Transcranial electromagnetic brain stimuli elicit a complex response in the electromyogram of active human hand muscles. Relatively weak stimuli evoke a short-latency primary response via a presumably monosynaptic corticospinal path. This is followed by a silent period that is terminated by a second peak at a latency of 50-80 ms. The responses evoked in single motor units in flexor digitorum profundus (FDP) were recorded. Responses were elicited at the second-peak latency only in trials in which no primary response was elicited in that unit, and only when the stimulus was given during the first half of the interspike interval (ISI). When given during the second half of the ISI, the same stimulus evoked a primary response but no second peak response. Stronger stimuli suppressed the second peak by evoking a primary response in more trials. Having discharged at about 20 ms latency, the parent motoneurone was unable to discharge again at second-peak latency, 30-60 ms later. The response at second-peak latency was not modified by disengaging both FDP and the extensors of the distal interphalangeal joint. Hence, this response is not secondary to a stretch reflex provoked by activation of the finger extensors, nor is it the result of a cutaneous signal resulting from movement of the finger. The latencies suggest that the corticospinal volley evokes a beta-motoneurone-mediated twitch in FDP muscle spindles, which elicits an afferent volley that activates the motoneurone reflexly. The first 100 ms or so of the silent period is due to the realignment of the first post-stimulus spike in most trials to corticospinal latency; i.e. this is not necessarily the result of an inhibitory or disfacilitatory process. Still stronger stimuli increase the duration of the ISI in which the stimulus is given, indicating the presence of an inhibitory/disfacilitatory process.