Introduction

The past decade or so has seen an increasing interest in developing universal models of phonological structure and representation. Research in this area has produced several theoretical concepts that seem to be universally relevant. One such concept is the mora, which is now claimed to play a crucial role not only in “mora-timed” languages, such as Japanese, but also in the phonological description of English and other languages, e.g. in accounting for stress patterns and compensatory lengthening (Hyman, 1985; Hayes, 1989). Despite this popularity, one can point out several crucial differences in the role played by the mora in Japanese and in other languages (for a review, see Kubozono, 1992, 1993).

However, the mora's role in Japanese has been studied largely from the viewpoint of speech production, through analysis of speech timing, speech errors, accent assignment rules, morphological patterns, etc., while little is known about its relevance to speech perception. With this background, this paper explores the role of the mora in speech perception in Japanese, by examining the ways in which auditory stimuli are segmented by native speakers of the language.

I shall first report the results of a series of experiments involving word blends, which demonstrate that the strategy used by native speakers of Japanese in segmenting monosyllabic words is crucially different from that used by native speakers of English. I shall argue that this striking difference can best be explained if the mora is assumed to be a basic phonological unit in Japanese.