Asymmetric coding of reward prediction errors in human insula and dorsomedial prefrontal cortex

The signed value and unsigned salience of reward prediction errors (RPEs) are critical to understanding reinforcement learning (RL) and cognitive control. Dorsomedial prefrontal cortex (dMPFC) and insula (INS) are key regions for integrating reward and surprise information, but conflicting evidence for both signed and unsigned activity has led to competing proposals for the nature of RPE representations in these brain areas. Recently, the distributional RL theory (dRL) has been used to explain RPE coding diversity in the rodent midbrain by proposing that dopaminergic neurons have differential sensitivity to positive and negative RPEs. Here, we use intracranially recorded high frequency activity (HFA) to show that this asymmetric scaling strategy captures RPE coding diversity in human dMPFC and INS. We found neural populations responding to valence-specific positive and negative RPEs, as well as unsigned RPE salience, which are spatially interleaved within each region. Furthermore, directional connectivity estimates suggest a leading role of INS in communicating positive and unsigned RPEs to dMPFC. These findings support asymmetric scaling across distinct but intermingled neural populations as a core principle in RPE coding, expand the scope of dRL, and reconcile longstanding theoretical debates on the role of dMPFC and INS in RL and cognitive control.