To report methods for performing epiretinal electrical stimulation with microfabricated electrode arrays and determining perceptual thresholds on awake human volunteers during acute surgical trials. Four hypotheses were tested: (1) epiretinal stimulation can be performed during acute experiments without obviously damaging the retina or degrading vision or the health of the eye; (2) perception can be obtained 50% of the time in blind patients with charge densities below published safety limits; (3) the minimal charge needed to induce perception would be higher in patients with more severe retinal degeneration; and (4) threshold charge would be lower at shorter stimulus durations. Five subjects with severe blindness from retinitis pigmentosa and one with normal vision (who underwent enucleation of the eye because of orbital cancer) were studied. Electrical stimulation of the retina was performed on awake volunteers by placing a single 250-microm diameter handheld needle electrode or a 10-microm thick microfabricated array of iridium oxide electrodes (400-, 100-, or 50-microm diameter) on the retina. Current sources outside the eye delivered charge to the electrodes. Assessment of damage was made by observing the clinical appearance of the eyes, comparing pre- and postoperative visual acuity, obtaining retinal histology in one case, and comparing perceptual thresholds with published safety limits. No clinically visible damage to the eye or loss of vision occurred. Even at sites removed from stimulation, histology revealed swollen photoreceptor inner and outer segments, which were believed to be nonspecific findings. Percepts could not be reliably elicited with 50-microm diameter electrodes using safe charges in one blind patient. With the two larger electrodes, only the normal-sighted patient had thresholds at charge densities below 0.25 and 1.0 millicoulombs (mC)/cm(2) for 400- and 100-microm diameter electrodes, respectively, which is one seemingly reasonable estimate of safety derived from the product of charge per phase and charge density per phase. In blind patients, thresholds always exceeded these levels, although most were close to these limits in patient 6. The range of charge density thresholds with the 400- microm electrode in blind patients was 0.28 to 2.8 mC/cm(2). The normal-sighted patient had a threshold of 0.08 mC/cm(2) with a 400-microm electrode, roughly one quarter of the lowest threshold in the blind patients. Strength-duration curves obtained in two blind patients revealed the lowest threshold charge at the 0.25- or 1.0-ms stimulus duration. Threshold charge densities in severely blind patients were substantially higher than that in a normal-sighted patient. Charge densities in blind patients always exceeded one seemingly reasonable estimate of safe stimulation. The potential adversity of long-term stimulation of the retina by a prosthesis has yet to be determined.