Objective: Unbelted occupants may increase the risk of injury for other occupants in a motor vehicle collision (MVC). This study evaluated the association between occupant restraint use and the risk of injury (including death) to other vehicle occupants. Design: A population based cohort study. Setting: United States. Subjects: MVC occupants (n = 152 191 unweighted, n = 18 426 684 weighted) seated between a belted or unbelted occupant and the line of the principal direction of force in frontal, lateral, and rear MVCs were sampled from the 1991-2002 National Automotive Sampling System General Estimates System. Offset MVCs were not included in the study. Main outcome measure: Risk ratios and 95% confidence intervals for injury (including death) for occupants seated contiguous to unbelted occupants compared to occupants seated contiguous to belted occupants. Risk ratios were adjusted for at risk occupant's sex, age, seating position, vehicle type, collision type, travel speed, crash severity, and at risk occupants' own seat belt use. Results: Exposure to unbelted occupants was associated with a 40% increased risk of any injury. Belted at risk occupants were at a 90% increased risk of injury but unbelted occupants were not at increased risk. Risks were similar for non-incapacitating and capacitating injuries. There was a 4.8-fold increased risk of death for exposed belted occupants but no increased risk of death for unbelted occupants. Conclusions: Belted occupants are at an increased risk of injury and death in the event of a MVC from unbelted occupants.