Justifying Kant’s Ideality Thesistpa.hss.nthu.edu.tw/committee/tpaseminar/2002/200208.pdf · 3 diagnosing Kant’s epistemological argument: his final assessment of the ideality

1

Justifying Kant’s Ideality Thesis Terence Hua Tai

Department of Philosophy, National Chung Cheng University, Taiwan Sun Yat-Sen Institute for Social Sciences and Philosophy, Academia Sinica, Taipei

October 10, 2002

§1. Two Recent Interpretations: Allison and Guyer

It is an all too familiar thesis put forward by Kant in the Critique of Pure Reason1 about the limitation of human cognition that we can have cognition of “appearances” but never of “things in themselves.” But it has been unclear how this distinctively Kantian thesis (the “limitation thesis,” for short) is to be vindicated in the Critique, or even how it is to be understood. Worse still, the thesis is intricately tied up with another equally Kantian but even more difficult thesis, first stated in the Aesthetic, about the so-called “transcendental ideality” of space and time, a thesis about how things in themselves must be acknowledged to be from a certain standpoint (see A28/B44, A36/B52). For it is precisely things in themselves toward which the limitation thesis requires us to take up an attitude of epistemological modesty. Yet, in its barest, Kant’s ideality thesis seems to say, and has been generally interpreted as saying, that things in themselves are not in space and time. Readers of the Critique have been puzzled over how, after yielding to the unpretentious stance advocated by the limitation thesis, we can ever be in a position to claim knowledge of things in themselves that they are neither spatial nor temporal. One might wonder how the glaring contradiction between the two theses could escape Kant’s eyes.

Commentators have attempted different ways to deal with this tension in Kant’s doctrine, and it may be hard to classify them. But it seems safe to say that there are “compatibilists” on the one hand who seek to dispel the appearance of contradiction between the two theses, and “incompatibilists” on the other who contend that Kant is at most entitled to the limitation thesis which in fact precludes the kind of knowledge asserted by the ideality thesis about things in themselves. Still other commenta-tors—call them “neutralists”—are inclined to a Strawsonian approach, suggesting that, whether the two theses are compatible or not, the limitation thesis can be “dis-entangled” from Kant’s transcendental idealist indulgence. In this paper I shall pro-pose an interpretation that ventures not only to reconcile the ideality thesis with the limitation thesis in a way that has gone unnoticed, but also to defend it against some unsympathetic and formidable criticisms. Problems engaging other interpretations and difficulties confronting them are instructive. So, to begin with, let me go over some prominent examples. I shall focus this section on compatibilists and incompatibi-lists—the latter are in fact opponents of the ideality thesis, while the former may or may not be. A recent Strawsonian neutralist will be discussed in §3.

Henry Allison in Kant’s Transcendental Idealism and Paul Guyer in Kant and the Claims of Knowledge are both compatibilists, but they differ in important ways. Allison interprets the ideality thesis as an analytic truth, but Guyer interprets it as

1 Quotations in this paper from the Critique of Pure Reason will follow Paul Guyer and Allen Wood’s translation (1998). “A” and “B” will be used to stand for the pagination of the first and second editions respectively.

2

synthetic a priori. This difference leads to further differences concerning how the ideality thesis is to be justified, and how it is related to the limitation thesis.

According to Allison, the ideality thesis is “really a consequence of Kant’s claim that space and time are epistemic conditions” (Al:114);2 moreover, “[Kant’s] doctrine that we can know things only as they appear, not as they are in themselves” (the limitation thesis) is interpreted as equivalent to the claim that our cognition is subject to “epistemic conditions” such as space and time—certain “universal and necessary conditions in terms of which alone the human mind is capable of recognizing some-thing as an object at all” (Al:9). Allison subsumes these conditions under two groups: the “sensible” and the “intellectual.” For him, the Aesthetic and the Analytic are intended to yield the conclusion that (representations of) space and time on the one hand and the categories on the other are, respectively, sensible and intellectual conditions of our objective cognition. As interpreted by Allison, this conclusion leads at once to two corollaries: first, the limitation thesis, which is equivalent to it; and second, the ideality thesis, which is “really a consequence of Kant’s claim that space and time are epistemic conditions.” If so, the ideality thesis is really a consequence of the limitation thesis, and the two theses must be perforce compatible.

According to Guyer, the ideality thesis is synthetic a priori, and it is to count, not as “first-order synthetic a priori knowledge by means of intuitions and concepts,” but as “philosophical knowledge by means of argument” (G:336). This enables Guyer to account for how the limitation thesis can be compatible with the ideality thesis: there is no inconsistency in saying both that things in themselves are unknowable by means of intuitions and concepts, and that we have an item of “philosophical knowledge,” attainable by means of argument, that things in themselves are neither spatial nor temporal. But if the ideality thesis is a kind of “philosophical knowledge by means of argument,” what argument does Kant have in support of it? Guyer adduces three “arguments”: the metaphysical, the theological, and the epistemological, the third being “the most fundamental” (G:354) in that it contains “Kant’s deepest reason” for the ideality thesis (G:349). Further, in Guyer’s view, the ideality thesis is not estab-lished after the limitation thesis, but before it: Kant maintains the latter thesis “because of his prior acceptance of the premise that [things in themselves] are not spatial and temporal” (G:336; emphasis added).

Thus, although both Allison and Guyer see no discrepancy between the two theses, they propose diametrically different accounts of how they are to be justified: for Allison, the ideality thesis is a consequence of space and time being “epistemic condi-tions,” and to say that space and time are such conditions is already part of what it means to say that “we can know things only as they appear, not as they are in them-selves” (the limitation thesis); Guyer then turns the order of proof the other way round, holding that Kant arrives at the limitation thesis by appealing to the ideality thesis, synthetically rather than analytically construed, and that Kant has argued for the latter thesis independently of the former.

Allison and Guyer come up with opposed evaluations of the ideality thesis which in turn lead to opposed evaluations of the limitation thesis. Whereas Allison tries to defend the ideality thesis, Guyer tries in the end to demolish its grounds, especially by

2 Throughout this section, page references to Allison (1983) and Guyer (1987) in the text will be put in parentheses and preceded by “Al” and “G” respectively.

3

diagnosing Kant’s epistemological argument: his final assessment of the ideality thesis is that it is “the consequence of an exceedingly immodest interpretation of the necessity of synthetic a priori propositions” (G:369; emphasis added). As mentioned just now, Guyer holds that Kant maintains the limitation thesis because of his prior acceptance of the ideality thesis. His unfavorable assessment of the ideality thesis threatens Kant’s grounds for the limitation thesis as well. Guyer goes so far as to argue that Kant’s “synthetic a priori theory of experience” itself may even be taken to support certain beliefs about things in themselves. As he puts it in the form of a query,

Why does [Kant] suppose that the genuine spatiality of things in themselves is incompatible with our a priori knowledge of spatiality rather than being our best explanation of it, given our a priori knowledge that spatiality is a necessary condition on our perception of objects? [G:363; original emphases]

Not only does the outcome of Kant’s transcendental investigation in the Aesthetic fail to set up boundaries of human cognition of objects, but it may even be used to justify an inference to the “best explanation” in terms of “the genuine spatiality of things in themselves”—of how things in themselves really are.

Guyer asks the question cited above in the course of considering how Kant is to rebut the so-called “neglected alternative objection” (or NAO, for short). Kemp Smith formulates this objection as follows:

Kant recognizes only two alternatives, either space as objective is known a posteriori, or being an a priori representation it is subjective in origin. There exists a third alternative, namely that although our representation of space is subjective in origin, space is itself an inherent property of things in themselves.3

The NAO assumes it to be logically possible that things in themselves may be in space and time. One who raises the objection might argue that the ideality thesis is incompatible with realization of such a logical possibility. This seems to be Stephan Körner’s argument:

It is always logically possible that what we perceive under the form of space and time is so ordered independently of our perception. It is quite possible that what a person sees through his irremovable spectacles as, let us say, pink, is also pink in fact, and would be seen so even if per impossibile the spectacles were removed.4

For Körner, the limitation thesis implies that our “spectacles” are irremovable, which in turn implies that we can never know whether or not the logical possibility in ques-tion, alleged by the NAO to be always there, is realized. But, as Körner might con-tinue, if realization of the logical possibility cannot be ruled out, the ideality thesis is false. One who argues in this way is an incompatibilist about the two theses.

Allison thinks Kant has in fact ruled out the logical possibility invoked by the NAO: given that space, for example, is a form of our sensibility, it cannot be claimed that “it is possible both for space to be such a form and for things as they are in themselves to 3 Kemp Smith (1962): 113; cited in Allison (1983): 111-2. For a historical survey of debates over the neglected alternative objection to the transcendental ideality of space and time, see Allison (1983): 111-4, and Guyer (1987): 363, and 461, note 20. 4 Körner (1955): 17; cited in Allison (1983): 114.

4

be spatial” (Al:114). Here, regardless of how Kant is to argue for it, the essential point against the NAO is that we know a priori, not that the logical possibility invoked by the objection is not realized, but, rather, that what is alleged to be possible is in fact (logically) impossible in the first place: as Allison elsewhere puts it, “a representation that functions in the manner described (as an epistemic condition)…cannot (in any sense) apply to things as they are in themselves.”5

Guyer also thinks Kant has not overlooked the possibility invoked by the NAO, even though he thinks the way Kant tried to rule it out is fallacious. According to Guyer, Kant thought the allegedly neglected alternative does not arise because it is incompatible with “our a priori knowledge of space and time, particularly with the necessity of this knowledge” (G:363). The a priori knowledge of space and time implies (a) that spatiality and temporality are “necessary” features of objects per-ceived by us; but (b) if they were features of these objects considered as things in themselves, they could only be contingent, not “necessary”; hence, (c) they cannot be features of things in themselves (see, especially, G:364). This is part of the epistemo-logical argument Guyer attributes to Kant for the ideality thesis. Unfortunately, according to Guyer, the two occurrences of ‘necessary’ in (a) and (b) are equivocal: the first means “de dicto necessity” (that, necessarily, if we perceive an object, it is in space and time), whereas the second means “de re necessity” (that if we perceive an object, it is necessarily in space and time). It is only because Kant mistakes the kind of necessity involved in (a), the necessity of our a priori knowledge of space and time, for de re necessity that he infers from (a) and (b) to (c). Hence, Guyer’s inexorable criticism, cited earlier, that the ideality thesis is “the consequence of an exceedingly immodest interpretation of the necessity of synthetic a priori propositions.”

