The thermal boundary layer beneath continental cratons extends into the Earth's mantle to depths of at least 200 km. It has been proposed that chemical depletion of the lithospheric mantle during partial melting offsets the effect of increased density from conductive cooling, resulting in neutral buoyancy with respect to the underlying asthenosphere. Mineral compositions of garnet peridotite xenoliths in the Kaapvaal craton give equilibration temperatures and pressures that define a continental conductive geotherm intersecting a mantle adiabat with a potential temperature of 1300¿C at ~60 kbar. We calculated normative densities for a low-temperature garnet and spinel peridotite xenolith suite using Mg#. At their temperatures and pressures of equilibration, all the low-temperature peridotites are positively buoyant with respect to the convecting mantle, which is inconsistent with the hypothesis of a neutrally buoyant thermal boundary layer. To account for the possibility that pressure, temperature, and mineral proportions may have varied over time, equilibrium solidus mineral assemblages for the low-temperature xenoliths over a range of pressures and temperatures were generated by free energy minimization using the program Perplex (http://www.perplex.ethz.ch). The equilibrium solidus densities for xenolith compositions along a 40 mW/m2 conductive geotherm were compared with the density of pyrolite along a mantle adiabat with a potential temperature of 1300¿C. These density calculations show that most of the xenoliths are positively buoyant with respect to asthenospheric mantle (pyrolite) at their temperatures and pressures of equilibration, confirming the results from the normative density calculations. Also, at the onset of the accretion of the Kaapvaal craton, when the thermal boundary layer was thinner and hotter than today, these Kaapvaal peridotites would have been positively buoyant with respect to the convecting mantle at shallower pressures. Therefore we propose that peridotite at the base of the Kaapvaal plate, in the mantle thermal boundary layer, was and is positively buoyant. Combined with evidence from the geoid and geomorphology, which suggests that cratons are isostatically compensated and neutrally buoyant, our results imply that there must be dense layers within cratonic crust or upper mantle that offset the positive buoyancy of depleted cratonic mantle peridotites.