The target hypothesis, dynamic specification and segmental independence  

Lindblom, B.

In Syllable Development: The Frame/Content Theory and Beyond (in press). Edited by B. L. Davis and K. Zajdó. Mahwah N.J.: Lawrence Erlbaum Associates.

Abstract

The first part of this chapter takes a critical look at two theoretical notions in phonetcs: targets and dynamically specified gestures. The conclusion is that the differences between 'static' and 'dynamic' and between 'targets' and 'gestures' are more apparent than real. The evidence reviewed converges on an updated variant of the classical target theory of speech production, that is a model whose motor output is controlled by a series of timeless spatial specifications (targets, via points) and which uses information on targets and situational factors to derive the best movement path between those goals according to a 'cost' criterion. This interpretation gives static targets a role to play at the input level but does not make them incompatible with the dynamically rich signal necessary for robust perceptual processing. The second part presents EMFO (emergent phonetic organization), a simplified but formalized and quantitative algorithm of phonetic learning. EMFO is used to explore two hypotheses about the nature of phonetic units, specifically the developmental end-state of their motor representations: (i) Maximal pre-specification of movement dynamics or Gestalt coding; each pattern is learned as a whole. (ii) Minimal pre-specification or target-based coding; each pattern is learned by finding the minimum input needed to produce that pattern. The latter strategy was found to promote the learning of phonemically coded input sets, whereas Gestalt coding offered no advantage for such sets. EMFO provides further evidence for an updated target theory. Despite its simplicity, it indicates how segmental independence can arise behaviorally, rather than from genetic pre-specification. It suggests that the re-use of discrete phonetic attributes is promoted by a motor organization capable of generating movement between arbitrary spatial goals. This ability provides the mechanism of phonetic recombination. Target specifications are the emergent products of the learning process and serve as the constituents on which the recombination operates.


PERILUS | All issues | 2005 issue | Previous | Next | Phonetics at Stockholm University