In order to reduce the number and severity of injuries to child occupants in car accidents, a great number of child restraint systems have been developed over the past years. Such systems must be adapted to the anthropometric characteristics of children and provide good protection; to achieve this, a knowledge of child tolerance to impact is required, but at present very little biomechanical data relating to children is available, especially for children in the first years of life. As the design, evaluation and certification of child restraint systems is performed with dummies and several dummy types are available- a relationship between dummy and expected child reactions must be identified.This paper, based on the work performed within the framework of the International Task Force on Child Restraint Systems*, proposes a comparison between child dummies and cadavers involved in identical experimental collisions, and restrained with child restraint systems. This comparison involves both kinematic and physical measurements when available, taking into account injuries observed on cadavers. It is known that the safety performance of child restraint systems is impeded by a lack of knowledge of the biofidelity of child dummies. From a limited number of tests, indications are given concerning the representativeness of the dummies and the consequences for the design of child restraint systems.