Performance stability, as assessed by trial-to-trial variance in a choice reaction time (RT) task, was evaluated as a measure of stimulant effects on performance during sleep deprivation. Administration of methylphenidate, pemoline, and a placebo began 16 hr into a 64-hr sleep-deprivation protocol. Performance stability deteriorated significantly, especially during the circadian nadirs. In absolute terms, sleep deprivation increased trial-to-trial variance more than it increased the mean correct RT. In addition, this measure demonstrated differing effects of the 2 drug regimens. Pemoline, at a dose of 37.5 mg every 12 hr, significantly reduced the overall average effects of sleep loss on performance stability during the first 24 hr of drug administration. Pemoline also reduced circadian-related instability in performance throughout the study. Methylphenidate, at a dose of 10 mg every 6 hr, counteracted circadian-related instability in performance during the first 24-hr period of drug administration (16-40 hr of sleep deprivation) but not during the second 24-hr period (40-64 hr of sleep deprivation). Methylphenidate did not significantly affect the overall average effects of sleep loss on performance stability. Thus, trial-to-trial variance appears to be a valuable measure for elucidating stimulant effects during sleep deprivation.