I am extremely grateful to both Sorin Baiasu and Edward Kanterian for the care and generosity with which they have engaged with the chapter on Kant in my book. Each of them raises concerns about my contention that Kant, in distinguishing between good metaphysics and bad metaphysics, finds himself embroiled in what by his own reckoning must count as the latter. I think their concerns take us to the very crux of the matter.

Let us begin with the idea that there is synthetic a priori knowledge, an idea on which of course so much of Kant’s edifice rests and which I think serves as an epitome of his predicament. Consider an item of knowledge that Kant would count as a case in point, say a geometrician’s knowledge that the combined number of vertices and faces of a convex polyhedron is two more than the number of its edges. And consider how Kant would argue for its syntheticity. He would try to show that mere scrutiny of the concepts involved, however thorough, could never, by itself, yield any such insight. This is reminiscent of the way in which a contemporary philosopher might argue for the extra-logical character of the knowledge. Such a philosopher might first identify what he or she took to be its logical form, expressed as some schema of the predicate calculus, and then show that mere scrutiny of this logical form could never, by itself, yield any such insight. So far, so similar. But there is a crucial difference. The logical form of the knowledge does not determine its content. The concepts involved in the knowledge, by contrast, do determine its content. That is, they determine how things must be for the knowledge to be true. Where the contemporary philosopher could proceed by providing some false reinterpretation of the logical schema that had nothing to do with geometry, Kant would have to reckon with some alternative to this very truth about spatial entities, either an alternative in which spatial entities are otherwise or an alternative in which there are no spatial entities at all. For if no such alternative existed – if no such alternative existed, mind, not just if no such alternative were realized – then no such alternative would remain to be ruled out after the concepts involved had been duly scrutinized. And how does that fall short of saying that mere scrutiny of the concepts involved could yield insight into this truth? But here is the rub. Given that Kant would also take the geometrician’s knowledge to be a priori, and given his conception of a priori knowledge, he is committed to thinking that, from the human standpoint, no such alternative does exist. It follows that, by his own lights, he could not show that the knowledge in question is synthetic without abandoning the human standpoint. That is, assuming he could not show that the knowledge in question is synthetic by adopting the standpoint of some other creature instead, he could not do so without entertaining thoughts, however tenuous, about things in themselves.Footnote ¹

This might be all very well, if these thoughts did not constitute synthetic a priori knowledge in their own right. But I claim in my book that, for Kant, they do. This means that he is committed to the existence of synthetic a priori knowledge about things in themselves, the very thing that he repudiates, the very thing, indeed, whose vain pursuit he would count as a prime symptom of bad metaphysics (cf. Critique of Pure Reason, Introduction, §III). It follows that Kant’s account of synthetic a priori knowledge, so far from helping him to keep bad metaphysics at bay, as it is intended to, has exposed him as a practitioner of it.

Baiasu has several concerns about this onslaught on Kant. One is a concern about my suggestion that Kant’s commitment to thoughts about things in themselves would be all very well if these thoughts did not constitute synthetic a priori knowledge. This is a comparatively minor concern, since it leaves the onslaught itself unchallenged. But it still merits a response from me. My reason for making this suggestion is not that I take Kant to accede to the possibility of thoughts about things in themselves that do not constitute knowledge. I do take Kant to accede to that possibility, as indeed do most commentators, but in this context I am more concerned with another possibility to which I take him to accede: the possibility of thoughts about things in themselves which, though they constitute knowledge, constitute knowledge that is analytic. I am more concerned with this possibility in this context since I take it to be a consequence of Kant’s approach to these issues that the thoughts in question do in fact constitute knowledge. In Baiasu’s lengthy note 4 he expresses reservations about whether Kant accedes to the possibility of analytic knowledge about things in themselves. In support of my contention I cite the final paragraph of the chapter on the distinction between phenomena and noumena in Critique of Pure Reason (A258–9/B314–15).Footnote ² But Baiasu urges that the immediately preceding paragraph supports its denial.

Among the many points to be made in response to this, the main one is that I do not intend ‘knowledge about’ in a very robust sense.Footnote ³ In the paragraph that I cite Kant says that an analytic assertion ‘takes the understanding no further, and since it is occupied only with that which is already thought in the concept, it leaves it undecided whether the concept even has any relation to objects’. I intend ‘knowledge about’ in a sense weak enough to conform with this. Thus I take the knowledge that aunts are female to be analytic knowledge about aunts, the knowledge that mermaids have fishes’ tails to be analytic knowledge about mermaids, and the knowledge that things in themselves are things irrespective of how they are given to us to be analytic knowledge about things in themselves. In the preceding paragraph, the one that Baiasu cites, Kant is talking about an exercise of the understanding that ‘determines an object’, something that requires the cooperation of sensibility. In the examples of analytic knowledge above there is no such determination.

