The cognitive and neural bases of processing talker variability in speech perception

Abstract

Talker variability is the principal source of phonetic variability in speech signals, resulting in a lack of consistency in acoustic-to-phonetic mapping. Previous studies have repeatedly demonstrated that listeners incur additional processing costs in order to successfully extract phonetic information in the face of such significant variability across talkers. These costs manifest as lower accuracy and slower response time in speech perception tasks and increased neural response in auditory brain regions in mixed- relative to single-talker settings. However, it is unknown how talker adaptation processes unfold over time and how they interact with the amount and nature of information preceding the target speech. Moreover, neuroimaging results alone cannot establish causal roles for auditory regions in talker adaptation. The first series of behavioral experiments investigates the effect of preceding context on talker adaptation by comparing response times in auditory word identification tasks between single- and mixed-talker conditions while manipulating the following attributes of carrier speech preceding the target word: its duration, its amount of talker-specific phonetic detail, and its temporal continuity to the target word. Results indicate that the duration of the carrier, but not the richness of detail within it, has a significant effect on talker adaptation, and that temporal continuity between the context and the target word facilitates talker adaptation. Extending these findings, the next study examines how processing talker variability improves as a function of the duration of continuous speech from a talker. Results demonstrate that the facilitatory effect of immediately preceding speech on talker adaptation linearly increases up to 600 ms, but longer exposure to continuous speech from a talker had no further facilitatory effect on processing in mixed-talker contexts. The last study investigates the causal role of left superior temporal gyrus (STG) in processing talker variability by applying high-definition transcranial direct current stimulation (HD-tDCS) to left STG while participants performed an auditory word identification task. Neurostimulation of left STG selectively decreased the facilitatory effect of immediately preceding context on talker adaptation. Discussed in light of models of speech perception, these studies together suggest a role for stimulus-driven auditory attention in behavioral phenomena known as talker adaptation.