We present new apparent wandering (APW) paths for the latest Carboniferous to Middle Jurassic of Europe-northern Asia, North America, and Gondwana. The paths have been constructed by ordering the paleopoles into a time sequence and calculating running averages. Globally, this statistical analysis cannot be carried back before the latest Carboniferous, because the paleomagnetic record becomes inadequate. The polar paths have all been rotated along with the land masses from which they have been observed into the classical configuration of Wegener (Pangea A) as quantified by later workers. The mean paleopoles agree in the Early Jurassic but disagree in the latest Carboniferous, Parmian, and Early Triassic. This result is insensitive to reasonable choices made regarding the arrangement of the continents within Gondwana; it is true either for results from Gondwana as a whole or from Africa and South America alone. This disagreement in APW paths suggests (but does not prove) the Pangea A lasted for only a short interval of time in the Early to Middle Jurassic (190--170 m.y.) and perhaps the latest Triassic (200 m.y.). Alternatively, we may say either that the paleomagnetic record is biased by sampling inadequacies or by systematically incorrect age assignments, or that the earth's field has been non-dipolar. We regard these explanations as less likely, although they cannot be discounted. The polar paths can be reconciled by rotating Gondwana anticlockwise by about 35 ¿ relative to Laurasia, placing Africa beneath Europe, and South America beneath eastern North America. This configuration is called Pangea B. Pangea B could be a valid reconstruction for the latest Carboniferous and Early Permian (approximately 290--260 m.y.). Pangea B provides a stastically satisfactory resolution of the paleomagnetic problem. It also provides a symmetrical arrangement of the continents bordering the Appalachian-Hercynian fold belt for the Late Paleozoic (which Pangea A does not), it provides a satisfactory framework for the early evolution of the Gulf of Mexico, and it provides an explanation of the absence of geological evidence for a Tethys ocean in the Paleozoic. The transition between B and A cannot be accurately reconstructed from the paleomagnetic data, because the resolution of the record is as yet inadequate, but it presumably occurred by a 3500-km minimum dextral megashear Laurasia and Gondwana during the late Permian and Triassic (approximately 250--200 m.y.), without the formation of ocean between them, and was accompanied by lesser motions between North America and Europe. The sense of megashear is the same as that invoked by Arthaud and Matte (1977) on geological grounds, but the magnitude and timing are different. Our interpretation favors the idea that continental drift is a continuous process, not confined to Jurassic and later times, but extending backwards into the Triassic and Permian. We suggest that Pangea was not an immobile configuration as envisaged by Wegener and most subsequent workers but that it evolved more or less continuously.