Introduction

Many key issues in speech perception revolve around two questions: first, what is the size of the basic phonological unit of perception (e.g. word, phoneme, feature); and, second, how general are the psycho-acoustic properties that are associated with those units. Answers to these questions vary widely. For example, the standard distinctive feature theory of Stevens and colleagues (Stevens 1989; Stevens and Keyser 1989; Stevens and Blumstein 1981) opts for units of very small size and great generality. This theory states that the transduction from signal to symbol takes place through a universal set of small units, the distinctive features, each of which is associated with well-defined psychoacoustic properties. At the opposite end of both the size and generality scales, researchers such as Goldinger (1997; see also Johnson 1997 and Pisoni 1997) suggest that the transduction from signal to symbol is mediated by an ‘episodic lexicon’, involving memory traces of exemplars (i.e. individual tokens of individual words of individual speakers). In this approach the basic phonological units are large (word-sized) elements of low generality.

My own research on speech perception (Nearey 1990, 1992, 1997, in press) has focused on phoneme-sized elements as the basic units of symbolic transduction. The acoustic patterns associated with these units are relatively general, but are clearly less so (and allowably more language-specific – yet a third sense in which my position is weaker) than Stevens's universal feature set (see Nearey 1998).