1.1 There is no single philosophical problem that is "the problem of free will". There are rather a great many philosophical problems about free will.' In this book I shall solve one of these problems and worry another at great length.

The problem I solve is the problem of "fatalism" or "future contingencies". This problem will be dealt with in Chapter II, which is a more or less self-contained essay. That chapter might have been left out of the book with almost no impairment of the argument of the remainder. It is included because it does, after all, bear on the question whether we have free will. And perhaps the fact that what it says is right will go some way toward making up for its irrelevance to the other parts of the book.

The problem of future contingencies is an old and honourable philosophical problem, but it is not one of the great central problems of philosophy. The main topic of this book is one of the great central problems: the problem of free will and determinism.

1.2 It is difficult to formulate "the problem of free will and determinism" in a way that will satisfy everyone. Once one might have said that the problem of free will and determinism — in those days one would have said 'liberty and necessity' — was the problem of discovering whether the human will is free or whether its productions are governed by strict causal necessity. But no one today would be allowed to formulate "the problem of free will and determinism" like that, for this formulation presupposes the truth of a certain thesis about the conceptual relation of free will to determinism that many, perhaps most, present-day philosophers would reject: that free will and determinism are incompatible. Indeed many philosophers hold not only that free will is compatible with determinism but that free will entails determinism. I think it would be fair to say that almost all the philosophical writing on the problem of free will and determinism since the time of Hobbes that is any good, that is of any enduring philosophical interest, has been about this presupposition of the earlier debates about liberty and necessity. It is for this reason that nowadays one must accept as a fait accompli that the problem of finding out whether free will and determinism are compatible is a large part, perhaps the major part, of "the problem of free will and determinism".

I shall attempt to formulate the problem in a way that takes account of this fait accompli by dividing the problem into two problems, which I will call the Compatibility Problem and the Traditional Problem. The Traditional Problem is, of course, the problem of finding out whether we have free will or whether determinism is true. But the very existence of the Traditional Problem depends upon the correct solution to the Compatibility Problem: if free will and determinism are compatible, and, a fortiori, if free will entails determinism, then there is no Traditional Problem, any more than there is a problem about how my sentences can be composed of both English words and Roman letters.

One of the main theses of this book is that the correct solution of the Compatibility Problem does not imply the nonexistence of the Traditional Problem; therefore my division of the problem of free will and determinism into two is no idle exercise. But before I say more about this division of the problem and about the ways in which I shall use it to organize this book, I shall explain what I mean by free will and determinism in sufficient detail to forestall certain possible misunderstandings.

1.3 In Chapter III, I shall formulate the thesis of determinism with as much precision as I am able to give it. Our present purposes will be served by a short, preliminary account of what is meant by determinism. Determinism is quite simply the thesis that the past determines a unique future. But let us see what that might mean.

Presumably, at any given moment there are many "possible futures", many ways in which the world might go on, Or at least this is true if we understand 'possible' in a sufficiently liberal way: there are certainly many "picturable" or "conceivable" or "consistently describable" futures. But many of the futures that are possible in this sense are impossible in another: they are physically impossible. For example, though I can picture to myself what it would be like for there to be a total eclipse of the sun this afternoon, though I can say without contradicting myself that a total eclipse of the sun will be visible this afternoon, there is an obvious sense in which this future I might imagine or describe is physically impossible. To say this is not to say that it is contrary to the laws of nature — if we may allow ourselves this piece of terminology — that there should be an eclipse this afternoon, for the laws of nature do not by themselves dictate when particular events like eclipses shall occur.' To say that a "possible future" containing an imminent eclipse is physically impossible is rather to say that given the past, the actual past, and the laws of nature, no eclipse will occur; that the past and the laws of nature together rule out any possibility of an eclipse this afternoon. Those who, like me, do not object to talk of "possible worlds" may think of the matter this way: while there are presumably possible worlds in which the laws of nature are the actual laws and in which there is an eclipse this afternoon, there is no possible world in which (i) the laws of nature are the actual laws, (ii) the past is the actual past, and (iii) there will be an eclipse this afternoon.

This example shows that there is a clear sense in which certain "imaginable", "conceivable", or "consistently describable" futures are physically impossible. Determinism may now be defined: it is the thesis that there is at any instant exactly one physically possible future.' There must, of course, be at least one physically possible future; if there is more than one, if at some instant there are two or more ways in which the world could go on, then indeterminism is true.

