Figure 2.

Performance of local spatial analysis (LSA) on simulated event-related potential (ERP) data containing a central widespread component and a lateralized local component. Left: EEG topographies were generated as a widespread component at Cz plus a local negative component at C3. Right: by exploiting the information contained in the trial-by-trial variability of the simulated data, LSA returns a scalp topography virtually identical to that of the original local component. Commonly used stationary filters (vertex reference, average reference, surface Laplacian) failed to highlight the local component. Only the contralateral difference returned a negativity in the left hemisphere, although it also created a spurious positivity in the contralateral hemisphere. Color axes represent voltage (µV).