Ninety children were assigned randomly to one of three groups for premedication with oral midazolam 0.5 mg.kg-1, diazepam 0.25 mg.kg-1 with droperidol 0.25 mg.kg-1, or trimeprazine 2 mg.kg-1. On arrival at the anaesthetic room, anxiolysis was satisfactory in 26 out of 29 (90%) children who received midazolam compared with 23 out of 29 (79%) who received diazepam-droperidol and 18 out of 29 (62%) who received trimeprazine (P < 0.05); at induction of anaesthesia these proportions were 24 out of 29 (83%), 16 out of 29 (55%) and 11 out of 29 (40%) respectively (P < 0.001). When individual groups were compared, anxiolysis was significantly greater in the midazolam group compared with the trimeprazine group on arrival in the anaesthetic room (P < 0.05) and significantly greater in the midazolam group than in either the diazepam-droperidol or the trimeprazine groups at induction of anaesthesia (P < 0.05 and P < 0.001 respectively). There were no significant differences in times to early recovery between the groups (25.4, 24.4 and 28.5 min). Analysis of behavioural questionnaires completed two weeks after hospitalization showed a trend towards fewer postoperative behavioural disturbances in children who received midazolam or diazepam-droperidol compared with trimeprazine (47 and 44% vs 75%); when the results for the benzodiazepine-containing premedicants were combined, the difference between these groups and trimeprazine was statistically significant (P < 0.05).