As motor vehicle emissions have been reduced to meet requirements of the clean air acts, they have become low enough to be difficult to measure accurately. This is especially the case for hydrocarbons, because after warm-up, there are fewer hydrocarbons emitted from a modern vehicle's tailpipe than in the surrounding air. It is therefore important to correctly compensate for the ambient hydrocarbon levels of the air used to dilute the collected exhaust. In estimating the accuracy of the federally required testing procedures, previously published error analyses have examined the effects of random errors. This study examines the systematic errors inherent in the CVS (Constant Volume Sampling) technique specified in federal regulations, estimates their sizes, and proposes a method using proportional ambient sampling whereby they can be avoided.