take the course which is easiest at each moment, according to the character of space-time in the particular region where they are; this course is called a geodesic.
Now it will be observed that I have been speaking freely of bodies and motion, although I said that bodies were merely certain strings of events. That being so, it is of course necessary to say what strings of events constitute bodies, since not all continuous strings of events do so, nor even all geodesics. Until we have defined the sort of thing that makes a body, we cannot legitimately speak of motion, since this involves the presence of one body on different occasions. We must therefore set to work to define what we mean by the persistence of a body, and how a string of events constituting a body differs from one which does not. This topic will occupy the next chapter.
But it may be useful, as a preliminary, to teach our imagination to work in accordance with the new ideas. We must give up what Whitehead admirably calls the "pushiness" of matter. We naturally think of an atom as being like a billiard-ball; we should do better to think of it as like a ghost, which has no "pushiness" and yet can make you fly. We have to change our notions both of substance and of cause. To say that an atom persists is like saying that a tune persists. If a tune takes five minutes to play, we do not conceive of it as a single thing which exists throughout that time, but as a series of notes, so related as to form a unity. In the case of the tune, the unity is æsthetic; in the case of the atom, it is causal. But when I say "causal" I do not mean exactly what the word naturally conveys. There must be no idea of compulsion or "force", neither the force of contact which we imagine we see be tween billiard balls nor the action at a distance which was formerly supposed to constitute gravitation. There is merely an observed law of succession from next to next. An event at one moment is succeeded by an event at a neighbouring moment, which, to the first order of small quantities, can be calculated from the earlier event. This enables us to construct a string of events, each, approxi-
