Objective To develop a tractive spinal cord injury model in rats with a novel spinal distractor so as to supply the rel iable animal model for researching the pathological mechanism and rehabil itation treatment of tractive spinal cordinjury. Methods A novel spinal distractor was prepared based on previous study. Sixty adult Sprague Dawley rats (weighing 250-300 g) were randomly divided into 5 groups, 12 rats in each group. T12-L3 spinal structures in the rear area were exposed and then T13-L2 spinal cords were revealed via dual laminectomy and kept integrity. In group A, a novel spinal distractor was placed without distraction; in groups B, C, D, and E, the T12-L3 spines were tracted with a novel spinal distractor which put on transverses process of T12-L3 vertebrae. During the tractive period, the somatosensory evoked potential (SEP) was used to monitor spinal cord function. The SEP ampl itudes descended 50% and kept distracting for 5 minutes in group B and for 10 minutes in group C, and descended 70% and kept distracting for 5 minutes in group D and for 10 minutes in group E, respectively to establ ish the tractive spinal cord injury model of T11-L2. The improved combine behavioral score (ICBS) was recorded at 1 and 7 days after injury in 6 rats of each group. The T13-L2 spinal tissue specimens were harvested for the morphological observation by HE and Nissl’s staining and for neurons counting. Results In group A, the ICBS score was 0 at 1 and 7 days after operation, showing significant difference when compared with the scores of the other groups (P lt; 0.05). The ICBS scores of groups D and E were significantly higher than those of groups B and C (P lt; 0.05). Edema and hemorrhage were observed in spinal cord surface and normal morphological structures were destroyed at different extent in groups B, C, D, and E at 1 day. There were adherence and congestion between spinal cord surface and peripheral issue without luster at 7 days, and dura depression was observed at the injury section, especially in group E. Necrosis and dissolution occurred in some neurons, and Nissl body structure dissolved or disappeared in groups B, C, D, and E. The neuron counting gradually decreased in accordance with the aggravation of injury in groups B, C, D, and E, showing significant difference when compared with group A (P lt; 0.05). Significant differences in neuron counting were found among groups B, C, D, and E (P lt; 0.05). Conclusion The tractive spinal cord injury model in rats can be successfully establ ished with novel spinal distractor, and the model establ ished by SEP ampl itude descending 70% and keeping distracting for 10 minutes is more suitable for study in tractive spinal cord injury.