Visual potentials evoked by pattern reversal (PRVEPs) were studied in 64 normal subjects (age range 7 to 15 years) and in 15 patients with primary generalized epilepsy (age range 8 to 13 years), 10 of whom were without anticonvulsant medication. Most of them were studied during sodium valproate (VPR) therapy and some during carbamazepine (CBZ) medication. A Quadristim set (Alvar) was used to present checkerboard patterns on a TV monitor, to amplify the EEG signals and to average and plot the evoked potentials. The potentials were elicited by binocular full-field 2/s checkerboard reversals, recorded from an electrode 4 cm above the inion, and analyzed for latency, amplitude and waveform. Our PRVEP measurements examined peak latency of positive P2 (or P100) component and trough-to-peak amplitude on N1P2 wave complex. The degree of similarity between pairs of PRVEP plots were determined by Pearson correlation coefficient r for an analysis time of 150 ms. In most of our patients, no pronounced influence of the disease itself on the parameters and waveform of the normal PRVEP pattern was demonstrated if anticonvulsant drugs were not taken. In patients who were under complete seizure control, the anticonvulsant did not change the PRVEP morphology as well. The PRVEP abnormality was most pronounced in patients who were taking anticonvulsant medication, but whose seizures were poorly controlled. This pattern distortion can be revealed by the correlation coefficient, but not by other PRVEP parameters. Therefore, this coefficient may be useful as a sensitive and objective measure both of PRVEP distortion and PRVEP improvement. Our results give further evidence that nondemyelinating disorders, but with synaptic transmission defects, can produce measurable changes in PRVEP morphology.