Task structure tailors the geometry of neural representations in human lateral prefrontal cortex

Area of Science:

• Cognitive Neuroscience
• Neuroimaging
• Computational Neuroscience

Background:

• The lateral prefrontal cortex (lPFC) is crucial for flexible task representation.
• Understanding the principles governing lPFC representational geometry is key to cognitive neuroscience.
• Task structure significantly influences neural representations.

Purpose of the Study:

• To investigate how the brain's lateral prefrontal cortex (lPFC) encodes information for tasks with different structures.
• To uncover the principles shaping representational geometry in the lPFC across varying task complexities.

Main Methods:

• Functional magnetic resonance imaging (fMRI) was employed to measure brain activity.
• Pattern analyses were used to decode neural representations.
• Participants performed two distinct categorization tasks: one flat and one hierarchical.

Main Results:

• The lPFC encodes task-relevant information using tailored geometries of intermediate dimensionality.
• Representational geometries enhance the separability of task-relevant variables and abstract subsets of information.
• In flat tasks, a global axis abstracts categories; in hierarchical tasks, a global axis abstracts context.

Conclusions:

• The lPFC dynamically tailors its representational geometry based on task demands.
• These findings reveal generalizable principles of neural representation for cognitive flexibility.
• Task-specific geometries optimize information processing in the lateral prefrontal cortex.