Distinct neural oscillations predict individual differences in feedback-guided cognitive flexibility.

Cognitive flexibility is the ability to keep or update contextual cognitive representations. The underlying neural mechanisms and how it is influenced by individual sensitivity to feedback are not known. We measured cognitive flexibility under uncertainty with the Wisconsin Card Sorting Test and brain activity with magnetoencephalography. Cognitive flexibility had large interindividual variability as indexed by the learning speed. Using a behavioral sequential learning model, we show that this variability in cognitive flexibility is predicted by individual sensitivity to feedback and anticipation and exploration tendencies, and we uncover their brain oscillatory signatures. Individual learning speed was predicted by suppressed alpha-beta (7-32 Hz) and increased broad-band gamma (32-143 Hz) amplitudes along with concurrent large-scale alpha (7-13 Hz) desynchronization. In contrast, we found no evidence for theta-band amplitudes enhancing performance directly, but rather an inverted relationship with rule certainty across a trial series. Importantly, these oscillatory sub-processes were explained by individual levels of distinct feedback sensitivity. These results provide a novel account on neural sub-computations underlying flexible feedback-guided contextual maintenance and updating of cognitive rule representations. Significance statement Cognitive flexibility is the ability to keep or update cognitive representations of context and allows us to make decisions under uncertainty. The ability to learn information from external cues has a large inter-individual variability, of which neural mechanisms are largely unknown. Here, we combine for the first time behavioral measures from the Wisconsin Card Sorting Task with MEG data and apply a sequential learning model to titrate the neural subcomponents underlying performance differences. We show that individual variability in learning under uncertainty is associated with differences in sensitivity to contextual cues, revealing distinct spectral profiles for learning from positive versus negative cues. Our findings demonstrate that cognitive flexibility relies on neural oscillations which are specific to the feedback cue type.