Determinism in this sense must be carefully distinguished from what we might call the Principle of Universal Causation, that is, from the thesis that every event (or fact, change, or state of affairs) has a cause. It is far from obvious what the logical relations that hold between these two theses are. I doubt, for example, whether the Principle of Universal Causation entails determinism. In order to deduce the latter from the former, we should need at least three premisses:

(1) if an event (or fact, change, state of affairs, or what have you) has a cause, then its cause is always itself an event (or what have you) and never a substance or continuant, such as a man;

(2) if an event (or what have you) A was the cause of an event B, then it follows, given that A happened and given the laws of nature, that A "causally necessitated" B, that B could not have failed to happen;

(3) every chain of causes that has no earliest member is such that, for every time t, some event in that chain happens earlier than t.

It is easy to see why each of these premisses is necessary for the deduction of determinism from the Principle of Universal Causation.

Suppose that the Principle of Universal Causation is true and suppose that premiss (1) is false. Suppose, that is, that the doctrine of immanent or agent causation is true." Suppose, to be more specific, that a certain change occurs in an agent, Tom, and Tom himself is the cause of this change, and no earlier state of affairs necessitated this change. Then the thesis of determinism is false. But our description of this case is internally consistent, for it does not entail that any event is without a cause. In particular, the change in Tom has a cause: Tom himself. The cause of this change, it is true, has no cause, but this does not entail the falsity of the Principle of Universal Causation, since its cause is not an event or change but a man, that is, a continuant.'

Suppose that the Principle of Universal Causation is true and suppose that premiss (1) is true and suppose that premiss (2) is false. That is, suppose that every event is caused by some earlier event or events but that these earlier causes do not necessitate, but merely produce, their effects. (That this supposition is consistent with our concept of causation—that is to say, with the concept of causation, for every concept is the concept it is and is not some other concept—has been argued by Professor Anscombe in her inaugural lecture.' I shall present and defend similar arguments in Chapter IV.) I think it is easy to see that, if these suppositions are correct, then, while every event has a prior cause, the past nevertheless does not determine a unique future.

Suppose that the Principle of Universal Causation is true and suppose that both premiss (1) and premiss (2) are true and suppose that premiss (3) is false. Then every event is caused by an earlier event that necessitates it; nevertheless, determinism might be false, for as Lukasiewicz pointed out, there might be a pair of times, t1 and t2, such that (i) a certain event A happens at t2, (ii) A is the final member of an infinite chain of causes, and (iii) every member of this chain occurs later than t1.' In that case, we could consistently suppose that the past (up to and including ti) and the laws of nature do not together determine whether A shall or shall not happen at t2, and do not, therefore, determine a unique future.

It follows that to deduce determinism from the Principle of Universal Causation we must assume (1), (2), and (3) as premisses. I doubt whether all three of these propositions are true. I am particularly doubtful about (2).' Therefore, I doubt whether universal causation entails determinism.

I am uncertain what to say about the question whether determinism entails universal causation. Could there be, e.g., an explosion that was not caused by any earlier events but which was none the less inevitable, given the past and the laws of nature? I think that anyone who answers immediately "Of course not!" reveals that a certain picture, or definition, or theory of causation has a very firm grip on him (which is not to say that he is wrong). But any real discussion of this question would lead us needlessly, for we need not answer it, into a discussion of causation, something I shall avoid whenever it is possible.

Having distinguished determinism from the Principle of Universal Causation, let us return to our examination of determinism. Our definition of determinism presupposes some understanding of the notion of a law of nature. I should like to define 'law of nature' in its turn, but I do not know how. I will mention, however, some constraints on an adequate definition of this concept and some necessary conditions for its application. These remarks, though they will not constitute a definition of 'law of nature', will at least show how my use of this term differs from that of some other writers. I should require any definition of 'law of nature' to have the following three consequences:

(i) the phrase 'is a law of nature' is a real predicate: it is typically and properly used in ascribing a certain property to certain objects (unlike, say, 'exists', according to Kant, or 'is good', according to R. M. Hare);

(ii) the objects that have this property are sentences or propositions (non-linguistic entities expressed by sentences) or whatever it is that are the bearers of truth-value: anything that is a law is also either true or false;

(iii) whether a proposition or sentence is a law is independent of what scientists or others happen to believe or happen to have discovered: a proposition, if it is a law, is unchangeably and objectively so, just as, according to the prevailing view of mathematics, a proposition, if it is a theorem, is unchangeably and objectively so, whatever mathematicians or others happen to believe or happen to have proved.

