Efficient and dynamic neural geometry of value and modality encoding in the primate putamen for value-guided behavior

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• The basal ganglia efficiently process diverse reward values from multiple sensory inputs using limited neural resources.
• Convergence of sensory information in neurons raises questions about preserving modality-specific details and overall information quality.

Purpose of the Study:

• To investigate how bimodal value neurons in the macaque putamen represent and preserve information from tactile and visual reward inputs.
• To determine if shared abstract representations in these neurons balance neural efficiency with information fidelity.
• To explore the relationship between the dynamic neural geometry of these representations and value-guided behavior.

Main Methods:

• Recording neural activity from bimodal and unimodal value neurons in the macaque putamen.
• Analyzing population representations of value and modality information.
• Examining the dynamic shifts in neural geometry during value-guided tasks.
• Correlating neural dynamics with behavioral adaptation and learning.

Main Results:

• Bimodal value neurons in the putamen efficiently preserve both value and modality information through shared abstract representations.
• Population representations generalized across modalities and values, forming low-dimensional, dynamically shifting neural geometries.
• Faster transformation into shared-value representations in bimodal neurons correlated with well-adapted, value-guided behavior.
• This correlation was absent in unimodal value neurons.

Conclusions:

• Bimodal value neurons in the putamen are crucial for balancing neural processing efficiency and information fidelity.
• Shared neural representations and their dynamic changes facilitate cognitive states necessary for effective value-guided behavior.
• These findings highlight the sophisticated coding strategies employed by the basal ganglia for reward processing.