The early pages of long books generally get far more attention than the later pages. But most authors give us their most important messages later in the book. Some authors even tell you this about their own book. And in the Critique of Pure Reason, Kant is one of those authors. The very first sentence of the A Preface is: ‘Human reason has the peculiar fate in one species of its cognitions that it is burdened with questions which it cannot dismiss, since they are given to it as problems by reason itself, but which it also cannot answer, since they surpass every capacity of human reason’ (KrV Avii).Footnote ¹ This sentence is plainly about the Transcendental Dialectic, not about Kant’s positive account of human cognition found in the Aesthetic and Analytic that has attracted most interpretative and critical attention. So it is the Dialectic that the Critique is mainly about. Ian Proops is therefore to be commended for writing a book concentrating on the Dialectic, especially a book so thorough, careful, and rigorously argued (Proops Reference Proops2021; cited simply by page number throughout). My task here is to discuss Chapters 11 and 12, dealing with the Third Antinomy and Kant’s struggle with the metaphysical problem of freedom. But those two chapters are a nearly fifty-page stretch of Proops’ text, and my space here is sharply limited. All I can do is register some points of agreement and disagreement. I place emphasis on the points of agreement, both because they are important and also because they give us a different picture of Kant’s treatment of freedom from the one you find in many other writers.

Proops sees that Kant thinks theoretical reason can neither prove nor disprove the proposition that the human will is free. Kant needs to establish only the possibility of freedom from a theoretical standpoint so that he can open the way to his claim that we are free from a practical point of view. For even practical considerations could not justify assent to what we know to be theoretically impossible.

But what kind of possibility is it supposed to be within the limited power of theoretical reason to establish? Kant distinguishes logical possibility from real possibility. He is clear that we cannot prove the real possibility of freedom (A558/B586). In the concluding paragraphs of his discussion, he says that his discussion aims to show only that causality from nature and causality from freedom ‘do not contradict each other’ and that the one causality ‘does not conflict’ with the other (A557-8/B585-6). This sounds like logical possibility. Proops, however, offers us the interesting suggestion that Kant’s aim is to establish only their epistemic possibility (p. 280) – that is, as far as we can know the human will might be free and at the same time subject to natural causality.

Kant never speaks of epistemic possibility. It apparently does not belong to his theory of modality – at least not explicitly. Though epistemic possibility is distinct from logical possibility, he might well have included it under that heading. Epistemic possibility would fit well with Proops’ conclusion about the Third Antinomy itself: that the truth of the thesis would not entail the falsity of the antithesis and the truth of the antithesis would not entail the falsity of the thesis (p. 281). At first glance, this conclusion might seem hard to accept, since the thesis says that we must assume a causality of freedom in addition to that of nature, while the antithesis says that everything happens only from natural causes (A444-5/B472-3). Does not each of these propositions directly claim the falsity of the other? But it becomes easier to see that Proops is correct once we realise that each proposition can be taken as referring either to appearances or to things as they might exist in themselves. A transcendental realist might see the thesis and the antithesis as contradicting each other, but once we entertain Kantian transcendental idealism, the two sides of the antinomy might not contradict each other. They might even both be true, so that the truth of one would not entail the falsity of the other, and we might be incapable of knowing whether either is true or false.

Another point on which I agree with Proops is the way he understands Kant’s defence of the claim that the human will has transcendental freedom when the question is addressed from the practical point of view. He presents the conclusion that we are transcendentally free as following from four premises: ‘First, that ought implies can, second, that we sometimes fail to do what the moral law requires of us, third, that we know our obligations, and fourth, that practical freedom implies transcendental freedom’ (p. 324).

I think this is at least sometimes Kant’s reasoning, but it needs a bit of unpacking. If you look carefully at the contexts in which Kant asserts ‘Ought implies Can’, it amounts to his claim that rational agents have practical freedom: the capacity to act according to reason and resist sensible impulses. Thus ‘can’ refers only to what we might call motivational possibility: the capacity to choose as we ought in the face of contrary incentives. It never refers to the external means you might need to do what you ought to do. When Proops says that Kant’s third premise is that we know our obligations, I think he means that we are aware of having certain specific moral obligations; and the second premise says that we are sometimes aware of failing to meet them due to a weakness of our character. Together with ‘Ought implies Can’, these premises entail that whatever we choose to do, we could have chosen to do otherwise. For when we do not choose to meet our obligations, it remains true that we nevertheless could have chosen to meet them; and when we do meet them, it is still true that owing to our moral imperfections, there was always the possibility that we might have failed. This entails that what we do is up to us; it cannot be causally necessitated by anything outside us. This ability of ours presupposes the capacity to act, as Kant puts it, entirely von selbst – ‘from itself’ or ‘from oneself’. That is what Kant means by transcendental freedom, and that this is presupposed by practical freedom is Proops’ fourth premise. So given the four premises Proops attributes to Kant, they provide a valid argument that we have not only practical freedom but also transcendental freedom. The premises, moreover, are not justified theoretically but appeal only to what we must accept from a practical standpoint. So this argument proves transcendental freedom only from a practical point of view.

