Dopamine-dependent prediction errors underpin reward-seeking behaviour in humans.

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology

Background:

• Instrumental learning theories focus on using success and failure to refine future decisions.
• Reward prediction errors are crucial for updating action values in learning.
• Dopamine is implicated in animal learning by modulating cortico-striatal synaptic efficacy.

Purpose of the Study:

• To investigate the direct link between dopamine, striatal activity, and behavioral choice in humans during instrumental learning.
• To determine if dopamine modulates reward prediction errors in the human striatum.

Main Methods:

• Administered drugs that enhance (L-DOPA) or reduce (haloperidol) dopaminergic function in human subjects.
• Measured striatal activity and behavioral choices during an instrumental learning task.
• Utilized a standard action-value learning algorithm to model behavioral choices.

Main Results:

• The magnitude of reward prediction error in the striatum was significantly modulated by L-DOPA and haloperidol.
• Subjects treated with L-DOPA showed a higher tendency to select the most rewarding action compared to those treated with haloperidol.
• The learning algorithm accurately predicted behavioral choices when incorporating the magnitude of prediction errors.

Conclusions:

• Dopamine-dependent modulation of striatal activity explains how the human brain uses reward prediction errors for decision improvement.
• This study provides direct evidence linking dopamine, striatal function, and instrumental learning in humans.