A three-part program was undertaken to establish an appropriate means of evaluating child restraints in automobile crashes.A standard seat was designed to provide a reproducible test base on which to evaluate child restraint systems in dynamic testing. Developmental and evaluation data are presented, including child restraint performance tests. Results showed the standard seat to be a durable, repeatable, and economical test platform which provides a realistic base for evaluation of child restraint systems.Commercially available three- and six-year-old child dummies were evaluated for their anthropometric measurements and dynamic response characteristics in pendulum impact tests and simulated crashes in representative automobile-child seat restraint environments. Simulated crashes included 20 and 30 mph frontal and 20 mph side impacts on automobile and specially designed bench seats. Two types of child seats, the G. M. “Love Seat” and Chrysler “Mopar,” were selected for testing as representative of belt and padding restraint types currently in use.The three-year-old child dummies were found to be capable of providing repeatable measurements of the head and chest accelerations and head deflections in sled tests and to be sufficiently sensitive of detecting differences in the crash environments.Acceleration measurements on both six-year-old child dummies were found to contain resonances.An extensive sled testing was performed with child restraint systems to determine which of the two types of dummies simulate best the in-vivo response of primates and what criteria may be employed to measure the dummy response and the efficacy of child protection systems. Test results indicate that the mechanical test dummy, consisting of clearly defined adjustable body segments, is a better simulator than the lumped mass unadjustable type. Test results also indicate that current instrumentation technology for application to in-vivo subjects is not adequately developed to permit consistent measurements in terms of “G” levels, injury criteria, etc. A better correlative measure between primates and mechanical child dummies is the head excursion, in both the forward- and lateral-collision modes. This is, however, limited to the first collision phase and is not valid for the rebound phase. The tested shield-type child restraint exposed the occupant to potential risk of ejection during the rebound mode of the forward-collision phase and to substantial excursion in lateral collisions if the seat does not have provisions for limiting the lateral motion of the test subject.