Thirty healthy volunteers were randomly assigned to either a caffeine or a placebo group to investigate the alerting effects of caffeine at night. Subjects adhered to a simulated night-shift schedule for 5 consecutive nights. On the first 3 nights, 2 mg/kg caffeine was added to decaffeinated coffee at 2220 and 0120 hours for the caffeine group. On nights 4 and 5 both groups received placebo. Each night, subjects completed five 60-minute sessions of a computerized simulated assembly line performance task (SALT), a multiple sleep latency test (MSLT) and questionnaires. Daytime sleep was recorded in the laboratory between 0900 and 1700 hours each day following nighttime testing. Caffeine decreased physiological sleep tendency on the night shift compared with placebo; however, the two groups performed at equivalent levels on the SALT. On nights 4 and 5, when both groups received placebo, there were no differences between the groups on the MSLT, suggesting the absence of a discontinuation effect. There were no differences on daytime polysomnograms between the groups.