1. Précis of Kant on the Sources of Metaphysics: The Dialectic of Pure Reason

2. Reply to Paul Guyer

Paul Guyer takes issue with three aspects of my reconstruction of Kant’s account of the rational sources of metaphysics. First, he finds ‘the framework of the book slightly confusing’ with respect to the way I distinguish between Kant’s critique of speculative metaphysics and what I call the Rational Sources Account. Second, he points out two respects in which the concept of God plays a more important role for Kant’s account of metaphysics than I ascribe to it. And third, Guyer objects to my account of the highest good and the postulates based on it (in my brief postscript on Kant’s ‘practical metaphysics’). I agree with Guyer that none of these points amounts to a substantial disagreement. Still, I am grateful for the opportunity to clarify and further explain some aspects of my reading of Kant.

First, the general framework of my book: I distinguish two strands in the Transcendental Dialectic: (i) the critique of traditional speculative metaphysics and of its alleged proofs and (ii) the Rational Sources Account, which reconstructs this form of metaphysics as arising from the structure of rational thinking (discursivity, iteration, completion). The Rational Sources Account, as I understand it, consists of three sub-theses, and it is developed by Kant on four levels (which correspond to the four parts of the Transcendental Dialectic) in a way that follows a three-part pattern (a logical/real transition and a regulative/constitutive transition plus the assumption of transcendental realism; cf. the précis of the book above). I admit that this complex structure can be confusing. But I think that it is necessary to bring out the complexity of Kant’s argument in the Transcendental Dialectic.

Guyer mainly complains that it is difficult to keep the two strands of the Transcendental Dialectic apart, the critical and the reconstructive. But this may be due to the fact that the critical strand has long been so dominant that we automatically read some steps of the other strand, the Rational Sources Account, as critical. In particular, it is tempting to understand the tendency to use the principles and ideas of pure reason constitutively (and not just regulatively) as part of Kant’s critique of metaphysics. Thus, Guyer writes: ‘the Rational Sources Account is Kant’s derivation of the regulative principles of reason from its most basic operations, while the critical core of the Dialectic is its explanation that we transform the legitimate regulative principles of reason into illegitimate constitutive principles because of our natural but unwarranted attraction to transcendental realism’ (p. 6). But note that Guyer’s ‘critical core of the Dialectic’ consists of two logically distinct points: (i) the transition from the regulative to the constitutive use of reason and (ii) the claim that the constitutive use is illegitimate. While the second point is part of Kant’s critique of metaphysics, I think of the first point as part of the Rational Sources Account, since it is the regulative/constitutive transition that explains why we not only ask metaphysical questions and entertain metaphysical hypotheses, but tend to accept metaphysical theses and theories as true representations of reality. That this step is illegitimate is an additional and logically independent point Kant makes.

This means that we must distinguish two questions: (1) Why do we tend to use rational principles constitutively (Kant’s answer: because of a tacit assumption of transcendental realism) and (2) Are we justified in using rational principles constitutively? (Kant’s answer: no, because that would take us beyond the limits of possible experience). As I understand it, while the answer to the first question is part of the Rational Sources Account, the answer to the second is part of Kant’s critique of metaphysics. Note that the Rational Sources Account is meant to be compatible with traditional metaphysics, since it reconstructs the thinking that underlies it and shows that it does not arise from contingent features of human psychology or from individual mistakes, but rather from the structure of reason as such.Footnote ³

In sum, while Guyer is correct to say that the critical core of the Transcendental Dialectic lies in criticizing the transition from the regulative to the constitutive, this does not mean that this transition itself cannot be part of the Rational Sources Account. Once this is granted, I think it is possible to keep apart the critical and the reconstructive strands in the Transcendental Dialectic.

