The Proper Study of Mankind

by Isaiah Berlin

  Anti-determinists have naturally retorted that this merely pushed the problem a step backwards: the ‘self’ played its part, indeed, but was itself hopelessly ‘determined’. It may be worth going back to the origins of this controversy, for, as often happens, its earliest form is also the clearest. It came up, so far as I can tell, as a consequence of the interest taken by the early Greek Stoics in two, at first unconnected, ideas: that of causation, that is, the conception, new in the fourth century BC, of unbreakable chains of events in which each earlier event acts as a necessary and sufficient cause of the later; and the much older notion of individual moral responsibility. It was perceived as early as the beginning of the next century that there was something paradoxical, and indeed incoherent, in maintaining that men’s states of mind, feeling and will as well as their actions were links in unbreakable causal chains, and at the same time that men were responsible, that is, that they could have acted otherwise than in fact they did.

  Chrysippus was the first thinker to face this dilemma, which did not seem to trouble Plato or Aristotle, and he invented the solution known as self-determination – the view that so long as men were conceived of as being acted upon by outside forces without being able to resist them, they were as stocks and stones, unfree, and the concept of responsibility was plainly inapplicable to them; if, however, among the factors that determined behaviour was the bending of the will to certain purposes, and if, moreover, such a bending of the will was a necessary (whether or not it was a sufficient) condition of a given action, then they were free: for the act depended on the occurrence of a volition and could not happen without it. Men’s acts of will and the characters and dispositions from which, whether or not they were fully aware of it, such acts issued, were intrinsic to action: this is what being free meant.

  Critics of this position, Epicureans and sceptics, were not slow to point out that this was but a half-solution. We are told that they maintained that although it might be that the operations of the will were a necessary condition of what could properly be called acts, yet if these operations were themselves links in causal chains, themselves effects of causes ‘external’ to the choices, decisions and so on, then the notion of responsibility remained as inapplicable as before. One critic4 called such modified determination hemidoulia – ‘half-slavery’. I am only half free if I can correctly maintain that I should not have done x if I had not chosen it, but add that I could not have chosen differently. Given that I have decided on x, my action has a motive and not merely a cause; my ‘volition’ is itself among the causes – indeed, one of the necessary conditions – of my behaviour, and it is this that is meant by calling me or it free. But if the choice or decision is itself determined, and cannot, causally, be other than what it is, then the chain of causality remains unbroken, and, the critics asserted, I should be no more truly free than I am on the most rigidly determinist assumptions.

  It is over this issue that the immense discussion about free will that has preoccupied philosophers ever since originally arose. Chrysippus’ answer, that all that I can reasonably ask for is that my own character should be among the factors influencing behaviour, is the central core of the classical doctrine of freedom as self-determination. Its proponents stretch in unbroken line from Chrysippus and Cicero to Aquinas, Spinoza, Locke and Leibniz, Hume, Mill, Schopenhauer, Russell, Schlick, Ayer, Nowell-Smith and the majority of the contributors to the subject in our own day. Thus when a recent writer in this chronological order, Richard Hare, in one of his books5 distinguishes free acts from mere behaviour by saying that a pointer to whether I am free to do x is provided by asking myself whether it makes sense to ask ‘Shall I do x?’ or ‘Ought I to do x?’, he is restating the classical thesis. Hare correctly says that one can ask ‘Will I make a mistake?’ or ‘Will I be wrecked on the sea-shore?’ but not ‘Shall I make a mistake?’ or ‘Ought I to be wrecked?’; for to be wrecked or make a mistake cannot be part of a conscious choice or purpose – cannot, in the logical or conceptual sense of the word. And from this he concludes that we distinguish free from unfree behaviour by the presence or absence of whatever it is that makes it intelligible to ask ‘Shall I climb the mountain?’ but not ‘Shall I misunderstand you?’ But if, following Carneades, I were to say ‘I can indeed ask “Shall I climb the mountain?”, but if the answer – and the action – are determined by factors beyond my control, then how does the fact that I pursue purposes, make decisions and so forth liberate me from the causal chain?’, this would be regarded as a misconceived enquiry by the Stoics and the entire classical tradition. For if my choice is indispensable to the production of a given effect, then I am not causally determined as, say, a stone or a tree that has no purposes and makes no choices is determined, and that is all that any libertarian can wish to establish.

