Detailed analysis of P13/14 and N20 wavelets was performed for 62 normal subjects and patients with various lesions along the somatosensory pathway. A histogram of the latencies of all the identified P13/14 wavelets (measured from P13/14 onset) demonstrated three latency-groups, which were named P13, P14a and P14b subcomponents. The relationship between the three newly identified subcomponents and the conventional naming of P13 and P14 was inconstant, indicating the ambiguity of the latter. P14b was most prominent in the contralateral central region, and therefore a P15 positivity slightly after P14b was often recorded in the CPc-Fz and CPc-CPi leads (CPc and CPi are centroparietal electrodes contralateral and ipsilateral to the stimulation). P14b/P15 was lost even in patients with cortical lesions, and thalamocortical fibers were assumed for its origin. The CPc-Fz and CPi-Fz leads registered a low negativity named broad N13', suggesting frontal predominance of the overall P13/14 complex. Both P13 and P14a were identified in a patient with a pontine lesion, and a caudal brainstem origin for both was suspected due to the onset of two repetitive bursts of the ascending lemniscal volley. We refuted the presynaptic origin of the scalp P13 potential and pointed out that a prolonged and/or polyphasic P11 frequently observed in patients with high cervical lesions can be mistaken as scalp P13. A histogram of the latencies of all the identified negative wavelets of N20 in the CPc-Fz lead (measured from N20 onset) revealed five definite latency-groups, which were named N20a, N20b, N20c, N20d and N20e subcomponents. The highest peak of N20 actually corresponded to either N20b, N20c or N20d, and this uncertainty, which must be related to intracortical processes, resulted in a large instability of the N20 peak latency as well as the age and sex dependence of the N20 onset-peak interval, both of which were demonstrated by our preceding study (Sonoo, M., Kobayashi, M., Genba-Shimizu, K., Mannen, T. and Shimizu, T. Detailed analysis of the latencies of median nerve SEP components, 1: selection of the best standard parameters and the establishment of the normal values. Electroenceph. clin. Neurophysiol., 1996b, 100: 319-331). Negative subcomponents in the CPc-NC lead and positive subcomponents in the Fz-NC lead constituted mirror images of each other, which suggested that these subcomponents were generated within area 3b.