In the case of Allison, the NAO arises as long as it seems logically possible for things in themselves to be in space and time.6 Accordingly, to rebut the NAO, Kant must prove this to be really logically impossible (impossible “in any sense,” as Allison would have it). But it may well be that for the NAO to be effective, mere logical possibility is not enough. A distinction may be drawn between merely enter-taining the logical possibility in question and holding that such a possibility may as a matter of fact be realized. If the NAO does no more than the former, the ideality thesis is not threatened if it is synthetic a priori rather than analytic, so that it is in some weaker than logical sense of the term that the thesis affirms the “impossibility” of things in themselves being in space and time. In that case, Kant is free to acknow-ledge the logical possibility in question, while insisting that in (i) merely entertaining it, one ends up at most with an “empty” thought, and that this is far from (ii) holding that such a logical possibility may as a matter of fact be realized: for (ii) commits one, whereas (i) does not, to (iii) taking it to be in a stronger than logical sense possible for our representations of space and time to have application to things in themselves. If so, Allison has confronted Kant with a version of the NAO that is in fact too weak to

5 Allison (1996): 188-9, note 50; emphasis added. 6 Recounting his interpretation in Kant’s Transcendental Idealism, Allison in a more recent work points out that “[the fact that] a form of sensibility…involves an essential reference to mind…rules out the possibility of either a numerical or a qualitative identity between empirical space as the content of a form of sensibility and some putative transcendentally real space”; Allison (1996): 9, emphasis added. This is supposed to be an argument against the NAO, and its aim is, I take it, to prove that what is alleged by the NAO to be (logically) possible is in fact impossible.

5

refute the ideality thesis if, contrary to Allison, this thesis is to be construed synthetically.

I shall side with Guyer by proposing precisely a synthetic account of the ideality thesis. Taking the NAO to serve as a preliminary “litmus test” for the adequacy of any account of the ideality thesis, I shall understand it as committed to (ii) and so also to (iii). Kant had better accept rather than deny the logical possibility of things in them-selves being in space and time. But in opposition to (iii), and by exploiting whatever resources are available to him through his transcendental investigation into the possi-bility of a priori cognition of objects, Kant can argue against the NAO that it is in a weaker than logical sense impossible for our representations of space and time to have application to things in themselves—and, therefore, that things in themselves cannot, in a weaker than logical sense, be in space and time (the ideality thesis).

In looking this way at Kant’s route for the ideality thesis, I am suggesting that the thesis can be justified solely on grounds internal to Kant’s Copernican framework. In contrast, in the epistemological argument Guyer attributes to him for the ideality thesis, Kant is supposed to appeal to the premise that (b) if spatiality and temporality were features of things in themselves, they could only be contingent, not “necessary.” Kant does argue at times as if he endorses something like (b).7 But when he does so, he is actually stepping outside his framework to attack the pre-Copernican assumption that “all our cognition must conform to the objects” (Bxvi), arguing that nothing a priori, and therefore necessary, can be ascertained about space and time under such an assumption. Here Kant may be said to endorse (b) only as a claim external to his Copernican framework: granted that we cannot even surmise that the logical possi-bility of things in themselves being in space and time may as a matter of fact be realized, Kant must insist within his own framework that, contrary to (b), spatiality and temporality cannot even be considered contingent, let alone “necessary,” features of things in themselves. In brief, for Kant (b) only holds true within pre-Copernican frameworks, not within his own. Hence, if Kant’s argument for the ideality thesis did include (b) as a premise, then he would be advancing it without taking himself to have already proved that all attempts to demonstrate a priori cognitions of objects under the pre-Copernican assumption “have come to nothing” (Bxvi). But I think this is not how Kant would reach the ideality thesis: it can be justified only after he has under-mined the pre-Copernican assumption by proving that his transcendental method is the only possible method by which a priori cognitions of objects can ever be vindicated. It will be my chief aim in this paper to suggest a systematic account of Kant’s proof-procedure that leads to the ideality thesis, and to show that Guyer’s imputation of the faulty epistemological argument to Kant is mistaken.

§2. Justification in Three Steps: A Synopsis

I shall suggest that the ideality thesis is derived directly from another thesis of Kant’s, which is stated in the B-Deduction after he has reached a conclusion “of the greatest importance” that “[t]he categories consequently have no other use for the cognition of things except insofar as these are taken as objects of possible experience” (B147-8). The thesis stated after this conclusion is:

7 I shall consider one such instance in connection with space and time in §7, and a similar instance in connection with a priori concepts of objects in §6.

6

The pure concepts of the understanding are free from this limitation [to which representations of space and time are subject] and extend to objects of intuition in general, whether the latter be similar to our own or not, as long as it is sensible and not intellectual. [B148]

This is a thesis about the “further extension of [the categories unaccompanied by representations of space and time] beyond our sensible intuition” (ibid.; italics added). It is also a thesis according to which “we have an understanding that extends farther than sensibility problematically” (A255/B310). The only legitimate function of this further extension is to “limit the pretension of sensibility” (A255/B311), or to serve “for nothing but to designate the boundaries of our sensible cognition” (A288/B345). As I interpret it, the further-extension thesis is different from the limitation thesis: the latter is expressed by the conclusion that Kant says is “of the greatest importance” above, and it is a thesis which fixes limitation to be honored by both sensibility and the understanding; but the former is about the limiting of “the pretension of sensibi-lity” by the understanding.

According to the interpretation I shall propose in this paper, the limitation thesis, the further-extension thesis and the ideality thesis mark three succeeding steps in a continuous train of argumentation that appeals in various ways to Kant’s transcen-dental method and his Copernican assumption that “the objects must conform to our cognition” (Bxvi). On the face of it, I may appear to reject Guyer’s account of the limitation thesis as a sequel to the ideality thesis. But I think he is right to some extent. Kant’s difficult distinction, in terms of which the limitation thesis is couched, between “appearances” and “things in themselves” cannot receive its definitive form intended by Kant until he has proved the ideality thesis. Nevertheless, the limitation thesis (a conclusion “of the greatest importance”) is, in my view, proved earlier than both the further-extension and the ideality thesis. Moreover, I shall try to argue for a construal different from Guyer’s of the distinction between “appearances” and “things in them-selves”: Guyer takes Kant to “identify objects possessing spatial and temporal pro-perties with mere mental entities,”8 but I shall argue that the ideality thesis has no such “subjectivist” implication.

The limitation thesis is derived immediately after Kant has completed the final portion of the “transcendental deduction of the categories” (or “the Deduction,” for short). One of the conclusions Kant would draw at the end of the Deduction is that

(C) the categories “are related necessarily and a priori to objects of experience” (A93/B126).

And it divides into the following two parts:

(C1) that the categories “are related…a priori to objects of experience”;

(C2) that it is necessarily to objects of experience (i.e., to them alone) that the categories are related.

I shall take (C2) to imply the limitation thesis, and it is, strictly speaking, not so much a conclusion of the Deduction as a sequel to it. Kant’s transcendental method (or “TM” for short) is used chiefly to prove (C1). Such a proof belongs to the so-called

8 Guyer (1987): 335; last emphasis added.

7

“objective deduction” (see Axvi-xvii), or the Deduction proper. (C2) cannot be proved by the TM itself, but by an extrinsic feature that Kant thinks it has in relation to all other methods, especially those which adhere to the pre-Copernican assumption, for proving the possibility of a priori cognitions of objects: the TM is, as Kant repeatedly argues and proclaims throughout the Critique, the only possible method for proving this possibility. Insofar as the Deduction results from application of such a method, it is, as I shall put it, “externally unique.” It is this feature that helps Kant to prove (C2). I shall consider an argument actually offered in the A-Deduction for the external uniqueness of the Deduction, an argument which is part of a larger one for (C2).

Kant’s next step is to derive the further-extension thesis, stated, as we have seen, in the B-Deduction immediately after he has reached the limitation thesis. The further-extension thesis involves reference to (representations of) space and time as not entitled to the same extension enjoyed by the categories. To figure out how the thesis is ever justified, we need to compare the Deduction with a parallel deduction of space and time. Kant never officially presented this latter deduction or even explicitly mentioned it in the Aesthetic. I shall propose a reconstruction of it on the basis of a parallelism between the two deductions. Once the two deductions are put together for comparison, we can discern a disparity between the categories and representations of space and time regarding whether or not they count as a priori conditions of the possibility of sensible cognition in general. The further-extension thesis must rely upon implications of this disparity for its justification.

The further-extension thesis will in turn be used to justify a “problematic” and “negative” application of the categories to things in themselves considered as “nou-mena in the negative sense”—things insofar as they are not objects of our sensible intuition (cf. B307). This is supposed to be the only sense in which it is “not only admissible but even unavoidable” for us to “think” things in themselves (A256/B311; cf. B309). The further-extension thesis allows the categories to “extend to objects of [sensible] intuition in general, whether the latter be similar to our own or not, as long as it is sensible and not intellectual” (B148; emphasis added). Thus, the permissible “problematic” application of the categories to noumena in the negative sense is also application of them to things which are objects of some (logically) possible intuition of a sensible kind which may or may not be the same as ours. Kant’s TM could justify our application of the categories to objects of intuition with which a non-human species of rational creatures is endowed if only such a species were known to exist. In this case, inter-specific application of the categories is, as I shall put it, “TM-possi-ble.” But the same does not hold of representations of space and time. For Kant, the disparity alluded to above between the categories and representations of space and time may be said to come down to this, that inter-specific application is TM-possible for the categories but not for representations of space and time: even if some non-human species of rational creatures were known to exist with a sensible intuition that is no different in kind from ours, Kant’s TM can never be used to justify our application of representations of space and time to objects of their intuition—objects which it is nevertheless permissible and “even unavoidable” for us to “think,” in a “problematic” way, as noumena in the negative sense by means of the categories.

