The purpose of this study was to quantify the biomechanical response of the human thorax during dynamic shoulder belt loading representative of that seen in a severe automotive collision. Two post-mortem human surrogates (PMHSs) (one male and one female) were instrumented with 26 single-axis strain gages on the ribs, sternum, and clavicle. The thorax of each PMHS was placed on a custom spine support bracket designed to support the thorax on either side of the spinous process, thereby allowing free motion at the costovertebral joints. In addition, the support bracket raised the thorax above the flat base plate, which could otherwise constrain the deformation and motion of the posterior region of the rib cage. The thorax of each PMHS was then loaded using a custom table-top belt loading system that generated thoracic displacement rates representative of a severe automotive collision, 1.3 m/s for the male PMHS and 1.0 m/s for the female PMHS. The rib fracture timing data, determined by analyzing the strain gage time histories, showed that severe thoracic injury (AIS = 3) occurred at 16% chest compression for the male and 12% chest compression for the female. However, these values are well below the current thoracic injury criteria of 29% chest compression for the male and 23% chest compression for the female. This data illustrates that serious thoracic injury (AIS = 3) occurs at lower chest compressions than the current ATD thoracic injury criteria. Overall, this study provides critical data that can be used in the design and validation of advanced ATDs and finite element models, as well as the establishment of improved, more stringent thoracic injury criteria.