Talker variability, auditory attention, and speech processing efficiency

Abstract

A classic finding in speech perception is that speech is processed more efficiently from a single, continuous talker than from mixed talkers. Given enormous variation in the acoustic realization of speech, it is thought that talker adaptation is necessary to ascertain the mappings between talkers’ speech acoustics and listeners’ abstract phonological representations. However, a suite of empirical studies from our laboratory may suggest a predominately attention-based explanation for effects attributed to talker adaptation: using behavioral experiments, neuroimaging, and noninvasive brain stimulation, we have probed how speech processing efficiency under talker variability is affected by temporal, phonological, contextual, and expectational factors. We find that processing benefits from talker continuity are automatic and feedforward, depending on temporal continuity in the source of speech but not the amount of talker-specific phonetic detail or listeners’ expectations about source continuity. Correspondingly, processing costs from talker discontinuity occur even when phonetic contrasts are unambiguous across talkers but are insensitive to the magnitude of phonetic variability, amount of preceding exposure to a talker, or top-down expectation about discontinuity. We consider how domain-general models of attention and auditory streaming may parsimoniously account for these differences in the speech processing efficiency between single and mixed talkers.A classic finding in speech perception is that speech is processed more efficiently from a single, continuous talker than from mixed talkers. Given enormous variation in the acoustic realization of speech, it is thought that talker adaptation is necessary to ascertain the mappings between talkers’ speech acoustics and listeners’ abstract phonological representations. However, a suite of empirical studies from our laboratory may suggest a predominately attention-based explanation for effects attributed to talker adaptation: using behavioral experiments, neuroimaging, and noninvasive brain stimulation, we have probed how speech processing efficiency under talker variability is affected by temporal, phonological, contextual, and expectational factors. We find that processing benefits from talker continuity are automatic and feedforward, depending on temporal continuity in the source of speech but not the amount of talker-specific phonetic detail or listeners’ expectations about source continuity. Correspondi...