Multivariate analysis of EEG coherence: stability of the metric, individual differences in patterning and response to...

When coherence is computed among all pairs of a multichannel EEG, the resultant matrix can be de-structured with conventional multivariate analysis procedures to characterize the patterning of electrophysiologic information at each frequency band of the EEG across brain regions. Six right-handed young men had EEG data recorded from left and right frontal, temporal, parietal, and occipital locations during three visits to the laboratory at one week intervals. During each session, data were recorded during a relaxation period, an alert resting period, and as loud white noise was presented. Factor analysis of coherence matrices showed a model pattern of a large posterior factor and an anterior factor, both of which were right-lateralized, plus a residual factor that was most often left-lateralized. Four of the six men showed this pattern on each occasion, particularly for the delta and the theta bands. Two men showed variant patterns, which were also stable over time. A priori de-structuring of the matrix into partial multiple intrahemispheric and inter-hemispheric coherences also showed generally higher coherences on the right side, with the exception of left occipital inter-hemispheric values. Effects of arousal manipulations were not striking, but the factor patterns suggested increases in right hemisphere coherence for the alpha and beta bands during high arousal.