A prospective study of 441 rear seat occupants seen following road traffic accidents compared injury severity in the 411 unrestrained and 30 restrained subjects. Using ridit analysis, the injury severity score showed far less injury among restrained subjects (P = 0.0001). All 11 deaths and 176 of the 178 most severe injuries occurred in the unrestrained group. One restrained (3 per cent) subject was admitted overnight and 64 unrestrained (16 per cent) subjects required admission; two-thirds required a prolonged admission (mean = 13 days). Fractures, major bruises, abrasions and lacerations were almost confined to unrestrained passengers, 221 events compared with two among restrained passengers. Twenty-three unrestrained subjects were ejected, nine of whom were killed. We conclude that the use of rear seat restraints greatly reduces injury severity and should be compulsory for rear seat occupants as it is for front seat occupants.