associated spatial map. In order to perform group-level analysis, the ICs estimated from each subject are clustered by means of the self-organizing group ICA (sogICA) method [Esposito et al., 2005], selecting the number of clusters equal to that of the ICs. The clusters with reproducible maps across subjects are associated with the large-scale networks active during the task. They are scaled to z-scores [Zar, 1996] and thresholded at |z|>1.5, in order to delineate significantly activated brain regions [Mantini et al., 2009]. Moreover, in order to isolate the fMRI networks with time-locked activity, the network time-courses are averaged with respect to the rare event onset, and a canonical hemodynamic response function (HRF), generated using a gamma function (delay time, 2 s; rise time, 4 s; fall time, 6 s; undershoot, 0.2; restore time, 2 s), is fitted to the mean time-locked response. The fMRI networks with a coefficient of determination (r-square value) larger than 0.3 are assumed to be task-related networks.