Philosophers have on occasion proposed necessary conditions for a proposition's being a law of nature. A law, for example, is supposed to be true, to be contingent, to entail the existence of no particular (contingent) individual and to "support its counter-factuals" or "warrant inference to subjunctive conditionals". But even if these conditions are individually necessary for lawhood, they are certainly not jointly sufficient.9 Consider, for example, the proposition that all men who are deprived of vitamin C develop scurvy. This proposition is presumably both true and contingent. Moreover, we may suppose that it "supports its counter-factuals": every man, past, present, or future, is such that if he were deprived of vitamin C, he would develop scurvy.` But it hardly follows that this general proposition is a law of nature, for the fact that it supports its counter-factuals might depend on quite accidental circumstances. Suppose, for example, that there is such a thing as "vitamin X", which could be used as an effective replacement for vitamin C; but suppose that all the vitamin X there is is locked in a vault on Mars. If that is true, then while our proposition does in fact support its counter-factual instances, its supporting them depends upon the accidental circumstance that the sole supply of vitamin X is inaccessible to human beings. Or, again, we may imagine that all men are such that they would develop scurvy if they were deprived of vitamin C, but that if an accident involving certain radioactive materials had happened at a certain time and place, some of the witnesses would have had descendants whose bodies were capable of synthesizing vitamin C and who would therefore not develop scurvy under any conditions of diet. It seems to me to be obvious that our concept of a law of nature entails that the possession of lawhood by a proposition cannot depend on such accidental occurrences as these. Therefore, these examples show that the conditions we have been examining are not jointly sufficient for lawhood. I know of none that are—except, of course, sets of conditions that are trivially sufficient, such as sets involving the requirement that a law be a physically necessary proposition—and neither, I think, does anyone else. Nevertheless, 'law of nature' seems to be an intelligible concept and one we can't get along without if we wish to give a complete description of the world. Let me give a simple example of this that is quite independent of any problems about determinism.

I have recently read an article on the possibility of inter-sidereal travel in which the authors divide the unpleasant necessities of this sort of travel into two categories: those imposed upon the travellers by the ignorance of the designer of their vehicle, and those imposed upon the travellers by the laws of nature. This is an important distinction. Spaceships and other artefacts are doubtless never perfect. But certain disadvantages of intersidereal travel are not going to be removed by technological advance as the corresponding disadvantages of inter-continental travel were removed. Inter-continental travel, now a matter of hours, was once a matter of months or years. But intersidereal travel, if it should ever come to pass, will always be a matter of years or centuries. No technological advance could ever change this unfortunate fact, for it is a consequence of the laws of nature.

Now I have just said something about the way things are; what I have said is as much a part of a complete description of the world as are the most ordinary factual statements of the geographer or the historian. And I think it is plain that what I have said I could not have said without employing the concept of a law of nature, or, at least, without employing some essentially equivalent concept like the concept of physical necessity. Therefore the notion of a law of nature makes sense, even if no one knows how to explain it to one who has not yet acquired it. (How you and I did acquire it is a question for the epistemologist or the historian of science; I am content to point out that we have it.) And, therefore, my definition of determinism, though it may rest on an undefined concept, at least rests on an undefined concept we have. Let us now see what can be done about defining free will.

1.4 I use the term 'free will' out of respect for tradition. My use of the term is not meant to imply that I think there is such a "faculty" as "the will". When I say of a man that he "has free will" I mean that very often, if not always, when he has to choose between two or more mutually incompatible courses of action—that is, courses of action that it is impossible for him to carry out more than one of—each of these courses of action is such that he can, or is able to, or has it within his power to carry it out. A man has free will if he is often in positions like these: he must now speak or now be silent, and he can now speak and can now remain silent; he must attempt to rescue a drowning child or else go for help, and he is able to attempt to rescue the child and able to go for help; he must now resign his chairmanship or else lie to the members, and he has it within his power to resign and he has it within his power to lie.

'Free will', then, is to be defined in terms of 'can'. But how is 'can' to be defined? I am afraid I do not know how to define 'can', any more than I know how to define 'law of nature'. Nevertheless, I think that the concept expressed by 'can' in the examples given in the preceding paragraph—the concept of the power or ability of an agent to act—is as clear as any philosophically interesting concept is likely to be. In fact, I doubt very much whether there are any simpler or better understood concepts in terms of which this concept might be explained. There are, however, concepts with which the concept of human power or ability might be confused, either because they really are similar to the concept of power, or because they are sometimes expressed by similar words. Perhaps what I say in the sequel will be clearer if I explicitly distinguish the concept of power or ability from those concepts with which it might be improperly conflated.

