We recorded motor evoked potentials (MEPs) from deltoid, biceps brachii, abductor pollicis brevis and flexor carpi radialis muscles of 5 normal volunteers during transcranial magnetic stimulation. With the subjects at rest, an 8-shaped magnetic coil was used to deliver 30 stimuli to different scalp positions 0.5 or 1.0 cm apart. The variability in amplitude and latency of MEPs was studied as a function of the scalp position stimulated, the number of stimuli at each position, and the percentage of maximal peripheral M responses (%M) elicited. The results were used to estimate the optimal number of stimuli at each position and the optimal spacing of scalp positions for topographic mapping of the human motor cortex. The amplitude and latency variability of MEPs were higher when suboptimal scalp positions were stimulated. Consequently, a larger number of stimuli were required to determine representative MEP amplitudes at suboptimal positions. In addition, there was an inverse relationship between %M recruited by transcranial magnetic stimuli in different subjects and the variability in MEP amplitude and latency. Latency variability was less pronounced than amplitude variability. Optimal sampling conditions are required to produce the best topographic maps, particularly to show subtle reorganization patterns in the human motor cortex.