Post-intervention high-beta synchronization tracks stable-manifold alignment during stick-balancing learning

Learning to stabilize an unstable object may depend not only on the accuracy of active corrective interventions but also on when to suspend intervention and allow the object's intrinsic dynamics to evolve. We investigated learning-related changes in control dynamics and associated beta-band electroencephalographic (EEG) activity during virtual stick balancing. Ten healthy young adults practiced stabilizing a virtual inverted pendulum with a joystick over two consecutive days; scalp EEG was recorded from seven participants. For each control cycle, control quality was quantified as the distance (dist) between the pendulum state at the end of each active intervention phase and the stable manifold of the control-OFF subsystem. With practice, balance duration increased, dist decreased, and the proportion of time without active intervention increased. Time-frequency responses stratified by control quality showed that post-intervention high-beta event-related synchronization was most prominent in cycles with the smallest dist. Single-cycle linear mixed-effects models further revealed an inverse association between post-intervention high-beta power and dist after accounting for intervention magnitude and movement direction, as well as trial progress where applicable. This relationship was evident in both the initial Control and final Skilled task datasets. These findings indicate that stick-balancing practice improves the ability to terminate active intervention near the stable manifold and suggest that post-intervention high-beta synchronization tracks the quality of the transition from active correction to the intrinsic dynamics of the control-OFF subsystem.