The mismatch between phonological and phonetic representations

The division of the stream of speech into a string of discrete segments represents the number of places where contrast between the current speech event and others is possible. Discrete segments may be employed not only to represent phonological contrasts in speech events, however, but also in the plans for actually uttering them. Beyond isomorphism with phonological representations, the appropriateness of discrete segments at some level in the plans speakers employ to produce utterances is demonstrated by the fact that the vast majority of speech errors reorder entire segments rather than parts of segments or features (Shattuck- Hufnagel and Klatt 1979). This speech error evidence does not, however, reveal whether a segment-sized unit is employed at all levels in the plan for producing an utterance, nor what principles govern the coordination of articulators.

Since more than one articulator is moving or otherwise active at any point in time in a speech event, the notion of a segment implies a specification for how these independent movements are coordinated. However, articulatory records of speech events cannot be divided into discrete intervals in which all movement of articulators for each segment begins and ends at the same time, nor can these records be obtained by a simple mapping from discrete underlying phonological segments. The continuous and coarticulatory properties of actual speech events instead make it very unlikely that the plans for producing them consist of nothing more than a string of discrete segments (Ohman 1966, 1967; Fowler 1980; Browman and Goldstein 1985, 1986; but cf. Henke 1966; Keating 1985).