A two-dimensional space of linguistic representations shared across individuals

Area of Science:

• Neuroscience
• Cognitive Science
• Linguistics

Background:

• The brain's left hemisphere, particularly frontal and temporal regions, supports language processing.
• Previous research has not identified clear organizational principles or spatial distinctions within these language areas.
• Existing studies show similar responses across language areas during both controlled and naturalistic comprehension.

Purpose of the Study:

• To uncover finer-grained organizational principles of language processing within the brain.
• To identify a structured representational basis for linguistic information.
• To investigate how brain responses to language generalize across individuals.

Main Methods:

• Applied data-driven decomposition to ultra-high-field (7T) functional Magnetic Resonance Imaging (fMRI) data.
• Analyzed brain responses from eight participants listening to 200 diverse sentences.
• Utilized cross-validation to identify shared neural structures across participants.

Main Results:

• Identified two generalizable components explaining 32% of brain response variance to sentences.
• Component 1 correlates with linguistic processing difficulty.
• Component 2 correlates with the abstractness of meaning, with a stronger presence in temporal regions.

Conclusions:

• Language processing in the brain has an interpretable, low-dimensional, and spatially structured representational basis.
• Meaning abstractness and processing difficulty are systematically organized within frontal and temporal language areas.
• These findings advance the understanding of fine-scale linguistic representations and neural organization.