Effects of semantic context and regional dialect variation on speech intelligibility in noise.

Abstract

Speech intelligibility in noise is affected by the semantic content of the utterance; semantically predictable utterances are more intelligible than semantically unpredictable or anomalous utterances. However, dialect variation interacts with semantic predictability in production; talkers produce more dialect‐specific variants in semantically predictable contexts and more standard variants in semantically unpredictable contexts. The goal of the current study was to explore the interaction between dialect variation and semantic context in perception. Participants were asked to transcribe the final word of semantically predictable, unpredictable, and anomalous sentences produced by talkers from four regional dialects of American English and mixed with speech‐shaped noise. The listeners exhibited an intelligibility benefit in the predictable context relative to the other two semantic contexts, regardless of talker dialect. However, differences in intelligibility between the dialects were reduced in the unpredictable and anomalous contexts relative to the predictable context, suggesting greater online adaptation to dialect variation in the unpredictable and anomalous contexts than the predictable context. These results are consistent with the finding that dialect differences may be attenuated in production in low predictability contexts, and suggest that perceptual dialect adaptation is affected by fine‐grained intratalker acoustic variation.