The correspondence between the fluctuation of the P300 response and the activity of the fMRI networks was analyzed, estimating the similarity between the P300 reference time-course and the network time-courses. Specifically, the correlation coefficient rp was computed for each subject. Since the correlation coefficients of the ICs that composed a specific cluster were not normally distributed, they were converted to z values using Fisher's r-to-z transformation (Zar, 1996). They were averaged, and the resulting value was back-transformed to r-values. A minimum correlation coefficient level to ensure statistical significance was determined using the two-tailed Pearson test (p<0.05 corrected). Accordingly, only the fMRI networks with rp>0.20, were considered to be significantly related to the P300 response. In addition, we performed for each network a random-effect analysis on the z-scores by means of t-statistics, to measure the consistency of the correlations across subjects (Zar, 1996).
