Abstracts, Cognitive Neuroscience Society, p. 93, April, 2000



Scott Makeig-1, Sigurd Enghoff-25,

Tzyy-Ping Jung-34 and Terrence J. Sejnowski-234

1-Naval Health Research Center

2-Salk Institute

3-Howard Hughes Medical Institute

4-University of California San Diego

5-Danish Technical University

Multichannel event-related EEG data, viewed as linear sums of macropotentials generated in functionally independent and spatially fixed brain (or extra-brain) regions, are well-suited for blind decomposition by Independent Component Analysis (ICA). ICA produces signals reflecting the separate activities of the independent signal sources, plus maps giving their respective projections to the scalp surface. A stability analysis of ICA decomposition of event-related 31-channel EEG data collected during a visual selective attention task (Makeig et al., J. Neurosci, 1999), plus trial analyses of higher-density EEG data, suggests that high density arrays may be most suitable for EEG-based functional imaging using ICA, but only if high quality data are recorded at a sufficient number of channels. In general, the functional independence of ICA-decomposed sources must be tested for (a) physiologically plausibility and (b) distinct and precise relationships to behavior or other variables. Decomposition of single-trial 31-channel EEG data from a visual selective attention experiment produced components meeting both criteria. Results suggest a wealth of new information about event-related brain dynamics, including: (1) A new view of the interaction between human attention and fast motor responding. (2) New hypotheses about the further decomposition of the late positive complex (or P300) into EEG-based subcomponents. (3) New information about the relation of early (P1/N1) visual response components to ongoing EEG processes.