Given the frequency and severity of cervical spine injuries resulting from rollover crashes, it is critical to analyze the mechanism of cervical spine injury in this loading condition. In rollover crashes, roof-to-ground impacts can generate axial compression of the cervical spine, which can result in paralysis and death. This study was performed to compare injury type and severity between component and full body inverted vertex impact tests with post-mortem human surrogates (PMHS);​ a secondary aim was to determine how changes in vertebral kinematics resulted in changes in hear reaction loads. Five PMHS were suspended in an inverted seated position and then dropped from two heights to achieve 2 m/s (one subjects once and another twice) and 4.4 m/s (all subjects) at impact. The subjects were dropped on a padded five-axis load cell to record the reaction force from impact. Each PMHS was instrumented with three blocks (each containing three accelerometers and three angular rate sensors) rigidly mounted along the upper thoracic spine and on the head. Injuries were determined using both CT scans and dissection following testing. Vertical force traces from the load cell reflect a similar two peak shape seen in previous full-body and component tests. High-speed (1000 Hz) X-ray video analysis shows the neck retains in its initial orientation but becomes increasingly compressed during the loading portion of the first peak. At the first peak, the cervical spine begins to curve, putting the cervical spine into extension, with the center of curvature around C3 or C4, and continues into bending during the unloading of the first peak. The head then translates forward and the neck moves into flexion during the second peak. Each PMHS achieved a flexion injury in the upper thoracic spine or the lower cervical spine during the testing, which occurred during the second peak of the force trace, contradicting previous theories that injury occurs at the first peak, where maximum force occurs. These tests suggest that the direction of torso loading, impact velocity, and boundary conditions at the ends of the cervical spine all affect the kinematics during impact as well as the resulting injuries, and should all be taken into account when determining appropriate injury criteria and developing biofidelic ATDs to predict injuries in crash tests.