Basal ganglia neurons dynamically facilitate exploration during associative learning

Area of Science:

• Neuroscience
• Behavioral Neuroscience
• Computational Neuroscience

Background:

• The basal ganglia (BG) are implicated in associative learning, linking stimuli to rewarding responses.
• Reinforcement learning models describe BG function in evaluating and selecting profitable actions.
• The striatum integrates contextual cues and reward prediction errors, but response variability generation remains unclear.

Purpose of the Study:

• To investigate the mechanistic role of the globus pallidus internus (GPi) in generating response variability during associative learning.
• To explore how basal ganglia output influences exploration versus exploitation.

Main Methods:

• Single-neuron recordings from the globus pallidus internus (GPi) in nonhuman primates (Macaca mulatta).
• Task involved motor associative learning, requiring pairing stimuli with rewarding responses.
• Analysis focused on GPi neuron activity in relation to learning stages and behavioral choices.

Main Results:

• A subset of GPi neurons exhibited learning-related firing rate changes.
• Lower GPi firing predicted exploratory behavior, while higher firing predicted exploitative responses.
• GPi activity shifted from permissive to suppressive as learning progressed.

Conclusions:

• Basal ganglia output, specifically from the GPi, initially permits response exploration during associative learning.
• Increased GPi activity suppresses alternative responses, promoting exploitation of learned behaviors.
• This provides a mechanism for how the basal ganglia refine action selection during learning.