Since the ICA method permitted the separation of a number of functionally distinct coherence patterns for each dataset, we used the sogICA method, in order to produce group inferences in a multi-subject analysis (Esposito et al., 2005). By means of this second-level analysis, we could also find reproducible spatial patterns across subjects. Moreover, information in the time-domain was not discarded during the clustering process; rather, the time-courses of brain activation for the ICs belonging to the same cluster were grouped and jointly analyzed for measuring the correlation of the specific network with the P300 reference time-course.
