Abstract A large number of velocity models derived from a variety of seismic data and using different seismic techniques have been published for the Archaean and Proterozoic shields. Here, we focus on the structure beneath southern Africa, where velocity models derived from most regional seismic data find the thickness of the seismic lithosphere to be less than 200 km. In contrast, velocity models derived from teleseismic body-wave and long-period surface-wave data determine the seismic lithosphere to be as much as 400 km thick. We believe that this disagreement is due to the ways in which the various datasets average the velocity structure. Our analysis of regional seismograms from propagation paths largely confined to the stable region shows that the average thickness of the seismic lithosphere beneath southern Africa does not exceed 160 km. We compare the vertical S-wave travel time of our velocity model derived from regional seismic data and those models derived from teleseismic data and find no significant difference. We determine the in situ velocities and densities from nodules from beneath southern Africa using a recently derived geobarometer and geothermometer; these are in excellent agreement with the velocity found in the high-velocity lid from the analysis of the regional seismic data. The lithospheric model that best fits the nodule data has a mechanical boundary layer thickness of 156 km and a lithosphere thickness of 176 km. However, the shear-wave velocity decrease at the base of the lid does not correspond to a change in mineralogy. Recent experimental studies of the shearwave velocity in olivine as a function of temperature and period of oscillation demonstrate that this decrease can result from grain boundary relaxation at high temperatures at the period of seismic waves. This decrease in velocity occurs where the mantle temperature is closest to the melting temperature.