Ethanol fuel production is growing rapidly in the rural Midwest, and this growth presents potential environmental impacts. In 2002, the U.S. Environmental Protection Agency (EPA) and the Minnesota Pollution Control Agency (MPCA) entered into enforcement actions with 12 fuel ethanol plants in Minnesota. The enforcement actions uncovered underreported emissions and resulted in consent decrees that required pollution control equipment be installed. A key component of the consent decrees was a requirement to conduct emissions tests for volatile organic compounds (VOCs) with the goal of improving the characterization and control of emissions. The conventional VOC stack test method was thought to underquantify total VOC emissions from ethanol plants. A hybrid test method was also developed that involved quantification of individual VOC species. The resulting database of total and speciated VOC emissions from 10 fuel ethanol plants is relatively small, but it is the most extensive to date and has been used to develop and gauge compliance with permit limits and to estimate health risks in Minnesota. Emissions were highly variable among facilities and emissions units. In addition to the variability, the small number of samples and the presence of many values below detection limits complicate the analysis of the data. To account for these issues, a nested bootstrap procedure on the Kaplan-Meier method was used to calculate means and upper confidence limits. In general, the fermentation scrubbers and fluid bed coolers emitted the largest mass of VOC emissions. Across most facilities and emissions units ethanol was the pollutant emitted at the highest rate. Acetaldehyde, acetic acid, and ethyl acetate were also important emissions from some units. Emissions of total VOCs, ethanol, and some other species appeared to be a function of the beer feed rate, although the relationship was not reliable enough to develop a production rate-based emissions factor.