WHEN I am said to have done something of my own free will it is implied that I could have acted otherwise; and it is only when it is believed that I could have acted otherwise that I am held to be morally responsible for what I have done. For a man is not thought to be morally responsible for an action that it was not in his power to avoid. But if human behaviour is entirely governed by causal laws, it is not clear how any action that is done could ever have been avoided. It may be said of the agent that he would have acted otherwise if the causes of his action had been different, but they being what they were, it seems to follow that he was bound to act as he did. Now it is commonly assumed both that men are capable of acting freely, in the sense that is required to make them morally responsible, and that human behaviour is entirely governed by causal laws: and it is the apparent conflict between these two assumptions that gives rise to the philosophical problem of the freedom of the will.

Confronted with this problem, many people will be inclined to agree with Dr. Johnson: 'Sir, we know our will is free, and there's an end on't'. But, while this does very well for those who accept Dr. Johnson's premiss, it would hardly convince anyone who denied the freedom of the will. Certainly, if we do know that our wills are free, it follows that they are so. But the logical reply to this might be that since our wills are not free, it follows that no one can know that they are: so that if anyone claims, like Dr. Johnson, to know that they are, he must be mistaken. What is evident, indeed, is that people often believe themselves to be acting freely; and it is to this 'feeling' of freedom that some philosophers appeal when they wish, in the supposed interests of morality, to prove that not all human action is causally determined. But if these philosophers are right in their assumption that a man cannot be acting freely if his action is causally determined, then the fact that someone feels free to do, or not to do, a certain action does not prove that he really is so. It may prove that the agent does not himself know what it is that makes him act in one way rather than another: but from the fact that a man is unaware of the causes of his action, it does not follow that no such causes exist.

So much may be allowed to the determinist; but his belief that all human actions are subservient to causal laws still remains to be justified. If, indeed, it is necessary that every event should have a cause, then the rule must apply to human behaviour as much as to anything else. But why should it be supposed that every event must have a cause? The contrary is not unthinkable. Nor is the law of universal causation a necessary presupposition of scientific thought. The scientist may try to discover causal laws, and in many cases he succeeds; but sometimes he has to be content with statistical laws, and sometimes he comes upon events which, in the present state of his knowledge, he is not able to subsume under any law at all. In the case of these events he assumes that if he knew more he would be able to discover some law, whether causal or statistical, which would enable him to account for them. And this assumption cannot be disproved. For however far he may have carried his investigation, it is always open to him to carry it further; and it is always conceivable that if he carried it further he would discover the connection which had hitherto escaped him. Nevertheless, it is also conceivable that the events with which he is concerned are not systematically connected with any others: so that the reason why he does not discover the sort of laws that he requires is simply that they do not obtain.

Now in the case of human conduct the search for explanations has not in fact been altogether fruitless. Certain scientific laws have been established; and with the help of these laws we do make a number of successful predictions about the ways in which different people will behave. But these predictions do not always cover every detail. We may be able to predict that in certain circumstances a particular man will be angry, without being able to prescribe the precise form that the expression of his anger will take. We may be reasonably sure that he will shout, but not sure how loud his shout will be, or exactly what words he will use. And it is only a small proportion of human actions that we are able to forecast even so precisely as this. But that, it may be said, is because we have not carried our investigations very far. The science of psychology is still in its infancy and, as it is developed, not only will more human actions be explained, but the explanations will go into greater detail. The ideal of complete explanation may never in fact be attained: but it is theoretically attainable. Well, this may be so: and certainly it is impossible to show a priori that it is not so: but equally it cannot be shown that it is. This will not, however, discourage the scientist who, in the field of human behaviour, as elsewhere, will continue to formulate theories and test them by the facts. And in this he is justified. For since he has no reason a priori to admit that there is a limit to what he can discover, the fact that he also cannot be sure that there is no limit does not make it unreasonable for him to devise theories, nor, having devised them, to try constantly to improve them.

