Learning from feedback is independent from feedback visibility, but supported by aperiodic neural activity

Humans and other animals learn from feedback, by tending to repeat rewarded behavior and change behavior that receives negative feedback. Previous findings suggest that feedback does not need to be consciously perceived in order to be effective. Using continuous flash suppression, we presented participants with visible and invisible positive and negative feedback during a time estimation task while recording EEG. Behavioral results showed that both visible and invisible feedback significantly influenced time estimation error and adjustment in trial N+1, suggesting that subliminal reward information can be effectively utilized. Electrophysiological indicators (reward positivity, P3a, theta activity, aperiodic exponent) showed feedback-valence effects, but only when the feedback was visible. Performance in trial N+1 was successfully predicted by the aperiodic exponent only. These findings suggest that (1) behavioral control is independent from conscious perception of feedback signals; (2) the valence sensitivity of electrophysiological indicators is not informative for effective learning or the impact of feedback on behavioral control; and (3) the general neural state, as characterized by the aperiodic exponent, is predictive of the quality of learning from feedback, with steeper exponents providing the most supportive conditions for learning.