Three children with very frequent refractory epileptic seizures underwent a behavioral intervention consisting of symptom discrimination, countermeasures, contingent relaxation, and positive reinforcement for correct responses in a systematic replication series. The studies involved a 6-h nonintervention base rate, a 6-h treatment phase, and a 6-h nonintervention follow-up under laboratory conditions for each child. Neurophysiologic and behavioral measures of the effects of treatment were made using electroencephalogram (EEG)-video equipment. Effects of treatment were assessed by using a random sample of EEG-video sequences in base rate and follow-up. Results showed that no significant reduction of either seizure behavior or paroxysmal EEG activity was found subsequent to training in discrimination of early paroxysmal activity and/or sensations preceding seizures. Both seizure behavior and paroxysmal activity were significantly reduced in all three cases following intervention with an adapted countermeasure technique. No additional effects could be noted subsequent to the application of either contingent relaxation or positive reinforcement for correct responses. Paroxysmal EEG changes and seizure behavior were highly correlated. Reduction of the clinical manifestation or seizure response by behavioral manipulation was accompanied by a reduction of the total amount of paroxysmal activity as measured by the EEG.