1

In his recent book, The Evolution of Modern Metaphysics, Adrian W. Moore takes Kant to play a crucial role in the evolution of modern philosophy.Footnote ¹ Thus, as he notes at the beginning of the chapter on Kant (which follows those on Descartes, Spinoza, Leibniz and Hume, and precedes those on Fichte, Hegel, Frege, the early Wittgenstein, the later Wittgenstein, Carnap, Quine, Lewis, Dummett, Nietzsche, Bergson, Husserl, Heidegger, Collingwood, Derrida and Deleuze):

Although I am unable to do justice to this excellent chapter, I aim to discuss some important aspects of Moore’s critique of Kant. In this way, I also hope to question his view on Kant more generally. In the next section, I briefly present the main elements of Moore’s reading of Kant’s transcendental idealism and attempt some terminological clarification. In section 3, I introduce and reply to some of Moore’s preliminary objections to Kant. Even if my replies are successful, more serious criticism is presented in section 4, where I attempt to deal with three issues. Section 5 concludes this article by considering the likely motivation that prompts Moore’s reading of Kant; my claim is that this reading points to fundamental issues – not of Kant’s transcendental idealism, however, but of the traditional idealism his view has often been taken to represent.

2

According to Moore, Kant’s metaphysics is unsatisfactory (p. 140). In fact, it is not simply unsatisfactory, but ‘ultimately and profoundly unsatisfactory’ (ibid.). Moore defines metaphysics as ‘the most general attempt to make sense of things’ (p. 1). He concludes his discussion of Kant’s metaphysics with the claim that Kant’s most general attempt to make sense of things ‘does not itself, in the end, make sense’ (p. 141). In brief, his argument is the following:

Moore employs the familiar analogy of Kantian spectacles to elucidate the central question of Kant’s critical philosophy, that of the possibility of synthetic a priori judgements (p. 117).Footnote ² He also uses it to elucidate the question of the possibility of ‘real’ metaphysics for Kant. For, as Kant puts it, for us, metaphysics ‘is not concerned merely with analysing concepts that we make of things a priori and thereby clarifying them analytically, but we want to amplify our cognition a priori’ (B18).

Kant of course thinks that the pursuit of real or ‘good’ (p. 123) metaphysics requires his famous Copernican revolution in thinking. Thus, with respect to the intuitional component of cognition, he says:

Moore reads this as follows: when we have knowledge of something that is independent of us (and, he adds, therefore also of something independent of that knowledge), this knowledge is possible if we are ‘given’ the thing, if it affects us in some way; in this case, the way in which we are affected is sensory. But we can be affected or be given something in a particular way only because we have some capacities for reception. Through these capacities, we make a contribution to the form and structure of our experience (pp. 119–20).

This introduces the spectacles analogy: ‘It is as though we have native spectacles through which we view things’ (p. 120). Since these spectacles are native, we can know a priori how things appear through them. Such knowledge is synthetic, since it is not the result of analysing a concept. This knowledge is knowledge of the appearances of things, but derived only from our very capacity for such knowledge. So it is not knowledge of how things are in themselves, independently of our capacity for such knowledge. This amounts to a version of idealism, since ‘the objects of our knowledge, as they are known to us, have a form that depends on the knowledge itself’ (p. 120).

This idealism, Moore continues, ‘is not a matter of what we see through our spectacles’, but ‘a matter of our seeing through spectacles at all’ (p. 120). For instance, that the sun is larger than the moon is part of what we know about objects, but the sun’s being larger than the moon does not depend on our knowing it to be so. Yet both the sun and the moon are spatial objects and, according to Kant, space is part of our spectacles. In this sense, Kant’s version of idealism, transcendental idealism, Moore concludes, refers to the relation of our cognitions to the faculty of cognition, not to the relation of our cognitions to things: the word ‘transcendental’ signifies ‘not our knowledge of objects, but our knowledge of how we know them’ (p. 121).

One may of course wonder how to understand these claims, because, for Kant, what we know, the object, is constituted by elements of our knowledge. So there is one sense in which the object of our knowledge cannot be independent of that knowledge itself. For instance, when we know that the sun is larger than the moon, we know something about spatial objects and, according to Kant, these objects are spatial for us, because our cognitive process constitutes them as spatial.