(i) The concept of the power or ability of an agent to act is not the concept of moral or legal permissibility. I might say to a hard-hearted landlord, "You can't simply turn them out into the street", knowing full well that, in the sense of 'can' that is our present concern, he very well can. The popular retort, "Oh, can't I? Just you watch!" is, I should think, usually a play on these two senses of 'can', though sometimes it may be an expression of a genuine confusion between them. And in some few cases, cases typified by the use of words like 'I can't go through with it', it may be that no one, either agent or spectator, can say with any confidence whether can is being used to express the idea of power or the idea of permissibility.

(ii) The concept of the power or ability of an agent to act is not the concept of physical possibility, nor is power or ability entailed by physical possibility. This can be shown by a simple example. Suppose I have been locked in a certain room and suppose that the lock on the door of that room is a device whose behaviour is physically undetermined; it may come unlocked and it may not: there is a future consistent with both the actual past and the laws of nature in which an internal mechanism unlocks the lock and another such future in which it doesn't. Then it is physically possible that I shall leave the room. But it does not follow that in any relevant sense I can leave the room.

(iii) The concept of the power or ability of an agent to act is not the concept of epistemic possibility. Consider the sentence, 'Castro could have arranged for Kennedy's assassination'. Clearly there are at least two things that someone who spoke these words might mean by them. He might mean, 'For all we know, Castro did arrange for Kennedy's assassination', or he might mean, 'Castro had it within his power to arrange for Kennedy's assassination'. These senses are obviously quite different and the first is of no particular interest to us.

(iv) The concept of the power or ability of an agent to act is not the concept of causal power or capacity. We say that penicillin has the power to kill certain bacteria, that a hydrogen bomb is capable of destroying a large city, and that a certain computer can perform a thousand calculations per second. (These are statements about capacities that may be unrealized. The vocabulary of our talk about the realization of causal capacities and the vocabulary of agency similarly overlap: we talk about the action of hydrochloric acid on zinc and the action of an automatic pistol.) But this sort of talk is really very different from talk of the power of an agent to act, despite their common origin in the technical terminology of medieval Aristotelianism. Perhaps the best way to appreciate this difference is to examine some ascriptions of causal capacity to human agents and to contrast them with ascriptions of ability to human agents. We ascribe a capacity, rather than an ability, to an agent when we say he:

can digest meat (being an omnivore);

has absolute pitch;

has normal colour vision;

can understand French.

We ascribe an ability, rather than a capacity, to an agent when we say he:

can cook meat;

can sing correctly;

can name the colours;

can speak French.

Despite their superficial similarity, there is all the difference in the world between the sort of property that the predicates in the first list ascribe to an agent and the sort of property that those in the second list ascribe to an agent. Consider, for example, the last item in each list. For a man to have the capacity to understand French is for him to be such that if he were placed in certain circumstances, which wouldn't be very hard to delimit, and if he were to hear French spoken, then, willy-nilly, he would understand what was being said. But if a man can speak French, it certainly does not follow that there are any circumstances in which he would, willy-nilly, speak French. The concept of a causal power or capacity would seem to be the concept of an invariable disposition to react to certain determinate changes in the environment in certain determinate ways, whereas the concept of an agent's power to act would seem not to be the concept of a power that is dispositional or reactive, but rather the concept of a power to originate changes in the environment."

Three points need to be made about what I have said about capacities and abilities.

First, it does not pretend to be an analysis of the distinction between capacities and abilities. It is, rather, an argument by example for the existence of this distinction. I do not know how to give a general account of it. (The reader may have noticed that I rarely attempt to give any general account or analysis of a concept, being content, in problematical cases, to try to show that we have a concept answering to a certain description and to try to distinguish it from other, similar concepts. In general, I am suspicious of philosophers' "analyses" of concepts, which seem to me to be only rarely correct and almost always tendentious. In the course of this book, I shall frequently appeal to our understanding of various unanalysed concepts — such as the concept of an agent's power to act — in order to convince the reader that one of my premisses is true. But note that if I were to offer a philosophical analysis of these concepts, I should have to appeal to our pre-analytical understanding of them as part of my argument for the correctness of my analysis. The appeal to intuition must turn up at some level of discourse. I prefer making "nonce" appeals to intuition at specific points in the argument to making the very abstract and general appeals to intuition that are inevitable when one is defending a philosophical analysis. This rationale for my procedure is, of course, self-serving, since I almost never know of any plausible analysis of the concepts I employ. I should add that my definitions of terms — such as 'determinism' — are not supposed to be analyses of concepts but explanations of my own technical terminology.)

Secondly, I do not mean to imply that this distinction is, at least in any very straightforward way, supported by ordinary usage. There is certainly nothing wrong with saying that someone is able to understand French, and probably nothing wrong with saying that a certain king lacks the capacity to rule.

