The applicability of the federal safety standard that governs child restraints (FMVSS 213) has recently been called into question. Population- based estimates of the risk of injury to children in child restraints and a description of the patterns and mechanisms of injury are necessary to evaluate this standard and identify areas of needed improvement. A probability sample of children 12 to 47 months in crashes was identified in an on-going crash surveillance system (1998-2002) which links insurance claims data to telephone survey and crash investigation data. The risk of injury in forward facing child restraints (FFCRS) was estimated and a series of cases was examined using in-depth crash investigation to identify the mechanisms of these injuries. Although children in FFCRS are well protected in crashes, further reductions in serious injuries might be achieved by reducing head, neck, and extremity injury risks. These results have implications for the current efforts to upgrade the current FMVSS 213 and better protect child passengers.