But now suppose it to be claimed that, so far as men's actions are concerned, there is a limit: and that this limit is set by the fact of human freedom. An obvious objection is that in many cases in which a person feels himself to be free to do, or not to do, a certain action, we are even now able to explain, in causal terms, why it is that he acts as he does. But it might be argued that even if men are sometimes mistaken in believing that they act freely, it does not follow that they are always so mistaken. For it is not always the case that when a man believes that he has acted freely we are in fact able to account for his action in causal terms. A determinist would say that we should be able to account for it if we had more knowledge of the circumstances, and had been able to discover the appropriate natural laws. But until those discoveries have been made, this remains only a pious hope. And may it not be true that, in some cases at least, the reason why we can give no causal explanation is that no causal explanation is available; and that this is because the agent's choice was literally free, as he himself felt it to be?

The answer is that this may indeed be true, inasmuch as it is open to anyone to hold that no explanation is possible until some explanation is actually found. But even so it does not give the moralist what he wants. For he is anxious to show that men are capable of acting freely in order to infer that they can be morally responsible for what they do. But if it is a matter of pure chance that a man should act in one way rather than another, he may be free but he can hardly be responsible. And indeed when a man's actions seem to us quite unpredictable, when, as we say, there is no knowing what he will do, we do not look upon him as a moral agent. We look upon him rather as a lunatic.

To this it may be objected that we are not dealing fairly with the moralist. For when he makes it a condition of my being morally responsible that I should act freely, he does not wish to imply that it is purely a matter of chance that I act as I do. What he wishes to imply is that my actions are the result of my own free choice: and it is because they are the result of my own free choice that I am held to be morally responsible for them.

Again, the objection may be raised that we are not doing justice to the moralist's case. His view is not that it is a matter of chance that I choose to act as I do, but rather that my choice depends upon my character. Nevertheless he holds that I can still be free in the sense that he requires; for it is I who am responsible for my character. But in what way am I responsible for my character? Only, surely, in the sense that there is a causal connection between what I do now and what I have done in the past. It is only this that justifies the statement that I have made myself what I am: and even so this is an over-simplification, since it takes no account of the external influences to which I have been subjected. But, ignoring the external influences, let us assume that it is in fact the case that I have made myself what I am. Then it is still legitimate to ask how it is that I have come to make myself one sort of person rather than another. And if it be answered that it is a matter of my strength of will, we can put the same question in another form by asking how it is that my will has the strength that it has and not some other degree of strength. Once more, either it is an accident or it is not. If it is an accident, then by the same argument as before, I am not morally responsible, and if it is not an accident we are led back to determinism.

Furthermore, to say that my actions proceed from my character or, more colloquially, that I act in character, is to say that my behaviour is consistent and to that extent predictable: and since it is, above all, for the actions that I perform in character that I am held to be morally responsible, it looks as if the admission of moral responsibility, so far from being incompatible with determinism, tends rather to presuppose it. But how can this be so if it is a necessary condition of moral responsibility that the person who is held responsible should have acted freely? It seems that if we are to retain this idea of moral responsibility, we must either show that men can be held responsible for actions which they do not do freely, or else find some way of reconciling determinism with the freedom of the will.

