We investigated the reproducibility of measurements of maximal voluntary torque and maximal voluntary activation using twitch interpolation. On 5 days, each of 5 subjects performed 10 maximal voluntary isometric contractions of their elbow flexors. Single supramaximal stimuli were delivered over biceps brachii at the measured peak torque during each effort, and in the relaxed muscle 5 s later. A voluntary activation score was calculated from the size of twitches evoked by the stimuli (resolution < 0.15 Nm). Although all subjects were able to drive the stimulated elbow flexor muscles maximally in some trials, they did not do so in 75% of all contractions. Maximal voluntary torques did not vary significantly within a subject between sessions. There were consistent differences in the level of maximal voluntary activation between subjects (P < 0.01), but no differences in voluntary activation within an individual across days in 4 of 5 subjects. Failure to drive the stimulated elbow flexor muscles maximally was not associated with inadvertent co-contraction of the antagonist muscles. © 1995 John Wiley & Sons, Inc.