On my interpretation, the TM-impossibility, which is weaker than logical impos-sibility, of inter-specific application of representations of space and time is what directly justifies the ideality thesis, and in a way that makes it immune, I think, to the neglected alternative objection. Given that Kant’s TM and his Copernican assumption

8

are not at issue here, the TM-impossibility of inter-specific application of representa-tions of space and time is sufficient to justify Kant in claiming that things in them-selves considered as noumena in the negative sense are not in space and time. Proved in this way, the ideality thesis will have grounds derived from within Kant’s trans-cendental framework; and it will be both continuous and consistent with the results of his investigation into the possibility of a priori cognition of objects. More important, whereas the ideality thesis has to do with whether or not it is TM-possible for us to apply certain a priori representations in us inter-specifically to things in themselves, the limitation thesis requires that they be applied to appearances, objects of our intuition, and never intra-specifically to things in themselves. What I shall try to articulate in what follows is, for this reason, a compatibilist interpretation of the ideality thesis.9

§3. Transcendent Metaphysics and the Deduction

We have seen Kant describe the limitation thesis as “of the greatest importance” in the B-Deduction. It is, then, no surprise that he should be eager to highlight the thesis prefatorily as follows:

In the analytic part of the critique it is proved…[T1] that we have no concepts of the understanding and hence no elements for the cognition of things except insofar as an intuition can be given corresponding to these concepts, con-sequently [T2] that we can have cognition of no object as a thing in itself, but only insofar as it is an object of sensible intuition, i.e., as an appearance; from which follows [T3] the limitation of all even possible speculative cognition of reason to mere objects of experience. [Bxxvi]

(T2) asserts the limitation thesis; (T3) is stated only to emphasize that (T2) applies to “all even possible speculative cognition of reason.” (T2) follows from (T1), which is supposed to be proved in the Analytic, or more specifically in the Deduction. Thus, against Guyer’s account of the limitation thesis as endorsed by Kant because of his prior acceptance of the ideality thesis, we have here a clear textual evidence that the limitation thesis, (T2), follows not from the ideality thesis but from (T1). But a recent commentator would question the validity of this inference from (T1) to (T2).

Rae Langton has attributed to Kant what she calls “the thesis of receptivity,” that “we can have knowledge [or cognition] of an object only in so far as it affects us.”10 This seems to be a more concise way to express the essence of (T1). However, whereas Kant takes (T1) to be something that has to be proved, Langton takes the thesis of receptivity to be a “fundamental Kantian thesis,” and even an “empiricist

9 To be sure, Kant himself argues for the ideality thesis in the “Conclusions from the Above Concepts” in the Aesthetic, and does so by going directly from the limitation thesis to the ideality thesis. On the account I propose, however, these arguments are unsuccessful unless the further-extension thesis, to be established not in the Aesthetc but in the Analytic, is introduced to play an intermediate role in between the limitation thesis and the ideality thesis. Thus, it is lamentable that Kant should try to arrive at the ideality thesis as early as in the Aesthetic. My account of how the ideality thesis is to be established does not, therefore, agree entirely with Kant’s actual arguments. I thank Allen Wood for indicating the need to make this critical aspect of my account explicit. 10 Langton (1998): 23.

9

starting-point” or “assumption.”11 From this “fundamental” thesis she thinks Kant infers what she calls the “thesis of humility,” that we can never have knowledge of things as they are in themselves. This latter thesis may be regarded as tantamount to (T2), an expression of what I have dubbed “the limitation thesis.” Thus, Langton might appeal to the passage cited above to support attribution to Kant of the inference from receptivity to humility. However, after holding Kant to make this inference, Langton comes to challenge it. Referring approvingly to P. F. Strawson, who she thinks has identified the same inference and found it likewise dubious, Langton says, “Although Kant believes that humility follows from receptivity, it does not appear to do so. Strawson calls this a fundamental unargued premise of the Critique.”12

The inference from (T1) to (T2) does seem to be flawed, or at least in need of additional premises. But the way the inference is defective may also serve to cast doubt on Langton’s account of it. (T1), the thesis of receptivity, implies that we can only have knowledge of objects that appear to us. This rules out at once the possibility of knowledge of supersensible objects, and makes one wonder how, according to Langton, Kant can regard the thesis as an “empiricist starting-point” or “assumption” without begging the question of transcendent metaphysicians outright, which it is reasonable to suppose Kant would not do, given the anti-rationalist thesis (T3), as implied by the limitation thesis (T2). Langton owes us an explanation of why the receptivity thesis is a “starting-point” or “assumption” rather than, as Kant in the passage cited above holds it to be, something to be proved by more basic claims in “the analytic part of the critique.” Moreover, the receptivity thesis itself says nothing that shows why knowledge of objects that appear to us cannot be knowledge of them as they are in themselves—humility does not, therefore, follow from receptivity. But it may well be that if the receptivity thesis is indeed proved in the Deduction, some other claims appealed to in its proof can help validate the inference from it to (T2).

Langton is what I have called a “neutralist,” who takes a Strawsonian approach to the ideality thesis by trying to salvage the limitation thesis from it. As she argues,

Our ignorance of things in themselves [the limitation thesis] is supposed to follow from Receptivity, and not from any particular Kantian doctrine about space [the ideality thesis]…. Kant would deny that we have knowledge of things in themselves no matter what he thought about space: the conclusion about Humility is independent of any conclusion about space.13

This interpretive tactic gives Langton reason for detaching the limitation thesis (the thesis of humility, as she calls it) from the Aesthetic, but she tries to detach it also from the Deduction: “If humility is supposed [by Kant] to follow from receptivity, then it should be possible to explore [Kant’s grounds for the thesis of humility] with-out exploring in detail the arguments about space, time, and the categories.”14 As the passage cited from the Preface to the second edition of the Critique shows, Langton is right in attributing to Kant the inference from receptivity to humility. But the same passage also shows that, for purposes either of exegesis or of rescuing valuable Kantian insight from obscurities, she may be wrong in taking the inference out of its

11 See Langton (1998): 2, 44. 12 Langton (1998): 3. 13 Langton (1998): 211; emphases added. 14 Langton (1998): 3.

10

niche, the Deduction, to which I shall now turn in order to find out Kant’s grounds for the limitation thesis. As we shall see, the grounds the Deduction makes available to Kant are not only sufficient but also necessary for the limitation thesis: Kant’s infer-ence to it is therefore inseparable from the Deduction, contrary to Langton.

The receptivity thesis requires that something be an object of our senses if we can have knowledge of it. Such a thesis is controversial and cannot be plausibly taken by Kant as a mere assumption from the outset, for it precludes knowledge of super-sensible objects such as God and souls, the topics of traditional rational theology and psychology. Kant should be able to defend the thesis against transcendent metaphysics. The Deduction is explicitly held to serve this end, among others: it is taken by Kant to lead to the conclusion about “the limitation of all even possible speculative cognition of reason to mere objects of experience,” as we have seen. In whatever way Kant argues for this limitation thesis, he should not take it for granted that objects of human cognition cannot include supersensible ones.

A seemingly innocent alternative may be to follow from the outset a guiding principle indicated by Kant for the “transcendental deduction of all a priori con-cepts…namely this: that they must be recognized as a priori conditions of the possi-bility of experiences” (A94/B126). Perhaps, what needs to be assumed as an initial premise of the Deduction is that we have experience. It is, then, empirical rather than all cognitions of objects that are assumed to be cognitions of “appearances” and never of “things in themselves”; and this is so regardless of whether or not the same holds true of “all even possible speculative cognition of reason.” But, still, this does not seem to do justice to transcendent metaphysics. By assuming that we have experience, and by showing that the categories are “a priori conditions of the possibility of experiences,” it will be concluded at the end of the Deduction that (C) the categories “are related necessarily and a priori to objects of experience.” This conclusion is so understood by Kant that it rules out the possibility of a priori cognitions of super-sensible objects. A transcendent metaphysician may well retort that, by beginning with the assumption that we have experience, and by following the guiding principle cited above for proving (C), Kant’s procedure surreptitiously excludes the possibility of a priori cognitions of supersensible objects. I think the Deduction does, however, appeal to the claim that we have experience, and its conclusion is indeed (C). But Kant may have a defensible way to disarm the transcendent metaphysician’s suspicion. To show this, I propose to proceed backwards, by examining (C) first, to make sure what Kant is getting at with the Deduction. Then, in the next section, I shall suggest an account of Kant’s TM and how it can enable him to prove (C) in a way that even a transcendent metaphysician must accept.

A preliminary clarification of some terms involved in (C) is called for. “Objects” (Gegenstände) are to be understood here in what Strawson calls the “weighty” sense,15 in which they can be said to exist “corresponding to and therefore also distinct from” the concepts which are related to them (cf. A104). This is to be distin-guished from a much weaker sense of “object,” in which we “can…call everything, and even every representation, insofar as one is conscious of it, an object [Objekt]”

15 See Strawson (1966): 73. In the same way, Henry Allison speaks of “a ‘real’ sense of object, that is…an object in the sense of an actual entity or state of affairs (an object of possible experience)”; and he thinks that “Kant’s term for an object in this sense is Gegenstand”; Allison (1983): 135.

11

(A198/B234).16 Since “objects of experience” are objects in the weighty sense, the kind of “experience” in question here is to be understood accordingly in a strong sense in which it does not consist merely of sensible impressions connected through mechanism of association (“experience” subjectively construed) but amounts to empirical “cognition of objects” (“experience” objectively construed), “a composite of that which we receive through impressions and that which our own cognitive faculty…provides out of itself” (B1-2).

