The microstate analysis followed the standard procedure used in earlier work [29]. The selected EEG epochs were digitally band pass filtered from 2–20 Hz. Global Field Power (GFP), which quantifies the overall potential variance across the set of electrodes, was computed at each sample in time. Since topography remains stable around peaks of GFP and changes during the troughs, only topographies at momentary maxima of the GFP were further analyzed. As in previous work [36], four optimally fitted microstate class topographies were computed using a modified version of the K-mean clustering algorithm [41]. The algorithm searched four classes of microstate topography and assigned each EEG topography to one of these classes. This number of classes has previously been found to be optimal and was maintained for compatibility with the existing literature. Microstate class topographies were computed individually and averaged across subjects, using a permutation algorithm that maximized the common variance over subjects [36].
