To test the validity of nurses' estimates and patients' self-reports of the duration and quality of the patients' sleep, the authors compared these estimates with measures of sleep taken with the electroencephalogram. The authors found that nurses correctly estimated the amount of sleep of only two of the 14 subjects and tended to consistently overestimate. Patients showed no tendency to consistently overestimate or underestimate, but only four of the 14 patients correctly estimated the duration of their sleep. The authors argue that conclusions about sleep patterns that are based on observational methods will have to be reexamined.Abstract Teaser