In the passage cited at the beginning of this section, Kant in effect holds all cogni-tions of objects to be cognitions of nothing but “objects of experience.” To acquire such cognitions, one “must refer [intuitions] as representations to something as their object [Gegenstand] and determine this object according to them” (Bxvii). Kant seems to be using the word “determine” here in a way aptly explained by this succinct definition: “To determine means…to judge synthetically”;17 in other words, it is to make a synthetic judgment that holds true of an object (or objects). If so, the referring act involved in an objective cognition may be regarded as more basic than the determining act, in that the latter already presupposes that we have succeeded in referring some representations in us to something as their object. This connection between the referring and the determining act may be used to elucidate the connection between what Kant calls “objective reality,” i.e., “relation to an object” (see A109, A155/B194), and “objective validity”: the referring act results, at first, in the obtaining of the relation of some representation(s) in us to an object, and then the determining act brings about a cognition or judgment that can be, by virtue of the “relation to an object” made possible by the referring act, held true or false of the object; and when a cognition or judgment can be held true or false of an object, it is “objectively valid” of it (see B142).

Now, the act of referring one’s representations to something as their object requires employment of that special concept which Kant calls “the concept of an object in general.” The categories are just those component concepts that make up this concept (see B128), which is supposed to be what we must use if we are to attain any cognition of something as an object (whatever contingent features this object may happen to have), or if we are to regard representations in us as representing things and events in a unified objective world (whatever things and events they are). But in order for us to be certain that the categories as concepts of an object in general can indeed enable us to attain cognitions of objects, they themselves must be shown to have rela-tion to an object. (C) is intended to ascribe, with justification, this relation to an object to the categories and to confer upon this relation an a priori status.

But (C) as it stands is surely intended to do more, as may be seen from its two parts: (C1) that the categories “are related…a priori to objects of experience”; and (C2) that it is necessarily to objects of experience that they are related. Kant has previously defined a “transcendental deduction” as an “explanation of the way in which concepts can relate a priori to objects” (A85/B117). (C1) may be seen as what any transcen-dental deduction of the categories is primarily aimed at, namely, the conferring of a

16 Allison calls this “an extremely broad sense of ‘object,’ which encompasses anything that can serve as the subject in a judgment”; Allison (1983): 135. He thinks Kant typically uses the word “Objekt” when speaking of an object in this weak sense. 17 Cited from Kant’s What Real Progress Has Metaphysics Made in Germany since the Time of Leibniz and Wolff, in Humphrey tr. (1983): 69.

12

priori relation to an object upon the categories, though it is with respect to objects of experience, in particular, that (C1) confers such relation upon the categories. (C1) therefore threatens empiricists rather than transcendent metaphysicians: Kant’s chief reason for (C1) is that the a priori relation of the categories to objects of experience is required for the possibility of experience; and if (C1) is true, then the empiricists have erred in denying the possibility of a priori cognition of objects while conceding the possibility of experience.18 Here it is objects of experience of which Kant is supposed to have proved that we can have a priori cognition. But (C1) as it stands does not exclude the possibility, to which transcendent metaphysics is committed, of a priori cognition of supersensible objects.

(C2) excludes this possibility. Insofar as the categories are concepts of an object in general, it is an analytic truth that we can have knowledge of something as an object only if the categories relate to it. Then, given (C2), it follows that we can have know-ledge of an object only if it is an object of experience, an appearance for us. Since if something is to become an appearance for us it must affect our senses, it follows that we can have knowledge of an object only if it affects our senses. This is precisely the receptivity thesis (T1) that Kant says is proved in “the analytic part of the critique”; and I have explained just now how it can be derived from (C2). If so, it is exegetically erroneous for Langton to commit Kant to the receptivity thesis as a “starting-point” or “assumption.”

Before proceeding to show how Kant can argue for (C1) and (C2) in the Deduction in a way that even a transcendent metaphysician must accept—especially in a way that does not take the receptivity thesis for granted, I must make a comment on the distinction between “things in themselves” and “appearances.” The Deduction should not simply assume a construal of the distinction as one between a set of objects lacking spatial and temporal properties (given the ideality thesis) and another set of objects that are “merely in us” (see A129) though nonetheless subject to space and time as forms of our intuition. These may seem to be prima facie two distinct sets of objects: being what is “merely in me,” an “outer appearance” that I am aware of cannot be numerically identical to the “thing in itself” that it is supposed to corres-pond to. This is a non-identity construal of the distinction. Opposed to it is an identity construal of the distinction as one between numerically the same objects “taken in a twofold meaning” (Bxxvii), or considered from two different viewpoints.19 Now, whichever construal Kant settles on, it should be the outcome of an argument, rather than presupposed from the beginning of the Deduction. As suggested in §2, Kant can arrive at the ideality thesis only after he has established the further-extension thesis; accordingly, the non-identity construal, presumably couched in reference to the ideality thesis, can take effect (if it does at all) only after the further-extension thesis is established. On the other hand, I shall assume that the identity construal is appropriate at least at the point where the limitation of human cognition of objects is initially fixed by the results of the Deduction and a parallel deduction of space and time. But we should not forget the transcendent metaphysician’s belief in the existence of supersensible objects, which can in no way be “taken in a twofold

18 Thus, to be more precise, (C1) directly threatens only those empiricists who already concede our possession of empirical cognition of objects. Unless, however, the possibility of empirical cognition of objects is established in the first place, (C1) is no threat to a skeptic. 19 I have adopted Allen Wood’s terminology in speaking of “identity” and “non-identity” construals. See Wood’s unpublished paper “What is Transcendental Idealism?”.

13

meaning.” In order for Kant not to go against this belief dogmatically, let alone fallaciously, in an argument, the kind of “objects” the Deduction starts off with must not be confined to objects to which the identity construal applies: they may simply be taken to be objects of which it is possible for us to have cognition, empirical or a priori. Supersensible objects are not bluntly excluded by this initial understanding of “objects.”

§4. Kant’s Transcendental Method

I shall now explain how Kant can establish the limitation thesis in the Deduction by arguing for (C2) without assuming the receptivity thesis from the outset. Since (C2) is a sequel to (C1) in the order of proof, let us look at Kant’s argument for the latter first. According to Kant, if any representation is to relate to its object(s) a priori, this can only be grounded on the fact that “the representation alone makes the object possible” (see A92/B124-5), in which case “the representation is…determinant of the object a priori if it is possible through it alone to cognize something as an object” (ibid.). This may be seen as Kant’s initial statement of the TM by which a deduction of the categories is in principle to be produced. By appealing to such a method, one makes the following inference:

(1) The categories are (a priori) conditions under which alone is any cognition of something as an object possible, “as far as the form of thinking is con-cerned.”20

(2) It is possible for us to have cognition of something as an object.

Therefore, (3) the categories relate a priori to objects of our cognition.21

(3) puts one in a rightful position to regard the categories as having objective reality, their (a priori) relation to objects of our cognition. Applying Kant’s Copernican assumption that “the objects must conform to our cognition,” one can carry the inference one step further by concluding from (3) that the a priori judgments we make about objects by means of the categories must, by virtue of that assumption, hold true of those objects a priori, and consequently that the categories must have objective validity as well. Kant’s own definition cited earlier of a “transcendental deduction” suggests, however, that the primary aim of such a deduction of a priori concepts concerns their objective reality. Accordingly, for the sake of simplicity, I shall focus on the account sketched above of Kant’s TM as a method for proving (3).

(3) is synthetic a priori, and so cannot be derived exclusively from analytic pre-mises. (1) is analytic; hence, (2) must be synthetic. This means that the kind of possi-bility asserted by (2) cannot be logical possibility, for otherwise (2) would be analytic.

20 Kant makes a parallel claim that “through [the categories] alone is experience possible (as far as the form of thinking is concerned)” (A93/B126). The qualification added in parentheses is deemed neces-sary by Kant because experience is taken by him to involve an intuitional component, “through which something is given,” and a conceptual component, through which the something given is “thought” (see ibid., B146). Without that qualification, the categories alone cannot be said to “make experience possible.” The same qualification must be added to (1). 21 Throughout this paper, expressions such as ‘objects of cognition’ or ‘objects of experience’ should be taken to mean objects of possible cognition or experience, and not merely objects that are actually cognized or experienced.

14

Rather, (2) involves “real” or “transcendental” possibility,22 in which case we must prove (2) by proving in some way that we do have cognition of something as an object. So it seems reasonable to say that the part of the inference to (3) that is more urgently in need of defense, and more difficult to prove, is the synthetic premise (2). This point will become crucial to Kant’s defense of the particular way in which the TM must be carried out in practice.

When Kant actually applies the method, he does not, however, make the above inference to (3) but the following inference instead:

(1’) The categories are conditions under which alone is any empirical cognition of something as an object possible, “as far as the form of thinking is con-cerned.”

(2’) It is possible for us to have empirical cognition of something as an object.

Therefore, (3’) the categories relate a priori to objects of experience (same as (C1)).

This inference gets Kant to (C1), and is a straightforward application of his methodo-logical claim, mentioned earlier, that “[t]he transcendental deduction of all a priori concepts…has a principle toward which the entire investigation must be directed, namely this: that they must be recognized as a priori conditions of the possibility of experiences” (A94/B126).23 Guided by this principle, one is supposed to look for “a priori conditions of the possibility of experiences.” In contrast, the inference to (3) concerns more generally conditions of the possibility of cognitions of objects. Experi-ences are only a subset of cognitions of objects. Nevertheless, I think it is the inference to (3) that accurately depicts Kant’s TM in its original form.

At the beginning of the paragraph that immediately precedes his statement of the methodological claim cited above, Kant has made it clear that the only possible case in which the relation of a representation to its object can obtain a priori is a case in which “the representation alone makes the object possible.” Here Kant cannot, and in fact does not, simply stipulate that the object “made possible” in such a case must be an object of experience. If he did, then not only would it be as if by a mere fiat that he ruled out the possibility of any concepts relating a priori to supersensible objects (which are objects of non-empirical cognition if we do have cognitions of them), but he would also stultify his own attempt to prove that we have a priori cognitions of objects: these are objects of non-empirical cognition if we do have a priori cognitions of them. So we should expect Kant to endorse the inference from (1) and (2) to (3) as a general characterization of the TM by which the Deduction is in principle to be constructed. Kant’s locution, “the representation alone makes the object possible,” may then be understood in terms of (1) and (2) as describing a case in which a repre-

22 For this distinction in Kant between logical and “real” or “transcendental” possibilities, see A244/B302, and footnotes at Bxxvi, B302-3 and A596/B624. Kant presses this distinction by saying that “the deception of substituting the logical possibility of the concept (since it does not contradict itself) for the transcendental possibility of things (where an object corresponds to the concept) can deceive and satisfy only the inexperienced” (A244/B302). Accordingly, the interpretive point I am making here is that, rather than asserting that the concept of “cognition of something as an object” is not self-contradictory, (2) asserts that we really do possess what corresponds to the concept. 23 Along the same line, Kant says, “The possibility of experience is therefore that which gives all of our cognitions a priori objective reality” (A156/B195).

