Objective. To compare the safety and efficacy of midazolam given intranasally with diazepam given intravenously in the treatment of children with prolonged febrile seizures. Design. Prospective randomised study Setting Paediatric emergency department in a general hospital. Subjects. 47 children aged six months to five) ears with prolonged febrile seizure (at least 10 minutes) during a 12 month period. Interventions. Intranasal midazolam (0.2 mg/kg) and intravenous diazepam (0.3 mg/kg). Main outcome measures. Time from arrival at hospital to starting treatment and cessation of seizures. Results. Intranasal midazolam and intravenous diazepam were equally effective. Overall, 23 of 26 seizures were controlled with midazolam and 24 out of 26 with diazepam. The mean time from arrival at hospital to starting treatment was significantly shorter in the midazolam group (3.5 (SD 1.8) minutes, 95% confidence interval 3.3 to 3.7) than the diazepam group (5.5 (2.0), 5.3 to 5.7). The mean time to control of seizures was significantly sooner (6.1 (3.6), 6.3 to 6.7) in the midazolam group than the diazepam group (8.0 (0.5), 7.9 to 8.3). No significant side effects were observed in either group. Conclusion. Seizures were controlled more quickly with intravenous diazepam than with intranasal midazolam, although midazolam was as safe and effective as diazepam. The overall time to cessation of seizures after arrival at hospital was faster with intranasal midazolam than with intravenous diazepam. The intranasal route can possibly be used not only in medical centres but in general practice and, with appropriate instructions, by families of children with recurrent febrile seizures at home.