1. The magnetic field generated by a brief current in a 9 cm diameter flat circular coil varies rapidly with time and when applied over the human scalp it is capable of exciting motor structures subserving the small hand muscles. With a peak magnetic field at the centre of the coil of 0.9-1.6 Tesla, single scalp stimuli produced twitches in the right abductor digiti minimi (a.d.m.), first dorsal interosseous (f.d.i.) and adductor pollicis muscles. Compound muscle action potentials (c.m.a.p.s) and single motor units from these muscles were recorded. 2. The twitch force and corresponding c.m.a.p. were greatly enhanced by voluntary background contraction of the muscle and depended on the direction and intensity of the magnetic field. Response amplitude was maximal with the stimulating coil centred near the vertex but precise coil position was not critical. 3. When the orthodromic volley set up in the peripheral nerve by magnetic stimulation of the brain collided with a maximal antidromic volley set up by stimulation at the wrist, remaining electromyographic (e.m.g.) activity gave evidence of multiple firing of some spinal motoneurones, provided that the muscle exerted a slight voluntary background contraction and the stimulus intensity was above threshold for relaxed muscle. 4. When the muscle changed from total relaxation to a slightly contracted state the onset latency of the c.m.a.p. was shortened by about 3 ms without further change when the background contraction increased. In slightly contracted muscle, c.m.a.p. onset latency was little affected over a wide range of stimulus intensities and was unaffected by the position of the stimulating coil within an area of 6 X 6 cm over the vertex. 5. Single motor units, recorded with needle electrodes in the a.d.m. or f.d.i. muscle, could be caused to discharge at a constant latency by threshold brain stimuli. Different single motor units had latencies varying from 22.4 to 32.1 ms. The same motor unit was activated from different coil positions up to 7 cm apart. These motor units had the lowest thresholds for voluntary activation. Stronger stimuli caused the same motor unit to discharge 1.5 ms earlier. 6. The threshold for excitation of pathways to the a.d.m. muscle was reduced by voluntary contraction of the contralateral a.d.m. and by contraction of the ipsilateral f.d.i. muscles. Using these procedures, the c.m.a.p. of totally relaxed a.d.m. muscle showed a more than 2-fold amplitude increase, a shortening of onset latency to a similar degree as with contraction of the muscle itself, and single motor units in the a.d.m. muscle discharged 1.5 ms earlier.(ABSTRACT TRUNCATED AT 400 WORDS)