The influences of coil position and coil-nerve distance on compound muscle action potentials (CMAPs), recorded from the first dorsal interosseus muscle during magnetic stimulation of the brachial segment of the ulnar nerve, were studied in 10 healthy volunteers. A 14-cm coil was held tangentially to the skin with the center overlying the nerve. Mapping of the CMAP latencies and amplitudes was made as the coil was displaced laterally in steps of 1 cm and in planes 0-3 cm from the skin surface. Stimulation with the coil center positioned 3 cm laterally to the nerve with the coil current directed proximally yielded the largest amplitudes with minimal variability and the most constant relationship to electrically evoked CMAPs. In this position the interindividual and intraindividual reproducibility of the magnetically evoked latencies were at least as good as those of electric stimulation when coil-skin distance was less than or equal to 2 cm.