The chemical mass balance model has been used to separate non-methane hydrocarbon emission factors measured in the Cassiar tunnel study into exhaust and evaporative emission factors. The local gasoline composition has been used as a real-w orld surrogate profile for exhaust emissions and has been demonstrated to result in vastly improved model performance compared to the performance obtained w ith the use of an exhaust profile derived from dynamometer testing. Because of the approach used, the combustion and unburned gasoline components of exhaust emission gases could be estimated separately. Unburned gasoline w as found to comprise 63.4 (7.0% of exhaust gases for light-duty vehicles operating in steady-state driving conditions in this study. On-road benzene emissions w ere found to split 71%/27%/2% betw een the combustion, unburned gasoline, and evaporative sources. Evaporative nonmethane hydrocarbons w ere found to represent 10.3 (0.8% of the total on-road emission rate on average. The apportionment of total NM HC emission factors to exhaust and evaporative emission factors allow ed a detailed comparison to exhaust and on-road evaporative emission factors predicted by the M OBILE4.1C and M OBILE5C models.