The temporary ineffectiveness of motor vehicle emission controls at startup causes emission rates to be much higher for a short period after starting than during fully warmed, or stabilized, vehicle operation. Official motor vehicle emission inventories estimate that excess emissions during cold-start operation contribute a significant fraction of all hydrocarbon, carbon monoxide (CO), and nitrogen oxide (NOx) emissions from California vehicles. In an effort to verify these estimates under real-world conditions, vehicle emissions were measured in an underground parking garage in Oakland, CA, during March 1997. Hot stabilized emissions were measured as vehicles arrived at the garage in the morning, and cold-start emissions were measured as vehicles exited in the afternoon; the incremental, or excess, emissions associated with vehicle starting were calculated by difference. Composite emissions from ~135 vehicles were sampled during each of six morning and six afternoon periods. Measured stabilized exhaust emissions were 19 ± 2 g nonmethane hydrocarbons (NMHC), 223 ± 17 g CO, and 8.6 ± 1.3 g NOx per gal of gasoline consumed. Cold-start emissions of 69 ± 2 g NMHC/gal, 660 ± 15 g CO/gal, and 27.8 ± 1.2 g NOx/gal were measured for vehicles spending an average of ~60 sec in the garage after starting in the afternoon. Using second-by-second emissions data from California's light-duty vehicle surveillance program, average fuel use during cold start was estimated to be ~0.07 gal, and the cold-start period was estimated to last for ~200 sec. When cold-start emission factors measured in the garage were scaled to represent the full 200-sec cold-start period, incremental start emission factors of 2.1 g NMHC, 16 g CO, and 2.1 g NOx per vehicle start were calculated. These emission factors are lower than those used by California's motor vehicle emission inventory model (MVEI 7G) by 45% for NMHC, 65% for CO, and 12% for NOx. This suggests that the importance of cold-start emissions may be overstated in current emission inventories. Overall, the composition of volatile organic compound (VOC) emissions measured during cold start was similar to that of hot stabilized VOC emissions. However, the weight fractions of unburned fuel and acetylene were higher during cold start than during hot stabilized driving.