Saddle points between neighboring deep ocean basins are the sites of unidirectional flow from one basin to the next, depending on the source of bottom water. Flow in these sites appears to be topographically controlled so the interface between the bottom water and the water above adjusts itself to permit bottom water flow from the basin that contains a source of bottom water into the next. Examples in the Atlantic include flow in the Romanche Fracture Zone, the Vema Channel, the Ceara Abyssal Plain, the Anegada-Jungfern passage, and the Discovery Gap, but there are many more. Theoretical predictions of volume flux using a method that requires only conductivity-temperature-depth data archives and detailed knowledge of bathymetry near the saddle point are compared with volume flux estimates using current meters and/or geostrophic estimates for seven cases. The ratio of prediction to volume flux estimate ranges from 1.0 to 2.7. Some ocean straits that separate adjacent seas are also found to critically control bidirectional flows between basins. Theory of the influence of rotation on such critical flows is reviewed. Predictions of volume flux in eight cases are compared with ocean estimates of volume flux from traditional methods. ¿ 1998 American Geophysical Union