In forming any theory of the cause of terrestrial magnetism, the first question that arises is whether we are to consider the near coincidence of the geographical with the magnetic axis as accidental or significant. The secular variation argues for the second alternative, and scientific opinion has always favoured the view that there is a definite reason for the close approach of the magnetic to the geographical pole. The view that iron is responsible for the observed magnetic field has generally been put aside, because iron loses its magnetisation at temperatures lower than those which all are agreed must hold at moderate depths below the surface. But the objection raised on this ground disregards the possibility that the critical temperature of iron may be raised by pressure. It may be shown that if that temperature depends on molecular distance, and the ratio of compressibility to thermal dilatation is assumed to be independent of pressure, the rate of increase of temperature with the depth is less than the rate of increase of the critical temperature, so that iron would retain its magnetic properties. Experiments on the effects of pressure on the critical temperature of iron have been in progress during the last four years, but have not hitherto led to a decisive result. For the present we must, therefore, keep our minds open to the possibility that the iron contained in the earth is magnetisable. The alternative opinion that electric currents circulating inside the earth are the cause of electric currents is next examined. Though this view seems to be favoured by scientific opinion, it has to overcome formidable difficulties before it can be accepted. Are the electric currents permanent, or are they only survivals of a former state of things which is dying out? If permanent, we must account for the electromotive forces, and we know of no such forces which could act in the manner required. If the electric currents are survivals, their intensity when the earth crust first formed must have been at least 1012 times as strong as it is now, and the difficulty of accounting for their original production would be correspondingly increased. In the opinion of the author, the difficulties which stand in the way of basing terrestrial magnetism on electric currents inside the earth are insurmountable. The question whether the rotation of the earth may be responsible for its magnetic field is next examined, and different possibilities are considered. If the rotation of a sphere independently of the chemical nature of the substance were to produce a magnetic field, it is shown by the theory of dimensions that the field at the surface of the sphere would be proportional to the angular velocity and independent of its size. The effect of rotating bodies set into rotation in our laboratories would give, therefore, magnetic forces equal to those of the earth with one rotation per day, and with angular velocities of the order of one per second the effects would be so large that they could not have escaped detection. A second supposition, that a neutral molecule in its motion behaves as if it carried a charge, leads to the result that if terrestrial magnetism be due to such a cause, the effects to be expected in any laboratory experiment that we could propose are far too small to be detected. But the theory is finally negatived by the additional conditions which must be imposed to destroy the effects of translation independently of rotation. A distinction must here be made between possible magnetic effects of a rotatory body on a magnet which is at rest and on one the supports of which are fixed to the rotating body. Returning to the effects of rotation, the possibility is considered that rotation instead of directly determining magnetic intensity, determines magnetic force which may or may not cause magnetisation according to the nature of the body. This view deserves attention, because there is some theoretical foundation for it, inasmuch as molecules may be considered to behave in a manner analogous to that of a gyrostatic compass. The theory would also explain in a natural manner the secular variation by a precessional motion of a magnetic molecule. On the other hand, the theory would have to explain why the iron inside the earth becomes - by rotation - more strongly magnetised than the iron in our laboratories. There is, of course, always the possibility of some substance being subject to the effects of rotation in a much higher degree than iron. Such questions can only be settled by experiments which are now in progress. Other theories of the secular variation are considered. In conclusion, the question is raised whether the negative electron is subject to gravitation, and if so, whether in comparing the weight of a negative electron with that of the unit positive charge, we must consider the proportionality between gravitation and mass to hold. In Lorentz' theory of gravitation this is abandoned. It is pointed out how fundamental questions have a bearing on the problem of terrestrial magnetism.