A technique for producing graded spinal cord compression injuries in rats is described. A metal plate 2.2 x 5.0 mm in size is applied to the exposed spinal dura and loaded with weights. Neurological function is tested postoperatively on an inclined plane. Reading of the maximal angle of rat performance on this plane was found reproducible on comparison between different observers. Laminectomy per se had a minimal effect on the performance. Compression of 35 g for 5 min caused a pronounced but incomplete injury, with almost total recovery within 14 days and with no difference between animals on artificial respiration and those breathing spontaneously. Animals whose spine was fixed during the compression had a better outcome than those without such fixation.