A method is proposed to determine components of evoked scalp potentials, in terms of times of occurrence (latency) and location on the scalp (topography). The scalp field distributions were evoked by checkerboard reversal and were recorded simultaneously in 47 channels. Component latencies are defined as times of maximal values of the electrical power of the evoked field (a measure of the amount of field relief); this measurement is independent of the choice of the reference electrode. In 10 subjects, two evoked components were found consistently: at 100 and at 140 msec. Plots of scalp locations of the extreme field values (i.e., reference-free data) at the occurrence times of the components showed occipitally positive and anteriorly negative extreme values at 100 msec, and vice versa at 140 msec. The occipital extreme values were surrounded by steep field gradients suggesting occipital generator processes. The polarity reversal of the evoked field distributions between 100 and 140 msec was a quick, jump-like location change of the extreme values in the field. The locations of the extreme field values were stable for long periods around peak times of the power curve. During these periods, the shape of the field remained constant (assessed by the average standard deviation of voltages per electrode between successive field distributions), suggesting also a stable localization of the generating process in depth. The field distributions tended to be concentric around the extreme field values. The major characteristics of the observed scalp fields showed no wave fronts, and no continuous 'traveling' of extreme values over larger distances.