Observations of seasonal variations in the atmospheric O2/N2 ratio are reported at nine baseline sites in the northern and southern hemispheres. Concurrent CO2 measurements are used to correct for the effects of land biotic exchanges of O2 on the O2/N2 cycles thus allowing the residual component of the cycles due to oceanic exchanges of O2 and N2 to be calculated. The residual oceanic cycles in the northern hemisphere are nearly diametrically out of phase with the cycles in the southern hemisphere. The maxima in both hemispheres occur in summer. In both hemispheres, the middle-latitude sea level stations show the cycles with largest amplitudes and earliest phasing. Somewhat smaller amplitudes are observed at the high-latitude stations, and much smaller amplitudes are observed at the tropical stations. A model for simulating the oceanic component of the atmospheric O2/N2 cycles is presented consisting of the TM2 atmospheric tracer transport model [Heimann, 1995] driven at the lower boundary by O2 fluxes derived from observed O2 saturation anomalies in surface waters and by N2 fluxes derived from the net air-sea heat flux. The model is optimized to fit the observed atmospheric O2/N2 cycles by adjusting the air-sea gas-exchange velocity, which relates O2 anomaly to O2 flux. The optimum fit corresponds to spatially and temporally averaged exchange velocities of 24¿6 cm/hr for the oceans north of 31 ¿N and 29¿12 cm/hr for the oceans south of 31 ¿S. These velocities agree to within the uncertainties with the gas-exchange velocities expected from the Wanninkhof [1992] formulation of the air-sea gas-exchange velocity combined with European Centre for Medium-Range Weather Forecasts winds [Gibson et al., 1997] but are larger than the exchange velocities expected from the Liss and Merlivat [1986] relation using the same winds. The results imply that the gas-exchange velocity for O2, like that of CO2, may be enhanced in the open ocean by processes that were not systematically accounted for in the experiments used to derive the Liss and Merlivat relation. ¿ 1998 American Geophysical Union