Reward processing under illusion of control: The sensitivity of the feedback-related negativity to prediction errors is not altered when random outcomes are perceived as the consequence of one’s own actions

The feedback-related negativity (FRN) event-related potential has been associated both with the cortical processing of reward and salience prediction errors (RPEs/SPEs), and with behavioral adjustments that optimize performance. While the FRN is sensitive to response-feedback contingencies and to the level of perceived control over rewards and losses, it remains to be explored how this waveform is influenced by random outcomes during reinforcement learning (RL) while participants develop an illusion of control (IoC). We present novel analyses of data from a previous study (Csifcsák, Melsæter & Mittner, 2020), in which a group of healthy adults was intermittently exposed to compromised control over outcomes (“yoking”) during RL. Earlier, we reported effects on latent parameters of RL and oscillatory activity during decision-making, whereas now we analyzed whether the FRN was sensitive to the controllability manipulation. Forty-six participants were randomized to “control” or “yoked” groups, differing only in their level of control over rewards/losses in three out of nine blocks of an RL task. The FRN was analyzed both in terms of its sensitivity to outcome valence (preferred vs. non-preferred) and with respect to its association with single-trial RPEs/SPEs. Bayesian statistics confirmed comparable ratings of perceived control and success for yoked vs. control participants, indicating IoC in the yoked group. Although response accuracy was at chance level during compromised outcome controllability, the FRN was statistically indistinguishable between the two groups, as revealed by a multitude of analytical approaches. We conclude that under IoC, the FRN is not sensitive to (the absence of) response-outcome contingencies. Furthermore, the state of IoC induces a dissociation between the FRN and choice behavior. These findings suggest that the cortical analysis of rewards and losses, measured above the midfrontal region, is dominated by higher-order cognitive/affective states when predictions about future events are unreliable.