Semantic Embodiment: Decoding Action Words through Topographic Neuronal Representation within Brain-Constrained Network

Poster Session B - Sunday, April 14, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Maxime Carriere¹ (maxime.carriere@fu-berlin.de), Tomasello Rosario², Pulvermüller Friedemann³; ¹Freie Universität, ²Berlin School of Mind and Brain, ³Cluster of Matters of Activity

Previous neuroimaging research suggests that action related words semantically related to actions typically performed with the upper or lower extremities, e.g. “grasp” vs “kick”, activate different parts of the cortical motor system consistent with the somatotopic representation of these body parts. This differential activation could reflect differences in semantic processing between symbol types or, alternatively, epiphenomenal activity related to but following semantic processing, for example imagining a scene related to symbol meaning. To address this controversy, we here simulate semantic learning of action words in real world environments where word meaning is grounded in bodily experiences. We find that a brain constrained network with area structure imitating cortical motor, language and connector hub areas builds neuronal circuits, or cell assemblies (CAs), with different distributions for arm and leg related words. Activation of specific word forms replicates the differential activity of dorsolateral motor areas known from neurometabolic (fMRI) and the fast spreading of activity shown by related neurophysiological (MEG/EEG) studies. Our results indicate that these differential activations are not “epiphenomenal” but reflect fast access to semantic memory circuits that store symbol forms together with their meaning.

Topic Area: LANGUAGE: Semantic