The paths of the virtual geomagnetic poles (VGPs), which are north magnetic poles conventionally calculated from the axial dipole formula, describing the transitional fields during Cenozoic polarity reversals, are presented. They provide evidence that, at least on a number of occasions, the transitional field recorded by sediments is predominantly dipolar. The VGP paths are mostly confined to two longitudinal sectors which are 180° apart, the first path passing over the Americas and the second over Australia and East Asia. The mantle ‘guides’ the rotating dipole along these preferred paths. However, a much smaller number of records show that for certain reversals, a path midway between the two preferred paths (i.e. a path across the Atlantic and the mid–Pacific) is favoured. We regard this as an important key to the mechanism.
Two hypotheses to explain these observations have been put forward. The first hypothesis notes that the preferred paths correlate roughly with two seismic tomography anomalies in the lower mantle that have been interpreted as lateral temperature anomalies associated with mantle convection. It is then suggested that the diffusion of these thermal anomalies somehow controls the core dynamo and its mode of reversal. Others have already suggested that undulations on the core mantle boundary could exert a controlling influence on core motions and therefore the behaviour of the dynamo. The alternate view simply accepts the concept that the dynamo reverses by dipole rotation and that the D″ layer is hemispherically asymmetric chemically: its quasi–metallic conductivity under the Pacific hemisphere had long ago been proposed to explain the low geomagnetic secular variation and non–dipole field observed there. The torques, resulting from the electromagnetic induction in the D″ layer by the reversing dipole rotate the core about the Earth's axis until the VGP path lies along either of the bounding meridians of the shell. The data presented on transitional polar paths should provide a test between the two theories.