Gradient phonetic implementation of regressive voicing assimilation in Catalan heterosyllabic two- and three-consonant clusters

Abstract

Electroglottographic data for word-final obstruents in C#C and CC#C sequences with a word-initial voiced consonant indicate that regressive voicing adaptation is a categorical and thus assimilatory process for most Catalan speakers. Word-final obstruents are planned as voiced since they exhibit full voicing or else an initial voicing period which is longer than the voicing lag associated with the vowel preceding the cluster. Segmental duration may also be used by speakers for realizing word-final obstruents as voiced in C#C but not in CC#C clusters. The phonetic implementation of voicing assimilation proceeds gradually: voicing for word-final obstruents differs considerably among speakers and is less for fricatives than for stops and whenever the word-initial consonant is an obstruent or an approximant than when it is a nasal, a lateral or an alveolar trill. The study also reveals that those C1 realizations which are less prone to acquire voicing show more token-totoken variability in voicing and segmental duration and therefore appear to be less tightly controlled by speakers. In conjunction with data from other studies, these Catalan data suggest that speakers and languages with prevoiced stops may differ with respect to the more or less gradient phonetic implementation of the regressive voicing assimilation process in heterosyllabic consonant clusters.