Autosegmental-metrical phonology – Unpacking the boxes

Abstract Autosegmental-metrical phonology has shown itself to be a highly successful framework for the description, analysis and comparison of the prosody of many of the world’s languages. What has contributed to the success of this framework is the fact that there is widespread use of prepackaged units within the model – referred to as “complex primitives”. The intonation systems of languages are described as having edge tones and, in some cases, also (post-lexical) pitch accents. These are defined in terms of both their association properties and their cueing function within the prosodic system. Edge tones associate with an edge (or a tone bearing unit at the edge) and are a cue to the juncture between prosodic constituents. Pitch accents associate with a head (usually a stressed syllable) and are a cue to prominence. I shall argue that we need to unpack these definitions, providing evidence from Tashlhiyt Berber, Maltese and Italian, languages in which the association properties and cueing functions of intonational tones do not automatically line up in this way.