In 1992, a Round Robin was sponsored by the CRC's Emissions Analysis Round Robin Subcommittee, to provide an opportunity for automotive emissions laboratories to compare their analytical methodologies with those used in other laboratories. Compressed gas samples were provided to participants to test hydrocarbon methodologies, while liquid samples were used for alcohol and carbonyl analyses. The results of this study were published in SAE 950780 and SAE 941944.A second Round Robin study was conducted in 1995, using the same basic structure as the first study. The results of the carbonyl analyses have been published separately (SAE 971609). The purpose of this paper is to compare methods used for hydrocarbon speciation of emissions by gas chromatography. As in the 1992 study, cylinders of a synthetic exhaust were prepared by using a fuel base, and adding components that would be expected as typical combustion products. These cylinders were distributed among a variety of participating laboratories, which included automotive manufacturers, petroleum companies, regulatory agencies, and contract laboratories.Direct comparison of individual species' concentrations showed that inter laboratory variability was on the order of 5 - 10% C.V. (coefficient of variation,%) for all species that were present at concentrations >1 ppmC. This is lower than the 10 - 15% C.V. found in the previous round robin. For species with concentrations <1 ppmC but >0.1 ppmC, variabilities of 5 - 18% were observed, except for MTBE which showed greater variability. For species with concentrations < 0.1 ppmC, variability ranged from 12 - 80%, though detection is more difficult at these lower concentrations. Interlaboratory variability for total NMOG was 4.6%. This is lower than the 18 - 20% observed during the previous round robin.Compositional profiles provided an additional perspective of the data. These profiles were developed by expressing the results for groups of species, categorized by hydrocarbon type and number of carbons. Comparison of results from each laboratory with the average results of all laboratories allows participants to see trends in, or areas of, discrepancies with respect to other laboratories.