Esimbi Vowel Height Shift: Implications for Faith and Markedness*

In this paper I discuss the theoretical implications of a vowel height transfer in Esimbi for three current issues in the analysis of featural phenomena in Optimality Theory (Prince and Smolensky 1993). Esimbi exhibits a remarkable surface distribution of vowels: a greater number of vowel height contrasts occur in prefixes than in roots. On the face of it, this appears to belie the generalization that affixes are universally unmarked relative to stems, a relation which McCarthy and Prince (1994, 1995) propose to capture in Optimality Theory with a metaconstraint universally ranking faithfulness constraints for roots over faithfulness for affixes. Previous work by Stallcup (1980a,b) and Hyman (1988) has convincingly shown that the vowel height distribution in Esimbi is produced by a transfer of height features from root to prefix vowels. I argue that this outcome is actually driven by the root versus affix faithfulness metaconstraint in concert with a word-initial licensing constraint for marked vowel height. The height shift in Esimbi offers an interesting case of nontonal features behaving as true autosegments, independent of their segmental sponsors, a phenomenon which bears on two additional points of theoretical interest. First, the features that undergo the transfer act as autonomous structural elements, signaling a need to extend the correspondence relation to Esimbi vowel height features within Correspondence Theory (McCarthy and Prince 1995). This kind of relation does not fall under the usual IDENT[Feature] formulation of featural faith, which demands identical feature specification of segmental correspondents, rather it is expressed by a MAX[Feature] constraint, which evaluates correspondence between features themselves. Esimbi feature shift thus offers new support for the existence of MAX[Feature] constraints (see Lombardi 1995, 1998; Causley 1996; also discussion in Lamontagne and Rice 1995; McCarthy and Prince 1995). A related point concerns the constraint driving the mobility of height features. Because the shifted features originate in a different location from the one in which they surface, the licensing constraint motivating the transfer must be formulated as a requirement on the position of marked phonological structure, that is, it is an instance of Positional Markedness (Zoll 1996, 1998) — it cannot be expressed in terms of Positional Faithfulness, where faithfulness is specific to the context of the licensing position (Beckman 1997, 1998, and references therein). These analytical findings contribute to the wider debate on the extent to which nontonal features should be understood as having “autosegmental” status in phonological structure.