On-road vehicle emission rates of nonmethane hydrocarbons (NMHCs) were measured in two tunnels in Milwaukee, WI, in summer 2000 and winter 2001. Seasonal ambient temperatures in the Midwestern United States vary more widely than in locations where most studies of NMHC emissions from vehicle fleets have been conducted. Ethanol is the added fuel oxygenate in the area, and, thus, emissions measured here are of interest as other regions phase out methyl tertiary butyl ether and increase the use of ethanol. Total emissions of NMHCs in three types of tunnel tests averaged 4560 +/- 800 mg L(-1) fuel burned (average +/- standard error). To investigate the impact of cold start on vehicle emissions, samples were collected as vehicles exited a parking structure in subzero temperatures. NMHC emissions in the subzero cold-start test were 8830 +/- 190 mg L(-1) fuel-nearly double the tunnel emissions. Comparison of ambient data for the Milwaukee area with tunnel emissions showed the impact of seasonal differences in fuels and emissions on the urban atmosphere. Composition of fuel samples collected from area gas stations in both seasons was correlated with vehicle emissions; the predominant difference was increased winter emissions of lighter hydrocarbons present in winter gasoline. A chemical mass balance model was used to determine the contributions of whole gasoline and gasoline headspace vapors to vehicle emissions in the tunnel and cold-start tests, which were found to vary with season. Results of the mass balance model also indicate that partially combusted components of gasoline are a major contributor to emissions of aromatic compounds and air toxic compounds, including benzene, toluene, xylenes, napthalene, and 1,3-butadiene, whereas air toxics hexane and 2,2,4-trimethylpentane are largely attributed to gasoline and headspace vapors.