15

sentation alone makes something possible as an object of cognition for us. It is not, indeed, a case in which a representation “produce[s] its object as far as its existence is concerned” (A92/B125). Rather, it is a case in which a representation is known a priori to be the sole “contributor,” so to speak, to the possibility of something’s becoming an object of cognition for us.

Thus, according to the TM as Kant initially introduced it at A92/B124-5, if we have proved (1) and (2), we have thereby proved that the categories alone make their objects possible as objects of cognition for us, and we can infer from this that they have objective reality a priori. But when Kant actually applies the method in the Deduction, his argument becomes: if we have proved (1’) and (2’), we have thereby proved that the categories make their objects possible as objects of experience for us, and we can infer from this that they have objective reality a priori. This line of reasoning may be said to be what Kant has in mind when he says, “The conditions of the possibility of experience in general are at the same time conditions of the possibility of the objects of experience, and on this account have objective validity in a synthetic judgment a priori” (A158/B197). But it should not be left unquestioned that Kant’s TM in its original form can support only this particular application.

Once the TM is understood in terms of the inference to (3), and the inference to (3’) is thought of as seemingly just one possible application of it, one might argue, against Kant, for the possibility of transcendent knowledge of supersensible objects by making an inference that seems on the face of it to have a standing on a par with the inference to (3’), as follows:

(1”) The categories are conditions under which alone is any non-empirical cogni-tion of something as an object possible, “as far as the form of thinking is concerned.”

(2”) It is possible for us to have non-empirical cognition of something as an object.

Therefore, (3”) the categories relate a priori to objects of non-empirical cognition.

One might argue that if Kant’s method sanctions the inference from (1’) and (2’) to (3’), it ought to sanction the present inference as well. But Kant must disallow this latter inference, since it follows from (3”) that it is not necessarily to objects of experience that the categories relate, contrary to (C2). Thus, even if Kant’s actual application of the TM to the categories entitles him to infer (3’), he must find a way to resist the inference from (1”) and (2”) to (3”) if his favored inference to (3’) (same as (C1)) is to be carried one step further to (C2). But both (1”) and (2”) are, as they stand, unobjectionable to Kant. If he infers from (1’) and (2’) to (3’), what can justify him in disallowing the symmetrical inference from (1”) and (2”) to (3”)?

A reason available to Kant for resisting the inference to (3”) while retaining the inference to (3’) lies in a crucial difference between (2’) and (2”) as premises that must be proved before the possibility of a priori cognition of objects can in turn be proved. (2’) and (2”) are synthetic claims, and, having to be proved a priori in a transcendental investigation, they are to be treated as a priori as well. But they do not express a priori cognition of objects, but a priori cognition about cognition of objects that is supposed to be possible a priori: (2’) is about empirical cognition of objects, and (2”) about non-empirical cognition of objects. But whereas to prove (2’) is to prove the possibility of empirical cognition of objects, to prove (2”) is to prove the

16

possibility of a priori cognition of objects—the very possibility that Kant’s TM is designed to prove in the end. We can then see a difficulty that confronts the inference to (3”) but not the inference to (3’): although both are supposed to enable us to produce a justification of the a priori employment of the categories, the inference to (3”), but not the inference to (3’), requires that one who makes it have already proved the possibility of a priori cognition of objects. If so, Kant may reject the inference to (3”) by pointing out that the only non-fallacious way to apply the TM to the categories is to make the inference to (3’).24

§5. From External Uniqueness to Limitation

Kant’s rejection of the inference to (3”) appeals to the claim that the Deduction as based on the inference to (3’) is “unique” in the sense that it results from the only legitimate application of the TM to the categories. But there is another sense in which the Deduction may be said to be “unique”: it is “unique” to the extent that it results from the only possible method for proving the possibility of a priori cognition of objects. Let us call “uniqueness” of the Deduction in the first sense “internal,” and its “uniqueness” in the second sense “external.” Offhand, the internal uniqueness of the Deduction implies that the inference to (3”) cannot be a possible application of the 24 This interpretive claim may be challenged, however, through the following consideration: The inference to (3’) may be said to fall under what Guyer has identified as “versions of the deduction which assume the possibility of knowledge of objects from the outset and which attempt to show only that such knowledge presupposes a priori knowledge of the categories—arguments…targeted at empi-ricists but not at skeptics”; Guyer (1987): 91, emphasis added. But I think the inference to (3’) can be interpreted differently. Although one of its premises, (2’), asserts the possibility of experience, this possibility does not have to be taken for granted from the outset. The inference to (3’) from the analytic (1’) and the synthetic (2’) may be supplemented by a derivation of (2’) that appeals to what Henry Allison describes as “the crucial theme in the Deduction,” namely, “the correlation between self-cons-ciousness or the capacity to say ‘I’ (apperception) and the consciousness or experience of an objective, spatio-temporal world”; Allison (1996): 27, cf. Allison (1983): 144. The correlation Allison speaks of may be taken to amount to a mutual entailment between the so-called “principle of apperception” and (2’). How this principle is to be formulated and understood raises deep issues. But it seems to be what Kant asserts with such statements as “We are conscious a priori of the thoroughgoing identity of ourselves with regard to all representations that can ever belong to our cognition” (A116), or “The I think must be able to accompany all my representations” (B131). (2’) may then be derived from the principle of apperception, and so does not have to be simply assumed against a skeptic who puts our possession of cognition of objects in doubt. However, since according to my account (2’) is synthetic a priori, so must be the principle of apperception. I am therefore committed to taking side with inter-preters such as Guyer (1987) and Patricia Kitcher (1982) who think the principle is to be construed as synthetic a priori rather than analytic. Allison is a recent advocate of an analytic interpretation of the principle of apperception; see Allison (1983): 137, and “Apperception and Analyticity in the B-Deduc-tion” in Allison (1996). Earlier advocates include Jonathan Bennett and Robert Paul Wolff; see Bennett (1966): 103, and Wolff (1963): 105, 186. Things would go much easier if this principle, from which (2’) can be derived, was analytic. But if it is indeed synthetic a priori, it doesn’t seem to enjoy what Dieter Henrich calls “Cartesian evidence”; see Henrich (1976): 59. It is in need of elaborate defense especially against a Humean skeptic who sees no legitimate use of what Kant refers to as the concept of “the standing and lasting I (of pure apperception)” (A123). Worse still, if the principle is synthetic a priori, the question arises as to whether, like (2”) in the rejected inference to (3”), it already asserts the possibility of a priori cognition of objects, so that, appealing as it does to a premise that must be derived from the principle of apperception, the inference to (3’) must also be rejected, and for the same reason that it is committed to a petitio principii. In any case, if the principle is to be taken synthetically, as I think it is, a justification that can meet the challenge of a Humean skeptic is called for. In my view, insofar as the concept “I” is an a priori concept, it is reasonable to expect Kant to justify it by means of the TM, as he did the categories as “concepts of an object in general.” How this can be done is a topic that deserves separate and extensive treatment.

17

TM to the categories. But in order for Kant to establish the internal uniqueness of the Deduction, it is actually both necessary and sufficient that the Deduction be shown to be externally unique as well, as I shall now explain.

The external uniqueness of the Deduction commits Kant to claims such as the following:

I am not, to be sure, of the opinion that excellent and thoughtful men…, aware of the weakness of previous proofs, have so often expressed, that one can still hope someday to find self-evident demonstrations of the two cardinal proposi-tions of our pure reason: that there is a God, and there is a future life. Rather, I am certain that this will never happen. [A741-2/B769-70]

Suppose Kant was wrong here. Suppose one could produce “self-evident demonstra-tions” of God’s existence and the immortality of the soul independently of the TM, so that the Deduction was not externally unique. Then the possibility asserted by (2”) of a priori cognition of objects could be proved in a way that is not question-begging, and the inference to (3”) would be as legitimate an application of Kant’s method to the categories as the inference to (3’). Then the Deduction would not be internally unique either: a different transcendental deduction of the categories was available which justifies their a priori relation to objects of non-empirical cognition. The external uniqueness of the Deduction is therefore required for its internal uniqueness: if the former is lacking, so is the latter, in which case Kant must accept the inference to (3”) and, consequently, discard (C2).

If, on the other hand, the Deduction is indeed externally unique, then any appli-cation of the TM that seeks to prove the a priori relation of the categories to objects of non-empirical cognition is bound to be premised fallaciously on the very thing that presumably only the TM can demonstrate, namely, the possibility of non-empirical cognition of objects. In that case, the Deduction is also internally unique.

We can then see that the external (and so also internal) uniqueness of the Deduction enables Kant to argue as follows:

(i) We are justified, given the Deduction, in holding the categories to relate a priori to objects of experience.

(ii) We cannot be justified, given the external uniqueness of the Deduction, in holding the categories to relate a priori to things that are not objects of experience.

Therefore, (iii) it is necessarily to objects of experience that we are ever justified in holding the categories to relate a priori.

(iii) is just another way of saying that we are justified in believing (C2). In this way, Kant arrives at (C2) by means of the Deduction taken as externally unique. As explained in §3, (C2) implies the receptivity thesis, and the reason the latter does not by itself imply the limitation thesis is that even if we can have cognition only of objects that appear to us, the possibility remains open that we can have cognition of these objects both as they appear to us and as they are in themselves. (C2) rules out this possibility: given (C2), and given the analytic truth that only if the categories relate to something as an object considered in some way, can we have cognition of it as an object considered in that way, it follows that we can have cognition of objects

18

considered only as “appearances” (or “objects of experience”), not as “things in them-selves.” In this way, the limitation thesis can be derived.