Let me add a remark concerning a more specific point that Guyer makes in the first part of his comments. Guyer reads the ‘logical maxim’ (to seek the unconditioned for everything conditioned, A307/B364) as the regulative version of the ‘supreme principle’, which claims that there exists something unconditioned for everything conditioned. He implies that the ‘logical maxim’ therefore is not a logical principle ‘properly speaking’ (p. 3), since this would require that it abstract from objects and only concern cognitions. As I argue, however, the logical maxim is better understood as concerning not objects but conditioned cognitions, and thus requiring us to seek unconditioned cognitions (i.e. first principles), which is not only what Kant’s text suggests, but also allows for the ‘logical maxim’ to be a logical principle in Kant’s sense.

I now turn to Guyer’s ‘substantive suggestions’ (pp. 6ff). First, Guyer objects that God plays a more important role in Kant’s argument than I allow for: ‘The idea of God is thus not just the idea of one more unconditioned alongside others, but in Kant’s mind all roads lead to the idea of God’ (pp. 7–8). While I agree with this claim, I do not agree with Guyer’s reading of the passage on which he bases it. Kant writes that there are ‘three classes’ of transcendental ideas, ‘of which the first contains the absolute (unconditioned) unity of the thinking subject, the second the absolute unity of the series of conditions of appearance, the third the absolute unity of the condition of all objects of thought in general’ (A334/B391). Guyer glosses this as follows: ‘In other words, however the ideas of the unconditioned subject and world-whole might be reached, the idea of God is inevitably reached as the unconditioned condition of those’ (p. 7). If Guyer wants to say here that God is thought of as the unconditioned condition specifically of our substantial souls and the empirical world as a whole, I do not think that this is what Kant says in the quoted passage. The three kinds of the unconditioned are distinguished here according to the objects they condition, namely a subject’s representations in the first case, empirical phenomena in the second and objects of thought in general in the third. Of course, our souls and the world are also objects of thought and, according to the metaphysics Kant reconstructs here, conditioned by God. But it is not as if the idea of God was introduced specifically as the unconditioned condition of the objects of the other two ideas.

Next, let me briefly comment on a remark Guyer makes in passing: ‘Thus transcendental realism is blocked with regard to the first two antinomies, but allowed as an indemonstrable possibility for the theses of the last two’ (p. 7). I assume that Guyer here thinks of transcendental realism as the claim that there are things in themselves that ground empirical phenomena, since it is this claim that allows Kant to resolve the dynamical antinomies. But that is not what Kant himself calls ‘transcendental realism’ and what I mean by that term in my book (cf. Chignell’s comments and my reply for a discussion of transcendental realism). As I see it, transcendental realism (TR) consists in the identification of empirical objects with things in themselves (and, more specifically, with noumena in the positive sense). This identification, which Kant of course rejects, is a background condition of all four antinomies, since they all rest on a constitutive reading of the ‘supreme principle’, which in turn is motivated by TR. Given the constitutive reading of the ‘supreme principle’, it follows from the existence of conditioned empirical objects that something unconditioned exists. The antinomies then arise because this ‘unconditioned’ can be thought of either as a first unconditioned condition or as the totality of conditioned conditions (A417/B445). Since the resolution of the antinomies is not part of the Rational Sources Account (but rather part of the critical strand of the Transcendental Dialectic), I do not discuss it in detail in my book. But here is how I think my reading would play out when applied to the resolution of the antinomies: If we give up TR, we lose the implicit premise of the theses and antitheses, namely that there is something unconditioned. In the first two antinomies, which concern only relations within the empirical world, this means that both sides turn out to be false (since series of conditions in the empirical world are neither finite nor infinite); in the last two antinomies, which potentially concern relations between events in the empirical world and their intelligible grounds as well, this means that both sides can be true, since there may be unconditioned conditions of empirical objects and events that lie outside the empirical world.Footnote ⁴

Next, Guyer suggests that in the published introduction to the third Critique (5: 180), Kant introduces the idea of God not as resting on a form of transcendental realism (TR), as he does in the first Critique, but on a misapplication of transcendental idealism:

