There is increased concern about the effects of sleep deprivation on physician performance. We administered four standard tests of cognitive function to 23 university hospital house staff. Each physician served as his or her own control, and the tests were administered at rest, after a night on call, and after a night of sleep for recovery. The study was designed so that normal learning would minimize any deterioration in the post-on-call test performance. Statistically significant deterioration occurred in 3 of the 4 tests after a night on call. Even physicians acclimated to sleep deprivation on a regular, every-third-or-fourth-night basis showed functional impairment. The results have implications for patient care under conditions where house staff are stressed by sleep deprivation and prolonged fatigue.