I do not think that this is the only route by which Kant argues for freedom from a practical standpoint. In the Groundwork, he appeals to what we must presuppose when we make any judgement (G, 4: 448). There the idea is that even a theoretical judgement, if we continue to accept and endorse it as what we truly judge, must be regarded as up to us and not causally necessitated by any external mechanism. For if we came to see it as something forced on us by external causes, then we could no longer credit it as a judgement we can endorse as our own. We would see it, as Kant says, as ‘steering from elsewhere’. No doubt people do sometimes make judgements caused by such alien manipulation, but once they see that this is the origin of such judgements, they can no longer credit them as a successful exercise of their own reason. Thus any argument that all our judgements are causally imposed on us in this way would have to discredit its own premises and be self-undermining.

Kant’s readers, however, often see him as threatened by fatalism. Once in the Critique of Pure Reason and once again in the Critique of Practical Reason, Kant suggests that if we could know enough empirically about our own psychology, all our actions might be revealed as causally necessary and predictable with the certainty of a solar or lunar eclipse (KrV, A550/B578, KpV, 5: 99). But all too few who quote these eye-catching passages realise that Kant in fact denies that the supposed knowledge is even possible. As he says in a later work:

Proops is one of the few scholars who appreciate this point. Regarding all conditionals of the form ‘If we could know all the psychological causes leading up to our action, then we could predict the action and cognize it as necessary, like an eclipse of the sun or moon’, Proops sees that the antecedent would require intellectual intuition of the subject, as Kant says explicitly in the very next sentence of the second Critique. And of course, Kant denies that we can have such intellectual intuitions. Therefore, the antecedent in any such conditional is not only false, but impossible, or as Proops says, not merely ‘counterfactual but counterpossible’ (pp. 300–1). Therefore, Kant rejects, and is even required to reject, the thought that the natural causation of events in the world of appearance might entail that our empirical knowledge of people’s actions would result in the impossibility of practical freedom. Kant’s position thus requires theoretical scepticism about free will and therefore establishes the epistemic possibility of both transcendental freedom and practical freedom. But Kant denies that it is ever possible for us to demonstrate on theoretical grounds either that there is freedom or that there is not. Theoretically, fatalism is always an epistemic possibility. Only in a practical respect must we claim we are free.

It is also seldom appreciated – but this time Proops does not mention it – that late in the first Critique, in the Canon of Pure Reason, Kant asserts that if human beings rationally consider whether people have free will, not restricting themselves to either theoretical or practical considerations alone, but take everything into account, then we can prove empirically that normal human adults do have practical freedom. We can determine empirically through behaviour when a person has the capacity to make choices according to reason and not determined by empirical impulses. People who are too immature do lack this capacity, and some adults also lack it in whole or part due to illness or other impairment of their faculties (A802-3/B830-1). There are empirical questions about the nature and extent of our capacity to make rational choices. Some of the literature on free will discusses them in an intelligent and informed manner. The point not to miss is that Kant’s position leaves such empirical questions entirely open.

Kant realises, of course, that there always remains a metaphysical problem of how transcendental freedom, hence how practical freedom, could be possible in relation to the causal structure of nature. What appears to us to be freedom on the empirical level, he says, might ‘with regard to higher and more remote efficient causes’ be once again natural necessity, so that the practical freedom we necessarily presuppose and even investigate with empirical common sense might after all be an illusion. Freedom thus remains permanently an insoluble problem for theoretical reason (A803/B831). It can be neither proven, disproven, nor in any way comprehended, nor can we ever gain insight into even its real possibility (G, 4: 459). This means that as a question of transcendent metaphysics, Kant can have no theory at all about how the freedom that we presuppose in acting and investigating empirically can be reconciled with natural necessity. Proops shows he appreciates this crucial point when he says that Kant holds that from a practical standpoint, we must ascribe transcendental freedom to ourselves, but ‘I have no insight into how this remarkable capacity is possible’ (p. 313).

Proops discusses Kant’s arguments using a set of terms not found in Kant but frequently used by philosophers nowadays who discuss the issue of free will. I mean terms like ‘determinism’, ‘libertarianism’, ‘compatibilism’, and ‘incompatibilism’. To his credit, Proops seems wary of these terms, but I do not think he sufficiently appreciates how utterly hopeless they are in relation to the entire topic of free will. Proops apparently cannot resist the temptation to ask and answer questions about Kant’s position using this disastrous terminology. I think that is a mistake.

