Final vowel shortening in Luganda

A process by which long vowels are shortened in "final position" has been noted by a number of linguists, e.g. Ashton et al [1954], Tucker [1962], Cole [1967], Stevick [1969], Katamba [1974], Clements [1986]. It is generally assumed that this shortening is characteristic of word-ends such that the process can even serve as a criterion for phonological word division. Despite the attention given to final vowel shortening (FVS), the relevant facts have not been exhaustively described. In this descriptive account, we show that FVS is a much more complex phenomenon than the Luganda literature suggests. We observe, for instance, that FVS does not work the same on nouns as it does on verbs and that an empirically adequate analysis must take into account the source of such word-final length, e.g. underlying vs. derived. In our solution, FVS first applies at the end of a phonological word (PW) and then again at the end of a clitic group (CG). In order for the facts to fall out from this analysis, we argue that at the PW level (1) the final vowel of verb forms is not affected because it is extrametrical, i.e. "invisible" and (2) the second mora of a monosyllabic stem is not affected because it is accented.