The determination of palaeo-elevation has emerged in the past 15 years as an important tool for constraining physical processes that govern the formation of mountain belts. Rowley and Currie report palaeo-elevations for the Lunpola basin within the Tibetan plateau and claim that these elevations are incompatible with 'mantle-thickening models' for mountain formation. We show here that their data do not support this conclusion and, indeed, are consistent with its opposite. The Tibetan plateau could have risen by a kilometre or more as its dense lower lithosphere sank into the underlying mantle.