The relationship between the number of febrile episodes and recurrent febrile convulsions was studied prospectively in 289 children after their first febrile seizure. They were randomized to either short-term diazepam prophylaxis (n = 152) or to no prophylaxis (n = 137), and followed for 18 months. Among untreated children with many subsequent febrile episodes (greater than or equal to 4 per year) 29 of 37 (78%) had a recurrence vs. 17 of 100 (17%) with only few (less than 4 per year) feverish illnesses. The former group had a 4:1 chance of developing further febrile fits, compared with a 1:4 chance in the latter (P less than 0.0001). A similar pattern was observed in the prophylaxis group, but less recurrences were seen (30% vs. 6%, P less than 0.0001). By Cox regression analysis, the subsequent occurrence of many febrile episodes could be identified among several items, including young age at onset, as the adverse factor most highly associated with further febrile fits (P less than 0.0001).