Data from the Hybrid III and THOR (Test Device for Human Occupant Restraint), anthropomorphic test devices (ATD) currently available to test the Occupant Protection requirements, are not well correlated to low-injury risk, as these ATDs were designed for automotive use. Automotive research is directed at preventing severe injuries in very low probability events. NASA vehicles require a lower risk of injury because the vehicles will land every time, making that a high probability event. The objective of this study is to develop injury risk functions for the Hybrid III and THOR ATDs. Matched pair tests between postmortem human surrogates (PMHS) and each ATD will be used to determine ATD-specific injury criteria. The merit of the matched pair design is the one-to-one correspondence of the results from external loads to both surrogates. Injury outcomes from PMHS tests will be used with region-specific data, such as forces and moments either individually or in combination, to derive ATD-specific injury criteria. Specific Aims 1. Identify appropriate datasets for ATD comparison 2. Test Hybrid III 50th percentile male and THOR in same conditions as historical testing 3. Use historical human data to establish tolerance and injury risk focusing on lateral responses and sex differences 4. Use Bayesian analysis combined with survival analysis along with human tolerance to estimate injury risk. Use results of prior data mining and existing literature as prior distribution 5. Develop new Injury Assessment Reference Values (IARVs) based on the new statistical analysis. Historical human data will be selected from the Medical College of Wisconsin (MCW) database. The data will be selected based on loading dynamics and subject demographics. Once these data are selected, the Hybrid III 50th percentile male and THOR ATDs will be tested in identical conditions. A Bayesian analysis along with survival analysis will be used to relate the resulting ATD responses to improve injury risk predictions. The results of the Occupant Protection (OP) Data Mining and Modeling Task will be used as prior distributions.