Interdependence patterns of multi-frequency oscillations predict visuomotor behavior

We show that sensorimotor behavior can be reliably predicted from single-trial EEG oscillations fluctuating in a coordinated manner across brain regions, frequency bands and movement time epochs. We define high-dimensional oscillatory portraits to capture the interdependence between basic oscillatory elements, quantifying oscillations occurring in single-trials at specific frequencies, locations and time epochs. We find that the general structure of the element-interdependence networks (effective connectivity) remains stable across task conditions, reflecting an intrinsic coordination architecture and responds to changes in task constraints by subtle but consistently distinct topological reorganizations. Trial categories are reliably and significantly better separated using oscillatory portraits, than from the information contained in individual oscillatory elements, suggesting an inter-element coordination-based encoding. Furthermore, single-trial oscillatory portrait fluctuations are predictive of fine trial-to-trial variations in movement kinematics. Remarkably, movement accuracy appears to be reflected in the capacity of the oscillatory coordination architecture to flexibly update as an effect of movement-error integration.