The evolution of Keith Runcorn's ideas from the static, elastic Earth of Jeffreys to a dynamic, convecting planet are presented based on discussions as his colleague over 25 years from 1963 to his retirement in 1988. Keith reached the concept of a dynamic planet by way of polar wander and continental drift using palaeomagnetism with great zeal. It took some time for Keith to convince himself of the reality of continental drift. Once convinced, he became an evangelist converting others and enthusiastically pursuing possible mechanisms for explaining it, homing in on mantle convection. To establish the nature and history of mantle convection his interests ranged from the world rift system and satellite gravity anomalies to the radiometric age peaks. A great step forward occurred when we realised that a region of uprising convection was not necessary under all the rifts as had been commonly advocated in the 1960s and that the mid-Atlantic, African and Indian Ocean rifts could be equally well explained by one large region of upwelling mantle convection. It was also realised that the plate convergence zones (island arcs, trenches and deep focus earthquakes) were much better correlated with the satellite gravity anomalies and it was much easier to locate the possible regions of downwelling mantle convection. Now, seismic tomography helps to establish the nature of mantle convection and it appears that relations among the Earth's surface features, the geoid anomalies and peturbations of mantle seismic velocities are near to being established. In the next few years a far better and accurate picture of the geometry of mantle convection so enthusiastically advocated by Keith Runcorn is likely to be seen.