Now, the title of the “summary representation” that concludes the A-Deduction refers to the “correctness and unique possibility of this deduction of the pure concepts of the understanding” (A128). In light of the foregoing account of the Deduction as externally unique, it may be said that the “correctness” of the Deduction entitles Kant to (i), its “unique possibility” entitles him to (ii), and together they yield him (C2). This line of reasoning is sufficient to establish the limitation thesis without assuming the receptivity thesis right from the start. So interpreted, the limitation thesis is in fact a consequence of what lies deep in the heart of Kant’s critical project: the Deduction, and the special status of the TM as the only possible method for proving the possibi-lity of a priori cognition of objects. More important, it is essentially the external uniqueness, or “unique possibility,” of the Deduction that justifies the limitation thesis: as the threat from the inference to (3”) shows, the Deduction without external unique-ness, even if it leads to (3’) (same as (C1)), will not enable Kant to establish the further conclusion (C2) that excludes the possibility of the categories relating a priori to things considered otherwise than as objects of experience. The limitation thesis is unfounded unless the Deduction is externally unique.

§6. Kant’s Argument for External Uniqueness in the A-Deduction

But how does Kant argue for the external uniqueness of the Deduction? The “sum-mary representation” mentioned above is where we may naturally expect to find him doing so. At this juncture of the Critique, Kant has just completed the main proof of the Deduction, so that he is in a position to hold (C1), that the categories “are related…a priori to objects of experience.” Apparently in an effort to show the categories to relate a priori not only to objects of experience but also to them alone, Kant begins the “summary representation” by asserting,

(A) “If the objects with which our cognition has to do were things in themselves, then we would not be able to have any a priori concepts of them at all [or, in other words, it would not be possible for any concepts such as the categories to relate a priori to the objects with which our cognition has to do].” [A128]

But from the fact that the TM has enabled us to prove (C1), it follows that the consequent of (A) is false. Then, by modus tollens, we obtain the negation of the ante-cedent of (A) as follows:

(B) “[T]he objects with which our cognition has to do [are not] things in them-selves.”

Since (B) is an a priori and, therefore, necessary truth, it amounts to the limitation thesis, which is inferred here from (A) plus the fact that the TM has enabled us to prove (C1). The first thing Kant does in the “summary representation” is to argue for (A). As I shall now try to bring out, essential to this argument is an auxiliary argument for the external uniqueness of the Deduction.

Immediately after asserting (A), Kant supports it by means of the following reductio ad absurdum: First of all, assume that the antecedent of (A) is true and yet its consequent false; in other words, assume that “the objects with which our cognition has to do” are things in themselves, and that we can have a priori concepts of those

19

objects. Then, either “we take them [i.e., the concepts] from the object” or “we take them from ourselves.” In the former case, the concepts can only be empirical, con-trary to the supposition that they are a priori. In the latter, “from ourselves” case, we cannot justify the assumed a priori relation of those concepts to things in themselves, for “that which is merely in us cannot…be a ground…why [those a priori concepts] should not instead be empty [i.e., lacking relation to an object]” (A129). The two cases exhaust all possible alternatives, and in neither case can we have concepts that relate a priori to objects considered as things in themselves, contrary to the assump-tion we started with. Hence, if the antecedent of (A) is true, its consequent cannot be false; in other words, the conditional (A) is true.

It may be noted that Kant would be willing to concede in retrospect that the fore-going argument is invalid, since toward the end of the B-Deduction he offers a similar argument where he brings up a “middle way,” neglected in the first edition, between the “from ourselves” and “from the object” cases: “namely, that the categories were neither self-thought a priori first principles of our cognition nor drawn from experi-ence, but were rather subjective predispositions for thinking, implanted in us along with our existence by our author” (B167; emphasis added). This third alternative, a “from God” case, would invalidate the above reductio.

I shall, however, put aside the question whether in his argument for (A) Kant has enumerated all possible cases and demonstrated in each case that no concepts can be proved to relate a priori to objects as long as the objects are taken to be things in themselves. But Kant’s reason for denying a priori relation to things in themselves in the “from ourselves” case may be revealing. As he says,

If we take [the concepts] from ourselves, then that which is merely in us cannot determine the constitution of an object distinct from our representations, i.e., be a ground why there should be a thing that corresponds to something we have in our thoughts, and why all this representation should not instead be empty. [A129]

The “ground” that Kant says “that which is merely in us” cannot be, seems to be precisely what he has been looking for all along since he wrote an oft-cited letter to Marcus Herz in 1772. There he raised the question, “What is the ground of the rela-tion of that in us which we call ‘representation’ to an object?”25 Kant’s subsequent explanation to Herz in the letter makes it clear that the question concerns more specifically the ground of the relation of an a priori representation to an object. It is a ground why an a priori representation has something corresponding to it as its object, and why it “should not instead be empty.” Let us call this the “ground for non-empti-ness” of an a priori representation or concept.

Now, it may be asked, in the “from ourselves” alternative considered in his argu-ment for (A), what is it that Kant thinks cannot be a ground for non-emptiness of a priori concepts of objects? Is he arguing that if we have a priori concepts of objects considered as things in themselves, and if we “take them from ourselves,” then what has served as the ground for non-emptiness of the categories in the preceding Deduc-tion will no longer be such a ground, so that the categories relate a priori not only to objects of experience, but also to them alone? If Kant means to argue this, then in the

25 Zweig (1999): 133.

20

statement, “that which is merely in us cannot…be a ground why there should be a thing that corresponds to something we have in our thoughts,” he may be interpreted as using the expression “that which is merely in us” to refer to his own transcendental ground, claiming that it would no longer be a ground for non-emptiness if the objects of our cognition were things in themselves. In that case, Kant would be concerned, at least directly, with the internal uniqueness of the Deduction; for he would be arguing that his own transcendental ground for non-emptiness of the categories can be effective only under the condition that the objects to which these concepts relate are considered as appearances rather than as things in themselves. But then, we may ask, in case these objects are considered as things in themselves and, consequently, Kant’s transcendental ground for non-emptiness becomes ineffective, why can there not instead be some non-transcendental grounds? Since Kant has not, on the present account, disqualified other grounds than his own for non-emptiness of a priori con-cepts of objects, neither has he disqualified other methods for proving the possibility of a priori cognitions of objects. His denial of the possibility of a priori concepts of objects considered as things in themselves in the “from ourselves” case would then be a non sequitur.

A more reasonable account seems to be that “that which is merely in us” refers to a priori concepts of objects, concepts that are in the “from ourselves” case supposed to be solely “taken from ourselves”; and that in denying that it can be a ground for non-emptiness, Kant may be said to argue that we cannot prove the a priori relation of a concept to an object simply through mere analysis of the concept itself, as has been famously attempted in the ontological proof of God’s existence. So interpreted, Kant is making a point that he later expresses in the Preface to the second edition of the Critique when he gives his reason for initiating a Copernican “experiment” in meta-physics: “Up to now it has been assumed that all our cognition must conform to the objects; but all attempts to find out something about them a priori through concepts that would extend our cognition have, on this presupposition, come to nothing” (Bxvi; emphasis added). Accordingly, Kant is arguing in the “from ourselves” case for the external uniqueness of the Deduction, and his conviction that the Deduction has this special feature turns out, on this alternative account, to be a conviction in the inevi-table failure of any attempt to prove the possibility of a priori cognitions of objects in accordance with the assumption that such cognitions “must conform to the objects.” This pre-Copernican assumption would require that a priori cognitions of objects, although “taken from ourselves” and owing nothing to their objects for their existence and contents, must nevertheless conform to the objects. This amounts to requiring that a priori cognitions of objects conform, not only to the objects, but also to them considered as things in themselves, and that this be the case regardless of whether or not these objects can affect our senses (for the objects in the “from ourselves” case can in principle be either sensible or supersensible if they are taken to be things in themselves). For Kant, if we construe a priori cognitions of objects in this way, we can never prove that we have them, nor can we prove that we have concepts that relate a priori to objects. We have already seen essentially the same point made by Kant: (A) “If the objects with which our cognition has to do were things in themselves, then we would not be able to have any a priori concepts of them at all.” This brings us back to (A), which may now be viewed as Kant’s way of claiming the inevitable failure of any method for proving the possibility of a priori cognitions of objects insofar as it follows the hitherto paradigmatic assumption. Thus, in the historical

21

context of his time, for Kant to demonstrate the external uniqueness of the Deduction is for him to argue for (A).

There is, then, an enormous difference between Kant’s TM as (i) merely one possi-ble method among others for proving the possibility of a priori cognition of objects on the one hand, and as (ii) the only such method on the other. In case (i), the possibility is left open that the categories may relate a priori to things in themselves, or that, contrary to the limitation thesis, we may have a priori cognition of things in themselves. It is not so, as we have seen, in the case of (ii), where the Deduction is externally unique. Hence, this feature of the Deduction is not only sufficient but also necessary for the limitation thesis to sustain. Epistemological modesty is out of place as long as the Deduction is less than externally unique, as in case (i). Taken as what Kant would in practice uphold in an effort to establish the external uniqueness of the Deduction and to defeat thereby the pre-Copernican assumption as a guiding principle for proving the possibility of a priori cognitions of objects, (A) may be said to be an indispensable “catalyst” which, when combined with the Deduction’s initial conclu-sion (C1) (that the categories relate a priori to objects of experience), changes it into a cause for what Langton calls “Kantian humility” (see §3). Langton, however, thinks Kant’s grounds for humility can be explored independently of the Deduction. We can now see that this cannot be right if the external uniqueness of the Deduction is indispensable for the determination of the boundaries of human cognition.

§7. A Parallel Deduction of Space and Time

The TM is supposed to be applied to both the categories and representations of space and time, so that we should have not only the Deduction in the Analytic but also a parallel deduction of space and time in the Aesthetic. If the former deduction is exter-nally unique, so must be the latter, and we should expect Kant to derive from this a claim that parallels (C2) (that the categories relate to objects of experience alone).