  But no libertarian can in fact accept this. No one genuinely concerned by the problem constituted by the prima facie incompatibility between determinism and freedom to choose between alternatives will settle for saying ‘I can do what I choose, but I cannot choose otherwise than as I do.’ Self-determination is clearly not the same as mechanical determination. If the determinists are right (and it may well be that they are) then the sort of determination in terms of which human behaviour should be described is not behaviouristic, but precisely Chrysippus’ hemidoulia. But half a loaf is not the bread that libertarians crave. For if my decisions are wholly determined by antecedent causes, then the mere fact that they are decisions, and the fact that my acts have motives and not only antecedents, do not of themselves provide that line of demarcation between freedom and necessitation, or freedom and its absence, which the ordinary notion of responsibility seems, at least for libertarians, so clearly to entail. It is in this sense that Bacon’s followers claim too much.

  This may be seen from another angle which will bring us back to the relations of knowledge and liberty. The growth of knowledge increases the range of predictable events, and predictability – inductive or intuitive – despite all that has been said against this position, does not seem compatible with liberty of choice. I may be told that if I say to someone ‘I always knew that you would behave with wonderful courage in this situation’ the person so complimented will not suppose that his capacity for freedom of choice is being impugned. But that seems to be so only because the word ‘knew’ is being used, as it were, in a conventionally exaggerated way. When one man says to another ‘I know you well: you simply cannot help behaving generously; you could not help it if you tried’, the man so addressed may be thought susceptible to flattery, because of the element of complimentary hyperbole in the words ‘cannot help’ and ‘could not … if you tried’. If the words were intended to be taken literally – if the flatterer meant to be understood as saying ‘You can no more help being generous than being old, or ugly, or thinking in English and not in Chinese’ – the notion of merit or desert would evaporate, and the compliment would be transformed from a moral into a quasi-aesthetic one.

  This may be made clearer if we take a pejorative example: if I were to say of x ‘x can no more help being cruel and malicious than a volcano can help erupting – one should not blame him, only deplore his existence or seek to tame him or restrain him as one would a dangerous animal’, x might well feel more deeply insulted than if we lectured him on his habits on the assumption that he was free to choose between acting and refraining from acting as he did, free to choose to listen to our homily or pay no attention to it. The mere fact that it is my character that determines my choices and actions does not, if my character itself and its effects are due to ineluctable causes, render me free in the sense that appears to be required by the notions of responsibility or of moral praise and blame. Knowledge of the causes and conditions that determine my choice – knowledge, indeed, that there are such conditions and causes, knowledge that choice is not free (without analysis of this proposition), knowledge that shows that the notion of moral responsibility is wholly compatible with rigorous determinism, and exposes
libertarianism as a confusion due to ignorance or error – that kind of knowledge would assimilate our moral views to aesthetic ones, and would lead us to look on heroism or honesty or justice as we now do on beauty or kindness or strength or genius: we praise or congratulate the possessors of the latter qualities with no implication that they could have chosen to own a different set of characteristics.