Here is how we are to look at it. Moore claims that, for Kant, the fact that the sun is larger than the moon does not depend on our knowing it to be so. Contrast this with Moore’s other example: on Kant’s account, things appear through our spectacles as if, for every event in the world, there is a cause; this knowledge is derived from our capacity for such knowledge – it is not independent of that capacity. The difference is not simply that one type of knowledge is of an object that is independent of that knowledge, whereas the other type is of an object that is dependent. We also seem to have two distinct criteria of in/dependence at work: above we said that, for Kant, our knowledge that the sun is larger than the moon does not depend on our knowing it to be so; here, the claim is that our knowledge that any event in the world has a cause does depend on our capacity for knowing it.

The first criterion of in/dependence is relative to our knowing X, whereas the second criterion is relative to our capacity for knowing X. The kind of knowledge that we have when we say that any event in the world has a cause is knowledge of something that does depend on our capacity for knowledge, because (at least if Kant is right) this is knowledge of something that derives from our epistemic capacity. By contrast, our knowledge that the sun is larger than the moon does not derive merely from our capacity for knowledge – we need experience to know that the sun is larger than the moon.

As noted, Moore takes Kant’s ‘idealism’ to refer to some sort of relation of dependence between the objects of our knowledge and our knowledge. I will return to a discussion of this relation in the final section of the article. In the next section, I will begin to present and discuss some preliminary objections to Kant that Moore formulates.

3

Moore focuses next on Kant’s distinction between appearances or phenomena and things in themselves. Phenomena are the objects of our experience as perceived through the spectacles, whereas things in themselves are the objects of our experience independently from our spectacles and, hence, as they are in themselves. One of the central questions Moore raises is whether, on Kant’s account, we can know anything about things in themselves. First, he answers, Kant does not deny that we can have analytic knowledge about things in themselves:

Here Moore makes reference to A258–9/B314–15 to substantiate the claim that Kant does not deny that we have analytic knowledge about things in themselves. I find this puzzling, since at A258/B313–14 Kant seems to deny precisely the possibility of knowledge about things in themselves, whether analytic or not. For instance, for Kant,

Kant denies here the possibility of cognition of objects otherwise than as objects of possible experience. Moreover, he also denies cognition based on analysis only: ‘With us understanding and sensibility can determine an object only in combination’ (A258/B314). The chapter ‘On the ground of the distinction of all objects in general into phenomena and noumena’, where the quotations above are located, denies precisely the possibility of making use of a concept to refer to things in themselves – transcendentally, as Kant puts it.Footnote ³ It is unclear how we can have analytic knowledge about things in themselves, if the concept to be analysed as part of the analytic judgement cannot refer to things in themselves. These are only some worries, but I do not focus much on them, since, as we will see below, there are more challenging objections to consider.Footnote ⁴

Moore is certainly right that Kant makes use of logic to rebut transcendent metaphysicians. I think there are at least two such legitimate uses. One is by pointing to inconsistencies in the reasoning of these metaphysicians. A second is by pointing to claims that go beyond the particular universe of discourse specific for particular contexts – and this is more directly relevant for the arguments of transcendent metaphysicians. For instance, Kant distinguishes between a dialectical opposition (the world as infinite versus the world as finite or non-infinite) and an analytical opposition (the world as infinite and the world as not infinite), and argues that the former adds a determination to the world considered as a thing in itself (namely, its magnitude).

Now, according to Moore, it is not only analytic knowledge about things in themselves that Kant does not deny; he does not even deny synthetic knowledge about things in themselves. Moore offers the following examples: Kant says that it is absurd to claim that there is an appearance without anything that appears (Bxxvi–xxvii, A696/B724, Prolegomena, 4: 350–1), and he makes several claims about what things in themselves are not.

These claims which seemingly contradict what Kant says about the limits of our cognition have of course already been the object of debate in the literature.Footnote ⁵ Kant has been defended as making legitimate use of negative claims concerning things in themselves and as justifying the claim about the necessity of things in themselves as corresponding to appearances only as a problematic necessity, and the things only as possible objects of thinking.Footnote ⁶ Perhaps for this reason, Moore regards these as only potentially problematic: ‘is it simply that we are beginning to witness cracks in Kant’s edifice?’ (p. 134). I am therefore not going to say anything else about these issues, but will move on to the more serious worry Moore expresses in this chapter.

