The human pattern electroretinogram (PERG) is a contrast-specific potential presumedly reflecting the functional integrity of ganglion cells. Many studies have devised criteria that enable PERG measurements to distinguish established glaucomatous (hypertonic) eyes from normal controls. As there are relatively few reports concerning the reproducibility and reliability of the PERG, we studied the intraindividual variability of the PERG in 20 healthy subjects. Both transient and steady-state responses were recorded using a high-contrast (98%), black-and-white, counterphasing checkerboard pattern (average luminance, 80 cd/m2) generated by a television monitor (subtending angle, 13.8 degrees x 10.8 degrees) using three different check sizes (15', 30', and 60'). Recordings were performed in both eyes simultaneously at a 7-day interval under test-retest conditions. Responses of 30' spatial frequency were most consistent and resulted in a mean amplitude (+/- SD) of 2.18 +/- 0.95 microV (P50) and 4.00 +/- 1.69 microV (N95) for transient patterns and 1.84 +/- 1.25 microV for steady-state patterns. No statistically significant difference was observed between either right and left eyes, test and retest conditions or 1st- and 7th-day recording sessions for PERG parameters. In linear correlation analysis there was an adequate, positive correlation between the right and left eyes (r = 0.78); a weak correlation between test and retest conditions (r = 0.58); and no correlation between measurements made at a 7-day interval. As a consequence, we conclude that the follow-up of patients (e.g., glaucoma, ocular hypertension) by means of PERG is critical, especially when therapeutic consequences may be based on the physiological variability of a weak retinal signal.(ABSTRACT TRUNCATED AT 250 WORDS)