An accurate model of relative plate motions in Gondwana breakup is based on visual fitting of seafloor isochrons and fracture zones (FZ) from the Riiser-Larsen Sea and Mozambique Basin. Used predictively, the model precisely locates kinematic markers in the West Somali Basin, which allows the conclusion that the spreading centres in the West Somali and Mozambique basins and the Riiser-Larsen Sea formed parts of the boundary between the same two plates. The locations of FZ and less well-defined isochrons from neighbouring regions are also consistent with their formation on other lengths of this same boundary and with its relocation from the West Somali Basin and northern Natal Valley to the West Enderby Basin and Lazarev Sea during chron M10n. Small independently moving plates thus played no role in the breakup of this core part of Gondwana. In an inversion procedure, the data from these areas yield more precise finite rotations that describe the history of the two plates' separation. Breakup is most simply interpreted to have occurred in coincidence with Karoo volcanism, and a reconstruction based on the rotations shows the Lebombo and Mateke-Sabi monoclines and the Mozambique and Astrid ridges as two sets of conjugate volcanic margins. Madagascar's pre-drift position can be used as a constraint to reassess the positions of India and Sri Lanka in the supercontinent.