The origin of mantle hotspots is a controversial topic. Only seven ('primary') out of 49 hotspots meet criteria aimed at detecting a very deep origin (three in the Pacific, four in the Indo-Atlantic hemisphere). In each hemisphere these move slowly, whereas there has been up to 50 mm/a motion between the two hemispheres prior to 50 Ma ago. This correlates with latitudinal shifts in the Hawaiian and Reunion hotspots, and with a change in true polar wander. We propose that hotspots may come from distinct mantle boundary layers, and that the primary ones trace shifts in quadrupolar convection in the lower mantle.