The relapse rate after discontinuation of antiepileptic drug treatment was investigated in 146 children with epilepsy, in whom medication was withdrawn according to a predesigned protocol, after a seizure-free period of at least 2 years and normalization of the EEG. The cumulative probability of remaining seizure-free in this series was 74.5%. Three-quarters of the relapses occurred during the withdrawal period and in the 2 years thereafter. From multivariate analysis, the factors indicating a significantly higher relapse risk were seizures with a known cause and female sex. In primary generalized epilepsy, no factor significantly increased the likelihood of a recurrence. In partial epilepsy, significant factors predictive of recurrence were the presence of a neurological deficit (focal neurological signs and/or mental retardation), female sex, a positive family history for epilepsy, and the number of drugs necessary for control of the seizures. The present results are compared with the available literature data. It is argued that using multivariate analysis after elimination of EEG variables uncovers significant clinical predictive factors that in other studies may have remained hidden. Finally, it is argued that statistical analysis may be used to enable the clinician to predict the likelihood of recurrence in individual children with a given set of relevant predictive factors.