Contrast and the evaluation of similarity: Evidence from consonant harmony

This paper argues that phonological similarity is evaluated over contrastive, phonological features. Evidence for this position is provided through analyses of consonant harmony processes. Typological studies of consonant harmony (Hansson, 2001, 2010; Rose and Walker, 2004) have shown that similarity plays a role in consonant harmony patterning with only highly similar segments interacting as targets and triggers. In this paper, two types of cases provide evidence that the relevant properties determining interacting segments are contrastive phonological representations. In one type of case (e.g. Bumo Izon), segments that appear to share a phonetic class with participating segments fail to participate in harmony. In the second type of case (e.g. Dholuo and Anywa), consonant harmony patterns differ between languages with similar phonetic inventories. A pair of segments in one language participates in harmony, and is thus evaluated as similar by the grammar. In another language, a phonetically identical pair fails to participate in harmony. In these cases, neither phonetic properties nor the model of phonological similarity proposed in Frisch et al. (2004) can account for the differences between languages. I propose that the underlying, phonological specifications differ from language to language accounting for the differences in the similarity evaluation.