A second concern that Baiasu has is that the knowledge to which I take Kant to be committed might anyway be analytic, in which case I have no reason, on my own account, to think that the commitment is a problem for him. Here Baiasu is picking up on the way in which I introduce the supposed problem. I claim that ‘[the very judgement that our metaphysical knowledge … is synthetic and a priori] must itself, presumably, count as an item of synthetic a priori knowledge’ (Moore Reference Moore2012: 138). Baiasu suggests, in opposition to this, that Kant defines metaphysical knowledge in such a way that the judgement that our metaphysical knowledge is synthetic and a priori counts as analytic (p. 476).Footnote ⁴

But the analyticity to which Baiasu draws attention – if that is what it is – does not trouble me. My claim that ‘[the very judgement that our metaphysical knowledge … is synthetic and a priori] must itself, presumably, count as an item of synthetic a priori knowledge’ is ambiguous. It can be read de dicto or it can be read de re. The analyticity would be a threat to it only on the de dicto reading. But it is intended de re. My thought is this: the judgement, of our metaphysical knowledge, that it is synthetic and a priori, must itself count as an item of synthetic a priori knowledge.

Baiasu has a third concern, which is by far the most important and the most substantial. This relates to the idea that it is impossible, on Kant’s view, to make such a judgement without abandoning the human standpoint and hence without entertaining thoughts about things in themselves. I quote a passage from Critique of Pure Reason in which I claim that Kant ‘all but concedes’ this (Moore Reference Moore2012: 139; the quoted passage is at A27/B43). Baiasu proposes a very different reading of this passage (p. 473). Where I take Kant to be contrasting a claim that we can make, from the human standpoint, about outer appearances with a claim that we can make, from no such standpoint, about things in themselves, Baiasu suggests that Kant is contrasting a claim that we can make, from the human standpoint, about outer appearances with a claim that we can make, still from the human standpoint, about all appearances – inner as well as outer.

If the passage were taken in isolation, this ingenious interpretation would certainly look at least as compelling as mine. But in the larger context from which the passage is taken I think the interpretation is simply unsustainable. As Baiasu correctly anticipates, I see a problem for the interpretation in the following sentence, later in the same paragraph:Footnote ⁵ ‘Our expositions … teach us the reality … of space in regard to everything that can come before us externally as an object, but at the same time the ideality of space in regard to things when they are considered in themselves through reason’ (A27–8/B44, emphasis in original). Baiasu has two responses to this. One is that the sentence in question can be read in a way that supports his own more charitable exegesis. The other is that it cannot be read in a way that supports mine.

It can be read in a way that supports his own exegesis, Baiasu claims, because it can be understood as adverting to things in themselves only ‘in the negative sense that Kant [specifies] as legitimate’, namely the sense whereby ‘we attribute to things in themselves only properties included in the presupposition that the object has as a property nothing belonging to sensible intuition’ (p. 474). I have two counter-responses to this. First, this reading still casts doubt on the idea that in the earlier passage Kant is concerned with some contrast between outer appearances and appearances more generally. Secondly, and perhaps more significantly, even if, in this sentence, Kant is to be understood as adverting to things in themselves only in the negative sense in question, that is already, I think, bad enough. This relates to the material at the very end of section 3 of Baiasu’s essay, where he himself raises the question whether our making negative claims of this kind about things in themselves might already offend against Kantian scruples. I raise the same question in my book and go on to ask whether we are beginning to witness cracks in Kant’s edifice (Moore Reference Moore2012: 134). Baiasu cites this passage and suggests that I am highlighting what I take to be a potential problem for Kant. In fact I take it to be more than a potential problem for Kant. I do not see how we can make even negative claims of this kind from the human standpoint. From the human standpoint, no object ‘has as a property nothing belonging to sensible intuition’.

