We examined the effect of oral midazolam premedication on postoperative behaviour. Seventy children (ASA Physical Status 1 and 2; aged 1-10 yrs) were assigned randomly in a prospective, blinded fashion to receive either midazolam 0.5 mg.kg-1 (maximum 10 mg) or placebo. Behaviour assessments were made prior to medication, during induction of anaesthesia and 15 min following arrival to recovery room. The baseline behavioural evaluation scores were not significantly different. The children receiving midazolam cried significantly less during induction (P < or = 0.02). At one week follow-up, eight of 35 subjects receiving placebo had experienced adverse behaviour changes (nightmares, night terrors, food rejection, anxiety, negativism); 19 of 35 of the midazolam group experienced these changes (P < or = 0.02). At four week follow-up, most behaviour changes had resolved. Children given preoperative oral midazolam were less likely to cry and fight while being anaesthetized, and preoperative sedation was associated with increased incidence of adverse postoperative behaviour changes.