Compensation for assimilatory devoicing and prosodic structure in German fricative perception

Abstract

An important source of phonetic variation in German fricatives is progressive voice assimilation: the lenis fricatives /v/ and /z/ are devoiced after /t/ across word boundaries. This process is gradient and moderated by prosodic structure: fricatives are more devoiced after smaller prosodic boundaries. We present three phoneme identification experiments, investigating how listeners deal with assimilatory devoicing and its prosodic conditioning. Fully voiced, partially devoiced and completely devoiced fricatives had to be identified as fortis or lenis in different segmental (assimilation versus non-assimilation context) and prosodic (after a word versus a phrase boundary) environments. Results indicate that 1. listeners compensate for assimilatory devoicing in judging partially devoiced fricatives more often as lenis in assimilation context than in non-assimilation context; 2. prosodic structure plays a role in compensation for assimilation: more devoiced fricatives are more often judged as lenis after word boundaries than after phrase boundaries in assimilation context, and 3. the influence of prosody is constrained by lexical effects: we found prosodic conditioning of compensation for the devoicing of /v/, contrasting with /f/, but not of /z/. These findings suggest that an on-line prosodic analysis of spoken language contributes to the resolution of lexical ambiguity arising from progressive voice assimilation.