What is autosegmental phonology?

Question

I am an armchair music theorist and trying to read about John Goldsmith's theory of autosegmental phonology.

Can someone summarize the basic principles behind his theory for a linguistic layman?

Answer 1

Understand it against the theoretical background that an utterance is composed of a uniform set of some 20 phonetic features, one per segment, in a "solid" matrix, so an utterance with 30 segments has 600 feature specifications. Autosegmental phonology holds that the segment actually subdivides into autonomous "autosegments", for example (in Goldsmith's dissertation version) there is a "segment" submatrix and a "tone" submatrix. This defines two "tiers", the elements of which can be related in a fairly relaxed way.

So certain segments might associate to one tone, or two or three tones; or no tones (consonants typically do not ever associate to a tone; vowels might not, under certain circumstances). A tone might not associate to anything (in which case it is "floating"), or it might associate to 1, or any number of segments.

His theory happens to start with the assumption that tones don't associate to segments in underlying forms and that associations come about via rules, but later developments indicated that that was not crucial or indeed viable as a universal assumption. So later research developments focused on explicating the derivational relation between tones and their segments. Additionally, Goldsmith focuses on tone and only presses the frontiers to other features such as nasality lightly (although actually proposes complete autonomization of every feature, at the end of the dissertation) – later word led to the theory of "feature geometry" which holds that all features are autonomous and related to each other in a complex manner.