4

Recall Moore’s claim that the only real sense we can make of things is whatever sense we can make of them by looking through our spectacles – call this the Real Metaphysics Claim (henceforth RMC); recall also his claim that we cannot make real sense of the RMC. This means that Moore assumes the RMC is not or cannot be the result of our looking through our spectacles. And yet Kant is quite clear that we can only look at things through our spectacles.

If the RMC is such that we cannot make real sense of it, this must be because of the kind of claim it is. What kind of claim is it? We can see this more clearly by looking at another formulation of the same problem. The question is, again, whether Kant’s position concerning the limits of our knowledge is consistent with some of the claims he makes, which, Moore argues, suggest that we can have knowledge about things in themselves. And yet unlike the previous examples of claims Kant makes about things in themselves, in this case, the contradiction seems to be systematic and affecting Kant’s system as a whole. For, this time, it is the very claim that we can only have metaphysical knowledge in the form of synthetic a priori judgements that Moore suggests goes against Kant’s view that we cannot have synthetic knowledge of things in themselves:

On Moore’s account, from our (human) standpoint, the a priori conditions of our experience cannot be other than they are. They are part as it were of our spectacles. But if they are instruments which make it possible for us to have substantial knowledge, then this knowledge is not to be expected from the instruments (spectacles), but from what we get to learn with their help (what we see through the spectacles). Hence, from the human standpoint, these a priori conditions cannot make a substantial contribution to our knowledge.Footnote ⁷ Nevertheless, Kant does claim that these a priori conditions do make a substantial contribution to our knowledge, because he sees the knowledge as synthetic. If from our human standpoint they cannot make a substantial contribution to our knowledge and, yet, Kant claims they do, then he commits himself to a perspective beyond the human standpoint. It is from this new perspective that our synthetic a priori knowledge would be seen as substantial (p. 139). Moore thinks the following claim by Kant confirms this:

Kant may be thought to confirm here that we have access to a perspective which goes beyond the limitation of sensible intuition. From this perspective, some claims have, for us, universal and unconditional validity. Yet, the ‘us’ here is no longer the group of human beings with the epistemic spectacles provided by a priori intuitions and concepts, since such human beings, on Kant’s own account, have only access to objects of sensible intuition.

Again, there would be a lot to discuss here, but I mention only three concerns. First, does Kant’s acknowledgement that synthetic a priori judgements are constitutive of our experience mean that they cannot make substantial contributions to our cognition? For Kant, one criterion for deciding whether something makes a substantial contribution to our knowledge is the principle of non-contradiction; if denying a judgement leads to a contradiction, then the judgement is analytic and, hence, not substantial. But Kant’s claim is that the necessity associated with these non-substantial judgements is distinct from the necessity of synthetic a priori judgements. Why suppose that the latter must also be non-substantial?

To put this differently: one might perhaps suggest, following Moore, that, as constitutive of experience, a synthetic a priori judgement is already presupposed by experience and, hence, is not informative or substantial. But this is simply not Kant’s view of substantial judgements – Kant does not assume that the only substantial judgements are those which are not necessary for experience; on the contrary, experience is made possible by principles which are synthetic and a priori true. If substantial judgements were only those which were contingent for our experience, then the possibility of synthetic a priori judgements would be denied from the start. It would follow, then, unsurprisingly, that we can talk about synthetic judgements that claim to be a priori only in relation to things in themselves.Footnote ⁸ Kant would seem committed to knowledge of things in themselves in his pursuit of synthetic a priori knowledge. But this self-stultification can easily be avoided, since we have a good Kantian alternative: necessary judgements which are a priori in virtue of their reference to our a priori intuitions.

Secondly, consider Kant’s comment that all outer appearances are next to one another in space; according to Moore, Kant here abandons the human standpoint and regards certain judgements from the perspective of things in themselves. He contrasts there two kinds of validity (A27/B43):

1. - That of the proposition P1: ‘All things are next to one another in space’, which is conditionally valid under the limitation that these things be taken as objects of our sensible intuition.

