Primary damage to the blood-spinal cord barrier (BSCB) is a nearly universal consequence of spinal cord injury that contributes significantly to the overall pathology. The in vivo tissue-level thresholds for mechanical disruption of the BSCB were identified by comparing the results of spinal cord contusions produced with weight drop injury to a finite element analysis (FEA) of the experimental model. The extent and severity of primary, physical disruption of the BSCB was quantified in adult rats five minutes after graded trauma induced with the Impactor weight drop model of spinal cord contusion. The volume of extravasation of three markers of distinct size – fluorescently labeled hydrazide (~730Da), fluorescently labeled bovine serum albumin (~70kDa), and immunohistochemically labeled red blood cells (~5µm) was evaluated in both the gray and white matter. Extravasation volumes increased with increasing drop height and decreasing species size, and were greater in gray matter than in white matter. A three-dimensional finite element analysis of the weight drop model was performed and validated with the in vivo experimental peak displacement of the spinal cord at two loading conditions. The peak compression of the model was within ten percent of the experimental results. A parametric analysis revealed that the model was most sensitive to changes in the viscoelastic properties of the spinal cord. The finite element model provided temporal and spatial profiles of mechanical parameters that were used to identify tissue-level thresholds for BSCB injury using logistic regression analysis. Thirteen mechanical parameters, including measures of stress, strain, and strain rate were investigated as predictors of BSCB injury. Maximum principal strain (LEP) was considered the best predictor of injury in the gray matter, while von Mises strain (LEVM) was the best predictor of injury in the white matter, although the LEVM thresholds for white matter included relatively substantial error compared to the thresholds for gray matter. The results can be used to improve means and measures of preventing spinal cord injury in humans, define loading conditions for in vitro models of injury, and design new experimental models that produce specific patterns of injury.