Kant writes in the Antinomies section of the first Critique: ‘nothing is left except to reflect on the origin of this disunity of reason with itself, on whether a mere misunderstanding might perhaps be responsible for it, after the elucidation of which perhaps both sides will give up their proud claims, but in place of which reason would begin a rule of lasting tranquility over understanding and the senses’ (B492–3).Footnote ¹

The prospect and promise of this peaceful reign is repeated a number of times in the Doctrine of Method. Since reason’s internal conflict rests on a misunderstanding, once this is removed, peace will ensue, even if at the price of diminished metaphysical pretentions. But has Kant fulfilled his promise? Has he guided us into the safe haven of good metaphysics, or is there some bad metaphysics in Kant’s account of good metaphysics? Is the root of reason’s division against itself really based merely on a misunderstanding? Finally, can Kant help us to identify any bad metaphysics in our own time?

The distinction between good and bad metaphysics plays a central role in Adrian Moore’s engaging chapter on Kant in The Evolution of Modern Metaphysics. As he explains, Kant agreed with Hume that much of what had been hitherto considered metaphysics had to be thrown out, for failing to make sense or for failing to be warranted by experience (p. 109).Footnote ² So the questions urged upon us by our metaphysical drive were either pseudo-questions or unverifiable. Moreover, old metaphysics had not even managed to agree on a procedure to attempt an answer to such questions, unlike mathematics (p. 110).Footnote ³ But this is where Kant’s agreement with Hume ends. For the sceptic Hume threw the baby of metaphysics out with the dirty bath-water of pseudo-metaphysics. Indeed, seen from Kant’s point of view, Hume offered some bad metaphysics himself, for he was unable to account for metaphysical issues such as the principle that every event in nature has a cause, or for mathematical necessity (p. 111). As Kant points out in the Doctrine of Method, Hume’s own scepticism had no proper justification.Footnote ⁴ Hume was a ‘geographer of human reason’ who aimed to quell the burning questions of metaphysics by locating their source beyond the limits of human reason, without being able to account for those very limits (B788).

In what follows, Moore develops Kant’s account of synthetic a priori knowledge and the correlated doctrine of transcendental idealism. This doctrine needs to embrace seemingly conflicting motives. If knowledge is knowledge of something independent of it, how can it be a priori? If it is a priori, how can it be synthetic, i.e. not purely conceptual? If it offers us metaphysical insight, how is it not a form of old metaphysics about experience-transcendent things? Moore argues that Kant manages the remarkable and original feat of bringing these motives into a coherent whole. He explains transcendental idealism in terms of the metaphor of the (‘native’) spectacles, which affect only how we see things, not which things we see (pp. 121–2). While these spectacles don’t allow us to know how things are in themselves, in accounting for the nature of these very spectacles in terms of synthetic a priori propositions (obtained by transcendental proofs of the conditions necessary for having the experience we do), Kant is involved in the most general attempt to make sense of things, which, on Moore’s definition, is just what metaphysics is (pp. 123–4).Footnote ⁵

Metaphysics thus understood is simply a priori knowledge of what we can know, involving an analysis of our cognitive faculties, undertaken in the Aesthetic and Analytic. Kant explains this conception of metaphysics in The Architectonic of Pure Reason (B870ff.). Of course, in its disunity against itself reason is not troubled by this sort of metaphysics, but by questions about the existence of God, the immortality of the soul, the freedom of the will and the origin and nature of the world. Our elementary metaphysical instinct is not concerned with knowing what we can know, but knowing the ultimate essence of things, of the greatest things. As Moore puts it: ‘metaphysicians most deeply aspire to … synthetic a priori knowledge of things in themselves’ (p. 125), i.e. ‘synthetic a priori knowledge without appeal to intuitions’ (p. 124). Since we humans can only have intuition-based knowledge, it follows that metaphysicians aspire to attain something contradictory, synthetic a priori knowledge which is not synthetic.