  This world view, if it became generally accepted, would mark a radical shift of categories. If this ever occurs, it will tend to make us think of much of our present moral and legal outlook, and of a great deal of our penal legislation, as so much barbarism founded on ignorance; it will enlarge the scope and depth of our sympathy; it will substitute knowledge and understanding for attribution of responsibility; it will render indignation, and the kind of admiration that is its opposite, irrational and obsolete; it will expose such notions as desert, merit, responsibility, remorse, and perhaps right and wrong too, as incoherent or, at the very least, inapplicable; it will turn praise and blame into purely corrective or educational instruments, or confine them to aesthetic approval or disapproval. All this it will do, and if truth is on its side, it will benefit mankind thereby. But it will not increase the range of our freedom. Knowledge will render us freer only if in fact there is freedom of choice – if on the basis of our knowledge we can behave differently from the way in which we would have behaved without it – can, not must or do – if, that is to say, we can and do behave differently on the basis of our new knowledge, but need not. Where there is no antecedent freedom – and no possibility of it – it cannot be increased. Our new knowledge will increase our rationality, our grasp of truth will deepen our understanding, add to our power, inner harmony, wisdom, effectiveness, but not, necessarily, to our liberty. If we are free to choose, then am increase in our knowledge may tell us what are the limits of this freedom and what expands or contracts it. But only to know that there are facts and laws that I cannot alter does not itself render me able to alter anything: if I have no freedom to begin with, knowledge will not increase it. If everything is governed by natural laws, then it is difficult to see what could be meant by saying that I can ‘use’ them better on the basis of my knowledge, unless ‘can’ is not the ‘can’ of choice – not the ‘can’ which applies only to situations in which I am correctly described as being able to choose between alternatives, and am not rigorously determined to choose one rather than the other. In other words, if classical determinism is a true view (and the fact that it does not square with our present usage is no argument against it), knowledge of it will not increase liberty – if liberty does not exist, the discovery that it does not exist will not create it. This goes for self-determinism no less than for its most full-blown mechanistic-behaviourist variety.

  The clearest exposition of classical self-determinism is probably that given in his Ethics by Spinoza. Stuart Hampshire represents him,6 it seems to me correctly, as maintaining that the fully rational man does not choose his ends, for his ends are given. The better he understands the nature of men and of the world, the more harmonious and successful will his actions be, but no serious problem of choice between equally acceptable alternatives can ever present itself to him, any more than to a mathematician reasoning correctly from true premisses to logically unavoidable conclusions. His freedom consists in the fact that he will not be acted upon by causes whose existence he does not know or the nature of whose influence he does not correctly understand. But that is all. Given Spinoza’s premisses – that the universe is a rational order, and that to understand the rationality of a proposition or an act or an order is, for a rational being, equivalent to accepting or identifying oneself with it (as in the old Stoic notion) – the notion of choice itself turns out to depend upon the deficiencies of knowledge, the degree of ignorance. There is only one correct answer to any problem of conduct, as to any problem of theory. The correct answer having been discovered, the rational man logically cannot but act in accordance with it: the notion of free choice between alternatives no longer has application. He who understands everything understands the reasons which make it as it is and not otherwise, and being rational cannot wish it to be otherwise than as it is. This may be an unattainable (and perhaps even, when thought through, an incoherent) ideal, but it is this conception that underlies the notion that an increase in knowledge is eo ipso always an increase in freedom, that is, an escape from being at the mercy of what is not understood. Once something is understood or known (and only then), it is, on this view, conceptually impossible to describe oneself as being at the mercy of it. Unless this maximal rationalist assumption is made, it does not seem to me to follow that more knowledge necessarily entails an increase in the total sum of freedom; it may or may not – this, as I hope to show, is largely an empirical question. To discover that I cannot do what I once believed that I could will render me more rational – I shall not beat my head against stone walls – but it will not necessarily make me freer; there may be stone walls wherever I look; I may myself be a portion of one; a stone myself, only dreaming of being free.

  There are two further points to be noted with regard to the relationship of freedom and knowledge:

  (a) There is the well-known objection, urged principally by Karl Popper, that the idea of total self-knowledge is in principle incoherent, because if I can predict what I shall do in the future, this knowledge itself is an added factor in the situation that may cause me to alter my behaviour accordingly; and the knowledge that this is so is itself an added factor, which may cause me to alter that, and so on ad infinitum. Therefore total self-prediction is logically impossible. This may be so: but it is not an argument against determinism as such (nor does Popper so represent it) – only against self-prediction. If x can predict the total behaviour of y, and y predict the total behaviour of x (and they do not impart their prophecies to one another), that is all that determinism needs. I cannot be self-consciously spontaneous; therefore I cannot be self-consciously aware of all my states if spontaneity is among them. It does not follow that I can never be spontaneous; nor that, if I am, this state cannot be known to exist while it is occurring, although it cannot be so known to me. For this reason I conclude that, in principle, Popper’s argument does not (and is not meant to) refute determinism.