According to Proops, as I read him, Kant is a determinist, he is not a libertarian, and he is a compatibilist, albeit of an unusual sort – not like either Leibniz or Hume. It is not so much that I disagree with these categorizations as that I regard this entire way of framing the issues as misleading and an invitation to confusion and error. Kant himself never uses any of this terminology, either to embrace it or reject it. And philosophers who do use it, whether to describe their own positions or the positions of those with whom they are arguing, seem to me to leave the issue of free will worse off. Even if we do not actually fall into error using this terminology, I think interpreting Kant’s position on freedom using it therefore leaves our understanding of Kant worse off.

Proops resists calling Kant a libertarian. Well, let us ask: What is a libertarian? It is someone who believes the human will is free. Kant surely does hold that the human will is free. Or perhaps a libertarian is someone who asserts freedom of the will in the face of those truths and arguments that lead others to deny freedom of the will. Kant famously does that too. Kant even claims we have transcendental freedom, the capacity to begin a state or a series of states von selbst: to act entirely independently of external influences of any kind. Proops admits this is Kant’s position. So in any of these ways, Kant could be described as a libertarian, as is done by many whom Proops cites. So why does Proops disagree with them? His reason seems to be that for them libertarianism is the denial of determinism, and Proops insists that Kant is a determinist. So that forces us to ask (again in the same unfortunate terminology): ‘What is determinism?’

It is certainly correct that Kant argues in the Second Analogy that all alterations occur in accordance with the law of cause and effect (B232). This law says that in any alteration, the consequent state follows the antecedent state in accordance with a necessary law. Or: the antecedent state determines the consequent state necessarily. Is not that what we mean by ‘determinism’? We might think so. Yet in the course of his resolution of the Third Antinomy, when he is trying to show that natural causality might not conflict with freedom, Kant says (in a passage Proops himself quotes):

This passage certainly does not sound like an endorsement of determinism. It may even sound like a blunt denial. But here we must keep in mind Kant’s aim, and his strategy, for resolving the Third Antinomy. Kant’s aim, as we have seen, is only to show that freedom and natural causality do not conflict, that as far as theoretical reason can establish, both together are possible – not really possible, but possible either logically, or at least epistemically. Here Kant appeals to his basic claim in resolving the dynamical antinomies: both the thesis and the antithesis might be true. Or more specifically: the antithesis might be true in the world of appearance, while the thesis might be true when things are regarded in themselves or as noumena or in an intelligible world.

It is important to appreciate, however, that in his resolution of the Third Antinomy, Kant could not possibly be committed positively to affirm any of the things he says about the intelligible world, noumenal causality, or noumenal freedom. Kant argues that noumena are in principle unknowable, and even their real possibility or impossibility is unknowable. A noumenon is, as Kant tells us at the end of the Amphiboly, one of the four kinds of nothing: a mere ens rationis, an empty concept without an object (A292/B348). All Kant intends to claim – all he is entitled to claim – is the mere logical, or even only the epistemic, possibility that the will might be causally determined when things are regarded as phenomena but nevertheless transcendentally free when they are considered as noumena. Therefore, what Kant says about noumenal causality in resolving the Third Antinomy cannot possibly be taken as an explanation, or even a genuine concept, of transcendental freedom in its relation to natural causality. Kant’s critics (when they mistakenly attribute such a fantastic metaphysical account to him) are entirely correct when they point out that a supposedly free will that must act outside space and time in a supernatural noumenal world could not be anything like the freedom we ascribe to ourselves. It yields no concept of an arbitrium liberum sed sensitivum, a free but sensitive power of choice, as Kant says ours is (A534/B562). As to how our free power of choice is possible, Kant says: ‘Freedom cannot be comprehended, nor even can insight into it be gained’ (G, 4: 459). Once again, as Proops puts it: ‘I know I have the capacity that constitutes transcendental freedom; I have no insight into how this remarkable capacity is possible’ (p. 313). This should tell us what to think about all those interpretations of Kant that try to use some theory of noumenal freedom to shed light on human action. We should dismiss them straightway and not entertain them at all, even from a practical point of view. It should also tell us what to think about Proops’ own musings about how noumenal action operates, about ‘intra-realm and inter-realm causality’, ‘inter-realm compatibilism and intra-realm incompatibilism’, ‘phenomenal determinism and noumenal libertarianism’, or how we should answer Ralph Walker’s worries that Kant’s theory might make you and me morally responsible for the Lisbon earthquake, American slavery, and the Holocaust. Proops cites replies to such worries that others have proposed. But my view is that all claims about noumenal causation are at most speculations Kant plays with in order to establish mere logical or epistemic possibilities. Beyond that, they are contemptible nonsense which should not detain our thinking for even an instant. My view about all the literature that does this could be better formulated using Hume’s famous words: ‘Commit it then to the flames, for it can contain nothing but sophistry and illusion!’