Although no deduction of space and time is officially presented in the Aesthetic, Kant does make such a parallel claim when he argues (D2) that representations of space and time “apply to objects only so far as they are considered as appearances” (A39/B56). Just as we have distinguished (C2) from (C1), so we may also distinguish (D2) from (D1) that representations of space and time “apply [a priori] to objects…as far as they are considered as appearances.” And just as Kant would prove (C1) by proving that through the categories alone is experience possible, “as far as the form of thinking is concerned” (and that experience is indeed possible for us), so, too, he would prove (D1) by proving that through space and time alone is experience possible, as far as the form of intuiting is concerned.26 Finally, just as he would go on to prove (C2) by appealing to the external uniqueness of the Deduction, so, too, he would go on to prove (D2) by appealing to the external uniqueness of the deduction of space and time, to the fact that this deduction results from application of the TM taken as the only possible method for proving the possibility of a priori cognitions of objects.

26 For Kant, whereas space is the “formal condition” of outer intuition, “time is an a priori condition of all appearance in general, and indeed the immediate condition of the inner intuition…and thereby also the mediate condition of outer appearances” (B50-1). The following schematic account of Kant’s application of the TM to space and time will leave aside this detailed qualification of them as formal conditions of our sensible intuition.

22

Now, if (D1) is true, then, given Kant’s Copernican assumption, objects considered as appearances must conform to our representations of space and time; and if (D2) is also true, then it is only objects considered as appearances, not as things in themselves, that must conform to them. Thus, through its external uniqueness, which gets us from (D1) to (D2), the deduction of space and time can itself (i.e., without reliance on a deduction of the categories) determine their boundaries. Moreover, given the external uniqueness of the deduction of space and time, it is in virtue of nothing other than the fact that space and time are conditions under which alone objects can appear to us that these objects must conform to our representations of space and time; for this reason, Kant takes them to be “merely conditions of sensibility.” Thus, avers Kant,

[Representations of space and time] determine their own boundaries by that very fact (that they are merely conditions of sensibility), namely that they apply to objects only so far as they are considered as appearances. These alone are the field of their validity, beyond which no further objective use of them takes place. This reality of space and time, further, leaves the certainty of experiential cogni-tion untouched; for we are jus as certain of that whether these forms necessarily adhere to things in themselves or only to our intuition of these things. [A39/B56; emphases added]

Here (D2), that representations of space and time “apply to objects only so far as they are considered as appearances,” is only held to justify the claim that beyond their boundaries “no further objective use of [representations of space and time] takes place”—a claim which does not by itself justify us in claiming to know that things in themselves are not in space and time (the ideality thesis). But Kant does not rest content with the weaker claim.

Space and time come to take on a “merely subjective” character when they are held to be “merely subjective conditions” of sensibility, with the implication clearly intended by Kant that we know that things in themselves are not in space and time. Thus we find him declaring,

[I]f we remove our own subject or even only the subjective constitution of the senses in general, all relations of objects in space and time, indeed, space and time themselves would disappear, and as appearances they cannot exist in them-selves, but only in us. [A42/B59]

The “merely subjective” character of space and time seems to be taken by Kant to amount to their transcendental ideality (see A28/B44, A36/B52). But, if I am right, (D2) can justify neither of these features attached to space and time. How, then, are they justified in the Aesthetic?

In the final section of the Aesthetic, we find Kant taking himself to have conclu-sively demonstrated the “merely subjective” character of space and time:

It is therefore indubitably certain and not merely possible or even probable that space and time, as the necessary conditions of all (outer and inner) experience, are merely subjective conditions of all our intuition…. [A48-9/B66]

This claim is supported by an argument which precedes it. We must now examine this argument to make sure how Kant arrives at the ideality thesis. Kant draws the above conclusion by appealing directly to this conditional:

23

(P) “If, therefore, space (and time as well) were not a mere form of your intui-tion that contains a priori conditions under which alone things could be outer objects for you…, then you could make out absolutely nothing syn-thetic and a priori about outer objects.” [A48/B66]

That one “could make out absolutely nothing synthetic and a priori about outer objects” may be interpreted as what Kant would hold to be sufficiently falsified by the fact that we can prove (D1), that representations of space and time have “objective” application a priori to appearances (the initial result of a deduction of space and time). Thus, granted that the consequent of (P) is false, so is its antecedent, whose denial is that space and time are “mere forms” of sensibility. Apparently, that space and time are such “mere forms” is taken by Kant to be another way of saying that they are “merely subjective conditions of all our intuition.” And if for space and time to be “merely subjective conditions” is indeed for them to be transcendentally ideal, then, in the end, application of modus tollens to (P) yields the ideality thesis.

But how does Kant establish (P) itself? The argument for (P) started earlier in the same paragraph with the announcement of an investigation:

(Q) “[I]f it were to be supposed that space and time are in themselves objective and conditions of the possibility of things in themselves, then it would be shown, first, that there is a large number of a priori apodictic and synthetic propositions about both, but especially about space, which we will therefore here investigate as our primary example.” [A46/B64]

Kant’s strategy for this investigation is to embark on a reductio ad absurdum to demonstrate that if the antecedent of (Q) is true, then we cannot prove that we have synthetic a priori cognitions about space and time. That we cannot prove this is what the consequent of (P) asserts, so that what the reductio strategy enables Kant to demonstrate is that if the antecedent of (Q) is true, so is the consequent of (P). Once this is demonstrated, Kant can derive (P) by assuming that the antecedent of (P), that “space (and time as well) were not a mere form of your intuition,” implies the ante-cedent of (Q), that “it were to be supposed that space and time are in themselves objective [etc.].” But does the antecedent of (P) really imply the antecedent of (Q)?

Consider the antecedent of (Q): It seems plausible to view it as intended by Kant to represent the position of those who embrace the pre-Copernican assumption as a guiding principle that has dominated all previous attempts to prove the possibility of synthetic a priori cognitions, including those about the spatiality and temporality of objects. If so, then, strictly speaking, the antecedent of (Q) is not that “space and time are [not] in themselves objective and conditions of the possibility of things in them-selves,” but, rather, that it is not to be supposed—as it would be if one followed the pre-Copernican assumption—that space and time are “in themselves objective [etc.].” But to say that this is not to be supposed is not to say that its opposite, that space and time are not “in themselves objective [etc.],” is to be supposed. Agnosticism about whether or not space and time are “in themselves objective [etc.]” still lurks as an open option even if we agree that the antecedent of (Q) is to be rejected. But as long as the agnostic option remains, the ideality thesis is not established even if the falsity of the antecedent of (Q) entitles Kant to deny the antecedent of (P), claiming that space and time are “mere forms” of our intuition. For given the agnostic option, space and time may be, for all we know, “in themselves objective [etc.].”

24

Does Kant, then, really intend the claim that space and time are “mere forms” of our intuition to amount to the ideality thesis? If he does not, then inference to the fal-sity of the antecedent of (P) does not count as inference to the ideality thesis after all. But if he does, then the antecedent of (P) does not really imply the antecedent of (Q): abandonment of the pre-Copernican supposition that “space and time are themselves objective [etc.]” does not by itself lead to endorsement of the ideality thesis (for agnosticism is still alive). In this case, Kant’s argument for the ideality thesis, asserted when he proclaims, “It is therefore indubitably certain…that space and time are merely subjective conditions of all our intuition,” is founded on a premise, namely (P), that he fails to justify by means of the reductio strategy mentioned above. Either way, Kant has not proved the ideality thesis in the final section of the Aesthetic.

Kant’s argument for (P) may be compared with his argument for (A) in the “sum-mary representation” in the A-Deduction (see §6). Both are part of a project to defeat the pre-Copernican assumption that has steered the controversy over the possibility of a priori cognitions of objects. And both concern the question whether there are limits to the kind of a priori relation to an object that has been proved, by means of the TM, to be possessed by the categories and representations of space and time. Taking Kant’s Copernican assumption into account, such a priori relation of a representation to an object not only means that the representation is not “empty,” but also confers upon it a special status of having its object(s) necessarily in conformity to it. Hence, limits to a priori relation to an object are also limits to what a priori cognitions of objects are supposed to be possible on the Copernican assumption. Kant’s strategy for drawing such limits is essentially the same in both arguments. Appealing to a reductio, he would first argue that if the incumbent assumption was adhered to, the initial conclu-sions (C1) and (D1) arrived at by means of the TM could not be proved at all; next, he would declare his proofs of (C1) and (D1) to be the only possible ones; and, finally, given this “unique possibility” of his proofs, their initial conclusions come to acquire a limiting nature: whereas it was initially proved that the categories relate a priori to objects of experience, now a reductio argument shows that they can have such relation only to them; and, whereas it was initially proved that representations of space and time “apply [a priori] to objects…so far as they are considered as appearances,” now a reductio argument shows that they can have such application only to appearances. This parallelism between Kant’s two reductio arguments may serve to show that his conclusion about space and time as “merely subjective conditions” or “mere forms” of sensibility does not follow from (P) plus the correctness of the deduction of space and time up to the initial conclusion (D1), if it is loaded with the implication that “things in themselves” are not in space and time.

On Kant’s Copernican model, as mentioned above, once it is established that the categories and representations of space and time have appearances as their objects, appearances must conform to them; and Kant’s arguments for (A) and (P) are meant to underscore a constraint of the model, namely, that it does not justify us in thinking the a priori representations in question also represent how things really are in themselves. To think in this way is to relapse into the pre-Copernican model. As if wary that it might be difficult for his readers to leave this old model behind, Kant is eager to remind them of this constraint of the Copernican model by saying,

We can accordingly speak of space, extended beings, and so on, only from the human standpoint. If we depart from the subjective condition under which alone we can acquire outer intuition…then the representation of space signifies

25

nothing at all. This predicate is attributed to things only insofar as they appear to us…. [A26-7/B42-3]

That “the representation of space signifies nothing at all” if one relapses into the pre-Copernican way of thinking and considers objects of human cognition as things in themselves, does not by itself mean that things in themselves have no spatial pro-perties. It can only be a way of revoking objective reality (and objective validity as well) from the representation of space if one does not abide by the Copernican way of conceiving of how a priori cognitions of objects are possible. So is the case with the representation of time.

Assuming, then, that the boundaries to which the categories and representations of space and time must equally be subject were steadfastly observed, Kant would seem to have no reason to transgress the boundaries himself by asserting the ideality thesis. It might appear that, whether one affirms or denies that things in themselves are in space and time, to do either is equally to think in the pre-Copernican way, and in disregard of the limitation thesis. So we may ask if Kant has any ground stemming from within his own Copernican framework for claiming that things in themselves are not in space and time.

