A transient high-dimensional geometry affords stable conjunctive subspaces for efficient action selection.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Flexible action selection is crucial for adapting behavior to changing contexts.
• Cognitive control mechanisms are hypothesized to separate similar neural states based on context.
• Stable neural representations are necessary for robust and time-invariant action selection.

Purpose of the Study:

• To investigate the neural geometry and dynamics underlying flexible action selection in the human brain.
• To determine how control representations constrain context-dependent behavior.
• To explore the relationship between neural state-space properties and performance in a flexible action selection task.

Main Methods:

• Electroencephalography (EEG) decoding methods were employed to analyze neural activity.
• Participants performed a context-dependent action selection task.
• A forced response procedure was used to probe neural trajectories during action selection.

Main Results:

• A transient expansion of representational dimensionality was observed before successful responses, separating conjunctive subspaces.
• Neural dynamics stabilized within a specific time window.
• Entry into this stable, high-dimensional neural state predicted individual trial performance.

Conclusions:

• The study establishes the critical role of neural geometry and dynamics in enabling flexible behavioral control.
• Findings suggest that transient increases in representational dimensionality and subsequent stabilization are key neural mechanisms for context-dependent action selection.
• The research provides insights into the brain's computational strategies for adaptive behavior.