The capability to predictively simulate occupant dynamics in vehicle rollover crashes using the Articulated Total Body (ATB/CVS) model was validated using the results of two controlled automobile rollover crash tests. The ATB model requires the occupant's inertial, geometric, and resistive joint torque properties, the vehicle interior geometry and motion, the contact characteristics for the occupant and vehicle interactions, and the seat belt characteristics. The validation was done by first simulating one test and adjusting the contact and belt properties to obtain good comparison with the test results. Then subsequent tests were simulated using the same properties, but changing only the input vehicle kinematics. Each occupant simulation used the standard Hybrid III data set and measured vehicle interior geometry. The vehicle kinematics were generated by simulating the vehicle dynamics with the ATB model. In one rollover, roof crush significantly affected the occupant's motion. To obtain an accurate occupant simulation, it was necessary to allow portions of the roof and windows to move and permanently deform. The motion time histories of the roof and window sections were derived visually from the test films. The simulation results included head, chest, and pelvis accelerations, and neck loads. Comparisons were made between these simulation results and the instrumentation data, and visual comparisons were made with the test films, validating the modeling technique.