It is no doubt with the object of effecting this reconciliation that some philosophers have defined freedom as the consciousness of necessity. And by so doing they are able to say not only that a man can be acting freely when his action is causally determined, but even that his action must be causally determined for it to be possible for him to be acting freely. Nevertheless this definition has the serious disadvantage that it gives to the word 'freedom' a meaning quite different from any that it ordinarily bears. It is indeed obvious that if we are allowed to give the word 'freedom' any meaning that we please, we can find a meaning that will reconcile it with determinism: but this is no more a solution of our present problem than the fact that the word 'horse' could be arbitrarily used to mean what is ordinarily meant by 'sparrow' is a proof that horses have wings. For suppose that I am compelled by another person to do something 'against my will'. In that case, as the word 'freedom' is ordinarily used, I should not be said to be acting freely: and the fact that I am fully aware of the constraint to which I am subjected makes no difference to the matter. I do not become free by becoming conscious that I am not. It may, indeed, be possible to show that my being aware that my action is causally determined is not incompatible with my acting freely: but it by no means follows that it is in this that my freedom consists. Moreover, I suspect that one of the reasons why people are inclined to define freedom as the consciousness of necessity is that they think that if one is conscious of necessity one may somehow be able to master it. But this is a fallacy. It is like someone's saying that he wishes he could see into the future, because if he did he would know what calamities lay in wait for him and so would be able to avoid them. But if he avoids the calamities then they don't lie in the future and it is not true that he foresees them. And similarly if I am able to master necessity, in the sense of escaping the operation of a necessary law, then the law in question is not necessary. And if the law is not necessary, then neither my freedom nor anything else can consist in my knowing that it is.

Let it be granted, then, that when we speak of reconciling freedom with determinism we are using the word 'freedom' in an ordinary sense. It still remains for us to make this usage clear: and perhaps the best way to make it clear is to show what it is that freedom, in this sense, is contrasted with. Now we began with the assumption that freedom is contrasted with causality: so that a man cannot be said to be acting freely if his action is causally determined. But this assumption has led us into difficulties and I now wish to suggest that it is mistaken. For it is not, I think, causality that freedom is to be contrasted with, but constraint. And while it is true that being constrained to do an action entails being caused to do it, I shall try to show that the converse does not hold. I shall try to show that from the fact that my action is causally determined it does not necessarily follow that I am constrained to do it: and this is equivalent to saying that it does not necessarily follow that I am not free.

If I am constrained, I do not act freely. But in what circumstances can I legitimately be said to be constrained? An obvious instance is the case in which I am compelled by another person to do what he wants. In a case of this sort the compulsion need not be such as to deprive one of the power of choice. It is not required that the other person should have hypnotized me, or that he should make it physically impossible for me to go against his will. It is enough that he should induce me to do what he wants by making it clear to me that, if I do not, he will bring about some situation that I regard as even more undesirable than the consequences of the action that he wishes me to do. Thus, if the man points a pistol at my head I may still choose to disobey him : but this does not prevent its being true that if I do fall in with his wishes he can legitimately be said to have compelled me. And if the circumstances are such that no reasonable person would be expected to choose the other alternative, then the action that I am made to do is not one for which I am held to be morally responsible.

A similar, but still somewhat different, case is that in which another person has obtained an habitual ascendancy over me. Where this is so, there may be no question of my being induced to act as the other person wishes by being confronted with a still more disagreeable alternative: for if I am sufficiently under his influence this special stimulus will not be necessary. Nevertheless I do not act freely, for the reason that I have been deprived of the power of choice. And this means that I have acquired so strong a habit of obedience that I no longer go through any process of deciding whether or not to do what the other person wants. About other matters I may still deliberate; but as regards the fulfilment of this other person's wishes, my own deliberations have ceased to be a causal factor in my behaviour. And it is in this sense that I may be said to be constrained. It is not, however, necessary that such constraint should take the form of subservience to another person. A kleptomaniac is not a free agent, in respect of his stealing, because he does not go through any process of deciding whether or not to steal. Or rather, if he does go through such a process, it is irrelevant to his behaviour. Whatever he resolved to do, he would steal all the same. And it is this that distinguishes him from the ordinary thief.

But now it may be asked whether there is any essential difference between these cases and those in which the agent is commonly thought to be free. No doubt the ordinary thief does go through a process of deciding whether or not to steal, and no doubt it does affect his behaviour. If he resolved to refrain from stealing, he could carry his resolution out. But if it be allowed that his making or not making this resolution is causally determined, then how can he be any more free than the kleptomaniac? It may be true that unlike the kleptomaniac he could refrain from stealing if he chose: but if there is a cause, or set of causes, which necessitate his choosing as he does, how can he be said to have the power of choice? Again, it may be true that no one now compels me to get up and walk across the room: but if my doing so can be causally explained in terms of my history or my environment, or whatever it may be, then how am I any more free than if some other person had compelled me? I do not have the feeling of constraint that I have when a pistol is manifestly pointed at my head; but the chains of causation by which I am bound are no less effective for being invisible.