I agree with Guyer’s reading of the passages from the third Critique and would only like to point out that, on my understanding of TR, TR and Kant’s own transcendental idealism are not contradictory, but merely contrary theses. The most general form of TR I discuss in my book is the claim (put roughly) that things are as we necessarily represent them to be (TR_rep). In the hands of the traditional metaphysician, this becomes the thought that the world must be as reason makes us expect it to be, a thought Kant himself rejects. But note that Kant’s own transcendental idealism can be seen as a restriction of TR_rep to empirical objects: While it is false to assume that the world in general must be the way we necessarily represent it, according to transcendental idealism it is true that the empirical world corresponds to the necessary features of our representing objects (space, time, categories). But then, the thought from the third Critique that Guyer highlights (all laws as necessary come from some understanding that imposes them on its objects) is not an extension, or misapplication, of transcendental idealism, but rather a form of TR, according to which the necessary features of representing objects correspond to the necessary features (laws) of reality.

Finally, Guyer turns to my brief postscript on Kant’s practical metaphysics, including the postulates of pure practical reason. I claim that because the postulates are theoretically undecidable, theoretical reason on its own would have to reject them; but because practical reason makes their acceptance necessary, theoretical reason must ‘give in’. First, Guyer objects that theoretical reason does not have to ‘give in’, because undecidability is not rejection. I take him to mean that theoretical reason does not have to reject the postulates, since given their undecidability it can remain neutral. Thus it does not have to ‘give in’ to practical reason, because it is not opposed to the postulates in the first place.

But I think that this underestimates Kant’s commitment to the traditional evidentialist view, found for instance in the ancient sceptics, Descartes and Hume, that if some proposition is theoretically undecidable, because we do not have conclusive evidence for or against it, the only rational response is suspension of judgement. Thus, considered from a purely theoretical point of view, we cannot rationally accept the postulates (5: 120). In this sense, theoretical reason must give in to practical reason: Since accepting the postulates is rationally necessary, theoretical reason must curb its inherent evidentialist tendency by allowing that we accept the postulates even though, on purely theoretical grounds, they would not be rationally acceptable.

Second, Guyer suggests that Kant’s view of the highest good in the first Critique, according to which it can be realized only in an afterlife, in later works gives way to a more plausible view, according to which the highest good is the telos of historical progress and can be realized, or at least approximated, in this world. I agree that we can find both conceptions in Kant’s work and that the latter is by far the philosophically more attractive.

3. Reply to Andrew Chignell

Andrew Chignell focuses his comments on the formulations of transcendental realism (TR) I introduce in chapter 5 and on my criticism of TR in chapter 9. In sections 2 and 3 of his comments, he asks how the different formulations of TR I work with are related and how they can be derived from Kant’s own formulation of TR as the identification of empirical objects with things in themselves (which he calls ‘TR_things’). I will first offer a general response to Chignell’s concerns and then answer two more specific questions he raises in this connection. Finally, I will briefly address the points he raises in sections 4 (concerning the supreme principle and appearances) and 5 (concerning my critique of TR).

First, on my reading of the Transcendental Dialectic, transcendental realism explains the phenomenon Kant calls ‘transcendental illusion’, which at the most general level consists in mistaking subjective principles and ideas for objective ones (A296–7/B353–4). I base this reading on Kant’s claim that transcendental idealism is the ‘key’ to the resolution of the antinomies (A490/B518), which implies that transcendental realism explains the transcendental illusion involved in the antinomies. Since Kant does not offer any other explanation for the transcendental illusion involved in the paralogisms and the theological arguments, but seems to assume a unified account of transcendental illusion, it is plausible to start from the ‘hypothesis’ (KSM, p. 140) that transcendental realism explains transcendental illusion throughout the Transcendental Dialectic.

