In a prospective randomised study, 289 children admitted consecutively to hospital with their first febrile seizure were allocated, by date of admission, to short term diazepam prophylaxis (n = 152) or to no prophylaxis (n = 137) and followed for 18 months. In untreated children, five major risk factors for recurrent febrile convulsions were identified: age 15 months or less at the time of the first febrile seizure, epilepsy in first degree relatives, febrile convulsions in first degree relatives, a first complex febrile seizure, and day nursery care. The 18 month recurrence rate was 80 to 100% if three to five risk factors were present, 50% if two factors were identified, 25% where one factor was found, and 12% if there were no predictors. During prophylaxis the recurrence rate was uniformly low (mean 12%) in all risk groups. In high (three or more factors) and intermediate (two factors) risk children prophylaxis provided effective seizure control and reduced the recurrence rate from 80%, or more, to 12% and 50% to 12%, respectively. In children with one risk factor 50% of all recurrences were prevented (25% to 12%). Prophylaxis was ineffective in very low risk children (12% to 12%).