We performed topographic mapping of somatosensory responses to median nerve stimulation delivered at 2, 5 and 10 Hz. Parietal N20 was significantly attenuated in 10 Hz somatosensory evoked potentials (SEPs), while central P22 diminished between 2 and 5 Hz, remaining stable thereafter. The single component most affected by increasing stimulus rate was N30, which abated by more than 50% in 10 Hz SEPs, as compared with basal responses. N30 attenuation disclosed the existence of an earlier negative component, N24, which appeared as a notch on the N30 ascending slope in 2 Hz SEPs, but became a well-defined peak at higher stimulus rates. The N24 negativity was not significantly modified by stimulus rate; it had a parietal counterpart (P24) with the same peak latency and identical behavior during the experimental procedure. Both P24 and N24 could be differentiated from central P22 on the basis of topographical distribution and response to stimulus frequency. P22 topography could be the result of a radially oriented generator, while P24/N24 appeared as the two poles of a neural source tangential to the scalp. P27 was seen in 40% of the subjects only; it is suggested that P27 is itself a composite potential to which the generator of N30 could contribute in part. We conclude that there is no single "optimal" stimulation rate for SEP recording. On the contrary, combination of different frequencies of stimulation should enhance the diagnostic utility of this technique by allowing a more selective assessment of overlapping activities.