Geomagnetic polarity reversal remains one of Nature's most enigmatic phenomena. Dynamo theory admits solutions in pairs with reversed magnetic fields B and -B, but detailed calculations are required to understand how the field can change sign. Theory also admits separate solutions with different symmetry across the equatorial plane, the symmetric (ES) ''quadrupole'' and the antisymmetric (EA) ''dipole'' solutions, which may be important in the reversal processes and which offer a simple framework for interpreting small paleomagnetic data sets. Ordinary secular variation leads to very large changes in the magnetic field over several centuries and could easily develop into full reversal; the theory of secular variation, which is relatively well developed, may therefore help in understanding reversals. Other clues to geomagnetic reversals come from the Sun, whose magnetic field reverses every 11 years. Paleomagnetic data show the Earth's magnetic field reverses every million years or so, with each transition taking about a thousand years, during which the intensity may fall by as much as 1 order of magnitude. Reversal frequency undergoes a modulation on the long timescale (107 years) of mantle convection, and there have been two long intervals in the past with no reversals. Such behavior is typical of a highly nonlinear dynamical system, but the very long timescale of changes in reversal frequency, and its close proximity to the overturn time of mantle convection, suggests some control of the dynamo by the mantle.

Short-term phenomena, such as change in the length of day and secular variation, have been studied extensively for evidence of core-mantle interactions, and we may draw on this body of evidence in order to understand long-term effects. Three physical mechanisms have been proposed: topographic, electromagnetic, and thermal, with the last two being most significant for long-term effects. Symmetries allow the dynamo to generate an EA field, with the major component a dipole, but lateral variations on the core-mantle boundary may lead to magnetic fields with no symmetry, reflecting the structure of the boundary anomalies. Changes in reversal frequency on the mantle convection timescale could arise either from changes in total heat flux from the core to the mantle or from instabilities associated with lateral variations at the core-mantle boundary. Neither mechanism is well understood, but the former involves significant changes to the Earth's overall heat budget, whereas the latter must always arise as a natural consequence of deep-mantle convection. Recentmeasurements of transition fields show pole paths that lie close to two preferred longitudes near 90¿W and 90¿E; if substantiated, the result would provide the first definitive evidence of long-term mantle control of the geomagnetic field. Further evidence suggests that the geomagnetic pole during stable polarity also lies along these two longitudes and that magnetic flux at the core surface tends to concentrate along the same longitudes, as does the present field. A simple theory is proposed relating changes in the core field to apparent transition paths measured at the Earth's surface. The model show that longitude confinement of the transition paths can occur for quite complicated core fields and that surface intensities can drop by 1 order of magnitude, on average, simply because of the reduction in length scale of the transitional field. Simple transition paths may be an indication of some organization of flux at the core surface but not of large-scale or small-amplitude core fields.