Neural evidence for voice-specific representations during listening and silent reading

Poster Session D - Monday, April 15, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Kate Revill¹ (krevill@emory.edu), Lynne Nygaard¹; ¹Emory University

Previous studies (Alexander & Nygaard, 2008) have shown evidence for voice-specific auditory imagery during silent reading. Neuroimaging studies have also demonstrated activation in temporal voice areas (TVA) during silent reading (Yao et al 2011) but have not been designed to determine whether the imagery is talker-specific or instead represents a generic voice. This study investigates the neural correlates of voice representation and talker-specific auditory imagery using naturalistic materials. Prior to scanning, participants heard two male talkers read text passages from autobiographies of Mark Twain and Benjamin Franklin. Talker and author were counterbalanced across participants. The talkers’ voices differed in mean pitch and speaking rate, and participants were able to identify the talkers following initial exposure. During scanning, participants (N = 20) read passages they were told were written by the two talkers, interspersed with additional listening passages from each talker. Separate functional localizer tasks were used to identify areas involved in auditory language processing (Scott et al 2016) and voice processing (Pernet et al 2015). An SVM classifier showed robust above chance discrimination of the two talkers’ voices during listening intervals in multiple ROIs across the language network, including bilateral TVA. Several ROIs in the left temporal lobe additionally showed evidence of above-chance discrimination of talker identity during silent reading of passages that were ‘authored’ by the two talkers but that participants had never heard spoken aloud. These results suggest that voice-specific representations are active during silent reading and constitute a form of auditory imagery during reading.

Topic Area: LANGUAGE: Other