Palmar stimulation was used to assess median nerve conduction across the carpal tunnel in 61 control patients and 105 patients with the carpal tunnel syndrome. With serial stimulation from midpalm to distal forearm the sensory axons normally showed a predictable latency change of 0.16 to 0.21 ms/cm as the stimulus site was moved proximally in 1 cm increments. In 47 (52 per cent) of 91 affected nerves tested serially, there was a sharply localized latency increase across a 1 cm segment, most commonly 2 to 4 cm distally to the origin of the transverse carpal ligament. In these hands, the focal latency change across the affected 1 cm segment (mean +/- SD: 0.80 +/- 0.22 ms/cm) averaged more than four times that of the adjoining distal (0.19 +/- 0.09 ms/cm) or proximal 1 cm segments (0.19 +/- 0.08 ms/cm). In the remaining 44 (48 per cent) hands, the latency increase was distributed more evenly across the carpal tunnel. Unlike the sensory axons the motor axons were difficult to test serially because of the recurrent course of the thenar nerve, which may be contained in a separate tunnel. The wrist-to-palm latency was significantly greater in the patients with carpal tunnel syndromes than in the controls for sensory (2.18 +/- 0.48 ms v 1.41 +/- 0.18 ms) and motor axons (2.79 +/- 0.93 ms v 1.50 +/- 0.21 ms). Consequently, there was considerable difference between the carpal tunnel syndromes and controls in SNCV (38.5 +/- 7.5 m/s v 57.3 +/- 6.9 m/s), and MNCV (28.2 +/- 4.5 m/s v 49.0 +/- 5.7 m/s). In the remaining distal segment, however, there was only a small difference between the two groups in sensory (1.48 +/- 0.28 ms v 1.41 +/- 0.22 ms) and motor latency (2.15 +/- 0.34 ms v 2.10 +/- 0.31 ms). The exclusion of the relatively normal distal latency made it possible to demonstrate mild slowing across the carpal tunnel in 36 (21 per cent) sensory and 40 (23 per cent) motor axons of 172 affected nerves when the conventional terminal latencies were normal. Sensory or motor conduction abnormalities were found in all but 13 (8 per cent) hands. Without palmar stimulation, however, an additional 32 (19 per cent) hands would have been regarded as normal.