In order to make this hypothesis work, I ask what exactly Kant means by ‘transcendental realism’, which he officially introduces at A490/B519 as the identification of, or lack of distinction between, appearances and things in themselves (see also A369). A satisfactory interpretation of what Kant means by transcendental realism (TR) in the context of an explanation of transcendental illusion must satisfy three conditions: (C1) It must explain how the traditional metaphysician, by identifying empirical objects with things in themselves, is committed to TR in the relevant sense. (C2) It must explain how the tacit assumption of TR in that sense can plausibly serve as an explanation for transcendental illusion. (C3) It must be such that TR can plausibly be attributed (as a tacit commitment) to ‘universal human reason’.

In KSM, I argue that the best candidate for playing this threefold role is TR_C, which says that there is a necessary correspondence between the principles of reason and the principles that structure reality. The idea is that, given a tacit assumption of TR_C, one will naturally assume that the subjective principles of human reason (which, according to Kant, can be properly used only regulatively) hold for reality itself (and thus can be used constitutively).

Chignell primarily questions whether TR_C satisfies C1, that is, whether TR_C is suitably related to Kant’s official definition of TR, and he raises doubts about several of the steps by which I arrive at TR_C. He assumes that my aim is to show that TR_C follows from, or can be derived from, Kant’s generic definition of TR as identification of appearances and things in themselves, and he questions the cogency of that derivation. I now see that some of my formulations indeed invite such a reading and I am grateful to Chignell for forcing me to clarify this point. In fact, my aim is not to derive one or more formulations of TR, but merely to explain how Kant could have thought that the identification of empirical objects and things in themselves could commit one to TR_C. Note that Kant does not have to claim that his generic formulation of TR (empirical objects are things in themselves) and TR_C are equivalent; it would be sufficient for my purposes if he held that the two formulations would appear to be equivalent to the traditional metaphysician (or, even weaker, that the traditional metaphysician is committed to TR_C insofar as she does not distinguish between empirical objects and things in themselves).

Now why these two claims should appear to be equivalent requires a complicated explanation (cf. KSM, pp. 140–8), but the main idea is that, in our context, things in themselves are thought of as noumena in the positive sense, which are the objects of a divine perspective, which in turn are the elements of a rational order, thus resulting in the claim that empirical objects are elements of a rational order which necessarily corresponds to the principles of reason. I do not claim that this is a cogent derivation of TR_C from Kant’s generic formulation of TR or even that TR_C is a plausible reading of what Kant meant at A490/B519 by ‘realism in the transcendental signification’, but merely that it was plausible for Kant to assume that in the hands of the traditional metaphysician, and given a set of background assumptions she is likely to accept, TR would amount to something like TR_C.Footnote ⁵ Thus, condition C1 is satisfied.

Moreover, I argue that only TR_C and not TR in its generic form (identification of empirical objects with things in themselves) can explain transcendental illusion (C2). So I take it that TR_C satisfies at least two of the three criteria. Concerning the final condition – TR as part of common sense or universal human reason – I think that it is the most problematic. What needs to be shown here is that there is a tacit commitment to TR_C rooted in human reason (KSM, 148–50). To see that this is not at all implausible, consider that reason is a cognitive faculty, aimed at cognizing reality. Therefore, if reason makes us expect reality to be a specific way, for instance to exhibit a systematic order, we tend to assume that reality really is that way. Of course, according to Kant that assumption is unwarranted, but it appears natural enough. If, as I would like to claim, a tacit commitment to TR_C is our best explanation for this natural tendency, then Kant was right to attribute TR_C to universal human reason.

Thus my general response to the questions Chignell is raising in sections 2 and 3 of his comments is this: it is not my aim to derive TR_C from Kant’s generic formulation of TR, but rather to argue that TR_C is the best candidate for a unified explanation of transcendental illusion throughout the Transcendental Dialectic. I do not think that his concerns undermine that aim.

