Using surface electromyography the reflex response of flexor carpi radialis elicited by forcibly dorsiflexing the wrist was compared with that elicited by applying vibration percutaneously to its tendon. This was done both in patients with Parkinson's disease and in normal subjects. The reflexes were elicited on top of a pre-existing voluntary contraction of the muscle of about 20 per cent maximum. The responses in parkinsonism were qualitatively similar to the normal, but differed quantitatively in certain respects. The response to 'stretch' of the muscle by wrist dorsiflexion normally continued at a high level up to at least 80 ms from the beginning of the movement, commonly with an apparent separation into 'short' and 'long' latency responses. On average, the later components of the response were enhanced in parkinsonian patients in comparison with the normals, confirming other workers' findings; they were also prolonged. The short-latency responses were unchanged. Vibration, in contrast, elicited solely a short-latency response with the initial reflexly-evoked augmentation of EMG activity coming to an end 40 to 50 ms from the beginning of the stimulation, even though the vibration was continuing. Such an absence of the later components that were so prominent with stretch was found whatever the size of the initial short-latency response evoked by vibration, including when it was comparable to that evoked by stretch in the same subject. This purely short-latency vibration response was on average unchanged in parkinsonism. The findings support the hypothesis, already advanced for the long flexor of the thumb, that the long-latency components of response are largely attributable to a spinal excitatory action of the spindle group II afferents with the delay arising from the slowness of their conduction. They are not readily compatible with either of the two major alternative hypotheses, namely the 'long-loop' (or transcortical) hypothesis and the 'resonance' hypothesis, both of which attribute the late response, as well as the initial response, to the spindle Ia afferents. The enhancement of the later components of response in parkinsonism thus now seems likely to be due to an increase in the postulated spindle group II excitatory action, possibly related to a reduction in opposing inhibition, rather than to any change in the reflex excitability of the higher centres on Ia activation. However, the rigidity of parkinsonism cannot be uniquely ascribed to an enhancement of group II action, because over the population as a whole clinically similar degrees of rigidity could be accompanied by quite different long-latency responses, and vice versa.