1. In the long flexor of the thumb the latency of the stretch reflex and of other manifestations of servo action is some 45 msec, roughly double the latency of a finger jerk. 2. Tendon jerks are feeble or absent in the long flexor of the thumb even in subjects with brisk long-latency stretch reflexes in this muscle. This, and other facts, suggests that the nervous mechanism of the tendon jerk is different from that of the stretch reflex. 3. A muscle that has feeble tendon jerks may show a late component in the response to a tendon tap, with a latency similar to that of the long-latency stretch reflex. 4. On the hypothesis that the excess latency of the stretch reflex over that of a tendon jerk is because the stretch reflex employs a cortical rather than a spinal arc, the excess would be expected to be larger in magnitude for the long flexor of the big toe and smaller for the jaw closing muscles. This is confirmed, 5. An alternative hypothesis that the long latency of stretch reflexes in thumb and toe is because they are excited by slow-conducting afferents is made improbable by the finding that stretch reflexes with an equal or greater excess latency are also found in proximal arm muscles. 6. The long-latency stretch reflex in proximal muscles was seen most distinctly in a healthy subject who happened to have feeble or absent tendon jerks. In ordinary subjects there is often a large, short-latency, presumably spinal component of the stretch reflex in proximal muscles; and short-latency responses to halt and release are also seen, The significance of this spinal latency servo action in proximal muscles remains to be explored. 7. The Discussion argues that the available data on conduction time to and from the cerebral cortex are compatible with the hypothesis that the long-latency component of the stretch reflex uses a transcortical reflex arc, and that none of the experiments described in the present paper are inimical to this view.