The data obtained from numerous palaeomagnetic measurements made during the past decade have shown that while the geologically younger rock formations are magnetized in directions close to that of the present earth's magnetic field, the remanent magnetic polarizations of older rocks depart markedly from this pattern. These observations are widely held by many workers to suggest that the main continental land masses have undergone movements relative to one another during the past. The present paper gives an account of a new analysis of the available data, making a minimum number of theoretical assumptions about the ways in which the rocks became magnetized and about the origin of the geomagnetic field. The results of this analysis strongly support the supposition that the observed wide divergence between the directions of the remanent magnetic vectors of older rocks and that of the present earth's field is systematic, and not a result of purely random processes occurring throughout geological time. The most reasonable explanations of the phenomenon appear to be that (a) the directions of magnetization of the earlier rocks have been changed by some widespread physical or geological processes since the time of their formation, (b) the earth's magnetic field has had strong multipolar components in past geological ages, (c) a relative drift of the continents across the earth's mantle has occurred. Of these hypotheses, (c) appears to be the most plausible. On the tentative assumption that the rock magnetic results can be explained by continental drift, it is possible to estimate the ancient latitude and the orientation relative to the earth's rotational axis, of each continent, although by palaeomagnetic measurements alone changes in relative longitude cannot be revealed.