Neuronal oscillations predict un-cued task switching in the Wisconsin card sorting test

Cognitive flexibility is the fundamental ability to rapidly adapt behavior to the environment, even when the contextual rules are implicit and change suddenly. Such type of task switching is commonly measured with the Wisconsin Card Sorting Test (WCST), in which cards need to be sorted by a changing rule that is not explicit to participants and thus needs to be deduced from feedback. Crucially, this ability to adapt to changing task rules in the absence of instructive cues shows large individual variability of which neural mechanisms have remained unknown. Preparative un-cued task switching has previously been connected to prefrontal theta and parietal alpha band amplitude modulations. However, it has remained unknown whether these local oscillation dynamics or large-scale network interactions would enable cognitive flexibility and explain the interindividual performance differences therein. Here, to test this possibility, we collected MEG data during the WCST from N = 26 participants. To assess individual variability in cognitive flexibility, we calculated the number of late errors as a proxy for learning speed and implemented a new sequential learning model for estimating the impact of feedback sensitivity on individual performance. To investigate the neural dynamics, we computed local and network oscillation dynamics from individually source-reconstructed data. Higher switching accuracy was predicted by local gamma (52–96 Hz) band amplitude increases and parietal alpha (7–13 Hz) and beta (13–32 Hz) band suppression as well as sustained frontoparietal alpha band (9–15 Hz) network desynchronization. Individuals with higher sensitivity to positive feedback showed larger alpha– and beta-band amplitude modulations and alpha network desynchronization, whereas sensitivity to negative feedback was mainly associated with larger gamma amplitudes. Our results demonstrate that individual variability in un-cued task switching is due to differential learning speed and sensitivity to feedback, the neural basis of which relies on distinct local and large-scale oscillatory network mechanisms.