Stretching human muscles with a mechanical device gave rise to multiple peaks in the rectified and averaged electromyogram. In the first dorsal interosseous the latency of the first peak (M1) was 32.4 +/- 2.4 ms (SD) and the latency of the second peak (M2) was 55.1 +/- 11.3 ms, in both cases measured from the time of the stimulus to the take-off point of the peak. Often a third peak (M3) was seen, having a considerably longer latency. The origin of peak M1 was considered to be in the stretch reflex arc because of its latency and its invariable association with muscle movement. Peak M2 was due to stimulation of afferent terminals in the skin and/or subcutaneous tissues by the mechanical device producing the muscle stretch. The conduction velocity of the pathway involved in the generation of the M1 component is the same as that for M2. This implies that central processing in the spinal cord delays the M2 response. The M2 mechanism does not involve a transcortical (long-loop) pathway because in foot muscles the M1-M2 delay remains the same as is found for hand muscles, although M1 latency is prolonged (to 39.4 +/- 6.2 ms for extensor digitorum longus). This indicates that there is not time for M2 impulses to traverse a pathway any longer than that passing to and from the spinal cord.