This paper examines the crash involvement and occupant safety record of sport utility vehicles (SUV’s) in comparison with those of cars, vans, and pickups. The investigation is based on the crash experience of the cited vehicles on the US roads, from 1988 to 1996, as compiled in the National Highway Traffic Safety Administration’s (NHTSA’s) records.In comparison with passenger car occupants, SUV occupants are exposed to significantly lower crash severities in planar crashes and they are younger. SUV occupants are underrepresented in crashes with other vehicles, but they are more frequently exposed to rollovers (8.7% vs. 33%). The overall belt use rate in SUV’s is about the same as for passenger cars. The overall belt use rate in rollover crashes is much lower than in planar crashes -51% for rollover vs. 62% for planar.The rate of seriously or fatally injured occupants per 100 involved in tow-away crashes is used in this study to assess the relative injury risk for different classes of vehicles. Overall injury risks are 1.42 for belted and 5.02 for unbelted occupants. In rollovers, the overall injury risks are 2.33 for belted and 10.11 for unbelted. Belt use has a large influence on the protection of occupants in both planar and rollover crashes. However, the benefits of belt use are the largest in rollover crashes.In comparing SUV’s and passenger cars, there is no statistical difference in overall injury risks for the combined belted and unbelted populations. However, when examining the belted populations, the injury risks of the SUV occupants are lower than those of passenger car occupants - 0.90 vs. 1.37 in planar crashes, and 1.39 vs. 2.84 in rollovers. For belted occupants, there is no statistical difference in the injury risks for SUVs in rollovers (1.37), and passenger cars in planar crashes (1.39).The unbelted population involved in SUV rollovers accounts for more than half of the serious and fatal injuries in SUV’s. The belted occupants in SUV’s have the overall lowest injury rates of all populations examined.