Akaysha C. Tang, 123 Dan Phung, 12 Barak A. Pearlmutter, 23 Robert Christner 4

Blind source separation (BSS) decomposes a multidi- mensional time series into a set of sources, each with a one-dimensional time course and a xed spatial dis- tribution. For EEG and MEG, the former corresponds to the simultaneously separated and temporally over- lapping signals for continuous non-averaged data; the latter corresponds to the set of attenuations from the sources to the sensors. These sensor projection vec- tors give information on the spatial locations of the sources. Here we use standard Neuromag dipole- tting software to localize BSS-separated components of MEG data collected in several tasks in which visual, audi- tory, and somatosensory stimuli all play a role. We found that BSS-separated components with stimulus- or motor-locked responses can be localized to physio- logical and anatomically meaningful locations within the brain.