What then of the claim that the sentence in question cannot be read in a way that supports my more hostile exegesis? This is based on the idea that the very appeal to spatial properties must, for Kant, be made from the human standpoint. Of course, at one level, this idea need not concern me. It is open to me to attribute the dissonance, not to faulty exegesis on my part, but to faulty philosophy on the part of Kant, whom I can always accuse of internal inconsistency. However, the point that Baiasu is making is an extremely interesting and important one, and it would be completely unsatisfactory to let the matter rest there. In fact, ironically, I want to defend Kant against this particular accusation of internal inconsistencyFootnote ⁶ and suggest that an appeal (of sorts) to spatial properties can be made from beyond the human standpoint.

For these purposes I shall assume something like what Baiasu calls, in line with recent common practice, ‘the two-standpoint interpretation of things in themselves’. I shall make this assumption not just because that is the kind of interpretation that I favour but also because, as Baiasu himself urges, it is the kind of interpretation which, in this context, makes most trouble for me. For it really does seem to cast the sheer use of spatial vocabulary as enough to mark someone as speaking from the human standpoint. But here I invoke an analogy. Imagine a B-theorist in the philosophy of time who claims that the sheer use of the words ‘past’, ‘present’ and ‘future’ is enough to mark someone as speaking from a particular temporal point of view. I think we can readily make sense of such a theorist if he or she also claims:

and denies that that claim (P) is from any temporal point of view. There are various ways of absolving such a theorist of any internal inconsistency, perhaps the most straightforward of which would be to insist that, contrary to appearance, (P) does not involve the use of the words ‘past’, ‘present’ and ‘future’, but rather involves something akin to their mention: that is, it calls attention, if not to the words themselves, then to their use, or to the concepts that they are used to express, or some such. I do not deny that there is far more to be said about this, and in fact I shall say some more about it towards the end of this essay. But unless there is reason to suppose that the B-theorist is in internal trouble, this suffices, I believe, to indicate how my interpretation of Kant’s sentence can survive Baiasu’s objection to it.

This whole discussion relates to a fundamental problem that I see for Kant’s transcendental idealism, encapsulated in what I call in my book the Limit Argument: it is impossible for us to draw a limit to what we can make sense of, in the way in which the transcendental idealist attempts to, because this requires that we make sense of the limit, which in turn requires that we make sense of what lies on both sides of it (Moore Reference Moore2012: 135).Footnote ⁷ The form that the transcendental idealist’s attempt to draw such a limit takes is to identify a priori conditions of our knowledge and to cast them as conditions of what we can know, in such a way that what we can know can be seen to depend, for some of its essential features, on our capacity to know it. The form that the problem takes, as I think we have just been witnessing, is that this leaves the transcendental idealist with no satisfactory account of how we assimilate transcendental idealism itself: of how we see this dependence of what we can know on our capacity to know it, given the very limit that the dependence is supposed to impose. Baiasu is quite right to complain about some incautious formulations of Kant’s transcendental idealism in my book whereby I suggest that the dependence concerns what we actually know, not just what we can know (§2). I thank him for this corrective. But I hope and believe that, in each case, a more careful formulation is available that does not conflict with anything else I say. At any rate, this fundamental problem for any attempt to draw a limit to what we can know – or to what, in one good sense of the phrase, we can make sense of – is what seems to me pivotal to any critique of Kant’s critique. This is a good cue for me to turn to Kanterian’s essay.

One thing that I think both Baiasu and Kanterian help us to appreciate is the extent to which Kant’s critique is an attempt to expose difficulties that we confront when we become self-conscious about our own limitations. This self-consciousness, Kant believes, leaves us dissatisfied, with the result that we try to transcend those limitations – precisely what we cannot do. But Kant’s critique is itself, of course, grounded in self-consciousness about those same limitations. In a way, then, I am out-Kanting Kant. I am suggesting that his own self-consciousness about those limitations has led him to try to transcend them. In so far as good metaphysics is a matter of maintaining a healthy self-consciousness about one’s own limitations, and bad metaphysics a matter of lapsing into an unhealthy self-consciousness about them, this relates back to what I said in the opening paragraph of this essay: Kant, precisely in distinguishing between good metaphysics and bad metaphysics, finds himself embroiled in the latter. And it is plain that, if we are not to suffer the same fate in our efforts to diagnose his problem, we must proceed very carefully indeed.Footnote ⁸