So at last, what about ‘compatibilism’? Is Kant a compatibilist? He does accept a universal principle of natural causality governing all alterations in the phenomenal world, and yet nevertheless maintains that we experience ourselves as agents having practical freedom, which presupposes transcendental freedom. Is that any form of compatibilism?

Over 40 years ago I wrote a paper whose title was ‘Kant’s Compatibilism’. Was I arguing that Kant is a compatibilist? I have since then also written papers whose titles were ‘Kant’s Deism’ and ‘Kant’s Historical Materialism’. But all three papers aimed not at ascribing the named view to Kant but instead putting the title itself in question. Proops praises me for suggesting, as he thinks, that Kant was trying to show ‘the compatibility of compatibilism and incompatibilism’. But what I said was this: ‘When we consider all Kant’s views together, it is tempting to say that he is trying to show not only the compatibility of freedom and determinism but also the compatibility of compatibilism and incompatibilism’ (Wood Reference Wood1984: 74). However, I also said that Kant’s views do not fit into the customary pigeonholes, and the free will issue is no exception (Wood Reference Wood1984: 73). And as we all know – or at least we should know it if we want to avoid making a mess of our lives – there are many things in life that we may be tempted to do which we must absolutely never do. I would classify saying that Kant is trying to show either the compatibility of freedom and determinism or the compatibility of compatibilism and incompatibilism as among those treacherous temptations that we absolutely must resist. I now think that when we discuss the philosophical problem of free will, we would be wise to expunge the terms ‘determinism’, ‘libertarianism’, ‘compatibilism’, and ‘incompatibilism’ from our vocabulary. Even if we can avoid errors while using them, they make it harder to do so.

I used to say, whenever the topic of free will came up, that the only conceivable solution to it would be some form of compatibilism. I said this because, on the one hand, denying free will is a non-starter: it is not morally reprehensible (as some suggest), but it is incoherent and self-undermining. It is another disastrous notion to think that the free will issue itself is about moral responsibility. If freedom has anything at all to do with moral responsibility, that is only because moral responsibility presupposes agency and free will is about agency. On the other hand, it is another non-starter to defend freedom of the will based on some supernaturalist causality (including the ridiculous caricature that ascribes to Kant a theory of free will in a timeless noumenal world). That is a rank superstition of which any self-respecting person should be ashamed. So rejecting both these non-starters, all that seems to be left is compatibilism. But having said this much, I would then add pointedly that, unfortunately, no form of compatibilism is defensible. Over time, however, my view has evolved even further. I would now define ‘compatibilism’ as any lie that some dishonest philosopher tries to use to make the annoying problem of freedom go away. But it is only honest to admit that it can never go away. It is insoluble.

My chief objection, therefore, to the standard terminology – and especially to the terms ‘compatibilism’ and ‘incompatibilism’ – is that this terminology takes for granted that you have to choose between the standard options, and further that you must have a ‘theory’ about how freedom is or is not compatible with causal determinism. But no theory of this is possible. That is why disputes about free will are endless and all pretended solutions to the metaphysical problem of freedom are dishonest.

People do make choices; they have what Kant calls ‘practical freedom’. For roughly the reasons I have ascribed to Kant, it is evident that whenever we do make a free choice, we are committed to holding that we can make this choice. But to acknowledge this commitment I do not first need a philosophical theory that explains how I can have the capacity to make choices. Therefore, we are neither required nor entitled to have any theory – either compatibilist or incompatibilist – about how free choice is possible.

This is also the position I would ascribe to Kant. For the everyday purposes of action and psychology, the practical freedom we must presuppose when we choose is something Kant takes to be provable empirically. There is overwhelming empirical evidence that we often do choose and that our choices are solely up to us. But when it comes to the transcendental freedom (beginning a state von selbst) that practical freedom presupposes, Kant denies that we can have any theory about how it is possible.

So here is what I think. Like what people call the ‘hard problem of consciousness’, the possibility of transcendental freedom is not about the world we know but rather about our access to that world. But how we have access to the world is not something to which we have access. That is why there are questions we cannot dismiss but also cannot answer. That is why the topic of freedom remains for philosophers a permanent source of controversy, puzzlement, and frustration – hence also of illusion and dishonest deception.