Molecules with neurotrophic activity are being evaluated for treatment of retinitis pigmentosa in animal models. In particular, great interest has been focused recently on erythropoietin (Epo). Evidence of its neurotrophic activity comes mainly from data demonstrating photoreceptor protection in a rodent light-damage model through systemic administration of a recombinant form of this hormone. Our goal was to test whether Epo retinal gene transfer can rescue or delay photoreceptor cell death. We delivered adeno-associated viral vectors encoding Epo intraocularly and, for comparison, intramuscularly to one light-induced and two genetic models of retinal degeneration. Intraocular Epo gene transfer resulted in sustained hormone expression in the eye, which was undetectable systemically. In contrast, Epo intramuscular gene transfer resulted in hormone secretion in the circulation, which was not detected in ocular fluids. The protein secreted from muscle and retina is of the same molecular weight as a commercial recombinant human Epo. Interestingly, following systemic but not intraocular Epo delivery, morphological photoreceptor protection was observed in the light-damage and rds/peripherin (Prph2) models of retinal degeneration. In the light-damage model, the morphological rescue was accompanied by a significant electrophysiological improvement of photoreceptor function. In contrast, no photoreceptor rescue was observed following Epo gene transfer in the rd10 model. This suggests that different apoptotic mechanisms, with varying sensitivities to Epo, occur in different retinal degeneration models. In conclusion, our data support Epo as a neuroprotective agent in some, but not all, retinal degenerations. Further, rescue is observed in specific models after systemic but not intraocular Epo gene transfer.