Thirdly, I have been making a conceptual distinction. No ontological conclusions should be drawn from the existence of this distinction. Nothing I have said entails that the abilities of agents are not in some sense "reducible to" or do not "supervene upon" the causal capacities of the agents—or of some parts of agents, such as organs, cells, or atoms—and their environment. Here are two analogous cases that may make this point clearer: the concept of a number is not the concept of a set, but that does not mean that statements about numbers cannot be reduced to statements about sets or that there are numbers in addition to sets; the concept of a mental event is not the concept of a physical event, but that does not mean that statements about mental events cannot be reduced to statements about physical events or that there are mental events in addition to physical events.

I repeat: the purpose of this discussion of abilities and capacities, unlike the purposes of the discussions of most other philosophers who have made this distinction, has been conceptual and not ontological. My only purpose has been to state as clearly as I can what concept it is I shall be using 'can' and `able' and 'power' to express. It is this sense, of course, that these words bear in the above definition of free will. And it is free will as defined in the present section that I shall argue is incompatible with determinism as defined in the previous section.

(v) The concept of the power or ability of an agent to act is not the same as the concept of a skill or an accomplishment. In the preceding discussion of abilities and causal capacities, I used the predicate 'can speak French' as an example of a predicate that expresses the power of an agent to act. I did this because 'can speak French' stands in instructive opposition to the capacity-predicate 'can understand French'. But there is more to be said about 'can speak French'. Suppose that Jean-Paul, a valiant member of the Resistance, has been captured by the Germans and bound and gagged. Can he speak French? Well, to be able to speak French is to be able to speak, and. he can't speak because he is gagged: so he can't speak French. On the other hand, if the German commander ordered all prisoners who could speak French brought before him, he would be unlikely to look approvingly on the action of the subordinate who produced only the ungagged French-speaking prisoners ("For the others, Herr Oberst, cannot speak simpliciter, and, a fortiori, cannot speak French").

Clearly there is a distinction to be made between a skill, accomplishment, or general ability, on the one hand, and, on the other, the power to exercise it on a given occasion. This is true despite the fact that the same words might be used in both kinds of situation (`can speak French'; 'can move her left arm'; 'can play the flute'). That is not to say that there may not be a close conceptual connection between the two. I should think, - in fact, that a statement ascribing a skill or other general ability to an agent is probably equivalent to some statement asserting that, under certain conditions, that agent has the power to perform acts that fall under certain descriptions. But I shall not pursue this question, since it is not relevant to our present concerns. It is plain that the 'can' that figures in discussions of free will and determinism is not the `can' of skill: the thesis of determinism may or may not be relevant to the question whether someone on a particular occasion can or cannot speak French; it is certainly irrelevant to the question whether that person is a French-speaker.

1.5 It is in these senses that I shall understand 'free will' and `determinism'. In Chapters III and IV, I shall argue that free will is incompatible with determinism. It will be convenient to call this thesis incompatibilism and to call the thesis that free will and determinism are compatible compatibilism. I have no use for the terms 'soft determinism', 'hard determinism; and `libertarianism'. I do not object to these terms on the ground that they are vague or ill-defined. They can be easily defined by means of the terms we shall use and are thus no worse in that respect than our terms. Soft determinism is the conjunction of determinism and compatibilism; hard determinism is the conjunction of determinism and incompatibilism; libertarianism is the conjunction of incompatibilism and the thesis that we have free will." (I should like to appropriate libertarianism' as a name for the thesis that we have free will, but the incompatibilistic associations of this term are too firmly ingrained in too many readers for this to be wise. I have toyed with the idea of using `elutherianism' for this purpose, but reason has prevailed. I have finally and reluctantly settled on `the free-will thesis'.) I object to these terms because they lump together theses that should be discussed and analysed separately. Even having them on hand is a permanent temptation to conflate the Traditional Problem and the Compatibility Problem. They are therefore worse than useless and ought to be dropped from the working vocabulary of philosophers.

There is one other term that commonly figures in discussions of free will and determinism that I shall avoid: 'contra-causal freedom'. This term is highly ambiguous and, moreover, in accepting incompatibilism the believer in free will commits himself to accepting none of the things it might mean.

'Contra-causal freedom' might mean the sort of freedom, if freedom it would be, that someone would enjoy if his acts were uncaused. But, as we shall see in Chapter IV, that someone's acts are undetermined does not entail that they are uncaused. (This point was briefly touched on in Section 1.3.)

