ABSTRACT
Among the noninvasive techniques available for human functional neuroimaging, the electroencephalogram (EEG) provides a measure of brain electrical activity with millisecond temporal resolution and centimeter spatial resolution (Nunez, 1981). Modern EEG systems typically feature 128 electrodes positioned over the scalp, giving a global view of brain dynamics that is complementary to anatomical images. A significant challenge is how best to represent the rich dynamical information in these large, complicated data sets. Traditionally, the methods employed by experimental psychology and neuroscience have been motivated by the need to consolidate these data across experimental conditions and subject groups, usually to facilitate hypothesis testing with either the t or F statistic. As we see here, although such averaging allows standard statistical tools to be used, it is employed at the expense of the true complexity of the data.
