A model that provides reproducible, submaximal yet sufficient spinal cord injury is needed to allow experiments leading to development of therapeutic techniques and prediction of clinical outcome to be conducted. This study describes an experimental model for spinal cord injury that uses three different volumes of balloon inflation and durations of compression to create a controlled gradation outcome in adult dogs. Twenty-seven mongrel dogs were used for this study. A 3-french embolectomy catheter was inserted into the epidural space through a left hemilaminectomy hole at the L₄ vertebral arch. Balloons were then inflated with 50, 100, or 150 レl of a contrast agent at the L1 level for 6, 12, or 24 h and spinal canal occlusion (SCO) measured using computed tomography. Olby score was used to evaluate the extent of spinal cord injury and a histopathologic examination was conducted 1 week after surgery. The SCO of the 50, 100, and 150 レl inflations was 22-46%, 51-70%, and 75-89%, respectively (p < 0.05). Olby scores were diminished significantly by a combination of the level of SCO and duration of inflation in all groups. Olby scores in the groups of 150 レl-12 h, 150 レl-24 h, and 100 レl-24 h were 0.5, 0, and 1.7, respectively. Based on these results, a SCO > 50% for 24 h, and > 75% for 12 h induces paraplegia up to a week after spinal cord injury.