Functional disconnection of associative cortical areas predicts performance during BCI learning

Brain-computer interfaces have been largely developed to allow communication, control, and neurofeedback in human beings. Despite their great potential, BCIs perform inconsistently across individuals. Moreover, the neural processes activated by training that enable humans to achieve good control remain poorly understood. In this study, we show that BCI skill acquisition is paralleled by a progressive reinforcement of task-related activity and by the reduction of connectivity between regions beyond those primarily targeted during the experiments. Notably, these patterns of activity and connectivity reflect growing automaticity and predict future BCI performance. Altogether, our findings provide new insights in the neural mechanisms underlying BCI learning, which have implications for the use of this technology in a broad range of real-life applications.