The standard procedure for determining non-methane hydrocarbon (NMHC) emissions is to subtract measured methane emissions from total hydrocarbon emissions measured by flame ionization detector. The results of this method were compared to the results of direct GC speciation of hydrocarbon emissions. For gasoline vehicles using an all-hydrocarbon fuel, the two methods demonstrate nearly perfect correlation, with a linear regression coefficient near 1.0, and R2 = 0.999. The correlation using reformulated gasoline is only slightly worse. For natural gas vehicles, however, the correlation was poor, with R2 < 0.30. This poor correlation is attributed to the high methane content of natural gas, which results in NMHC emissions being very low compared to the level of methane. Both the total hydrocarbon and methane measurements contain some error, and the resulting combined error in the NMHC concentration is of the same order as the concentration itself. It is concluded that the subtraction method produces unacceptable errors in determining NMHC emissions from natural gas vehicles, and should not be used for this purpose.