Sodium valproate enteric-coated tablets were used in this double-blind, randomized, cross-over study of 16 patients with juvenile myoclonic epilepsy comparing 1000 mg and 2000 mg VPA daily in b.i.d. administration with 6 months of observation on each dose. Myoclonic, absence, and generalized tonic-clonic seizures were registered separately. Subjective side-effects were monitored, and a computerized neuropsychologic test battery was performed on each dose. There was no significant difference in seizure frequency between the two doses. Only 25% of the patients were seizure free throughout the study despite concentrations well within the normally proposed therapeutic range for VPA. During the higher dose, 37.5% of the patients had an improved seizure control, but 25% of the patients had an increase in seizure frequency compared to the lower dose. However, there was no correlation between VPA concentrations and subjective side-effects or neuropsychologic test results. Our observations point out the possibility that the common strategy of increasing plasma levels in difficult-to-treat patients until side effects occur should perhaps be reconsidered, but this suggestion needs further confirmation.