Reversal learning and dopamine: a bayesian perspective.

Area of Science:

• Neuroscience
• Cognitive Science
• Behavioral Economics

Background:

• Reversal learning involves inhibiting previously rewarded actions.
• Dopamine and orbitofrontal cortex lesions are linked to reversal learning deficits.
• Previous studies often used animals with limited reversal experience, confounding learning of the task structure itself.

Purpose of the Study:

• To investigate reversal learning strategies in monkeys with extensive experience using a probabilistic two-arm bandit task.
• To analyze the effects of dopamine manipulation on reversal performance using a Bayesian approach.
• To clarify animal strategies and the impact of specific drugs on decision-making during reversals.

Main Methods:

• Developed a Bayesian analysis approach for behavioral data.
• Utilized a probabilistic two-arm bandit reversal learning task with monkeys.
• Administered haloperidol, levodopa (l-DOPA), and placebo to assess drug effects.

Main Results:

• Monkeys exhibited rapid switching between choices at reversal points, deviating from gradual reinforcement learning (RL) value updates.
• Haloperidol administration led to a stronger prior expectation of reversal occurrence compared to l-DOPA or placebo.
• This enhanced prior was appropriate given the animals' extensive experience with mid-block reversals.

Conclusions:

• Bayesian analysis effectively clarifies animal strategies in reversal learning.
• Haloperidol influences the integration of prior knowledge with current evidence during choice reversals.
• Findings suggest dopamine modulates the weighting of prior expectations in decision-making.