Thirty-six normal male subjects underwent total sleep deprivation for 48 hours, were then administered either placebo, 5, 10, or 20 mg of d-amphetamine, and sleep deprived for an additional 12 hours. Sleep deprivation produced a significant reduction in sleep latency, as well as marked decrements in cognitive performance and self-ratings reflecting vigor and fatigue. Amphetamine reversed these effects in a dose-related way but the pattern and persistence of the reversal varied across measures. After 20 mg, sleep latency normalized for several hours, but then declined. Behavioral effects tended to follow the pattern of sleep latency. On cognitive tasks, 20 mg produced a sustained return to normal performance in an attentional arithmetic task and a gradual improvement in a verbal reasoning task. The partial temporal dissociation among sleep latency, behavioral, and cognitive effects suggests that varying doses of amphetamine may have time-related differential neurochemical effects or that various dimensions of arousal and alertness may be differentially sensitive to amphetamine.