A general role for medial prefrontal cortex in event prediction

Area of Science:

• Computational neuroscience
• Cognitive neuroscience

Background:

• The predicted response-outcome (PRO) model posits medial prefrontal cortex (mPFC) learns action-outcome contingencies.
• Recent data suggest mPFC signals predictions and errors even without action contingency.

Purpose of the Study:

• To generalize the PRO model beyond action-contingent outcomes.
• To demonstrate the generalized PRO model's broader explanatory power for mPFC function.

Main Methods:

• Computational modeling using a generalized PRO framework.
• Simulations to test model predictions against existing mPFC data.

Main Results:

• The generalized PRO model successfully predicts outcomes irrespective of action contingency.
• Simulations show the model accounts for human ERP, fMRI, and macaque single-unit data.
• The model's predictions align with a wider range of mPFC research findings.

Conclusions:

• mPFC learns to predict salient events generally, not just action-contingent outcomes.
• This generalized model provides a unified framework for mPFC, linking it to reinforcement learning, Bayesian learning, and cognitive control.