Leibniz would have been surprised to be told that in his metaphysics he was really aiming for synthetic a priori knowledge, given that Leibniz took metaphysics to investigate necessary truths, which are proven by the method of analysis (i.e. the proof of identities by means of the progressive substitution of complexes with their simple elements).Footnote ⁶ And Frege, in the nineteenth century, will not only be surprised by the suggestion, but explicitly deny that his ontological investigation into the nature of numbers is based on synthetic a priori truths, or aims for such truths; to the contrary, number statements are logically derivable from the laws of logic, and hence analytic, but still about numbers, which are non-physical and non-psychological entities. Our knowledge of numbers is analytic, but does rest on our grasp or intuition of logical entities like judgeable contents, functions, extensions (Frege Reference Frege1979: 7, Reference Frege1984: 300; Gabriel Reference Gabriel2000: 31–2). That we have a logical, and thus non-sensible, intuition was also affirmed by Russell (cf. his ‘awareness of universals’) and Husserl (cf. his doctrine of categorial intuition).Footnote ⁷

We may wonder what Kant might reply to these rationalists. Presumably, he would reply that they misunderstand what they are doing. Leibniz has simply misclassified at least some necessary truths as analytic, as Kant argues for example with respect to the principle of sufficient reason, a pillar of Leibniz’s metaphysics (B246). Frege was guilty of a similar misclassification; few would argue today that Hume’s Principle, required for the original logicist project, is an analytic truth. Kant would surely also object to the possibility of an ‘awareness of universals’ and of ‘categorial intuition’. And what would Kant say about logical entities enjoying high ontological currency today, e.g. propositions, possible worlds, functions from possible worlds to propositions, and so on? Presumably he would argue that talk of knowledge of these entities is based either on a confusion between (no matter how abstract and complex) forms of cognition and the objects of this cognition, or, alternatively, on a confusion between a category and a merely regulative idea, i.e. on what Kant calls a ‘transcendental subreption’ (B537, B610–11). The former alternative is entailed by his doctrine of the categories as mere forms of thought, which are ‘empty concepts of objects’ when lacking experiential input (B148–9, 288, 306ff.). The second alternative is worked out in the Dialectic. Take, for instance, the concept of a possible world, the concept of the set of ‘all possible worlds’,Footnote ⁸ the concept of ‘absolutely everything’,Footnote ⁹ or indeed something apparently so innocuous as Frege’s Latin letters in his concept-script, ranging over ‘all’ judgeable contents (see Frege Reference Frege1972: §11) – all these look suspiciously like Kantian ideas. About ideas Kant says that there cannot be any appearance found which represents them in concreto (B595, cf. Prolegomena, §40, 4: 327). And surely, we cannot represent ‘all possible worlds’ in concreto; there is simply no possible experience in which these can be given. It seems legitimate to ask today’s metaphysicians for a new ‘transcendental deduction’, an account of the possibility of our alleged knowledge of these entities. If such a deduction cannot be given, we would have good reason to believe that such entities are fictions, in Kant’s sense of ‘Hirngespinste’ (see e.g. B269, B415). In the Transcendental Ideal (B599) Kant claims that our concept of an object is guided by the so-called principle of thorough determination, according to which we presuppose of any object that for all possible predicates F of things and their negations ¬F, that F or ¬F must be true of the object (B599–600). This principle, according to Kant, involves the concept of the totality of all possibility, and that is an idea (B601). But an idea does not give us any knowledge about any putative object falling under it. While contemporary metaphysics is hardly concerned with proofs about God or the immortality of the soul, its assumption that we can refer to and acquire seemingly non-empirical knowledge about things like possible worlds etc. can still be classified as dogmatic, following Kant. Indeed, it can be classified as ‘bad metaphysics’ in the way specified by Moore, i.e. as an enterprise ‘failing to have a suitable warrant in experience’, or worse, ‘failing to make sense’ (p. 109). Could we then not revive Kant’s critical doctrines of the categories and ideas, stripped of his more problematic views, and bring them to bear upon contemporary metaphysics? This is one sense in which we can pursue Moore’s question ‘How might we use [transcendental idealism]?’ (p. 141) despite the various problems the doctrine faces.

