Comparison of predicted GFP amplitudes to the observed GFP amplitudes at different noise gap widths shows that the function of GFP amplitude reduction is significantly different during selective attention from that predicted by a pure gain mechanism.

<p>The grand average GFP amplitude function during Ignore task was fixed to the GFP amplitudes of end-point stimuli types during Attend tasks. This was done across all subjects in order to obtain prediction of GFP amplitude for the in-between stimulus types. Positive value means that the observed value was bigger than the prediction. The vertical grey bars denote subject-wise differences, sorted by size, while the horizontal black bar is the average GFP difference (n = 10 for each stimulus type). The values during Ignore task are used only as a reference, as they are zero-mean due to the comparison algorithm.