The cross-equatorial cell (CEC) in the Indian Ocean is a shallow (z > -500 m) meridional overturning circulation, consisting of northward flow of southern-hemisphere thermocline water, upwelling in the northern hemisphere, and a return flow of surface water. In this study, several types of ocean models, varying in complexity from a 1.5-layer analytic model to a state-of-the-art general circulation model (GCM), are used to investigate CEC structure and its dynamics. Pathways are illustrated by tracking model drifters from the northern-hemisphere upwelling regions, both forwards in time to follow the surface pathways and backwards in time to follow the subsurface flows. In the subsurface branch, cross-equatorial flow occurs via a western-boundary current, where strong horizontal mixing can alter the sign of its potential vorticity. In contrast, surface pathways cross the equator in the interior ocean at almost all longitudes. Sources of CEC water are flow into the basin in the southeastern ocean, subtropical subduction, and the Indonesian Throughflow. The models differ in which source is most prominent, a consequence of their different parameterizations of vertical-mixing processes and basin boundary conditions. The surface, cross-equatorial branch is driven by the annual-mean component of the zonal wind stress tau_x. It is predominantly antisymmetric about the equator with westerlies (easterlies) north (south) of the equator, and so is roughly proportional to latitude y. The resulting negative wind curl drives a southward Sverdrup flow across the equator. For a tau_x that is exactly proportional to y, the Ekman pumping velocity is identically zero; as a consequence, no geostrophic currents are generated by the wind, and the Sverdrup transport is equal to the Ekman drift. In GCM solutions, the southward, cross-equatorial flow occurs just below the surface (z< -100 m), typically beneath a northward surface current, so that there is a shallow, cross-equatorial "roll". This feature is the direct (local) response to southerly cross-equatorial winds, and its basic dynamics are linear. Because it is so shallow, the roll has little impact on the CEC heat transport. It does, however, influence where surface drifters can cross the equator; they cross near the eastern boundary where the roll is weak.