16 Jan 2007

Bob Ladd

Note: This talk will be in 1.01, 14BP

I argue, broadly in sympathy with Hale and Reiss's notion of "substance-free phonology", that much of the autosegmentally-inspired discussion of "precedence" in phonology is based on the fundamental error of considering the temporal properties of actual phonetic events rather than a more abstract notion of order. My argument is based primarily on the phonological status of tone.

Tone has been treated in two quite distinct ways in phonological theories, as a feature (e.g. Jakobson, Fant & Halle (JFH)) and as a phoneme (e.g. the Chinese tradition). The JFH view of tone is part of a more general theory of phonology that idealises the stream of speech as an ordered string of phonemes, each of which is a bundle of unordered features. Actual duration, in this view, comes into play only when the abstract formal string is given phonetic realisation. The Chinese traditional view of tone implies a somewhat different idealisation, formally speaking a partial ordering: there are relations of precedence between some pairs of elements (initials precede finals) but no defined precedence relation between others (finals and tones are part of the same syllable, but neither precedes the other). Importantly, however, both idealisations allow us to distinguish abstract formal strings from phonetic realisation: the Chinese view can be formalised not as simultaneity, which implies real time, but only as absence of precedence, which need not.

This distinction is the key to understanding developments in autosegmental phonology. Autosegmental phonology grew out of the attempt to reconcile African tonal phenomena with the JFH idealisation, and the original notion of an "autosegment" looks like an attempt to escape from the restrictions of a totally ordered abstract phonology. However, based on Goldsmith's metaphor of the orchestral score, autosegmental phonology soon began to treat the non-synchronisation of tones and syllables as comparable to the non-synchronisation of phonetic gestures more generally. This immediately thrust phonologists into the realm of real time, and into the set of issues that have occupied autosegmental phonology ever since. Such phenomena, in my view, are properly the realm of models based on continuous mathematics, like Browman and Goldstein's Articulatory Phonology.

If we do not equate autosegments with features, we can reconcile the original insights of autosegmental phonology with the formal rigour of SPE and the empirical coverage of Articulatory Phonology. If we accept that tones are like phonemes rather than features, and we try to formalise the intuition behind the original notion of the autosegment, then tone can be treated as involving partial ordering or undefined precedence. This in turn means we can revert to something like the SPE conception of the relation between phonology and phonetics: phonology is formal, abstract and symbolic (i.e. "substance-free"); phonetics deals with concrete physical events in real time.

[back to PWorkshop]

<owner-pworkshop@ling.ed.ac.uk>