Decomposing Parts of Speech: Neural Encoding of Semantic and Grammatical Dimensions and Their Combinational Structure

Semantic root and grammatical function independently predict neural responses during naturalistic comprehension, with interaction effects adding extra variance and prototypicality sensitivity in two of three MEG datasets.

• Why it matters: Part-of-speech labels mix meaning, grammatical role, and their combinations, leaving unclear whether brain activity tracks either dimension alone or their mapping structure.
• What they did: Croft’s typological framework decomposed PoS into semantic root type and grammatical function, fitting time-resolved MEG encoding models on ~10 hours of narrative speech from three participants.
• The result: Root effects were mainly left-lateralized, function effects were more bilateral, and interaction effects clustered in left temporal/temporoparietal cortex; prototype nouns and verbs showed distinct coefficient time courses, especially in left superior temporal cortex.

AI summary of the abstract · may contain errors · not medical advice · read the paper.