Base-Identity and Uniform Exponence: Alternatives to Cyclicity

This paper proposes to reinterpret cyclic effects not as input-output relations between separate levels (McCarthy & Prince 1993, Kenstowicz 1994) but rather--building on the proposals of Kraska (1995)--as similarity constraints between and among morphologically related words (cf. similar proposals independently developed by Benua 1995, Burzio 1995, McCarthy 1995, Steriade 1994, among others). In particular, a constraint of Base-Identity is proposed that evaluates candidates for a complex word [X+Y] on the basis of their similarity to the immediate constituents [X] and [Y] to the extent that the latter occur as independent outputs. In the first section we review previous approaches to cyclicity in the OT framework and propose an alternative in terms of the Base-Identity requirement between independent output forms. In the second section we discuss cases (Italian /s/-voicing, Korean cluster simplification and default accent) in which the availability of a constituent of the derivative as an independent word plays a crucial role. We then examine examples where the similarity effect is sensitive to the gross prosodic structure of the base (Palestinian Arabic, Dutch). The final section reviews similarity effects that do not fall under the rubric of identity to the base as a separate word. For these we posit a constraint of Uniform Exponence (Flemming 1995) that evaluates sets of morphologically related words for segmental and prosodic similarity. Various instances of paradigm leveling (Spanish, Swiss German, Latin, Polish) are analyzed in these terms. The paper closes with an overview of the range of observed similarity effects.