'Contra-causal freedom' might mean the sort of freedom, if freedom it would be, that someone would enjoy if in acting he violated the laws of nature, that is, if he worked miracles—even if only small miracles, local to the motor centres of his brain. Now I am not one of those philosophers who think that miracles are conceptually impossible. It seems to me that if God created ex nihilo a spinning object, then the proposition we call 'the law of the conservation of angular momentum' would be false. Yet, it seems to me, it might be a law of nature for all that. I think I understand the notion of a supernatural being, that is, the notion of an agent who is superior to and not a part of Nature (this enormous object that the natural sciences investigate), and I think that the falsity of a proposition counts against its being a law of nature if and only if that falsity is due entirely to the mutual operations of natural things, and not if it is due to the action of such an "external" agent upon Nature. But it does not follow from this perhaps rather quaint thesis about the concept of miracle that we can perform miracles, for there is no reason to suppose we are supernatural beings. And even if we are supernatural beings, that we are is not a consequence of the joint truth of the freewill thesis and incompatibilism. If these two theses are true, then determinism is false, and, moreover, one's free choices are undetermined: if I have a free choice between A and B, then my doing A is consistent with the past and the laws of nature and so is my doing B. Incompatibilism, therefore, entails that neither my freely doing A nor my freely doing B would "violate" a law of nature. It follows that it is sheer confusion to attribute a belief in contra-causal freedom, in the present sense, to the incompatibilist who believes in free will.

Finally, "contra-causal freedom" might be attributed to an agent if that agent has it within his power to act contrary to the laws of nature; that is, if the agent is able to perform certain acts whose performance would be sufficient for the falsity of certain propositions that are in fact laws of nature. But the incompatibilist, whether or not he accepts the free-will thesis, does not believe in contra-causal freedom in this sense. Quite the contrary: it is because the incompatibilist believes that we do not possess this power with respect to the laws of nature that he believes free will requires indeterminism. The precise sense in which this is true will be evident from an inspection of the arguments for incompatibilism that will be presented in Chapter III.

Incompatibilism, therefore, may perhaps be described as the thesis that free action is "extra-causal"; to say it is the doctrine that free action is "contra-causal" can only lead to confusion.

1.6 Incompatibilism can hardly be said to be a popular thesis among present-day philosophers (the "analytic" ones, at any rate). Yet it has its adherents and has had more of them in the past. It is, however, surprisingly hard to find any arguments for it. That many philosophers have believed something controversial without giving any arguments for it is perhaps not surprising; what is surprising is that no arguments have been given when arguments are so easy to give. Perhaps the explanation is simply that the arguments are so obvious that no one has thought them worth stating. If that is so, let us not be afraid of being obvious. Here is an argument that I think is obvious (I don't mean it's obviously right; I mean it's one that should occur pretty quickly to any philosopher who asked himself what arguments could be found to support incompatibilism):

If determinism is true, then our acts are the consequences of the laws of nature and events in the remote past. But it is not up to us what went on before we were born, and neither is it up to us what the laws of nature are. Therefore, the consequences of these things (including our present acts) are not up to us.

I shall call this argument the Consequence Argument. In Chapter III, I shall argue for incompatibilism by presenting three very detailed versions of the Consequence Argument. Or, if you like—how does one count arguments, anyway?—I shall present three detailed arguments each of which is suggested by the Consequence Argument.

In Chapter IV, I shall examine three arguments for compatibilism: the "Paradigm Case Argument", the "Conditional Analysis Argument", and an argument I shall call the Mind Argument because it has appeared so often in the pages of that journal."

The Paradigm Case Argument attempts to establish the compatibility of free will and determinism by an examination of cases of human action that, it is alleged, serve as paradigms for the teaching of the meanings of words like 'free'.

The Conditional Analysis Argument maintains that statements ascribing to human agents the power or ability to act otherwise are to be analysed as disguised conditionals, which, when their disguise is removed, can be seen to be compatible with determinism.

The Mind Argument proceeds by identifying indeterminism with chance and by arguing that an act that occurs by chance, if an event that occurs by chance can be called an act, cannot be under the control of its alleged agent and hence cannot have been performed freely. (In this brief statement of the Mind Argument I have left entwined three strands of reasoning that I shall disentangle in Chapter IV.) Proponents of the Mind Argument conclude, therefore, that free will is not only compatible with determinism but entails determinism.