The answer to this is that the cases I have mentioned as examples of constraint do differ from the others : and they differ just in the ways that I have tried to bring out. If I suffered from a compulsion neurosis, so that I got up and walked across the room, whether I wanted to or not, or if I did so because somebody else compelled me, then I should not be acting freely. But if I do it now, I shall be acting freely, just because these conditions do not obtain ; and the fact that my action may nevertheless have a cause is, from this point of view, irrelevant. For it is not when my action has any cause at all, but only when it has a special sort of cause, that it is reckoned not to be free.

But here it may be objected that, even if this distinction corresponds to ordinary usage, it is still very irrational. For why should we distinguish, with regard to a person's freedom, between the operations of one sort of cause and those of another? Do not all causes equally necessitate? And is it not therefore arbitrary to say that a person is free when he is necessitated in one fashion but not when he is necessitated in another?

That all causes equally necessitate is indeed a tautology, if the word 'necessitate' is taken merely as equivalent to `cause': but if, as the objection requires, it is taken as equivalent to 'constrain' or 'compel', then I do not think that this proposition is true. For all that is needed for one event to be the cause of another is that, in the given circumstances, the event which is said to be the effect would not have occurred if it had not been for the occurrence of the event which is said to be the cause, or vice versa, according as causes are interpreted as necessary, or sufficient, conditions: and this fact is usually deducible from some causal law which states that whenever an event of the one kind occurs then, given suitable conditions, an event of the other kind will occur in a certain temporal or spatio-temporal relationship to it. In short, there is an invariable concomitance between the two classes of events; but there is no compulsion, in any but a metaphorical sense. Suppose, for example, that a psycho-analyst is able to account for some aspect of my behaviour by referring it to some lesion that I suffered in my childhood. In that case, it may be said that my childhood experience, together with certain other events, necessitates my behaving as I do. But all that this involves is that it is found to be true in general that when people have had certain experiences as children, they subsequently behave in certain specifiable ways; and my case is just another instance of this general law. It is in this way indeed that my behaviour is explained. But from the fact that my behaviour is capable of being explained, in the sense that it can be subsumed under some natural law, it does not follow that I am acting under constraint.

If this is correct, to say that I could have acted otherwise is to say, first, that I should have acted otherwise if I had so chosen ; secondly, that my action was voluntary in the sense in which the actions, say, of the kleptomaniac are not; and thirdly, that nobody compelled me to choose as I did : and these three conditions may very well be fulfilled. When they are fulfilled, I may be said to have acted freely. But this is not to say that it was a matter of chance that I acted as I did, or, in other words, that my action could not be explained. And that my actions should be capable of being explained is all that is required by the postulate of determinism.

If more than this seems to be required it is, I think, because the use of the very word 'determinism' is in some degree misleading. For it tends to suggest that some person has arranged it, then the proposition is false. But if all that is meant is that it is possible, in principle, to deduce it from a set of particular facts about the past, together with the appropriate general laws, then, even if this is true, it does not in the least entail that I am the helpless prisoner of fate. It does not even entail that my actions make no difference to the future: for they are causes as well as effects; so that if they were different their consequences would be different also. What it does entail is that my behaviour can be predicted: but to say that my behaviour can be predicted is not to say that I am acting under constraint. It is indeed true that I cannot escape my destiny if this is taken to mean no more than that I shall do what I shall do. But this is a tautology, just as it is a tautology that what is going to happen is going to happen. And such tautologies as these prove nothing whatsoever about the freedom of the will.