Stem selection and OT

Abstract

A long-standing issue in studies of generative phonology and morphology concerns the proper treatment of stem and affix allomorphy**. This class of alternations is often divided into three types — phonological, grammatical, and lexical — depending on the kind of information which determines the variation in the shape of the stem or affix involved (cf. Bauer 1988, Katamba 1993). In ideal situations at least, it is reasonable to expect that phonological allomorphy will be analyzed in terms of thoroughly phonological mechanisms, grammatical allomorphy will be handled by some interaction of the relevant morphosyntactic and morphophonological processes, and lexical allomorphy will involve listing idiosyncratic information in lexical entries which will supersede otherwise applicable morphological or phonological operations by some version of lexical blocking (Kiparsky 1973, Aronoff 1976, Janda and Sandoval 1984, Anderson 1992, among many others). And in textbook cases of these three types, these assumptions seem quite uncontroversial. Thus, for the regular plural marker -/z/ in English, the identification of its three surface variants (z, s, Iz) in terms of purely phonological processes of cluster devoicing and epenthesis works perfectly well. Likewise, in cases involving choice of, for example, DAT vs. ACC suffixes in Latin on head Ns of NPs occuring as complements of Verbs like persua:de:re ‘to persuade’ (√lmeo: amico:(DAT)/*meum amicum(Acc) persua:de:re ‘to persuade my friend’) as opposed to Verbs like ba:sia:re ‘to kiss’ (*meo: amico:/ √meum amicum ba:sia:re ‘to kiss my friend’), any reasonable account of inflectional morphology is going to provide appropriate spell-out rules to permit the generation of the desired endings. Finally, in cases of full suppletion as with the English past tense form went, the most reasonable analysis appears to be that (i) this form is listed as the special past tense form in the lexical entry for the Verb go, and (ii) the presence of this lexical form blocks the application of the regular Past Tense Rule, so that the otherwise expected form *goed does not arise.