Representations Reconsidered (ii): Autosegmental Phonology

Abstract

In the discussion of the standard model given in the interlude, we saw (see p. 155) that, where phonological representations are linear sequences of segments, and phonological properties are transmitted from one segment to another, we are unable to capture certain sorts of generalisation, such as the Akan vowel harmony generalisation: that roots are marked as either [+ATR] or [−ATR] and all affixes, whether prefixes or suffixes, agree with (harmonise with) the root in their value for this feature. We will now address this issue of the limitations inherent in a phonological model which allows only for operations on linear sequences of segments, where those segments are represented as unordered bundles of features.