Frontal theta links prediction errors to behavioral adaptation in reinforcement learning.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Action monitoring research identifies frontocentral voltage deflections (feedback-related negativity, FRN) to negative feedback.
• The FRN is hypothesized to represent neural responses to prediction errors in reinforcement learning.
• Single-trial FRN analysis is challenging with traditional ERP methods, but theta oscillations offer a viable alternative.

Purpose of the Study:

• To investigate the single-trial relationship between neural activity and expectation violation during reinforcement learning.
• To explore the role of theta oscillations in medial prefrontal cortex in response to prediction errors.
• To link single-trial theta activity to behavioral adaptation, specifically changes in reaction time.

Main Methods:

• Recorded electroencephalography (EEG) during a probabilistic reinforcement learning task.
• Utilized a Q-learning computational model to calculate single-trial reward prediction errors.
• Analyzed single-trial theta oscillatory activity in relation to prediction errors and reaction time changes.

Main Results:

• Medial and lateral frontal theta activities were found to correlate with the magnitude of reward prediction errors (both positive and negative).
• Medial frontal theta activity was associated with reaction time slowing after errors, reflecting adaptation.
• Lateral frontal theta activity was linked to reaction time speeding for correct choices, suggesting working memory involvement.

Conclusions:

• Theta oscillations in distinct frontal brain regions encode reward prediction errors.
• These neural signals are crucial for adaptive behavioral adjustments, with different frontal areas supporting different adaptation strategies.
• The findings provide a single-trial neural basis for understanding prediction error signaling and its role in learning and behavior.