Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Cortical Organization of Shared Neural Computations for Reading and Auditory Speech Perception
Poster E119 in Poster Session E, Thursday, October 26, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Also presenting in Lightning Talks E, Thursday, October 26, 10:00 - 10:15 am CEST, Auditorium
Terri L. Scott1, Matthew K. Leonard1, Edward F. Chang1; 1UCSF
Reading is a skill that does not naturally develop; it must be acquired by learning to associate novel visual input with neural systems previously devoted to auditory speech comprehension. Behavioral studies have shown that reading automatically activates some knowledge of the sound of a word, even in adult readers; however, it is unknown what types of functions the auditory speech perception system supports during proficient, naturalistic reading. Here, we used intracranial electrocorticography (ECoG) to investigate not only the spatial overlap of cortical activation during reading and listening to speech, but also the temporal properties of neural activity in these shared regions. Eleven participants undergoing clinical monitoring for intractable epilepsy listened to and read naturalistic sentences while ECoG signals were recorded from high-density electrode grids. We found electrodes with significant high-gamma amplitude (HGA) responses to both speech and reading across several frontotemporal brain areas, with the largest proportion localized to the superior temporal gyrus (STG). Individual subject analyses revealed that shared speech and reading responsive electrodes are consistently observed in posterior STG. We used non-negative matrix factorization to cluster common temporal features of HGA timecourses during reading and found that a key dynamical feature that distinguishes shared reading responses from modality-specific reading responses across brain areas is an increase in high-gamma activity over the course of the sentence. We found similar dynamics in these same electrodes during speech perception. Speech responses in shared electrodes show more gradual build-up of HGA over time compared to the larger population of electrodes with modality-specific speech responses. Furthermore, feature-based temporal receptive field modeling of HGA during speech perception revealed that shared electrodes, when compared to those with modality-specific responses, are less sensitive to stimulus driven features such as sentence onset and auditory envelope and more sensitive to word-level features such as lexical frequency and surprisal. Taken together, these results suggest that neural computations active during speech perception and reading occupy a space in the processing hierarchies of both tasks that has access to word-level and sentence-level representations. The most robustly and consistently identified area showing this effect was posterior STG, thus this work sheds new light on how sentence context may shape the dynamics of speech perception as well as the functional role auditory speech systems play during proficient reading.
Topic Areas: Reading, Speech Perception