1. Stretch of voluntarily activated human muscle results in a reflex response consisting of short-latency (M1) and delayed long-latency (M2) components. The mechanism of the M2 response remains the subject of controversy. The present study tested the universality of the hypothesis that the M2 response results from the transmission of low-threshold muscle afferent input travelling over a long-loop supraspinal pathway. Muscle reflex responses resulting from imposed stretch were obtained from the first dorsal interosseus (FDI), biceps brachii (BB), triceps brachii (TB) and triceps surae (TS) muscles. 2. Patients suffering from Huntington's disease (HD) show a selective loss of FDI-M2 responses, with sparing of the M1. This has been attributed to disruption of supraspinal pathways as a part of the disease pathology. Accordingly, HD has been used in the present study as a model to test the universality of the long-loop hypothesis: if this is so, then HD patients with an absent FDI M2 should also fail to show an M2 response in other muscles. 3. It is shown that a group of HD patients in whom the FDI-M2 response was absent or residual developed clear M2 responses in the TB, BB and TS muscles following stretch sufficient to invariably evoke this component in normal subjects. 4. It is thus concluded that longer-latency stretch reflex components are not invariably mediated over long-loop supraspinal pathways, but that this mode of control is dominant only in muscles, such as those of the hand, whose function depends largely on direct cortical control.