To return to Moore’s statement that ‘metaphysicians most deeply aspire to synthetic a priori knowledge of things in themselves’, we may wonder about another issue. If ‘noumenal knowledge’ is, in last instance, an impossibility, how is the very aspiration to it possible? More precisely, is this aspiration actually intelligible? It is true that Kant distinguishes between knowing and thinking (e.g. Bxxvi), and claims that we can think far more than we can know (p. 134). But even mere thinking has to obey some constraints, in particular the law of contradiction, for Kant writes: ‘I can think whatever I like, as long as I do not contradict myself’ (Bxxvi). Since metaphysics aspires to attain synthetic a priori knowledge which is not synthetic, it is contradicting itself. That would mean that its aspiration is unintelligible, if to aspire to X entails that one be able to think X (if asked ‘What do you aspire to?’, I should be able to say, and think, intelligibly, ‘X’).

Moore draws some helpful distinctions in this respect. He makes an initial distinction between the scope and limits of metaphysics, i.e. what metaphysics can justifiably achieve and what it cannot. The scope is revealed in the Analytic, the limits in the Analytic (through the transcendental deduction) and the Dialectic. But the greatness of Kant’s account consists not only in tracing the limits, but also in explaining ‘why we nevertheless feel the urge to make these attempts’ of transcending them (p. 125). Like Wittgenstein, he does not merely reveal our illusions, but also the logic behind them. As Moore explains Kant’s argument, the questions metaphysicians are tempted to ask always involve some idea of reason. Ill-conceived questions are those which involve ‘a confused amalgam of an idea of reason with some concept that can only be applied to objects of experience’ (p. 126). Such a question has no answer, in the intended metaphysical sense. He illustrates this nicely with the example of the confused concept of the physical universe as a whole (pp. 126–7). A well-conceived question contains no such amalgam, and has an answer, at the level of things in themselves (ibid.), but we lack the resources to know it. The question of the existence of God, or the immortality of the soul, might be examples of well-conceived questions.

Interestingly, Moore suggests different senses of the irresistibility of metaphysical questions. It is irresistible, because natural, for all of us to pose well-conceived questions. It is irresistible, because based on specific philosophical misunderstandings, for metaphysicians to further pose ill-conceived questions. In both cases, it appears, the root illusion persists (p. 129), but the major difference seems to be that natural reason expects to know the unknowable, while philosophical reason expects to know the nonsensical.

Is this a fair interpretation? If so, the question still remains as to how we can make sense of our aspiration to unavailable knowledge. How can ‘our hopes concern how things are in themselves’ (p. 132), if we do not have the slightest inkling about them? How can hope reach out to the noumenal realm, while knowledge cannot? Moore suggests that ultimately there are cracks in Kant’s transcendental idealism, a point he demonstrates in his discussion of Kant’s wonderful sphere analogy in the Doctrine of Method (pp. 136–7, cf. B76ff.). If the surface of the Earth is what we can make empirical sense of, then at any one point in time what we see, standing at a particular location, is limited by the horizon. This limit can of course be extended, and is called by Kant Schranke (ibid., also Prolegomena, §57, 4: 351). By contrast, to know what we can make sense of, in principle, is to know the dimensions of the Earth’s surface, and that is to know the Grenzen, boundaries, of our (geographical) knowledge, by means of some synthetic a priori reasoning, geometry (see B790). But as Moore argues, ‘we can legitimately refer to the limits of a globe only because we have access to a dimension other than the surface’s own two [dimensions]’ (p. 137). Similarly, transcendental idealism can only work if it can make sense of the limits it claims our knowledge has. But for this we need to make sense not only of what lies on our side of the limits, but also what lies beyond, in whatever attenuated sense. And Moore points out that transcendental idealism cannot make sense of what lies beyond, with its account of transcendental illusion as based on the ‘empty play of our concepts’ (p. 137, my italics).

Adding to Moore’s point, the problem can also be approached from a related aspect. In his classification of philosophy in The Architectonic of Pure Reason (B870ff.) Kant identifies as one branch of metaphysics ‘in the narrow sense’ the so-called physiology of pure reason (B873), which, as two of its branches, contains the pure cognition of the world as a whole and of God, two ‘transcendental cognitions’. These appear to be cognitions answering to well-conceived questions and are not the illusory cognitions discussed in the Dialectic.

