19, Eckman 2005) e. Compatibility with connectionism (Constraint systems of the OT type are attractive for implementation in terms of connectionist networks (Smolensky 1999, Dell et al. 1999, Seidenberg and MacDonald 1999). McCarthy 2002a points out in his FAQ section, however, that OT (excluding for the gradient constraint implementations by Boersma and Hayes) differs from connectionism in having strict domination.) f. Factorial typology derives from free ranking (“By assuming that all constraints have to be universal, OT severely restricts the degrees of freedom in model formulation in linguistics ♣ A shorter version of this paper was read at the LSA Annual Meeting in Atlanta on January 3, 2003. Thanks to Morris Halle, Andrea Calabrese, Andrew Nevins, Justin Fitzpatrick, Laurie Karttunen, John Frampton, Bill Idsardi, Eric Raimy, Ellen Kaisse, Donca Steriade, Cheryl Zoll, and the members of the MIT Phonology Circle for comments on earlier drafts.