2. - And that of the proposition P2: ‘All things, as outer appearances or intuitions, are next to one another in space’, which is valid universally and without the limitation above.

Consistently with Kant’s claim that we cannot have knowledge of things in themselves, can we not take him to be saying simply that, by indexing the statement concerning outer appearances being next to one another in space, the statement becomes universal precisely by reference to all appearances and not, self-stultifyingly, with reference to things in themselves?

Let me explain this further. Let us assume, as Moore suggests, that the domain of discourse for the first sentence is that of outer appearances, whereas the domain of discourse for the second is that of all things, including things in themselves. If this were so, then Kant would – perhaps unwittingly – commit himself to a knowledge claim concerning things in themselves, since he would assume a position which transcends the human standpoint. But one obvious question to ask is whether Kant would miss anything important by instead limiting the universe of discourse, as suggested above, to the domain of objects of possible experience? If not, then a charitable interpretation would surely allow this to him.

From his discussion of the two statements, as well as, more generally, from his expositions of space, Kant seems to derive the following conclusion:

Does the conclusion that space is transcendentally ideal presuppose a claim to knowledge of things in themselves? I think that Kant answers this negatively: as the way in which we receive impressions from outer objects, space is a form of one of our epistemic capacities, namely, a form of sensibility; if we consider an object without taking into account the specific way in which it is given to us, then, as it is in itself and independently from us, the object is not spatial; spatiality depends on our epistemic capacities and, hence, according to the definition of transcendental idealism presented in section 2, it is transcendentally ideal. However, this conclusion is derived only by considering things in themselves in the negative sense that Kant specified as legitimate (again, as discussed at the end of section 3): namely, we attribute to things in themselves only properties included in the presupposition that the object has as a property nothing belonging to sensible intuition. If this is correct, then we could read Kant as consistent: his view that the second statement is universally and unconditionally true does not require a perspective which transcends the human standpoint.

In fact it might even be that this ‘charitable’ interpretation is philosophically required in this context. Let us assume that the universe of discourse for P2 is indeed all things – including things in themselves. On one standard reading, what a statement does is to attribute certain properties to the entities which constitute the domain of discourse. It does this by using certain predicates. But there is a complication as to how we should understand this universe of discourse.

As is well known, there are at least two important interpretations of transcendental idealism, which yield distinct views of our universe of discourse.Footnote ⁹ According to the two-world interpretation, things in themselves and objects of experience are numerically different entities. As a result, our universe of discourse will consist of things in themselves (t₁, t₂, …, t_n) and distinct objects of experience (o₁, o₂, …, o_n). By contrast, the two-standpoint interpretation claims that things in themselves and objects of experience are not numerically distinct, but every object of experience (o_n), as conceived independently of our epistemic capacities, is a thing in itself (t_n). In this case, any relevant things in themselves would simply be objects of experience conceived independently from the human standpoint. So we should either regard this universe of discourse as populated by things in themselves or by objects of experience.

Now Moore adopts the two-standpoint account, although not without qualifications (p. 121, n. 36). Moreover, he claims that the second statement includes a commitment to a universe of discourse consisting of things in themselves. Yet this leads to an important difficulty, since, on Kant’s account, we cannot ascribe spatial predicates to things in themselves. Hence, P2 (‘All things, as outer appearances, are next to one another in space’) would not make sense, if by ‘things’ we meant things in themselves. Outer appearances are not things in themselves that have additional spatial properties; things in themselves are things considered independently from our epistemic capacities; as soon as we introduce spatial properties, we adopt the human standpoint and abandon the transcending standpoint of things in themselves. Hence, P2 simply says that ‘Objects of human experience (which may be inner or outer), as outer appearances, are next to one another in space’.

