Follow-up telephone interviews were conducted with 2,100 parents of children under the age of four who were involved in accidents in North Carolina in order to overcome two limitations of previous studies -- questionable restraint and injury information. The data obtained was weighted to make it comparable with the statewide accident population. Conclusions drawn are that unrestrained children are injured by contact with the vehicle interior or ejection. In contrast, properly restrained children are bruised by contact with the restraint system itself or cut by flying glass or loose objects. When severe head or fatal injury rates for children under four are compared across all levels of crash severity, safety belts are shown to be 59 percent effective, improperly used safety seats are 48 percent effective and properly used safety seats are 81 percent effective in preventing these injuries.