We report a study performed to compare the time and plasma drug concentrations necessary to achieve a similar state of sedation after midazolam premedication given by various routes in children of 2-5 years old. Children were randomly allocated to one of three groups to receive midazolam 0.2 mg.kg-1 given intranasally, 0.5 mg.kg-1 given orally or 0.3 mg.kg-1 given rectally. Sedation was measured regularly until venepuncture was possible in a cooperative child. At this time, a first blood sample was taken to measure plasma concentration, followed by another 10 min later. Anaesthesia consisted of intravenous propofol supplemented with regional analgesia. At recovery from anaesthesia, a third blood sample was taken. Adequate sedation occurred sooner (7.7, SD 2.4 min) with intranasal than oral (12.5, SD 4.9 min) or rectal (16.3, SD 4.2 min) midazolam. The initial blood levels were lower when the drug was given by the alimentary routes despite higher doses (146, SD 51 ng.ml-1 in 11.5, SD 3.9 min; 104, SD 34 ng.ml-1 in 21 +/- 6 min; and 93, SD 63 ng.ml-1 in 23.1, SD 3.5 min for the intra nasal, rectal and oral routes respectively). Duration of surgical procedures, and of propofol infusion, and recovery from anaesthesia was similar for the three groups. The only problem arose in a 30-month-old boy in the intranasal group who developed respiratory depression with a plasma midazolam concentration of 169 ng.ml-1. Intranasal midazolam is an excellent alternative for rapid premedication provided that respiratory monitoring is used.