Let me now turn to two of Chignell’s more specific questions concerning TR. First, he asks whether TR_pos (empirical objects are noumena in the positive sense) is not an obvious non-starter, given that it identifies empirical objects with noumena in the positive sense, which are non-sensible objects. I do not think that this is the case, because Kant’s conception of noumena in the positive sense allows for these objects also to have sensible properties, as becomes clear from the fact that human beings, considered as noumena in the positive sense, have an intelligible character, but also have an empirical character and other sensible properties (cf. A539/B567). In this way, empirical objects could be identical to noumena in the positive sense, which can be fully cognized only from a divine perspective. That is just what TR_pos claims.

Moreover, Chignell asks where the necessity operator in TR_C comes from. TR_C says that there is a necessary correspondence between the principles of human reason and the principles that structure reality. Let me first point out that this does not mean that, according to TR_C, our rational capacities are infallible, as Chignell suggests. After all, the correspondence in question may be necessary, but only partial; moreover, we can make mistakes in following the principles of our own reason (as we do in the case of logical fallacies and, according to Kant, in the case of transcendental illusion), even if these principles should necessarily correspond to reality.

Now one way for a traditional metaphysician to arrive at TR_C (including necessity) is by assuming that God created the world as a rational order and created human reason as a means for us to discover that order. This assumption is the foundational idea behind medieval scholasticism and it was still common in Kant’s own time. An alternative route to TR_C (including necessity) would be the Aristotelian claim that the forms of objects and the forms of true thoughts are identical. Again, it is not my claim that TR_C follows logically from Kant’s generic definition of TR, but merely to explain why the fact that the traditional metaphysician does not distinguish empirical objects from things in themselves could plausibly result in the assumption that the principles of human reason necessarily correspond to reality.

In section 4 of his comments, Chignell offers an answer to a question I raise but do not answer (KSM, pp. 152–6) about why, according to Kant, the ‘supreme principle of pure reason’ (roughly: if there is something R-conditioned, then there is the totality of its R-conditions, which is R-unconditioned) does not hold for appearances, which is a claim Kant is committed to in the context of his resolution of the antinomies. Chignell’s solution rests on the idea that, while appearances are ‘merely phenomenal’ in that all their properties qua appearances are mind-dependent, they nonetheless do not, according to Chignell’s definition of ‘mere phenomenality’, depend on our minds for their existence. This allows for the following solution:

But note, first, that this does not apply to the case of an unconditioned condition (since in that case the totality of conditions is not infinite); and second, even in the case of an infinite totality of conditions, it would only follow that the unconditioned does not exist in the domain of appearances, and not that it does not exist at all. But I take it that what Kant needs to reject is not just a version of the ‘supreme principle’ that allows us to infer, from the existence of something conditioned in appearance, the existence of something unconditioned in the domain of appearances. Rather, he must reject a version that allows us to infer the existence of something unconditioned (without further qualification), since it is that (stronger) version on which the antinomies rely. I do not see how Chignell’s conception of the ‘merely phenomenal’ character of appearances helps us understand how Kant can reject that stronger principle.

In closing, Chignell addresses my claim that transcendental realism (in the form of TR_C) can be rejected without accepting Kant’s transcendental idealism. He objects that my argument is successful only if TR_C contains the (as he sees it) unmotivated necessity operator. He thinks that transcendental realism is in fact true if it is understood as the claim that our rational faculties allow us to cognize reality, albeit fallibly and incompletely. I agree that such a view is unobjectionable. But note that this view does not give rise to transcendental illusion (as Chignell is of course well aware), so it is very different from the kind of view that underlies traditional metaphysics and is criticized by Kant. What matters for the traditional metaphysician, as Kant understands him, is that our rational faculties alone, independently from empirical input and thus completely a priori, allow us to cognize the fundamental structures of reality. As I understand Kant, he thinks that this commits the traditional metaphysician to something like TR_C. It is TR_C, not Chignell’s more moderate version of transcendental realism, of which I argue that it can be rejected without accepting Kant’s own transcendental idealism.

Let me close by repeating that I am truly grateful to Andrew Chignell and Paul Guyer for their comments, which contain more important questions and suggestions than my brief response can do justice to.