Autosegmental phonology is a theory of non-linear phonological representation. It was developed out of research in Generative Phonology at MIT in the mid and late 1970s, as a response to certain problems in the phonological theory of that time.
Autosegmental phonology was initially developed in response to the challenge of developing an adequate theory of tone. Its immediate source of inspiration was the work of Williams 1971 and Leben 1973; these were the first to introduce non-linear structures into generative phonology in their treatments of tone systems in West African languages such as Margi, Igbo and Mende. In the model proposed by these writers, underlying tones were represented on separate tiers from the feature matrices representing vowels and consonants; they were subsequently merged with these matrices by Tone Mapping Rules that applied in the course of derivation, creating single-tiered representations in surface structure.
The principal innovation of autosegmental phonology, as presented in Goldsmith 1976, was the idea that tone mapping rules do not merge tonal and segmental representations, but associate their elements by means of formal entities known as Association Lines. In this framework, phonological representations consist of parallel tiers of phonological segments, both tonal and segmental.
t t t t
H L H L
H=high tone L=low tone t=any tone-bearing unit (vowels or syllables)
Elements of each tier, called AUTOSEGMENTALS, are sequentially ordered; elements of adjacent tiers are simultaneous if and only if they are linked by association lines. In this model, all tiers remain independent throughout derivations: at no point is the tonal tier merged with segmental tier.
A further innovation of autosegmental theory is the set of universal principles termed Well-Formedness Conditions, which govern the multi-tiered structure of the representation. These principles not only define the set of theoretically possible inter-tier configurations; they also trigger the operation of a set of universal repair mechanisms, often termed Association Conventions, whenever configurations that violate them arise.
In subsequent work, autosegmental phonology underwent further development; by the mid-1980s it could be considered a fully general theory of phonological representation, radically different from the linear representational systems of more traditional approaches. The primary innovation of the generalized model has been the view that not just tone and other so-called ‘prosodic’ features, but all phonological feature are arrayed on separate autosegmental tiers. In this conception, which draws upon earlier research in Metrical Phonology and Prosodic Phonology.
X X X X Skeletal Tiers
Y Y Y Y
Z Z Z Z Substantive Tiers
W W W
Further developments in autosegmental phonology include the grid-based theory of stress proposed by Halle & Vergnaud 1987, and the model of intonation and prosodic structure developed by Pierrehumbert & Beckman 1988. For a more recent overview and several new proposals, see Goldsmith 1990. As remarked by Van der Hulst & Smith 1982, progress in autosegmental phonology has owed much to its ‘problem-solving efficiency’- i.e., its success in finding solutions for previously unsolved representational problems, and integrating them into a consistent, over-all theoretical framework.
EDLSTEIN JEFFEREY P. & MCGARY JANE etc. EDS. 1992. INTERNATIONAL ENCYCLOPEDIA OF LINGUISTICS. OXFORD UP.[ET1]