Fig. 2.

a Average trials in one subject for a single channel in each condition. The channel (in the occipital region) was selected as the one that shows the greatest difference in variance between the two conditions. The cue was presented at 0 s. b and c New average trials resulting from the application of CSP filters (described in part 3 of the “Data analysis” section). The filter was optimized for the period of time between the two dotted lines. Two different projections (first and last row (filters) of the projection matrix, Ramoser et al.2000) have been applied to each average trial. In b the filter (first projection) maximizes the separation in variances between conditions by increasing the variance of the free condition and decreasing the variance of the forced condition, in c (last projection) it increases the variance of the forced and decreases the variance of the free condition. Note that in b and c, trials are much more easily discriminated based on variance differences than they were originally in a