But what is this transcendental cognition exactly? Does it amount to mere thinking about God and the world? This does not seem to be true, if we follow Kant’s distinction (Bxxvi) between erkennen (to cognize, to know) and denken (to think): the Erkennen of an object requires that I can prove its possibility, i.e. the objective reality of the corresponding concept, while mere thinking does not require this. It may be replied that Kant does not require ideas to have objective reality. Indeed, this is implied in various places. Only categories have objective reality, and they do so only in combination with the conditions of sensibility, while ideas are ‘further removed from objective reality’ (B595). Since ideas are devoid of sensibility, if ideas had objective reality, there would be a pure, non-sensory use of concepts yielding knowledge of things in themselves (B249–50, also B194, 270). All that is needed is that ideas, e.g. the unconditioned, can be thought without contradiction (which is arguably one of the major selling points of transcendental idealism; see Bxx). Kant writes in the Analytic of Principles: ‘a pure use of the category is possible, i.e., without contradiction, but it has no objective validity’ (A253). And further: ‘this further extension of concepts beyond our sensible intuition does not get us anywhere. For they are then merely empty concepts of objects, … mere forms of thought without objective reality … Our sensible and empirical intuition alone can provide them with sense and meaning (Sinn und Bedeutung)’ (B148). In the curious section on nothingness (at the end of the Analytic, B346ff.) Kant characterizes such concepts without objects as entia rationis, mere fictions (Erdichtung).

So it appears that transcendental cognition involves mere thinking (see p. 134), coherently but vacuously, about God and the world, mere entia rationis. But this is still problematic.

To begin with, there is a difference between categories and ideas. Moore explains it in terms of different uses of the same thing. Ideas are simply uses of the categories freed of their experience-related content. However, despite Kant’s own assertions in this respect (B435), we need to be careful here, for this would turn ideas into empty categories, lacking objective reality. Kant also says, and needs to say, that ideas have (some sort of) objective reality. Otherwise they would be rather useless for the noble purpose he envisages for them. This is why he requires a transcendental deduction for them as well, to prove that they have some, however indeterminate, ‘objective validity, and don’t merely represent empty thought-entities (entia rationis ratiocinantis)’ (B697, a claim repeated in Prolegomena, §40, 4: 327). Even if we accept that this objective reality/validity points towards rational faith, we may still wonder what this objective reality exactly is. It is clear that ‘objective reality’ in this context cannot mean what it means in the context of the Analytic, for there it means ‘application to objects that can be given to us in intuition’ (B150–1) or it means, more generally, the feature a representation has when it is referring to an object, which requires that ‘the object must be able to be given in some way’ (B194). But what does ‘objective reality’ mean in the context of ideas? At the end of the Dialectic Kant makes a distinction between an object’s being given either absolutely or in the idea (B698). The objective reality of the categories relates to the former, that of ideas to the latter. In the latter case we use ideas to represent objects ‘indirectly’, not in order to know their properties, but only to understand their greatest systematic unity in reason and experience (B698–9). In doing so, we derive the objects of experience from ‘the imagined object of this Idea as its ground or cause’, which allows us to consider the world as if created by God, etc. (cf. also B706).

This is problematic in several respects. First, as Kant also argues, the objects of ideas are mere analogues of real things (B702, 706), which means that the latter are required to make sense of the former. But this would involve a circularity, since Kant also argues, as just seen, that we derive real objects (objects of experience) from the objects of ideas. Second, the object of the idea is described as an imagined object (B698) and as a ‘Something in general’ (B706). But the latter is the expression of my concept of an object, which is different from my concept of an imagined object.Footnote ¹⁰ Moreover, while I can ‘posit’ some real object in a context (e.g. assume that there is some unfamiliar object in my pocket), it is not clear how I can posit the object of an idea, which, given its utmost generality, is devoid of any context and is a real object (wirklichen Gegenstand) only as something in general (B705). My reason’s giving itself this supposed object (B709) is also not the same as my imagining it, for reason is surely distinct from imagination, since imagination is the representation of an object in intuition without its presence (B151).