Unless we adopt the two-world interpretation, which Moore (rightly, I think) avoids, outer appearances are a subset of possible objects of experience, but are not a subset of things in themselves. Thus the unconditional and universal character of the truth of P2 confirms that the domain of discourse is given by objects of human experience. For, as Kant says, ‘the source of all truth’ is the source ‘of the agreement of our cognition with objects’ (A237/B296). Yet, within the confines of theory (as opposed to practical philosophy), the only way for us to make sense of the notion of an agreement with objects is within the domain of possible experience. Hence, the passage quoted by Moore is not a confirmation of the self-stultification of Kant’s position.

A third concern is as follows. Moore claims that a problem for Kant’s position is that it cannot make real sense of itself. Given that, for Moore, metaphysics is the most general attempt to make sense of things, an important part of metaphysics has a self-reflective character: it attempts ‘to reflect on one’s own activity, and to try to make sense, in particular, of the sense that one makes of things’ (p. 7). But, just for this reason, the self-reflective account which is obtained in this way has a more general character than the account which explains how to make sense of things, and it is a meta-metaphysical enterprise.

Kant’s attempt to make sense of things is his transcendental idealism. If his meta-metaphysics cannot make real sense of transcendental idealism, then whatever philosophical advantages transcendental idealism may have in its disputes with alternative metaphysical theories will be seriously undermined. As we have seen, one form this issue can take is the suggestion that we cannot make real sense of the Real Metaphysics Claim. A more explicit form was also articulated in the charge introduced above: our metaphysical knowledge is synthetic and a priori; this claim must itself, presumably, count as an item of synthetic a priori knowledge; and yet this seems to be a judgement about things in themselves; if so, then we cannot really account for it.Footnote ¹⁰

But I wonder whether we need to accept that the claim that our metaphysical knowledge is synthetic and a priori is itself a synthetic a priori judgement.Footnote ¹¹ After all, with this additional degree of reflection, we seem to be well anchored above intuition within the realm of concepts, and so one would think within the realm of the analytic. Given that Kant seems to define metaphysical knowledge as knowledge which is both necessary and substantial, it follows that the claim under question cannot be synthetic.Footnote ¹² Even in this case, one important issue may remain, so I add here a final clarificatory comment. Consider Moore’s worry: Kant can account for the synthetic a priori item of knowledge that metaphysical knowledge is synthetic a priori only by making substantial knowledge claims about things in themselves. So far, I have put a lot of emphasis, in my evaluation of his criticism, on the following argument Moore seems to provide against Kant. From our human standpoint, the a priori conditions of our experience cannot be otherwise than they are – they are, as it were, part of our spectacles; therefore they cannot be expected to provide us with substantial, non-analytic knowledge. Yet Kant thinks that these a priori conditions are a basis for synthetic a priori knowledge; so if such knowledge is not possible from the human standpoint, then Kant must be committed to its possibility from a perspective beyond that of the human, and so with regard to things in themselves.

I have explained what I think is wrong with this argument. But I would like to make an observation: insofar as this is the argument on which Moore relies in order to show that the allegedly synthetic a priori claim ‘Metaphysical knowledge is synthetic a priori’ requires substantial knowledge of things in themselves, this is an argument which applies to any synthetic a priori claim whatsoever – whether it is a second-order claim about our metaphysical knowledge or a first-order claim about objects of our knowledge. One concern that remains is that, in this way, I do not capture the significance that Moore attributes to the fact that Kant’s problem has to do specifically with the second-order claim that we have metaphysical knowledge. Hence, the remaining worry is the following: when I reflect on the truth of my claim that metaphysical knowledge is synthetic a priori, because I do not make reference to objects of my knowledge, but talk about knowledge in general, I seem to need a perspective which is beyond the a priori conditions that make objects of knowledge possible for me, that is, beyond the human standpoint.

Consider again the example: according to Moore, P1 is true for outer appearances, whereas P2 is true universally. As we have seen, however, for Kant, certain claims we can make about appearances do not make sense when applied to things in themselves. For instance, whereas it makes sense to say that all appearances, as outer appearances, are next to one another in space, it makes no sense to say that all things (including things in themselves), as outer appearance, are next to one another in space.

