The kinematics of rear-end collisions based on published acceleration pulses of actual car-to-car collisions (10 and 23 mph) were reproduced on a crash simulator using anthropomorphic dummies, human cadavers, and a volunteer. Comparison of the responses of subjects without head support were based on the reactions developed at the base of the skull (occipital condyles). The cadavers gave responses which were representative of persons unaware of an impending collision. The responses of both dummies used were not comparable with those of the cadavers or volunteer, or to each other.An index based on voluntary human tolerance limits to statically applied head loads was developed and used to determine the severity of the simulations for the unsupported head cases. Results indicated that head torque rather than neck shear or axial forces is the major factor in producing neck injury.When the head was initially supported by a flat, padded headrest, all subjects gave comparable headrest loads. Using this configuration, the volunteer withstood an equivalent 44 mph simulation with only slight discomfort.With the head separated from the headrest by 2-3/4 in. the head load increased from 150 to 390 lb, but with additional padding, the load was increased to only 250 lb.Controlled seat back rotation decreased the magnitudes of the head loads and neck reactions for the supported and unsupported head cases, respectively.