The authors reviewed the administration of p.r.n. medications over 12 months in 49 child psychiatry inpatients. There were 1263 p.r.n. administrations (mean +/- SD = 25.6 +/- 36.7 administrations/patient). Only 7 (14%) of the 49 patients had none; 9 (18%) had more than 50. The number of p.r.n. administrations correlated with the dose of regular neuroleptics and with the time spent in seclusion. Age, race, sex, diagnosis, neurological disorders, IQ, and self-injurious behavior were poor predictors of the number of p.r.n. administrations. Antihistaminics accounted for 54% (683) of the p.r.n. administrations, neuroleptics for 24% (303), and chloral hydrate for 17% (214). Seventy percent (891) of the p.r.n. drugs were given for disruptive behavior. Only 32% (403) of the 1263 administrations were clearly effective, 14% (170) were ineffective, and 54% (690) were dubious. A controlled trial is needed to assess the efficacy of this ubiquitous psychiatric practice.