The protective effects of erythropoietin on spinal cord injury have not been well described. Here, the eukaryotic expression plasmid pcDNA3.1 human erythropoietin was transfected into rat neural stem cells culturedin vitro. A rat model of spinal cord injury was established using a free falling object. In the human erythropoietin-neural stem cells group, transfected neural stem cells were injected into the rat subarachnoid cavity, while the neural stem cells group was inject-ed with non-transfected neural stem cells. Dulbecco’s modified Eagle’s medium/F12 medium was injected into the rats in the spinal cord injury group as a control. At 1–4 weeks post injury, the motor function in the rat lower limbs was best in the human erythropoietin-neural stem cells group, followed by the neural stem cells group, and lastly the spinal cord injury group. At 72 hours, compared with the spinal cord injury group, the apoptotic index and Caspase-3 gene and protein expressions were apparently decreased, and the bcl-2 gene and protein expressions were noticeably increased, in the tissues surrounding the injured region in the human erythro-poietin-neural stem cells group. At 4 weeks, the cavities were clearly smaller and the motor and somatosensory evoked potential latencies were remarkably shorter in the human erythropoi-etin-neural stem cells group and neural stem cells group than those in the spinal cord injury group. These differences were particularly obvious in the human erythropoietin-neural stem cells group. More CM-Dil-positive cells and horseradish peroxidase-positive nerve fibers and larger amplitude motor and somatosensory evoked potentials were found in the human erythro-poietin-neural stem cells group and neural stem cells group than in the spinal cord injury group. Again, these differences were particularly obvious in the human erythropoietin-neural stem cells group. These data indicate that transplantation of erythropoietin gene-modified neural stem cells into the subarachnoid cavity to help repair spinal cord injury and promote the recovery of spinal cord function better than neural stem cell transplantation alone. These ifndings may lead to signiifcant improvements in the clinical treatment of spinal cord injuries.