1. Neural recordings were made from motor fascicles of the ulnar or radial nerves while the motor cortex was stimulated percutaneously using high-voltage electrical stimuli or transient magnetic pulses to determine whether human muscle spindle endings could be activated by such stimuli and, if so, whether this occurred before the recruitment of alpha-motoneurones. 2. In relaxed subjects, no evidence of muscle spindle activation could be detected in nine recordings of multiunit neural activity and four recordings from single spindle afferents using stimulus levels up to 600 V and 1.5 T. These levels produced a prominent twitch contraction of the intrinsic muscles of the hand and of forearm muscles. Passive stretch of the contracting muscle did not reveal a fusimotor action too weak to be detected under isometric circumstances. 3. With twenty-six single spindle afferents, the stimuli were delivered during a voluntary contraction of the receptor-bearing muscle. This served to 'focus' the effects of the stimulus on the relevant motoneurone pools and increased the probability that fusimotor neurones innervating the endings were active. 4. None of the twenty-six spindle afferents could be activated by stimuli subthreshold for alpha-motoneurones, even when the stimuli were delivered during passive stretch of the contracting muscle. With eighteen afferents, stimuli above threshold for alpha-motoneurones were delivered: twelve remained unaffected but the discharge of six altered. 5. Three afferents were activated at latencies of 35, 39 and 40 ms, respectively 16, 20 and 20 ms after the onset of the EMG potentials in the receptor-bearing muscles. This latency difference is too short to be attributable to activation of gamma-motoneurones: arguments are presented that the increase in spindle discharge could result from activation of beta-motoneurones. 6. The discharge of three afferents increased at latencies of 70, 75 and 85 ms, too early to be due to stretch on the falling phase of the twitch contraction of the receptor-bearing muscle. Responses at these latencies could involve activation of gamma- or beta-motoneurones. 7. These findings in human subjects suggest that transient stimulation of the motor cortex may effectively access fusimotor neurones.