The question of whether muscle spindle afferents might control human motoneurone activity on the basis of the "size principle" during voluntary contraction was investigated by recording the discharge of single motor units (n = 196) in wrist extensor muscles while stimulating the homonymous muscle spindles by means of tendon taps. The mechanical stimuli were delivered with a constant post-spike delay of 80 ms so that the resulting afferent volleys could be expected to reach the motoneurones towards the end of the inter-spike interval (mean +/- SD duration: 124.7 +/- 11.9 ms). In the six subjects tested, the response probability was found to be significantly correlated with the motor units' functional parameters. Differences in twitch rise times, twitch amplitudes, recruitment thresholds and macro-potential areas were found to account for 18%, 9%, 6% and 2% of the differences in the response probability observed within the whole population of motor units tested. These differences could not be due to differences in firing rate for two reasons: first, the motor units were found to discharge with a similar range of inter-spike intervals whatever their functional characteristics; secondly, the weak positive correlation observed between the response probability and the motor unit firing rate showed parallel regression lines between the late-recruited fast-contracting motor units and the first-recruited slowly contracting motor units, but the y-intercept was significantly higher in the latter case. This confirmed that the responses of the first-recruited slowly contracting motor units tended to be larger whatever the firing rates. In most of the pairs tested in the same experiment, the motor units which had the lowest recruitment thresholds, longest contraction times, smallest contraction forces or smallest motor unit macro-potentials tended to produce the largest responses, which also had the longest latencies. Taking the response latency to be an index of a motoneurone's conduction velocity and therefore of its size, the data obtained with this index and with other functional indices such as the twitch rise times and amplitudes, the macro-potential areas and the recruitment thresholds-can be said to be fully consistent with the "size principle", as previously found in anaesthetized animals. It can be inferred that the presynaptic inhibition which is liable to take action during voluntary contraction does not seem to alter the graded distribution of the muscle afferent projections to human wrist extensor motoneurones.