I shall argue in Chapter IV that these arguments fail. There are, I think, no other arguments for compatibilism that need be taken seriously. Let me give just one example of an argument that need not be taken seriously. Some philosophers have urged that to suppose that free will and determinism are in conflict is to confuse compulsion with determination by causal laws. For, these philosophers argue, to do something of one's own free will is to do that thing without being compelled to do it, and to behave in accordance with a deterministic set of causal laws is not to be compelled. I reply that this argument confuses doing things of one's own free will with having free will about what one does. These are not the same thing. The following case shows this. Suppose a certain man is in a certain room and is quite content to be there; suppose, in fact, that he wants very much to. remain in that room and that it would be difficult indeed to induce him even to consider leaving it. Then, I should think, he remains in the room of his own free will. But we can with perfect consistency go on to suppose that he has no free will about whether he leaves the room: suppose that, unknown to him, the only door is locked and that it is not within his power to open it.

Now someone might want to say that our imaginary agent did not remain in the room "of his own free will". I don't think this is right, but I will not argue the point. For if it is true that our agent did not remain in the room "of his own free will", then we cannot establish that a person does something "of his own free will" by establishing that his doing it is uncompelled. But the "compulsion" argument we have been considering certainly does depend on the premiss that one can so establish that a person has acted "of his own free will". Let us grant this premiss. The example of the man locked in the room shows that it does not follow from a person's doing something "of his own free will" that he can do otherwise. And thus it does not follow from the undoubted fact that we often do things of our own free will that what I have called the free-will thesis is true. And, therefore, the Compatibility Problem is not going to be solved by jejune reflections on compulsion.

1.7 If the arguments of Chapter III are correct, then incompatibilism is true. These arguments have premisses. Some of the premisses are more controversial than others. That is, some of the premisses of Chapter III will be accepted without question by the compatibilist and others he will want to argue about. Let us call the conjunction of these "controversial" premisses P. Suppose I am willing to grant that if any of my premisses are false, the false ones are conjuncts of P. The compatibilist and I will thus agree that if compatibilism is true, then P is false. One compatibilist has actually argued (in effect) that this proposition on which he and I agree entails that I am begging the question against compatibilism by assuming the truth of P.

It should suffice to point out that the situation in which this argument places the compatibilist and me is a perfectly symmetrical one: I am in a position to employ the same argument, mutatis mutandis, to prove the conclusion that his choice of premisses begs the question, and I should be as well justified in employing the argument against him as he is in employing it against me. Now why, I have asked myself uneasily, would anyone say something that can be so easily refuted? I think there are two possible answers. (i) The compatibilist thinks that compatibilism is prima facie true and that the burden of proof lies on the incompatibilist. (When a philosopher says, "The burden of the proof lies on you", he means, "You must deduce your conclusion from the truths of immediate sensory experience by means of an argument that is formally valid according to the rules of elementary logic, I on the other hand may employ any dialectical tactic I find expedient".) (ii) He thinks that the premisses from which he derives compatibilism are more plausible than those from which I derive incompatibilism.

As to the first of these possibilities, I deny that compatibilism is prima facie right and incompatibilism prima facie wrong. Quite the other way round, if you ask me. But I shall not assume that either of these propositions is prima facie right. I shall treat them as philosophical theses of equal initial plausibility, and this, it seems to me, is the only reasonable way to approach the Compatibility Problem.

In Section 4.3, I shall attempt to undermine the thesis that the premisses of the compatibilist are prima facie more plausible than my own by comparing the premisses of one argument for compatibilism, the Conditional Analysis Argument, with the premisses of one of the incompatibilist arguments presented in Chapter III. I think that the unprejudiced reader—if such can be found: very likely anyone willing to read as far as Chapter IV will have brought strong feelings about the Compatibility Problem to his reading of this book—will agree that if the premisses of my argument and the premisses of the Conditional Analysis Argument are considered independently of their implications for the Compatibility Problem, then the former will be seen to be considerably more plausible than the latter.

1.8 If the arguments of Chapters III and IV are cogent, then I shall have provided good reasons for thinking that free will and determinism are incompatible. Suppose I am right. Suppose these theses are incompatible. Which, if either, ought we to accept? As a first step towards answering this question, I shall, in Chapter V, address the question, "What would it mean to reject free will?" I shall argue that such a rejection would have at least two interesting consequences.

First, anyone who rejected free will could not consistently deliberate about future courses of action. This is so, I shall argue, owing simply to the fact that one cannot deliberate without believing that the things about which one is deliberating are things it is possible for one to do. How important one takes this consequence to be will, of course, depend on how important one thinks consistency is.

