Rough draft - please ask before citing - comments welcome! Lexical Exceptions in Optimality Theory

In this paper, I argue that only one of the proposed mechanisms is adequate on both of these counts: lexically specific constraints. Furthermore, I argue that these criteria necessitate the lexical indexation of both markedness and faithfulness constraints. The first of these claims contradicts the position of Anttila (2002), who argues that cophonologies can better meet criterion (1b) than lexically specific constraints. Here I make the case that co-phonologies and lexically specific constraints are notational variants in terms of the kinds of exceptionality they can capture, so long as both markedness and faithfulness constraints can be lexically indexed (Pater 2000; cf. Fukazawa, Itô and Mester). However, it appears that no extant version of the co-phonology model simultaneously satisfies both (1a) and (1b), whereas lexically specific constraints do so straightforwardly. To facilitate the comparison of the different models, I will use a small set of constraints to construct hypothetical instances of lexical exceptionality. I will relate each of these cases to general predicted types of exceptionality, and discuss how the typology is instantiated crosslinguistically. The constraints I adopt are the following: