Five late model vehicles equipped with representative emission control systems were used to determine the effect of benzene concentration of gasolines on evaporative and exhaust benzene emissions. The vehicle selection included three different fuel induction systems and two different exhaust emissions control systems. The test fuels consisted of 25 and 40% aromatic base fuels each at four benzene levels ranging from 0.02 to 4%. Evaporative and exhaust determinations included measurement of regulated emission components and benzene emission in each test segment. Benzene level in the fuel tank head space was also measured. In addition to the above evaporative and exhaust emission test program, exhaust samples were collected simultaneously before and after the exhaust emissions control system to determine engine-out and tailpipe-out emission rates as well as the catalyst conversion efficiency.Benzene levels in evaporative and exhaust emissions and tank head space vapor all increased linearly with increasing fuel benzene level. Only the tank head space measurements showed no benzene with benzene-free fuel. Tests with benzene-free fuel showed that the measured benzene evaporative and exhaust emissions were due to canister elution of benzene captured from previous test fuels and benzene formation during combustion, respectively. Vehicular benzene emission rates were computed from measured evaporative emissions, exhaust emissions, and estimated refueling loss. The fleet-averaged total benzene emission rates ranged from 10 to 31 milligrams per mile and were linearly related to benzene level in the test fuel. For 1.5 vol-% benzene fuel, the contributions to the total benzene emissions averaged about 70% exhaust, 20% evaporative, and 10% refueling.Comparison of benzene emission measurements before and after the catalyst showed that the benzene level was reduced about 74 to 95%, and the hydrocarbon reduction ranged from 82 to 91% among the five cars.The evaporative benzene carry-over study showed that evaporative benzene emissions were directly affected by the composition of prior test fuels for as many as two or three prior tests.