One of the principal aims of my book is to emphasize some of the deep similarities between Kant and Wittgenstein. Here already is a clear case in point. Just as Kant has a distinction between good metaphysics and bad metaphysics, so too Wittgenstein has a distinction between good philosophy and bad philosophy. (This is true of both the early Wittgenstein and the later Wittgenstein, but for the time being my focus will be on the latter.) Roughly speaking, bad philosophy is what results when self-consciousness about our own language leads us to misconstrue its grammar and tempts us to transgress some of the limitations imposed on our sense-making by that grammar, while good philosophy is the attempt to counteract bad philosophy. But this means that good philosophy must involve self-consciousness of just the same sort as the bad variety. It must involve the same attention to our language, the same observation of the concepts expressed through the grammar of our language, the same experimentation with those concepts, the same prodding and stretching of them.Footnote ⁹ And this in turn means that, if our attempts to practise good philosophy are not simply to issue in more bad philosophy, we must proceed very carefully indeed. The similarity with Kant is striking.

Proceeding very carefully indeed, for a Wittgensteinian, will be largely a matter of scrutinizing the concepts we use and trying, with as much clarity as possible, to articulate the results. The aim will be to display the grammar of our language while importing nothing into the exercise beyond what can be got out of it. And this is reminiscent of the process whereby, for Kant, we acquire analytic knowledge. The question naturally arises, then, to what extent acquiring analytic knowledge could likewise serve for a Kantian as a staple of proceeding very carefully indeed. For Kant himself, of course, to no real extent at all. This is because Kant sees good metaphysics as the pursuit of synthetic knowledge. Scrutiny of our concepts can assist in this pursuit, but it can never be a staple of it (cf. A191–2/B152).

Kanterian asks how Kant might reply to those many philosophers, both before him and after him, whose conception of their own metaphysical practice suggests that they would regard the acquisition of analytic knowledge as having much greater import than that. He makes a number of interesting suggestions about what Kant would say vis-à-vis these philosophers (pp. 485–6). I especially approve of his suggestion that Kant would accuse them of treating a priori concepts of reason, freed of whatever apparatus allows them to be knowledgeably applied to objects of possible experience – what Kanterian calls ‘ideas’ (p. 486) and what I call ‘ideas of reason’ (Moore Reference Moore2012: 126) – as though they could be knowledgeably applied to other objects instead.

It would lie beyond the scope of this essay to dwell further on Kanterian’s suggestions. There is one matter, however, on which I cannot resist a comment. I agree with Kant that mere scrutiny of our concepts cannot take us very far in practising good metaphysics, certainly nowhere near as far as we can go. But I disagree with him that good metaphysics is the pursuit of what he would classify as synthetic knowledge. This is for reasons that I have already indicated. The idea that good metaphysics is the pursuit of synthetic knowledge, combined with the idea, to which I subscribe, that it is an essentially a priori exercise, eventually undermines itself. (This is quite apart from any reservations that we might have about whether the distinction between analytic knowledge and synthetic knowledge can be satisfactorily drawn in the first place, an issue about which I have said nothing.) What then do I offer in its place? Well, one thing that I offer in its place is an enterprise that is neither the mere scrutiny of concepts nor their implementation in discovering how things are, neither the pursuit of analytic knowledge nor the pursuit of synthetic knowledge if one wants to put it in those terms – an enterprise that is not the pursuit of knowledge at all. I have in mind the creation of new concepts. That this is one of the principal aims of good metaphysics is itself one of the principal leitmotifs of my book.Footnote ¹⁰

But what about bad metaphysics? If bad metaphysics is in large part an attempt to breach barriers that are unbreachable, that is to say if bad metaphysics is in large part an attempt to do the impossible, and if, furthermore, it is our task, or one of our tasks, as good metaphysicians, to expose and diagnose instances of bad metaphysics, then we had better be clear about what it is to attempt to do the impossible. This is an important issue that Kanterian raises on pp. 487–8. of his essay. I have a little to say about this in my book. In particular, I remark on the way in which centuries of attempts to trisect an angle with ruler and compass testify to the possibility of attempting to do the impossible (Moore Reference Moore2012: 4).Footnote ¹¹ But simply to know that it is possible to attempt to do the impossible is not of much avail when it comes to trying to identify instances of the phenomenon. Worse still, there is the further complication, to which Kanterian alludes, that, in so far as the impossibility in metaphysics is a kind of unintelligibility, then it looks as though the description of what the bad metaphysician is trying to do, in any given case, must itself be unintelligible. But this suggests that we must participate in the bad metaphysician’s error in order even to characterize it. If this is right, it provides a further graphic illustration of how the self-consciousness that is integral to good metaphysics, in its battle against bad metaphysics, is in constant danger of fostering more of the (bad) same.

