STUDY OBJECTIVES To determine whether real-life driving would produce different effects from those obtained in a driving simulator on fatigue, performances and sleepiness. DESIGN Cross-over study involving real driving (1200 km) or simulated driving after controlled habitual sleep (8 hours) or restricted sleep (2 hours). SETTING Sleep laboratory and open French Highway. PARTICIPANTS Twelve healthy men (mean age +/- SD = 21.1 +/- 1.6 years, range 19-24 years, mean yearly driving distance +/- SD = 6563 +/- 1950 miles) free of sleep disorders. MEASUREMENTS Self-rated fatigue and sleepiness, simple reaction time before and after each session, number of inappropriate line crossings from the driving simulator and from video-recordings of real driving. RESULTS Line crossings were more frequent in the driving simulator than in real driving (P < .001) and were increased by sleep deprivation in both conditions. Reaction times (10% slowest) were slower during simulated driving (P = .004) and sleep deprivation (P = .004). Subjects had higher sleepiness scores in the driving simulator (P = .016) and in the sleep restricted condition (P = .001). Fatigue increased over time (P = .011) and with sleep deprivation (P = .000) but was similar in both driving conditions. CONCLUSIONS Fatigue can be equally studied in real and simulated environments but reaction time and self-evaluation of sleepiness are more affected in a simulated environment. Real driving and driving simulators are comparable for measuring line crossings but the effects are of higher amplitude in the simulated condition. Driving simulator may need to be calibrated against real driving in various condition.