A recent study found that the gold foil electrode produces large pattern electroretinogram amplitudes, but the test-retest reliability was low. In a three-center study, we observed that 90% of 29 patients who were tested with gold foil electrodes used three times appeared to have markedly lower amplitudes than when tested with new electrodes during the same session. Across study centers, the mean of the new electrode recordings was 3.78 microV (standard deviation, 1.13 microV), versus 2.93 microV (1.29 microV) for used electrodes. This 0.85-microV reduction (22%) was statistically significant (F = 7.10 p = 0.01). Electrodes used three times demonstrated an average change in the coefficient of variation of 14% (standard deviation/mean = coefficient of variation; new, 1.13/3.78 = 30%; used, 1.29/2.93 = 44%). Two of the study sites (Houston/Indianapolis) conducted test-retest pattern electroretinograms on a total of 18 patients and found the mean evoked potential to be 3.55 microV with new electrodes and 2.82 microV with used electrodes. The coefficient of variation for the test-retest data was 30% and 47% for new and used electrodes, respectively. Light microscopy showed small cracks on the surface of the electrode, with the number and configuration of the cracks varying in each electrode. The presence of cracks is further complicated by their proximity to the tear film. These sources of variation can result in significantly different impedances. We propose that constant flexion, as a result of patient blinking, causes cracks in the thin gold surface of the electrode. Used electrodes will produce lower pattern electroretinogram amplitudes and poor test-retest reliability.(ABSTRACT TRUNCATED AT 250 WORDS)