Palatalization and other non-local effects in Southern Bantu languages.

Palatalization in Southern Bantu languages presents a number of challenges to phonological theory. Unlike 'canonical' palatalization, the process generally affects labial consonants rather than coronals or dorsals. It applies in the absence of an obvious palatalizing trigger and it can apply non-locally, affecting labials that are some distance from the palatalizing suffix. The process has been variously treated as morphologically triggered (e.g. Herbert 1977, 1990) or phonologically triggered (e.g. Cole 1992). I take a phonological approach and analyze the data using the constraint-based framework of Optimality Theory. I propose that the palatalizing trigger takes the form of a lexically floating palatal feature cor (Mester and ltd 1989 Yip 1992 Zoll 1996). The study locates siSwati palatalization within its broader Southern Bantu context. In my analysis I show that the behaviour of the other selected Bantu languages' palatalization follows from an analysis parallel to siSwati. Palatalization in all the languages involves an attempt to link the V-Place cor to a labial in the passive, diminutive, and locative and in addition, to an alveolar in the diminutive. The cor either palatalizes the root-final labial, as in khipha > khijv'a 'remove', or may surface as in the passive, as in pha > phiwa 'give'. Realization of the cor feature is dependent on language-specific differences. The languages investigated are compared within an Optimality Theory framework, showing that their differences follow from universal constraint re-rankings. I also include in this study other processes that are related to long distance palatalization: vowel harmony/co-articulation and tonal phonology. These processes show us that long distance effects are not peculiar to palatalization, since vowel harmony/co-articulation may also involve non-adjacent segments. Tone shift and tone spread may result in tone being realized syllables away from its underlying source. The vowel facts and some of the tonal facts of siSwati have been investigated phonetically. It is argued that while Sesotho shows true vowel harmony, siSwati is still at the co-articulation stage. This difference is also modelled as re-ranking in an Optimality Theory grammar