coefficient of inertia of the aether, and on the dimensions of its nucleus, where cis the velocity of radiation: the increase of its gravitational energy would presumably in like manner depend on , where c' is the velocity of propagation of gravitation and is enormously greater than c. On neither ground does it appear likely that mass is to any considerable degree an attribute of gravitation.
117. The main additional result derived from this second-order discussion is that if we assume all molecular forces to be electric forces, motion of a material system through the aether alters its dimensions in a minute but definite manner. A scrutiny, on all sides, of the basis of this inference is of course desirable. As a preliminary it is to be noticed that the molecular forces on the action of which it depends are extremely great in comparison with any distributions of force arising from finite currents or electrifications produced in the system as a whole. In the comparison between the two identical systems, one at rest the other in motion, of the analogy above developed, their electrons occupy corresponding positions in their spaces at all times: thus at first sight it is only systems in which the electrons are absolutely at rest that can be thus compared. But even in the case of dielectric bodies at rest, though the molecules are fixed the electrons are revolving in the molecules: yet that does not sensibly affect the application of the correspondence. For the only difference thereby introduced in it is that the phases of the orbital motions of those molecules of the moving material system that are situated further in advance, in the direction of the movement of the system, are slightly accelerated in comparison with the corresponding phases in the fixed system. Now the permanent or secular relations between molecules, supposed far enough apart not to interfere in a structural manner with each other so as to form compound molecules, are independent of these relative phases: to obtain them we in fact replace each molecule by its steady secular equivalent in the Gaussian sense, as has to be
