Fig. 1.

Left: Schematic illustration of the generation of the spectral components that were used as cues (red columns) and outcomes (blue column). In each time step, all 104 spectral component cues from each of the three temporal windows (312 cues in total) predict each individual spectral component outcome. This leads to a total of 104 outcomes in each time step. Note: the spectral components are not to scale. Right: Schematic illustration of how a moving training window was used to generate time steps (‘ts’) for training. Each temporal window is 25 ms. The window moves to the right 10 ms in each time step. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)