Paleo-temperature data indicates that the Earth’s mantle did not cool at a constant rate over geologic time. The data are consistent with slow cooling from 3.8 to ~2.5 billion years ago with a transition to more rapid cooling extending to the present. This has been suggested to indicate a change in global tectonics from a single plate to a plate tectonic mode. However, a tectonic change may not be necessary. Multi-stage cooling can result from deep water cycling coupled to thermal mantle convection. Melting and volcanism removes water from the mantle (degassing). Dehydration tends to stiffens the mantle, which slows convective vigor and plate velocities causing mantle heating. An increase in temperature tends to lower mantle viscosity which acts to increase plate velocities provided that mantle viscosity offers resistance to plate motion. If these two tendencies are in balance, then mantle cooling can be weak. If the balance is broken, by a switch to net mantle rehydration, then the mantle can cool more rapidly. We use coupled water cycling and mantle convection models to test the viability of this hypothesis. Within model and data uncertainty, the hypothesis that deep water cycling can lead to a multi-stage Earth cooling is consistent with present day and paleo data constraints on mantle cooling. It is also consistent with constraints that indicate a change from net mantle dehydration to rehydration over the Earth’s geologic evolution. Probability distributions, for successful models, indicate that plate and plate margin strength play a relatively minor role for resisting plate motions relative to the resistance from interior mantle viscosity.