Paleogeography can be reconstructed using various crust- or mantle-based reference frames that make funda-mentally different assumptions. The various reconstruction models differ significantly in continental paleo-longitude, but it has been difficult to assess which models are more valid. We suggest here aLLSVP test, where an assumed correlation between present-day large low velocity shear-wave provinces and the paleogeography of supercontinent Pangea at breakup ca. 200 million years ago can be used to assess the relative accuracy of pub-lished reconstructions. Closest correlations between continental paleolongitude and the African LLSVP are ach-ieved with mantle-based reference frames (moving hotspots and true polar wander), whereas shallower crust-based reference frames are shown to be invalid. The relative success of mantle-based frames, and thus the importance of the depth of reference frame, supports the notion that mantle convection is largely vertical compared to the horizontal plate motion of tectonics.