Directionless Syllabification and Ghosts in Yawelmani

In optimality theory, the grammar evaluates candidate outputs in parallel against a hierarchy of ranked, violable constraints. One of the main advantages of this framework is its ability to capture nonlocal dependencies straightforwardly, obviating the need to build the ill-formed intermediate structures that are sometimes inevitable in bottom-up approaches. A major challenge for parallel evaluation then is to account for directionality effects, in particular with regard to syllabification where directionality has been intimately bound up with the painstaking insertion of prosodic structure one piece at a time either from right to left or left to right. One approach to this problem, building on an observation of Kirchner’s, has been to specify an edge of the word toward which the prosodic structure aspires, thereby encoding directionality in a way consistent wit the principles of optimality. Examples of this include McCarthy & Prince (1993a) and Padgett & Mester (1993). A different approach to the directionality problem, and the one I will take in this paper, is to seize the opportunity to see if the illusion of directionality might in fact be the result of more interesting properties of a language. It is in this light that I want to look at Yawelmani today. Noske (1985) and Archangeli (1991), building on the work of Kisseberth, Kuroda and Newman, have argued that the placement of epenthetic vowels in Yawelmani follows from right to left syllabification. I’ll show however that directional syllable building is not a necessary component of the Yawelmani grammar, and that the pattern of syllabification and epenthesis, as well as the behavior of the notorious ghost segments, follows simply from align constraints which act to preserve the shapes of the Yawelmani templates on the surface while minimizing violation of higher ranked constraints such as PARSE, FILL, and *STRUC.1