Secondly, and, I think, much more importantly, to deny the existence of free will commits one to denying the existence of moral responsibility. Until a short while ago, most philosophers would have taken this to be obvious. But if any of these philosophers had been asked to defend this obvious thesis, he would almost certainly have appealed to the following principle: a person can be held morally responsible for what he has done only if he could have done otherwise. In a recent remarkable article, however, Harry Frankfurt has presented convincing counter-examples to this principle. I shall devote the bulk of Chapter V to showing that even if Frankfurt is right, it is none the less true that moral responsibility is possible only if we have free will.

Now some philosophers will perhaps want to protest at this point in the argument that while I may indeed have shown that in some sense free will is incompatible with determinism, and while I may have shown that free will in some sense is logically necessary for moral responsibility, I have not shown that there is any single notion of free will that has both these features. Therefore, these philosophers may allege, I am not in a position to say that considerations having to do with moral responsibility can be used to show that we ought to accept the doctrine called 'free will' in Chapters III and IV and there shown to be incompatible with determinism.

I shall meet this possible objection in two ways. First, I shall ask the reader to examine the premisses of the arguments of Chapter III after they have been rewritten according to the following rule: at each place at which the words 'free will' occur in the premisses of these arguments, replace them with the words 'free will in just that sense of free will that is relevant to questions of moral responsibility'. I contend that the reader will find that the rewritten premisses are no less plausible than the original ones. Secondly, I shall present an argument for the incompatibility of moral responsibility and determinism that makes no mention whatever of free will, though it will be structurally identical with one of the arguments for the incompatibility of free will and determinism that occurs in Chapter III. I think this "direct" argument for the incompatibility of responsibility and determinism will have the following feature: it will clearly be sound if and only if the corresponding argument of Chapter III is sound. But if this is true, then it seems very unlikely that it is only some sort of "free will" that has nothing to do with moral responsibility that is shown in Chapter III to be incompatible with determinism.

The principal conclusion of Chapter V will therefore be that to reject free will—in just that sense of 'free will' in which we have earlier argued that free will is incompatible with determinism—is to reject moral responsibility.

In Chapter VI, I shall discuss the Traditional Problem, that is, the problem of finding out whether determinism is true, or whether the free-will thesis is true, or whether neither is true. I shall proceed by asking what reasons we have for thinking that determinism is true and what reasons we have for thinking that the free-will thesis is true.

There are two sorts of reason for believing in determinism. First, one might believe in determinism because one believes that science has shown determinism to be true. I shall argue at length that science has, if anything, shown determinism to be false. Secondly, one might believe that determinism is a truth of reason, on the ground that it is a logical consequence of the Principle of Sufficient Reason. I shall show that, whether or not determinism is a consequence of the Principle of Sufficient Reason, the Principle of Sufficient Reason must be rejected, since it entails the collapse of all modal distinctions.

In addition to asking what reasons might be brought in support of determinism, we shall also ask what reasons might be brought in support of a certain closely related but weaker thesis: the thesis that human beings, like typical digital computers, behave in ways that are entirely determined by their past states and their current "input". It is important to consider this thesis because: (i) it would seem to be incompatible with free will if and only if determinism proper is; (ii) unlike determinism proper it is not in even prima-facie conflict with current physical theory; and (iii) it is commonly contended that the empirical study of human beings has uncovered facts that strongly support it. I shall argue that this common contention is sheer bluff.

The conclusion of the argument whose course is summarized in the last few paragraphs is that neither physics nor pure reason supports determinism, and, moreover, that the scientific study of human beings does not support the thesis that the behaviour of human beings is "for all practical purposes" determined.

What reasons can be brought in support of the free-will thesis? It cannot, I think, be seriously maintained that we can know by some sort of introspection that we have or that we do not have free will. And neither can it be maintained that the empirical study of human beings is likely to show us that we have or that we do not have free will. The only relevant argument would seem to be this: if we do not have free will, then there is no such thing as moral responsibility; therefore, since there is such a thing as moral responsibility, there is such a thing as free will. (Moreover, since free will is incompatible with determinism, determinism is false.) The first premiss of this argument is defended in Chapter V in the way outlined above. In Chapter VI, we shall examine its second premiss, and I shall defend my use of this argument against the charge that for an incompatibilist so to argue amounts to his claiming to be able to prove that determinism — a thesis about the motion of matter in the void — can be shown to be false by a priori reflection on moral responsibility. I shall also examine a condition of certain philosophers (their having fallen under the spell of "scientism") that makes it psychologically very difficult for them to believe that such "tender-minded" arguments as this could possibly provide one with good reason to reject determinism.

I shall finally address the question, "What would you say if, after all, the progress of science did show that indeterminism was untenable?"