Methodological advances in encoding models of brain: Applying temporal response functions to magnetoencephalography for written text perception

Abstract

Recent advancements in cognitive neuroscience have expanded the tools available to study language processing beyond traditional event-related potentials (ERPs), and introduced new methods, such as the Temporal Response Function (TRF). TRF allows a nuanced investigation of brain dynamics by modeling neural responses as a convolution of stimuli with self-optimized TRF curves. While TRF has been successfully applied in auditory speech research, its application to written language processing remains unexplored. In this study, we introduce a novel approach to TRF analysis in reading using magnetoencephalography (MEG), leveraging its high spatial resolution. We employed the Rapid Serial Visual Presentation (RSVP) paradigm to present text word-by-word, avoiding eye-movement artifacts and enabling precise timing. By integrating predictors, such as word onset, word length, and semantic dissimilarity (SD), we explored both low- and high-level linguistic processing during reading. Our analysis of 17 participants revealed significant early neural responses within 150 ms post-word onset, associated with semantic processing, supporting the notion of rapid semantic integration in written text perception. This study serves as a proof of concept for using TRF in reading research, extending its utility from auditory to written language domains. Our findings contribute to understanding the neural mechanisms underlying reading and suggest potential applications for studying populations with reading impairments, such as dyslexia.