Advanced Tongue Root harmony in Setswana.

This thesis addresses certain issues in the analysis of Advanced Tongue Root (ATR) vowel harmony in Setswana, a language spoken in South Africa and Botswana. Setswana presents a particularly interesting problem as regards the analysis of harmonic patterns since there exists in the language evidence on the one hand that: (a) all vowels in a stem are preferentially either (+ATR) or $\rm\lbrack{-}ATR\rbrack,$ but (b) in some cases the requirements of harmony seem to require the cooccurrence of vowels with conflicting ATR values. The pattern of ATR harmony observed in the language presents a number of difficult problems for the adequate description of the harmonic process. These problems are outlined in the thesis. The hypothesis made in this thesis is that there exists a feature $\rm\lbrack{\pm}RTR\rbrack$ which is distinct from $\rm\lbrack{\pm}ATR\rbrack;$ specifically, (+RTR) is different from $\rm\lbrack{-}ATR\rbrack.$ The inclusion of this feature solves the problem of making a distinction between the phonologically "advanced" vowels (scI, $\upsilon\rbrack$ and (i, u). The analysis of ATR vowel harmony presented here will assume the general framework of Autosegmental Phonology, incorporating Underspecification and Grounding Condition theory. It will be argued that grounded conditions hold of varying domains, with some conditions holding of all representations, while other conditions govern only the output of specific rules. (Abstract shortened by UMI.)