Potential recording of brain activities always encounters the problem resulting from the activation of reference electrodes. Current source density (CSD) computation does not take reference sites into account and consequently may better localize the generator sources. In the past, several attempts have been made to record CSDs of the somatosensory evoked responses (SERs) following median nerve stimulation. In order to compare the generating mechanisms of SERs following median nerve and tibial nerve stimulation, the scalp CSD distributions of the median nerve SER and the tibial nerve SER were compared in 5 normal subjects. In the median nerve SER, far-field potentials such as P14 and N16 were abolished in the CSD records. N20, P25 and N35 showed almost identical CSD distributions, albeit P25 had a reversed polarity. By contrast, the tibial nerve SER showed similar distributions for P40 and P60 CSDs, but N50 had a different distribution from the others. In the potential records, P40 and P60 were distributed predominantly ipsilateral to the stimulus (paradoxical lateralization), whereas the P40 and P60 CSDs formed a dipole localized over the contralateral foot somatosensory area. N50 disclosed the same tendency, although it had a slightly different CSD pattern from that of P40 and P60. These findings suggest that the median nerve and tibial nerve SER components are not necessarily comparable and that under certain circumstances CSDs are better indicators of local electrical events than the corresponding potentials.