The risk of injury to child car occupants can be markedly reduced by their use of appropriate child restraints. These can be child seats with their own integral harness, child booster seats or booster cushions in association with adult seat belts or, if the child is old and large enough, by the use of adult seatbelts alone. However, the protection afforded can be negated if there is significant loading to the child and restraint through the car seat backrest. This paper describes analyses of accident data to demonstrate the occurrence of this effect in field accidents and presents the results of dynamic tests performed to explore the effect on different restraint types of limiting the load intrusion from the rear. Results of tests with child seats with integral harnesses show that head forward excursion is the main concern, the R44 limits being impossible to meet if the seat back is allowed to move as far forward as the R-point. Where the child is restrained by the adult belt, belt forces in excess of the injury tolerance for adults were exceeded, raising to very high levels for backrest movement up to 150mm ahead of the R-point.