Over the past twenty years or so, a good qualitative description of the ocean circulation has been built up through the use of two-dimensional wind-driven models. These correctly predict the position and sense ot circulation of the major oceanic gyres, and explain the strong western boundary currents. However, the only direct measurements available for quantitative comparison with theory are those of the transport of the Gull Stream. Observed values are considerably larger than those predicted by linear theory, but can be explained by nonlinear models. . Two-dimensional models have their shortcomings, however, because they take no account ot vertical structure. They cannot properly model effects of nonlinearity, bottom topography, or of motions driven by thermohaline effects. They overlook important aspects of the circulation such as the equatorial undercurrent, and upwelling. Three-dimensional numerical models are designed to include these effects, and successfully reproduce, in a qualitative fashion, the major features of the ocean circulation. The models may be used for ‘experiments’ to find possible effects of changing parameters (such as the width and depth of Drake Passage), and to investigate the relative importance of different regions. With presentday computers, however, there are resolution and other problems which limit the degree to which the real ocean’s parameters may be matched.