Impact of talker adaptation on speech processing and working memory

Abstract

Talker adaptation is known to facilitate immediate speech recognition. However, it is unclear whether and how talker adaptation also facilitates working memory for speech. Using electroencephalography (EEG) during a delayed recall of digit span task, we investigated whether talker adaptation facilitates working memory performance. We also investigated whether neural alpha oscillatory power reflects facilitatory effects of talker adaptation. Listeners encoded sequences of seven randomly ordered digits, recalling them after a 5-s retention. Digit sequences were spoken by either a single talker or multiple talkers, and were presented at either faster or slower speech rates (0-ms vs. 500-ms inter-digit intervals). Overall, listeners responded faster and more accurately for single-talker sequences compared to multiple-talker sequences. Especially for the faster presentation rate, listeners were more efficient (faster and more accurate) in recalling sequences spoken by a single talker. For the faster presentation rate, processing digit sequences spoken by single vs. multiple talkers also elicited reduced alpha power during both speech encoding and working memory retention. These results suggest that talker adaptation reduces cognitive effort during both speech encoding and memory retention, thereby producing more efficient working memory for speech information, especially when listeners process speech rapidly.