This study examined the effect of median width on the frequency and severity of crashes on homogeneous highway sections with a traversable (nonbarrier) median. Crash data from 982 highway sections in Utah and 2,481 highway sections in Illinois were used in the analyses. Log-linear models were used to assess the effects of median width and other roadway variables on the crash rate. The total crash rate appears to decline steadily with increasing median width from 0 to 33.5 m (0 to 110 ft). For Utah it declines by a factor of six, and for Illinois, it declines by a factor of 13. Over this range of median width, the rates of serious injury, all injury, and property-damage-only crashes also decline by up to a factor of 15. However, when constructing new highways, medians need to be at least 9.1 m (30 ft) wide to have a positive safety effect. Additional data will be needed to determine the median width at which a median barrier should be used.