We recorded motor evoked potentials (MEP) to transcranial magnetic stimulation from abductor pollicis brevis (APB), flexor carpi radialis (FCR), biceps brachii and deltoid muscles at rest and during slight voluntary activation. An 8-shaped coil connected to a Cadwell MES-10 magnetic stimulator was positioned over different scalp positions 1 cm apart. At least 24 stimuli were delivered at each location. Latencies of MEPs were compared with those obtained by electrical and magnetic stimulation during muscle activation. Progressively longer MEP latencies were obtained in 5 groups depending on the type and position of stimulation. The shortest latencies were obtained with (1) electrical stimulation during muscle contraction and (2) non-focal magnetic stimulation during muscle contraction; magnetic stimulation at rest produced longer latencies with stimulation of (3) an optimal position, (4) a suboptimal position, and (5) a non-optimal position. Mean latency differences between successive groups were 1.9, 2.0, 1.6, and 2.6 msec for APB. Similar latency differences were found for the other arm muscles. The results are compatible with the hypothesis that the different latencies evoked by stimulation at different scalp locations are determined by the summation at spinal motoneurons of excitatory postsynaptic potentials generated by successive numbers of I waves.