In order to be able to ascribe meaning to brain signals, the artificial neural network processes them in multiple layers (1). A visualisation of how the amplitude affects the decoding decision of the network shows the researchers what the algorithm is learning (2).