The apparent polar wander paths (= APWPs) for the northern and southern continents have been redetermined using revised time-scales and data. When plotted on the conventional Pangea, Pangea A, the paths are distinct but essentially linear and parallel from Late Carboniferous to Late Permian time. Both paths then undergo a sharp change in direction and converge so as to overlap in Early to Middle Jurassic time. If the Permo-Triassic magnetic field was an axial dipole, then paleomagnetic data show that Pangea was mobile during most, and possibly all, of the mid-Triassic to mid-Jurassic interval. Because of the uncertainties in pole ages and positions, and errors in measurements, the new APWPs are consistent with a range of Pangeas. In particular, the data are compatible with two paleomagnetic Pangeas: A2 and a new class of Pangea here named the D Pangeas. These resemble Morel and Irving's (1981) Pangea B and the C Pangeas of Smith et al. (1981): unlike the A2 Pangeas, the B, C and D Pangeas all require large (> 2000 km) relative motion between the northern and southern continents. Geological data require the transition from an earlier Pangea to Pangea A to take place along a transform zone, here named the Pangea Transform Zone, or PTZ. The PTZ is considered to be technically analogous to the Great Glen Fault of Scotland. Four types of possible PTZ are examined: passing through either the western or eastern end of the Brunswick Magnetic Anomaly of eastern North America, or north and south of Iberia. The geological data suggest that those Pangeas that require large displacements on PTZs that pass east of the Brunswick Magnetic Anomaly are improbable. B and C Pangeas are snapshots giving precise paleomagnetic reconstructions. D Pangeas fit long stretches of the APWPs as well as possible. A2 Pangeas fit the APWPs moderately well but also require the geological displacement to be smaller. The differences in displacement between the D and A2 Pangeas reflect geometrical relationships between Euler poles, PTZ curvature and the APWPs. The general geological evidence suggests that D Pangeas are less probable than A2 Pangeas but that they cannot be rejected. Transition from A2 to A is most likely to have occurred via a sequence of rotations about different poles rather than a single rotation. In detail, the conventional Pangea requires modification. Africa must be broken into at least two fragments along the Benue Trough and its northern continuation. Northwest Africa then fits more tightly into northern South America, which modifies the APWPs and reduces the gap between them. By including the effects of continental extension and using detailed geological data the gap can be reduced to one in which the A63 circles touch or overlap. Non-dipole fields or an expanding Earth need not be invoked to explain the mismatch of the APWPs on the conventional Pangea. The finite rotation required for the older Pangea to Pangea A is very close to Van der Voo and French's (1974) single rotation for A2 to A but is actually the sum of three smaller rotations that reflect the modification to Pangea A, continental extension and the postulated PTZ displacement. Of the possible PTZs, a PTZ parallel to the West African continental magnetic anomaly with a displacement of about 170 km gives a better fit of this anomaly to the East Coast magnetic anomaly of North America. Further refinements to Pangea A2 will depend on a better knowledge of stretching factors on the passive continental margins of the modified Pangea A and incorporation of rotations due to major faults such as the South Atlas Fault in Morocco.