Eight epileptic patients with mixed seizures refractory to medical control participated in a double-blind crossover study to determine the effectiveness of operant conditioning of the EEG as an anticonvulsant procedure. Baseline levels of seizures were recorded for four months prior to the beginning of treatment. Participants then received false (noncontingent) feedback for two months followed by an ABA-patterned training program lasting a total of ten months. Subjects were assigned to three treatment groups based on different schedules of EEG feedback. They were first trained (A1 phase) either to suppress slow activity (3 to 8 Hz), to enhance 12- to 15-Hz activity, or to simultaneously suppress 3- to 8-Hz and enhance 11- to 19-Hz activity. This was followed by a B phase, in which patients were trained to enhance slow activity (3 to 8 Hz). In the final phase (A2), the initial training contingencies were reinstated. Neuropsychological tests were performed before and after training, and changes in EEG activity as determined by Fast Fourier spectral analyses were analyzed. Five of eight patients experienced a decrease in their mean monthly seizure rate at the completion of feedback training as compared with their initial baseline level.