  (b) Stuart Hampshire, in the course of some recent remarks,7 advances the view that self-prediction is (logically) impossible. When I say ‘I know that I shall do x’ (as against, for instance, ‘x will happen to me’, or ‘You will do x’), I am not contemplating myself, as I might someone else, and giving tongue to a conjecture about myself and my future acts, as I might be doing about someone else or about the behaviour of an animal – for that would be tantamount (if I understand him rightly) to looking upon myself from outside, as it were, and treating my own acts as mere caused events. In saying that I know that I shall do x, I am, on this view, saying that I have decided to do x: for to predict that I shall in certain circumstances in fact do x or decide to do x, with no reference to whether or not I have already decided to do it – to say ‘I can tell you now that I shall in fact act in manner x, although I am, as a matter of fact, determined to do the very opposite’ – does not make sense. Any man who says ‘I know myself too well to believe that, whatever I now decide, I shall do anything other than x when the circumstances actually arise’ is in fact, if I interpret Hampshire’s views correctly, saying that he does not really, that is, seriously, propose to set himself against doing x, that he does not propose even to try to act otherwise, that he has in fact decided to let events take their course. For no man who has truly decided to try to avoid x can, in good faith, predict his own failure to act as he has decided. He may fail to avoid x, and he may predict this; but he cannot both decide to try to avoid x and predict that he will not even try to do this; for he can always try; and he knows this: he knows that this is what distinguishes him from non-human creatures in nature. To say that he will fail even to try is tantamount to saying that he has decided not to try. In this
sense ‘I know’ means ‘I have decided’ and cannot in principle be predictive.

  That, if I have understood it, is Hampshire’s position, and I have a good deal of sympathy with it, for I can see that self-prediction is often an evasive way of disclaiming responsibility for difficult decisions, while deciding in fact to let events take their course, disguising this by attributing responsibility for what occurs to my own allegedly unalterable nature. But I agree with Hampshire’s critics in the debate, whom I take to be maintaining that, although the situation he describes may often occur, yet circumstances may exist in which it is possible for me both to say that I am, at this moment, resolved not to do x, and at the same time to predict that I shall do x, because I am not hopeful that, when the time comes, I shall in fact even so much as try to resist doing x. I can, in effect, say ‘I know myself well. When the crisis comes, do not rely on me to help you. I may well run away; although I am at this moment genuinely resolved not to be cowardly and to do all I can to stay at your side. My prediction that my resolution will not in fact hold up is based on knowledge of my own character, and not on my present state of mind; my prophecy is not a symptom of bad faith (for I am not, at this moment, vacillating) but, on the contrary, of good faith, of a wish to face the facts. I assure you in all sincerity that my present intention is to be brave and resist. Yet you would run a great risk if you relied too much on my present decision; it would not be fair to conceal my past failures of nerve from you.’ I can say this about others, despite the most sincere resolutions on their part, for I can foretell how in fact they will behave; they can equally predict this about me. Despite Hampshire’s plausible and tempting argument, I believe that such objective self-knowledge is possible and occurs; and his argument does not therefore appear to me to lessen the force of the determinist thesis. It seems to me that I can, at times, though perhaps not always, place myself, as it were, at an outside vantage-point, and contemplate myself as if I were another human being, and calculate the chances of my sticking to my present resolution with almost the same degree of detachment and reliability as I should have if I were judging the case of someone else with all the impartiality that I could muster. If this is so, then ‘I know how I shall act’ is not necessarily a statement of decision: it can be purely descriptive. Self-prediction of this kind, provided that it does not claim to be too exact or infallible, and meets Popper’s objection, cited above, by remaining tentative, allowing for possible alterations of conduct as a result of the self-prediction itself – seems possible and compatible with determinism.