A technique was developed for the measurement of total nonmethane organic carbon (Cy) in the gas phase based on cryogenic collection, preseparation of CO, CH4, and CO2, conversion of all carbon to CH4, and measurement by flame ionization detector. Collection and recovery efficiencies were found to be essentially quantitative for C2-C7 hydrocarbons and better than 95 and 93% for methanol and formaldehyde, respectively. CO2 prevented the measurement of the C2 hydrocarbons in atmospheric samples, and system response to ambient water vapor limited the precision of the measurement at very low Cy. The measurement system had an estimated uncertainty of 10% at ambient mixing ratios of 100 ppb C, and a detection limit of between 5 and 7 ppb C. Cy was measured in Boulder, Colorado, during June of 1993 and at Chebogue Point, Nova Scotia, during the North Atlantic Regional Experiment intensive campaign, August to September 1993. Cy values ranged from several hundred ppb C to 11 ppb C in Boulder, and from 177 to below 5 ppb C at Chebogue Point, with medians of 55 and 11 ppb C, respectively. Cy was found to correlate with both CO and O3 at higher levels of those two pollutants. The sum of Cy and the C2 hydrocarbons was virtually always equal to or greater than the sum of the hydrocarbons. Several exceptions to this occurred when high HC values were observed in cannister samples but not Cy measurements. These instances were attributed to short-term perturbations from local sources. There was a net difference between total nonmethane organic carbon and the sum of hydrocarbons and carbonyls of 3.5¿7.8 ppb C overall. This difference appears to be broadly correlated with anthropogenic pollution. There does not appear to be a significantly large reservoir of reactive carbon unaccounted for by hydrocarbon or carbonyl measurements. ¿ 1998 American Geophysical Union