The authors report on an examination of the relationship, or correlation, between performance in the roof crush test (FMVSS 216, ) and the likelihood of injury following rollover for different car models. Specifically, it has been asserted that roof crush test performance is a valid indicator of the protection afforded passengers in a rollover accident. If this is correct, it would be expected that cars which perform relatively “better” in the roof crush test will also perform relatively “better” in protecting passengers in rollover accidents on the road. Performance in the roof crush test is measured by inches of roof deflection for application of a load through a rigid, unyielding plate placed in specific orientation with respect to the automobile roof. Protection of passengers in rollover accidents is measured by actual injury rates obtained from accident/injury data from the state of Washington collected for multiple model and accident years. Three statistical indicators (rank correlation statistic, sample correlation coefficient, and linear regression) have been applied to the roof crush test and rollover accident data for a collection of twelve different car models. This analysis reveals that there is no apparent relationship between roof crush performance, as measured by the roof crush test specified in FMVSS 216, and occupant protection, as measured by injury rates reported in the Washington state accident data base.