Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Neural encoding of articulatory features during naturalistic speech perception

Poster Session F - Tuesday, March 10, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Olivia Bizimungu¹ (), Laura Gwilliams², Sylvain Baillet^1,3; ¹McGill University, ²Stanford University, ³Université de Montréal

The role of the speech production system during speech perception has long been debated. Although previous findings support that phonetic perception recruits cortical articulatory representations, this interaction has largely been studied in unnatural listening paradigms. For example, finding that motor cortex activates somatotopically when a listener hears individual syllables, and perceptual judgements of isolated words can be biased according to the listener’s own orofacial configuration. It remains unknown, however, whether sensorimotor integration is specific to the active perception of basic linguistic stimuli, or whether naturalistic listening to continuous speech, with much richer predictive structure, involves similar processes. In the present study, we tested whether articulatory representations are encoded during naturalistic speech comprehension. We re-analyzed previously obtained magnetoencephalography (MEG) data from 11 healthy participants listening to various TED talks (total listening time = 60.27 min). Using temporal response functions (TRF), we modelled continuous brain responses to acoustic, articulatory, and predictive features derived from the speech input. Comparing the accuracy of multi-feature TRFs to acoustic-only models revealed a subset of MEG sensors that encode incoming speech sounds’ place of articulation. Crucially, articulatory encoding was strongest in frontal, central and parietal sensors, roughly corresponding to motor and somatosensory areas; and was spatially distinct from temporal sensors that preferentially tracked acoustics. In contrast, encoding of contextual predictions increased with acoustic sensitivity across MEG sensors. These findings expand our understanding of sensorimotor involvement in everyday listening and provide insight into the distinct features that drive neural speech processing.

Topic Area: LANGUAGE: Other