This too is something that I discuss in my book. My discussion connects with Wittgenstein once again, though this time specifically with the early Wittgenstein. A very similar predicament attends the early Wittgenstein’s attempt to draw the limits of reality, the limits of all that can be thought or said (Wittgenstein Reference Wittgenstein1961: passim). The predicament concerns how to combat a metaphysician who, in one way or another, tries to contest these limits. Suppose, for instance, that a metaphysician says, ‘Greenness is part of reality’. For Wittgenstein, that cannot be right. Since reality is all that can be thought or said, it must consist of facts, not of things (see Wittgenstein Reference Wittgenstein1961: 1.1). That grass is green is part of reality, because it is possible to think and to say that grass is green; but greenness is not part of reality, nor indeed is grass, because there is no such thing as either thinking or saying either greenness or grass. But the predicament is clear. In so far as a construction such as ‘thinks greenness’ is nonsense, then so too is a sentence such as:

Or, to put the point in a way that is itself no doubt under the sway of the metaphysician’s illusion: if there is no such thing as either thinking or saying something, then neither is there any such thing as either thinking or saying that there is no such thing as either thinking or saying that thing (cf. Moore Reference Moore2012: 244).

Now I promised earlier that I would return to my discussion of the sentence (P). And this is where it is once again relevant. For there are clear affinities between (G) and (P). Or rather, more specifically, there are clear affinities between (G) and the second clause of (P). This is reflected in the fact that the second clause of (P) might naturally be cast in the form: ‘There is no such thing as an event’s being past, present or future considered in itself.’ When I introduced (P), it was in the context of my suggestion, to which Baiasu objected, that Kant has to see himself as employing spatial vocabulary beyond the human standpoint. If part of Baiasu’s objection to this suggestion is that it is tantamount to the suggestion that Kant has to see himself as making play with a kind of nonsense, then I think that the affinities that we have just noted, between (G) and the second clause of (P), help to reinforce it. But I also think that I can stand by my earlier attempt to deflect the objection. And, concomitantly with that, I think that my earlier attempt to deflect the objection can help us to see a way through the current predicament too. Making play with nonsense (one example of which, but by no means the only example of which, is mentioning it) does not entail talking nonsense, and is certainly not a bar to communicative success.Footnote ¹²

This is also relevant to another very important part of Kanterian’s discussion. Often what is going on, when we engage in bad metaphysics, is that we are mistaking a legitimate regulative use of a concept for a legitimate constitutive use of it – where the mistake is compounded by the fact that the concept in question does not have a legitimate constitutive use. I talk a little about this in §7 of my chapter on Kant. But Kanterian considers the case where the reason why the concept does not have a legitimate constitutive use is that it is confused; and he raises the question of how, in that case, even a regulative use of the concept can be legitimate. As he puts it, ‘to make sense of an “as if X” construction, we need to be able to make sense of “X”’ (p. 490). Now although this is clearly of a piece with what we have just been considering, it may be altogether less obvious how the idea of making play with nonsense might help. But I think it does. It seems to me that phrases of the form ‘to proceed as if X’ can be used to refer to exercises of knowledge. It also seems to me that often, when they are so used, the knowledge in question is not knowledge that anything is the case. Rather it is inexpressible knowledge of how to cope with a range of situations in a certain way. The question then arises: what equips a phrase of that form to be used to refer to an exercise of that knowledge? And the answer, I submit, is: the fact that, if one were to attempt (unsuccessfully) to put the knowledge into words, then what would result would be whatever expression occupies the place of ‘X’ in the phrase. But this does not require that whatever expression occupies the place of ‘X’ in the phrase make sense.

This obviously merits much more extensive discussion.Footnote ¹³ Suffice to comment in conclusion that Kanterian, like Baiasu, has directed our attention to an aporia that seems to me fundamental to any attempt such as Kant’s to reckon with the limits of sense-making – and therefore fundamental to any attempt such as ours to reckon with any attempt such as Kant’s to reckon with the limits of sense-making. The remarks towards the end of this essay are intended to indicate that this aporia is as relevant to the work of Wittgenstein as it is to the work of Kant. And this in turn provides me with a welcome opportunity to reaffirm my conviction, which I try to develop throughout my book, that a good understanding of each of these two great thinkers cannot fail to be reinforced by a good understanding of the other.