Voicing and the Skeleton

The main goal of this paper is to show that the behaviour of voicing in obstruents can better be accounted for if we assume that voicing is represented by a specific skeletal configuration rather than by phonological features. It will be shown that previous analyses of voice phenomena run into serious problems, especially when it comes to the representation of voice assimilation and intervocalic voicing. The analysis outlined in the final section of this paper gives a representationally motivated account of voice phenomena in CVCV and manages to unite the effects of voice assimilation, word-final devoicing and intervocalic voicing. The representation proposed here also makes a number of predicitions, which may well prove to be correct: (i) languages with prevoicing in initial stops necessarily have regressive voice assimilation (ii) progressive assimilation and coda devoicing appear only in languages with a contrast based on different phonetic properties (e.g. aspiration).