Slide Slam O10
Using TMS to evaluate a causal role for right posterior temporal cortex in talker-specific phonetic processing
Sahil Luthra1, Hannah Mechtenberg1, Cristal Giorio1, Rachel M. Theodore1, James S. Magnuson1,2,3, Emily B. Myers1; 1University of Connecticut, 2Basque Center on Cognition Brain and Language, 3Ikerbasque - Basque Foundation for Science
Individual talkers differ substantially in how they produce speech sounds. For instance, talkers vary in the specific voice-onset time (VOT) they produce for voiceless stop consonants (/p/, /t/, /k/ in English; Allen, Miller, & DeSteno, 2003) and these differences are stable across linguistic contexts (Theodore, Miller, & DeSteno, 2009). Listeners are sensitive to such differences and can explicitly indicate what variation is typical of a talker (e.g., Allen & Miller, 2004). Theories suggest that maintenance of talker-specific phonetic details benefits speech perception, with perception guided by listeners’ knowledge of whether a given talker is likely to produce a specific phonetic variant. In a previous fMRI study, Myers and Theodore (2017) exposed listeners to two talkers: one who produced /k/ with short VOTs and one with relatively longer VOTs. Through exposure, listeners learned which variant was typical of each talker. In a subsequent in-scanner phonetic categorization task, listeners heard typical and atypical variants from each talker. A cluster in the right middle temporal gyrus (RMTG) was sensitive to how typical the variant was for each talker, consistent with a burgeoning literature suggesting that the right hemisphere may play a key role in conditioning phonetic identity on talker information (Luthra, 2021). The current work used transcranial magnetic stimulation (TMS) to test whether the RMTG plays a causal role in processing talker-specific phonetic variation (N=27). In each of three experimental blocks, listeners heard two talkers: one who produced a voiceless stop with short VOTs and one with relatively longer VOTs. During training, listeners made a four-alternative forced-choice to indicate both who was talking (e.g., Peter or Sheila) and what word they said (e.g., gain or cane). Prior to each training trial, listeners received TMS. A different stimulation site was used for each block, with sites at the RMTG, left MTG, and scalp vertex (control). At test, listeners heard speech from one of the two trained talkers and indicated which of two variants (short-VOT or long-VOT) was typical for that talker. During training, listener accuracy was at ceiling for talker identification, regardless of stimulation site. Though listeners were also near ceiling in their phonetic decisions, logistic regression analyses indicated a trend towards an effect of stimulation (p = 0.06). Specifically, listeners were less accurate in deciding which word they heard (e.g., gain or cane) after RMTG stimulation (96.3 ± 18.9; mean ± SD) compared to LMTG (98.8 ± 11.0) and control (97.9 ± 14.4) stimulation. However, stimulation did not influence test performance; regardless of stimulation site, listeners learned which phonetic variant was typical of each talker. The relatively modest effects observed here are consistent with the view that talker-specific phonetic processing is supported by the posterior temporal cortex bilaterally. When the RMTG was stimulated, listeners may have been able to rely on the left hemisphere to compensate (and vice versa). Thus, even though the RMTG is functionally recruited when listeners condition phonetic identity on talker information, impairing its function may only have modest consequences for talker-specific phonetic processing.