Effects of type, token, and talker variability in speech processing efficiency

Abstract

Phonetic variability across talkers imposes additional processing costs during speech perception, evident in performance decrements for mixed- versus single-talker speech. However, within-talker phonetic variation across different utterances is another, relatively unexplored source of variability in speech, and it is unknown how processing costs from within-talker variation compare to those from between-talker variation. Cognitive consequences of talker variability are also mostly measured from two-alternative forced-choice tasks, whereas naturalistic speech processing occurs in a much larger decision space. Do talker-variability effects scale when both the stimuli and the decision space are more complicated? Here, we measured response times in a speeded word identification task that factorially manipulated three dimensions of speech variability: number of talkers (one versus four), number of target word choices (two versus six), and number of talker-specific exemplars per word (one versus eight). Across all eight experimental levels, larger decision spaces led to significantly slower word identification. Word identification was also slower in conditions with mixed talkers and conditions with multiple exemplars. This pattern of interactions suggests complex processing relationships between type, token, and talker variability and provides preliminary evidence for how both within- and between-talker variability impose additional processing costs in speech perception.