Numerous studies have described frequency specific event-related desynchronization/synchronization (ERD/ERS) calculated by averaging across single-trial absolute power time-series. However, there are two main limitations of this approach. First, searching for univocal correspondences between specific sensory, motor or cognitive processes and specific space-frequency-time oscillatory activities gives only a fragmentary description. Second, these oscillatory features may not even exist in the individual trials.
Here, we use a combination of graph visualization and mixed models to show that the single-trial activities in a 48 dimensional spatial-temporal-spectral space across four cortical areas (frontal medial, parietal medial, and sensorimotor cortices), three movement phases (pre, during and post), and four frequency bands (θ, ⍺, β, γ) are tightly coordinated with each other. The structure of their interdependence network (effective connectivity) for the different task conditions (e.g. without or with an unexpected or expected visual perturbation) were very similar (high cross prediction among task conditions), yet unique as the task conditions could be discriminated with above chance level accuracy.
Furthermore, the single-trial activities are predictive of fine trial-to-trial variations in behavioral features such as movement error and duration. Remarkably, when movement error and duration are included as nodes in the effective connectivity network, the causal direction of the influence was from EEG oscillatory elements to movement duration (“control”) whereas “regulatory” connections from movement error to EEG oscillatory elements.
Altogether, our findings suggest that visuomotor cognition and behavior are supported by collective brain states regulating distributed sub-systems in an integrated manner, rather than by a multitude of segregated spatio-temporal-spectral regions, each undergoing independent oscillatory processes.