Voxelwise Encoding Models Show That Cerebellar Language Representations Are Highly Conceptual.

Area of Science:

• Neuroscience
• Cognitive Science
• Psycholinguistics

Background:

• The cerebellum's role in language understanding is increasingly recognized, challenging earlier theories focused on low-level processing.
• Recent studies indicate cerebellar involvement in complex cognitive tasks, necessitating a re-evaluation of its linguistic functions.

Purpose of the Study:

• To investigate the nature of language representations within the cerebellum using advanced neuroimaging techniques.
• To compare cerebellar language processing with cortical representations, particularly focusing on semantic hierarchies.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was employed with five participants (3 men, 2 women) passively listening to 5 hours of natural language stimuli.
• Voxelwise encoding models were fitted using feature spaces spanning from acoustic to high-level conceptual language processing.
• Model prediction performance was evaluated on separate blood-oxygen-level-dependent (BOLD) data to assess representational content.

Main Results:

• Cerebellar responses to language were predominantly explained by high-level conceptual features, not low-level acoustic or phonemic ones.
• The cerebellum exhibited a higher proportion of voxels representing social semantic categories (e.g., 'social,' 'people') compared to cortical areas.
• Conversely, the cerebellum showed lower representations for other semantic categories like 'mental,' 'concrete,' and 'place' words than the cortex.

Conclusions:

• Cerebellar language processing operates at a conceptual level, with a distinct preference for social semantic information.
• These findings suggest the cerebellum contributes to general cognitive processing rather than language processing exclusively.
• Cerebellar language representations differ significantly from cortical representations, highlighting its unique role in cognitive functions.