A KNOWLEDGE of past changes in the mean elevations of large continental areas is important for understanding both dynamic processes in the Earth’s mantle(1) and the evolution of the global climate(2). But virtually all methods for determining palaeoelevations are problematic(3), in part because changes in either elevation or climate can give rise to the same observed geological phenomena(4). Ideally, palaeoelevations would be inferred directly from estimates of palaeopressure, and it has recently been shown(5) that vesicles in basaltic lava flows preserve a record of atmospheric pressure at the altitude of emplacement; however, the elevations thus obtained have large errors (similar to 1.4 km), and the method can at present be applied only to lavas that have had a simple emplacement history(5). Palaeobotanical methods for estimating palaeoelevation have received much attention(6-12), but they rely on empirical temperature-elevation relationships, the uncertainties in which are difficult to evaluate for past climates. Here we describe an alternative palaeobotanical approach, based on energy conservation in the atmosphere, in which fossil leaf assemblages are used to infer enthalpy, rather than temperature. This approach is relatively insensitive to palaeoclimate, with an expected error in palaeoelevation of similar to 700 m.