Although rollover crashes represent a small fraction (approximately 3%) of all motor vehicle crashes, they account for roughly 22% of crash fatalities to occupants of cars, light trucks, and vans (NHTSA Traffic Safety Facts, 2005 (1)). Of the fatally injured occupants in rollover crashes, 57% were ejected (2). With the development of advanced airbag and sensing technologies, General Motors (GM) has introduced systems intended to help mitigate the risk of head and torso ejection during a rollover crash. The implementation of these systems was preceded by the development of a suite of rollover sensor laboratory tests designed to simulate several types of rollover initiations. Many of these tests were conducted with instrumented Hybrid III 50th percentile Anthropomorphic Test Devices (ATDs) seated in the front outboard seating positions. For tests in which an Injury Assessment Reference Value (IARV) (3) was exceeded, a methodology was developed to provide a detailed summary of the vehicle kinematics, timing of ATD contacts, ATD peak responses, and film observations. Using this procedure, GM was able to identify common trends of peak ATD responses relative to restraint use and rollover initiation type. IARVs were shown to be exceeded in all test types, with both belted and unbelted ATDs. Although exact ATD motion was unpredictable, test type did have some effect on the location of ATD contact. In addition, the location of contact by leading side ATDs was influenced more by test type than by restraint usage. IARVs were shown to be exceeded with the vehicle at a wide range of orientations. Any impact during which the motion of the ATD head was arrested prior to stopping the ATD body showed the potential for exceeding a neck compression IARV. This was true regardless of vehicle orientation, location of the head contact, or dynamic deformation of the vehicle's structure.