Dynamically shifting from compositional to conjunctive brain representations supports cognitive task learning.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Neural representations must adapt from novel to practiced task performance.
• The geometric changes in neural representations during this transition are not well understood.
• Existing theories lack a comprehensive explanation for cognitive task learning dynamics.

Purpose of the Study:

• To investigate the shift in neural representation geometry during cognitive task learning.
• To test the hypothesis that practice involves a transition from compositional to conjunctive representations.
• To explore the subcortical and cortical origins and spread of these representational changes.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was used to monitor brain activity during the learning of multiple complex tasks.
• Analysis focused on the geometry of neural representations, specifically the shift from compositional to conjunctive patterns.
• Behavioral data on task performance and cross-task interference were collected.

Main Results:

• Functional MRI data confirmed a dynamic shift from compositional to conjunctive representations during learning.
• This representational shift was correlated with reduced cross-task interference and improved behavioral performance.
• Conjunctive representations were observed to originate in subcortical regions (hippocampus, cerebellum) and subsequently spread to the cortex.

Conclusions:

• Practice-related learning involves a transition from compositional to conjunctive neural representations.
• This shift optimizes task-specific neural activity, reducing interference and enhancing performance.
• The findings extend memory systems theories by incorporating subcortical-cortical dynamics in cognitive task learning.