The object of this paper is to amend, in an important particular, the theory of ferromagnetic induction put forward by me more than 30 years ago, and to describe a new model. That theory was itself a modification of the earlier theory of Weber. To Weber is due the fundamental notion that a substance contains minute particles, each of which acts as a magnet, and that in the process of magnetising a ferromagnetic substance these are turned into more or less complete alignment. The ultimate magnetic particles use to be called “molecular magnets”: we now recognise them as attributes of the atom, not of the molecule, and (in all probability) they derive their magnetic moment from the circulation of electricity in electron orbits or in ring electrons. What turns is not the molecule nor the atom, but something within the atom. The characteristics which distinguish ferromagnetic substances from other paramagnetics are: (1) the much larger amount of magnetism they can acquire under the action of an impressed field; (2) the fact that the acquired magnetism tends towards a saturation limit when the field is progressively increased; (3) the fact that the acquired magnetism shows hysteresis with respect to variations of the field, except in certain small initial changes. Weber’s theory explained (1) and (2). My modification of it explained, in addition, (3) as an effect of the irreversible action which occurs when the equilibrium of a magnetic element becomes unstable through change in the externally impressed magnetic force, and it swings over, with dissipation of energy, into a new position of stability. The stability in both positions is sufficiently explained by magnetic forces only. In breaking away from one stable position it is deflected at first in a quasi-elastic (reversible) manner until the external force reaches a certain value at which the equilibrium is upset. The essence of hysteresis is the turning from one position of stability to another, through a region of instability. If the conditions are such that there is no unstable phase in the turning, then there is no dissipation of energy, and consequently no hysteresis. This occurs in very feeble magnetisation, when the deflections are reversible; it also occurs if the piece be caused to rotate in a field of great strength. J. Swinburne pointed out that, as a consequence of my theory, hysteresis should vanish when a cylinder of ferromagnetic metal is rotated in a very strong field, and this curious result was confirmed experimentally by F. G. Baily.