Gas exchange experiments have been performed in a wind-water tunnel filled with fresh water or seawater. Transfer velocities have been measured for nitrous oxide and argon in a range of wind speeds extending from 3 m/s to 14 m/s. The air-liquid interface was covered either with only normally developed wind waves or with mechanically generated waves in addition. For u9 m/s), a jump of the values of kL up to a factor of 3 is observed first for Ar and then for N2O, the ratio kL,Ar/ kLN20 being highly variable. This observation is interpreted as being the result of the onset of breaking waves which create bubbles through which mass transfer takes place. No difference for gas exchange either in fresh water or in seawater has been identified from the whole set of experiments. A theoretical model is developed to compute the contribution of bubbles to gas exchange. The transport equations, derived for a single bubble, take into account the diffusive mass flux exchanged between the bubble and the liquid, the hydrodynamic pressure, and the presence of two gaseous components inside the bubble. The equations used to represent the dynamics take care of the orbital motion of water associated with waves. A relation to link sources and spectra of bubble populations is derived. Numerical values for spectral parameters are taken from the literature. It is found that the results of experiments run at 9 m/s and 10 m/s are well explained for N2O and for Ar. A unique distribution of bubbles explains the excess of gas transfer associated with the presence of breaking waves. In a further step the theoretical treatment is extended to compute the transfer velocity kB for gas exchange through bubbles as a function of the concentration gradient in atmosphere and water for three gases, helium, argon, and nitrous oxide. It is shown that in wind tunnel conditions, kB is independent of the concentration gradient. This does not hold at the atmosphere-ocean interface. The transfer velocity kB is highly dependent on the gas concentrations in air and water when they are close to solubility equilibrium. It is demonstrated that equilibrium must occur at the ocean surface for supersaturation conditions. The higher saturation anomalies and the larger transfer velocities are computed for the less soluble gas. This conclusion is in good agreement with observations made at sea.