Abstract Where crust shortens horizontally and thickens, isostatic compensation of the thick crust requires that a mountain range or high plateau form. Yet, mantle lithosphere must also shorten horizontally, and if it thickens, it should provide a weight that maintains the range at a lower elevation than it would be if isostatic compensation resulted only from thickened crust. Being cold and dense, mantle lithosphere (or at least its lower part) should be convectively unstable, and hence should sink rapidly into the underying asthenosphere as downwelling sheets or plumes. Indeed, tomographic images of mountain ranges do not, in general, reveal a deep root of thickened mantle lithosphere, suggesting that some process removed it. Removal of mantle lithosphere should make the remaining lithosphere more buoyant, and when (if) mantle lithosphere is removed, the earth’s surface should rise. I exploit seismological images of upper mantle structure, numerical experiments of Rayleigh-Taylor instability (arising when a dense layer overlies a less dense one), and inferences of paleo-altimetry, among other geologic observations to argue that mantle lithosphere commonly is removed when mountain ranges are built, with unexpected consequences in some cases. I will use geological and geophysical data from Tibet, the Andes, the Tien Shan, etc. to illustrate apparent consequences of removal of mantle lithosphere. (Anyhow, if all of this interpretation proves to be nonsense, I hope to present interesting facts and well executed numerical experiments that others will find useful.)