The superimposed twitch technique is frequently used to study the degree of motor unit activation during voluntary effort. This technique is one of the preferred methods to determine the activation deficit (AD) in normal, athletic, and patient populations. One of the limitations of the superimposed twitch technique is its variability under given contractile conditions. The objective of this research was to determine the source(s) of variability in the superimposed twitch force for repeat measurements. We hypothesized that the variability in the activation deficit measurements may be caused by the timing of the twitch force relative to the onset of muscle activation, by force transients during the twitch application, by small variations in the actual force from the nominal target force, and by variations in the resting twitch force. Twenty-eight healthy subjects participated in this study. Sixteen of these subjects participated in a protocol involving contractions at 50% of their Maximal Voluntary Contraction (MVC) effort, whereas the remaining twelve participated in a protocol involving contractions at 100% of their MVC. Doublet twitch stimuli were superimposed onto the 50% and 100% effort knee extensor muscle contractions, and the resting twitch forces, voluntary knee extensor forces, and superimposed twitch forces were then measured. The mean resting twitch forces obtained prior to and following 8 sec of 50% of MVC were the same. Similarly, the mean superimposed twitch forces determined at 1 sec, 3 sec, 5 sec, and 7 sec into the 50% MVC were the same. The variations in twitch force were significantly smaller after accounting for the actual force at twitch application than those calculated from the prescribed forces during the 50% MVC protocol (p<0.05). Furthermore, the activation deficit and the actual force showed statistically significant negative correlations for the 50% MVC tests. The interpolated twitch torque determined for the maximal effort contractions ranged from 1 to 70%. In contrast to the protocol at 50% of MVC, negative correlations were only observed in five out of the twelve subjects during the 100% effort contractions. These results suggest that small variations in the actual force from the target force can account for the majority of the variations in the superimposed twitch forces for submaximal but not maximal effort contractions. For the maximal effort contractions, large variations in the superimposed twitch force exist due to undetermined causes.