1. Single-unit activity was recorded with the microneurographic technique from the radial nerve of attending human subjects. During active finger movements, impulses in spindle afferents from the extensor digitorum muscle were analysed along with joint movements, size of imposed load and EMG activity of the receptor-bearing muscle. 2. In a simple motor adaptation task the subjects were requested to perform ramp-and-hold movements of prescribed amplitudes and velocities at a single metacarpo-phalangeal joint. A test run consisted of a series of movement cycles when the flexor muscle was continuously loaded with a constant torque, immediately followed by cycles when this load was abruptly decreased during the flexion movement, producing a fast stretch of the receptor-bearing muscle. The subjects' task was to strive for movements of constant velocity and particularly to minimize the effect of the disturbance. In order to allow prediction on the basis of immediately preceding cycles, the disturbance was always injected at the same angular position in a number of successive cycles. 3. Motor adaptation was manifested as a successive decrease of the perturbation amplitude, usually associated with the development of a continuous and growing EMG activity in the parent muscle and a growing reflex response of long latency (60 ms). Short-latency reflexes were not seen. 4. The main mechanism accounting for the improved performance was a co-contraction of the agonist-antagonist muscle pair during voluntary movements, producing an increased muscular stiffness. The reflex did not contribute to the motor adaptation because it was not fast enough to curtail the perturbation. 5. The development and the growth of the reflex were not due to a growing fusimotor drive during adaptation, because spindle discharge actually decreased when the reflex increased. The size of spindle response was related to the amplitude of perturbation rather than to the amplitude of the reflex. These findings suggest that reflex modifications were due to central excitability changes which paralleled the muscle contraction. 6. Spindle firing rate during active movements was generally higher in disturbed cycles compared to undisturbed cycles, indicating a higher fusimotor drive. Since muscle contraction was present mainly in the former, this finding may simply represent a case of fusimotor activation along with skeletomotor activation. No indication of an independence between the two was found. 7. The findings lend no support for the view that the size of the stretch reflex in a behavioural task is adjusted by selective changes of the fusimotor drive.