One might reply that the problem concerning the objective reality of ideas cannot be fully solved using the resources of Kant’s theoretical philosophy, but requires an engagement with his practical philosophy. Indeed, it is in this context that Kant returns to the problem, speaking about the objective, but practical reality of ideas (e.g. Fortschritte, 20: 300), which he also describes as ‘moral-practical reality’ (Zum ewigen Frieden, 8: 416). Our ideas obtain this reality through the categorical imperative, which simply amounts to our acting as if their objects, God and immortality, were given, i.e. as if we had knowledge of God and our immortality (ibid.).Footnote ¹¹ This is no doubt a crucial move in Kant’s overall project of metaphysics, and we may wonder why Moore does not devote more discussion to it.Footnote ¹² But it is not clear that the move saves Kant from inconsistency. By compositionality, to make sense of an ‘as if X’ construction, we need to be able to make sense of ‘X’. Hence, to make sense of ‘as if we had knowledge of God and our immortality’, we need to be able to make sense of our having knowledge of God and our immortality, i.e. have a coherent conception of such a possibility. But we do not, for all (human) knowledge involves sensory intuition for Kant. We would have to be able to make sense of knowledge of an object which no experience can give (B709), of knowledge of something about which the question whether it is a substance, whether it has necessity, whether it has the greatest reality, etc., has no significance (B724). In other words, this would require us to make sense of the possibility of knowledge about an object about which the question which properties it has lacks significance, which looks like a straightforward contradiction.

The question of how and in what sense ideas have objective reality is not a minor one. Moore quite rightly says: ‘The whole complex machinery that drives Kant’s transcendental idealism … in fact serves to keep our most important hopes alive. “I had to deny knowledge, in order to make room for faith” (Bxxx)’ (p. 132). The reconciliation between ‘the demands of Christian morality and the demands of Newtonian mechanics is the most important, most profound, and the one to which Kant is most ardently committed’ (p. 132). Kant cannot and does not want to curb our aspiration for the higher, for, as he says of his secret metaphysical ally in the first Critique: ‘Plato noted very well that our power of cognition feels a far higher need than that of merely spelling out appearances according to a synthetic unity in order be able to read them as experience’ (B370).Footnote ¹³ But how is this rational faith to be based on and motivated by mere fictions, concepts without sense and meaning, empty forms of thoughts, especially if we know that they are fictional and empty? How do we even distinguish between two concepts without sense and meaning, and what distinguishes a concept without sense and meaning from a nonsensical concept, a non-concept? As above: how are we able to aspire to and hope for something whose concept is empty or worse? How is noumenal hope possible, if the noumenon is a mere limiting concept (B310)? How can one aspire to transcendence, if its very concept is a limiting one? The ideal of reason, i.e. God, is in last instance not even a noumenon, not even inscrutable and unknowable, since it is ‘not even given as a thinkable object’ (B642). As ‘a mere idea, it must find its seat and its resolution in the nature of reason’. At the same time reason maintains its drive for the infinite, and cannot be repressed to throw ‘a glance on the wonders of nature and the majesty of the world’s architecture – by which it elevates itself from magnitude to magnitude up the highest of all, rising from the conditioned to the condition, up to the supreme and unconditioned Author’ (B652; section on the physico-theological proof). Reason’s desire for the absolute appears to be struggling here against its own tendency to dissolve its own noble aim.

In conclusion, it appears that reason, even in Kant’s system, remains in disunity with itself. The question is whether this is so due to a ‘mere misunderstanding’, this time on Kant’s part, which, once removed, will offer ‘full satisfaction’ to reason’s cognitive drive, as Kant puts it in the last sentence of the Critique of Pure Reason (B856), or whether to a more permanent feature of human reason.Footnote ¹⁴ The drama of metaphysics took a particular twist with Kant, but did not end with him, as Adrian Moore’s subsequent chapters in his book so engagingly demonstrate.