To say that A is “necessarily” followed by B is thus to say no more than that there is some general rule, exemplified in a very large number of observed instances, and falsified in none, according to which events such as A are followed by events such as B. We must not have any notion of “compulsion”, as if the cause _forced_ the effect to happen. A good test for the imagination in this respect is the reversibility of causal laws. We can just as often infer backwards as forwards. When you get a letter, you are justified in inferring that somebody wrote it, but you do not feel that your receiving it _compelled_ the sender to write it. The notion of compulsion is just as little applicable to effects as to causes. To say that causes compel effects is as misleading as to say that effects compel causes. Compulsion is anthropomorphic: a man is compelled to do something when he wishes to do the opposite, but except where human or animal wishes come in the notion of compulsion is inapplicable. Science is concerned merely with what happens, not with what _must_ happen. When we look for invariable rules of sequence in nature, we find that they are not such as common sense sets up. Common sense says: thunder follows lightning, waves at sea follow wind, and so on. Rules of this sort are indispensable in practical life, but in science they are all only approximate. If there is any finite interval of time, however short, between the cause and the effect, something may happen to prevent the effect from occurring. Scientific laws can only be expressed in differential equations. This means that, although you cannot tell what may happen after a finite time, you can say that, if you make the time shorter and shorter, what will happen will be more and more nearly according to such-and-such a rule. To take a very simple case: I am now in this room; you cannot tell where I shall be in another second, because a bomb may explode and blow me sky-high, but if you take any two small fragments of my body which are now very close together, you can be sure that, after _some_ very short finite time, they will still be very close together. If a second is not short enough, you must take a shorter time; you cannot tell in advance how short a time you may have to take, but you may feel fairly certain that there is a short enough time.