1. We have investigated the loss of power seen during high-intensity exercise of human leg muscles such as might occur during sprinting. Subjects exercised the quadriceps and hamstring muscle groups using a Cybex dynamometer at an angular velocity of 90 deg s-1 once a second for 6 min. At 1 min intervals the quadriceps were electrically stimulated via the femoral nerve to produce an isometric contraction which was then released into an isokinetic shortening contraction at 90 deg s-1. 2. The extent of central fatigue was assessed by comparing the force of a voluntary isokinetic contraction with that elicited by electrical stimulation during isokinetic releases. Two subjects were repeatedly tested. In the first series of experiments, exercising the quadriceps of one leg, the instantaneous power fell to about 50% over the course of 2 min and remained constant for the rest of the exercise. For one subject the voluntary and electrically stimulated forces declined in parallel while for the second subject the voluntary force was 10% less than the stimulated force at the end of the exercise. These results show that central fatigue represented a minor factor contributing no more than one-fifth of the total loss of power in these circumstances. 3. In a second series of experiments the subjects alternately contracted the quadriceps and the hamstrings of both legs in an exercise which had a high rating of perceived exertion and entailed considerable respiratory and cardiovascular effort. The time course and proportionate loss of power were very similar to those seen with the one-leg exercise and neither subject showed evidence of significant central fatigue. The pattern of force loss was very similar for the hamstrings. We conclude that, for determined subjects, afferent feedback from muscles, tendons and joints or from the respiratory and cardiovascular systems does not have a major role in inhibiting voluntary activation of the quadriceps during heavy exercise. 4. In both series of experiments the power output during electrically stimulated isokinetic contractions was reduced to 50% of the initial value after 2 min of exercise while the isometric force, measured immediately before the release, fell to only 75%. This suggests that fatigue affects isometric and shortening contractions to different extents and the loss of power may be due to a combination of factors, only one of which is evident in the loss of isometric force.