Occupant protection in rear impact involves two competing challenges. On one hand, allowing a deformation of the seat would act as an energy absorber in low severity impacts and would consequently decrease the risk of neck injuries. However, on the other hand, large deformations of the seat may increase the likelihood of occupant ejection in high severity cases. Green et al., 1987 analyzed a total of 919 accidents in Great Britain. They found that occupant ejection resulted in a risk of severe injuries and fatalities between 3.6 and 4.5 times higher than those cases where no ejection was observed. The sample included single front, side and rear impacts as well as multiple impacts and rollover. The rate of belt use in the sample was 50%. While this analysis included all forms of impact scenarios, nevertheless, it highlights the relative injury severity of occupant ejection.Extensive literature search has found no full-scale rear impact tests involving Post Mortem Human Subjects (PMHS) conducted in a laboratory environment and resulting in ejection.This paper describes a total of 10 sled tests conducted on 3 belted PMHS using a simplified seat design composed of rigid plates assembled such that the angular and linear stiffness of the seatback (including the foam) was modeled. The initial angular position and the range of motion of the seatback, the size of the PMHS, the slack length of the seatbelt, the angular stiffness of the seatback, and the use of headrest were varied in the test matrix while the pulse was kept constant (triangular acceleration with a peak of 17 G at 30 ms and a duration of 95 ms).In the test series, the tests were not run randomly but the likelihood of occupant ejection was increased systematically until ejection occurred.PMHS seat ejection was observed only for the 95th percentile, initially positioned with a seatback angle relative to the vertical equal to 22°, a range of seatback angular motion equal to 44° and no headrest. Repeating the test under the same conditions but with the pretensioner fired did not prevent the ejection. In addition, the 50th percentile belted specimen was not observed to sustain rearward seat ejection under realistic conditions including the use of headrest.