A method of conducting computer experiments on ionic recombination in an ambient gas is described. Account may be taken of binary recombination, and its contribution to the total recombination coefficient may be separated. Recombination in molecular oxygen is treated as an example. It is found that the recombination coefficient for ambient gas (or termolecular) recombination by itself is as predicted by the quasi-equilibrium statistical method in the low density region and as predicted by the Langevin-Harper formula in the high density region. The slightly modified Natanson formula which was introduced to bridge the very wide gap between these two regions is not successful, the pattern of its failure being that it underestimates the influence which the mobilities of the ions have on the recombination coefficient. An unsuspected characteristic of binary recombination revealed by the investigation is that its rate coefficient depends on the density of the ambient gas. Before attempting to extrapolate laboratory data on the total ionic recombination coefficient to zero ambient gas density it is essential to take cognisance of the steep rise of the binary recombination coefficient in the very low density region.