As measurement capabilities in crash test dummies improve, new injury criteria should be considered to take advantage of these improvements. The THOR-NT dummy thorax has been designed with three-dimensional displacement measurement capability at four points in the chest. To correlate those measurements with injury, chestband displacements from Post Mortem Human Subjects (PMHS) tests corresponding to the THOR-NT chest displacement points were used to simulate thorax loading in a 2-D finite element model of the human thorax. The model, method and model validation were described in Campbell et al. (2005). In the current study, data from both upper and lower chestbands were used to predict rib fractures in the PMHS crash tests. Due to the close proximity of the two upper THOR-NT chest displacement points, some of the simulations did not adequately represent the PMHS loading. To improve the simulations, a new set of runs were created using wider chest displacement points to determine if they would be more successful in simulating injury. Rib stress and strain from the 2-D finite element model of the PMHS thorax were used to predict injury or noninjury in the PMHS tests. Statistical analysis using logistic regression was used to investigate a new thoracic injury criterion based on the finite element model simulations.