Sixty-three simulated frontal impacts using cadaveric specimens were performed to examine and quantify the performance of various contemporary automotive restraint systems. Test specimens were instrumented with accelerometers and chest bands to characterize their mechanical responses during the impact. The resulting thoracic injury severity was determined using detailed autopsy and was classified using the Abbreviated Injury Scale.The ability of various mechanical parameters and combinations of parameters to assess the observed injury severities was examined and resulted in the observation that belt restraint systems generally had higher injury rates than air bag restraint systems for the same level of mechanical responses. To provide better injury evaluations from observed mechanical parameters without prior knowledge of what restraint system was being used, a dichotomous process was developed. This process first determines, based on multiple chest deformation patterns, whether the restraint system is behaving belt like or bag like, and then uses either the separately developed belt or bag criteria, to estimate injury severity. This two-step process has shown improved predictive capabilities for this data set.