In the Amphiboly, Kant states explicitly that ‘the representation of an object as a thing in general is not merely insufficient but rather, without sensible determinations of it and independent of an empirical condition, contradictory in itself’ (A279/B335). Kant draws a distinction between a logical and a transcendental consideration of a judgement. The first looks at the judgement’s concepts independently of the objects they refer to and, hence, independently of the cognitive faculty for which they can be an object; the second, by contrast, looks at the concepts as referring to particular objects, which can either be merely objects of pure understanding or objects of sensibility. Hence,

Thus with regard to the second-order judgement that Moore takes to be problematic, a logical consideration would only regard it as a relation between concepts without concern for what these concepts refer to – whether appearances or things in themselves. From this perspective, we could perhaps say that it is analytic or synthetic. If, by contrast, we consider it transcendentally then we would need to think it under conditions of sensible intuition. If we were to regard this judgement as referring to things in themselves, then not only could we not have synthetic knowledge of these things in themselves, but assuming we had, we would not be able to regard this as a condition for the truth of this judgement as applied to appearances, as Moore suggests would be entailed by asserting the synthetic a priori truth of the judgement that metaphysical knowledge is synthetic a priori. For, according to Kant, ‘even if we could say anything synthetically about things in themselves through pure understanding (which is nevertheless impossible), this still could not be related to appearances at all, which do not represent things in themselves’ (A276/B332).

With this, I conclude the discussion of Moore’s objection of self-stultification. In the next and final section, I suggest one possible motivation for Moore’s objection.

5

In section 2 we saw that, according to Moore, Kant’s idealism refers to a relation of dependence between the objects of our knowledge and our knowledge of these objects. However, Kant’s transcendental idealism must be carefully distinguished from traditional idealism, a view from which Kant tried to distance himself, but with which he has been associated by critics ever since he published the first Critique. Transcendental idealism claims that the objects of our knowledge do not depend on what we know and so on ‘our knowledge’ of them, but on how we know them. In other words, the objects depend on the forms of our sensibility and on the categories of our understanding – they are constituted by a priori intuitions and concepts – but they are not reducible to them. Thus as we have seen, synthetic a priori judgements refer to facts which depend on our epistemic capacities, while empirical judgements are judgements about objects independent from our knowledge. By contrast, it is traditional idealism (usually associated with Berkeley) that takes our objects of knowledge to depend on our knowledge of them and, hence, to be reducible to them.

Contrast these forms of idealism with transcendental realism, according to which we can have synthetic knowledge of things in themselves. According to this position, the objects of our knowledge (things in themselves) are independent from our knowledge of them, since by definition things in themselves are things as they are independently of our epistemic capacities. Moore’s criticism of Kant thus relies on the claim that we must claim synthetic knowledge of things in themselves, and so be transcendental realists, if we are to be able to account for a central claim of transcendental idealism, namely that metaphysical knowledge is synthetic a priori.Footnote ¹³

Thus Moore’s subtle reading of Kant is cast in the form of an objection to Kant’s inconsistency, philosophically the strongest type of objection one can formulate. Yet, if my replies are correct, then perhaps Moore’s objections do not really point to the inconsistency of transcendental idealism, but to some strong reasons why traditional idealism is unacceptable. As we have seen, Moore’s argument is that, on Kant’s account, we should and can have access to things in themselves, since it is only by reference to this access that we can make sense of our alleged metaphysical knowledge of phenomena. This suggests that Moore regards Kant’s transcendental idealism as potentially collapsing into a traditional version of idealism.

To be sure, as Moore also notes (p. 141), this is an interpretation of Kant that has a long history and very powerful representatives.Footnote ¹⁴ If it turns out this reading of Kant is correct, then Moore’s argument concerning the possibility and need for synthetic knowledge of things in themselves may be supported indirectly, as a critical reaction to traditional idealism and an attempt to rescue transcendental idealism from collapsing into it.

Yet, although Moore’s criticism may be well motivated in this way, I have argued that it is eventually unwarranted. The starting point of Moore’s discussion is the question of the cogency of Kant’s metaphysics; in response I have argued that two of his arguments in support of the claim that, on Kant’s account, we can only account for our metaphysical knowledge as synthetic a priori from a perspective which transcends the human perspective are not warranted. If this is right, then the arguments are not sufficient to show that Kant’s transcendental idealism is, as Moore claims, unsatisfactory.Footnote ¹⁵