Melatonin 3 mg and secobarbital 100 mg assigned randomly were given to 40 psychiatric patients for sleep induction during EEG recording. Nine patients who did sleep naturally comprised a comparison group. EEGs were read blind; most were interpreted as mildly abnormal or within normal limits. No statistically significant differences between the three groups were observed in response to photic stimulation, hyperventilation or in frequency of paroxysmal variants. The electroencephalographer was able to identify the melatonin patients significantly more accurately than those who received secobarbital on the basis of lack of EEG manifestations of fast frequencies typical of barbiturate effects. Self-assessments of drowsiness, anxiety and performance on a perceptual-motor task were similar in the melatonin and secobarbital patients. However, the secobarbital group showed more impairment on a locomotion test than those who received melatonin or slept spontaneously. The results suggest that melatonin is a plausible alternative for EEG sleep sedation, especially for ambulatory patients.