§8. From Further Extension to Ideality

The question may be answered in a way that shows that the right place for Kant to justify the thesis of the transcendental ideality of space and time is where he has both (C2) and (D2) at his disposal to entitle him to claim “limitation of all even possible speculative cognition of reason to mere objects of experience” (Bxxvi). To see how Kant could consistently permit the boundaries of human cognition to be crossed over without being breached by the ideality thesis, we may note that the foregoing parallel-ism does not prevent Kant from treating the categories differently—in fact more favorably—than he does representations of space and time: he allows the former, but not the latter, a “further extension…beyond our sensible intuition” (B148; see also A253-4/B309). Perhaps, lack of this “further extension” on the part of representations of space and time can help explain why Kant should commit himself to the ideality thesis. I shall now try to find out how Kant would support the claim that the “further extension…beyond our sensible intuition” is enjoyed by the categories but not by representations of space and time, what this further-extension thesis amounts to, and how it might lead Kant to attribute transcendental ideality to space and time.

As we have seen, Kant’s method for justifying the relation of a priori represen-tations to an object requires, in principle, investigation into (i) conditions under which alone is it possible for us, human beings, to have cognition of something as an object. Because of the internal uniqueness of the Deduction (see §5), Kant’s actual appli-cation of the method only investigates (ii) conditions under which alone is it possible for us to have empirical cognition of an object. Now, for Kant, what conditions (ii) consist of are no more than the categories as the form of thinking on the one hand, and representations of space and time as the form of our sensibility or intuition on the other. But what, in principle, do conditions (i) consist of? The Deduction’s external uniqueness yields the conclusion that, whether empirical or a priori, human cognition can only be cognition of appearances (the limitation thesis); since such cognition is of a sensible kind, the answer to the question about conditions (i) must be that they consist of exactly the same things as do conditions (ii). Thus far, the categories and

26

representations of space and time go together. But they part company if we go on to ask whether they also make up (i*) conditions for any sensible cognition of objects, whether or not it is of a kind that is possible for us. For Kant, the answer is no: although human cognition can only be of a sensible kind, there may in principle be different kinds of sensible cognition; and while it can be known a priori that all of them must have the categories as conditions of “thinking an object” (for otherwise they would not be cognitions of objects), it can also be known a priori that it is not necessary that all of them also have representations of space and time as conditions of (sensibly) “intuiting an object” (for if this were necessary, then there would only be just one kind of sensible cognition of objects).

It does not follow from this disparity, however, that the categories relate, whereas representations of space and time don’t relate, a priori to objects of sensible cognition in general (objects of any sensible cognition whatsoever): for Kant, neither of them do. The categories do not relate a priori to objects of sensible cognition in general because the TM does not sanction the following inference even though its premises are held by Kant:

(1*) The categories are conditions under which alone is any sensible cognition of something as an object possible, “as far as the form of thinking is con-cerned.”

(2*) It is possible for us to have sensible cognition of something as an object.

Therefore, (3*) the categories relate a priori to objects of sensible cognition in general.

(1*) is analytic, and (2*) synthetic. Moreover, (2*) follows from what has figured as a premise in the Deduction, namely, that we do have empirical cognition of objects, plus the analytic truth that empirical cognition is of a sensible kind. But (3*) is supposed to justify relation of the categories not only to objects of empirical cognition, but also to things considered as objects of non-human sensible cognition. Hence, to infer (3*), we have yet to prove there to be non-human rational creatures in posses-sion of sensible cognition, which may or may not exemplify the same kind of sensible cognition as ours. But, it may be asked, what if such creatures do not exist? And even if they do, what if their cognition does not differ in kind from ours because its possi-bility is subject to the same (a priori) conditions as is the possibility of our cognition of objects (in spite of whatever biological differences there might be between our species and theirs)? I think Kant would reply to the two questions in the same way: it is both logically possible that non-human rational creatures exist and that they don’t, and both logically possible that their cognition differs in kind from ours and that it doesn’t. We simply have no a priori ground for determining which of the two possibi-lities in each pair has been realized rather than the other, nor for that matter have we any a priori ground for determining whether, as implied by (3*), the categories have relation even to things considered as objects of non-human sensible cognition. So, even if (1*) is accepted, (2*) does not justify us in moving on to (3*).

Contrasted with this failed inference to (3*) is a parallel, and also failed, inference in the case of representations of space and time as follows:

(1**) Representations of space and time are conditions under which alone is any sensible cognition of something as an object possible, “as far as the form of [intuiting] is concerned.”

27

(2*) It is possible for us to have sensible cognition of something as an object.

Therefore, (3**) representations of space and time relate a priori to objects of sensible cognition in general.

Like (1*), (1**) is analytic. But if both (1*) and (1**) are true, then there is only one logically possible kind of sensible cognition, which human cognition exemplifies, and in which we find the categories to be the form of thinking, and representations of space and time to be the form of intuiting. But, as mentioned above, for Kant it is both logically possible that cognition possessed by non-human rational creatures, if any, differs in kind from ours, and that it doesn’t. If both are logically possible, then it cannot be the case that ours is the only logically possible kind of sensible cognition. Hence, (1*) and (1**) cannot be both true. Kant accepts (1*); so he must reject (1**) and, therewith, the inference to (3**).

For Kant, the two inferences above are what we must appeal to in order to prove, by means of the TM, the a priori relation of the categories, and that of representations of space and time, to objects of sensible cognition in general. Since the TM is also taken by Kant to be the only possible method for proving the relation of any a priori representation to an object, it follows from the failure of the two inferences that, for him, neither the categories nor representations of space and time relate, with justifica-tion, to objects of sensible cognition in general a priori. But there is an important dif-ference here between the two sets of a priori representations. Although the categories don’t relate a priori to objects of sensible cognition in general, they could, if only non-human rational creatures were known to exist with their own cognitive apparatus. Not so for representations of space and time: given the falsity of (1**), they not only do not, but also can never relate to objects of sensible cognition in general even if such creatures were known to exit. It is worth exploring further implications of this difference for Kant’s critical doctrine about the boundaries of human cognition.

The fact that Kant would take (1*) in the first inference to be true and (1**) in the second false may be used to illustrate the disparity that I think he has recognized between the categories and representations of space and time. For Kant, if we are to justify the relation of the categories, or of representations of space and time, to objects of non-human sensible cognition at all, we must apply the TM by proving that these representations are conditions of the possibility of such cognition. Thus, insofar as it is not logically necessary that human intuition be the only kind of sensible intuition, (1**) is not analytically true, nor, for that matter, is it analytically true that space and time are conditions of the possibility of any non-human sensible intuition. As a result, the attempt to infer (3**) by means of the TM is doomed to failure, and we can never be justified a priori in applying representations of space and time to things considered as objects of non-human sensible intuition. Even so, Kant would insist—and think it of great importance to insist—that we can at least “think” objects of non-human sensible intuition by means of the categories. So “thought,” these objects fall, by definition, under what he calls “noumena in the negative sense”: “If by a noumenon we understand a thing insofar as it is not an object of our sensible intuition, because we abstract from the manner of our intuition of it, then this is a noumenon in the negative sense” (B 307). But we have seen that the inference to (3*) also fails. How, then, can we be justified in “thinking” objects of non-human sensible intuition by means of the categories?

28

It is in keeping with Kant’s Copernican way of thinking to surmise that if a non-human species of rational creatures were to exist, the objects of their sensible cognition would have to conform both to whatever is the form of their thinking and to whatever is the form of their sensible intuition. But given (1*), we know a priori that both their thinking and ours must have the same form, a form constituted by the categories; and, given the falsity of (1**), we also know a priori that their intuition and ours may have different forms. Nevertheless, since our thinking and theirs must have the same form, we know a priori that, in the imagined case, we can “think” objects of their intuition in the same way in which they “think” them: we can do so by abstracting from the relation of those objects to our intuition. In this way we are “thinking” objects as noumena in the negative sense: we know a priori that what we “think” by means of the categories could be objects considered as such noumena if only non-human rational creatures with their own intuition of objects were known to exist. Here we are applying the categories, unaccompanied by representations of space and time as forms of our sensibility, to objects beyond the sphere of “appear-ances” in a hypothetical or, in Kant’s term, “problematic” manner. As he puts it,

In the end…we have no insight into the possibility of…noumena [in the nega-tive sense], and the domain outside of the sphere of appearances is empty (for us), i.e., we have an understanding that extends farther than sensibility prob-lematically…. [A255/B310]

One might wonder why the same “problematic” extension cannot be accorded to representations of space and time. It might be thought that Kant denies a similar extension of representations of space and time to objects of non-human sensible intui-tion because he has taken human intuition to be sui generis. But isn’t it logically possible that some, or even all, non-human rational creatures share the same kind of sensibility with us, even though their species and ours may be physically constituted in radically different ways? If so, then, the falsehood of (1**) notwithstanding, repre-sentations of space and time could still apply to noumena in the negative sense if (some species of) non-human rational creatures were known to exist and such a logical possibility known to be realized. How would Kant reply to this objection?

Kant says of the categories that, unlike representations of space and time, they “extend to objects of intuition in general, whether the latter be similar to our own or not, as long as it is sensible and not intellectual” (B148; emphasis added). Here Kant uses the expression ‘objects of intuition in general’ to mean “objects of [sensible] intuition in general,” and leaves room for the logical possibility of sensible but non-human intuition. For Kant, then, it is indeed logically possible that there are other kinds of sensible intuition than that which, characterized by space and time as its forms, subsumes our intuition. But this logical possibility implies, as we have seen, the falsity of (1**). According to Kant’s TM, the falsity of (1**) is sufficient to dis-allow us ever to apply representations of space and time a priori, even if “prob-lematically,” to noumena in the negative sense, things that are not considered as objects of our intuition. In fact, an analogous conclusion would also be true of representations of space and time as they were used by a hypothetical non-human species of rational creatures who had the same kind of intuition as ours: given that Kant’s TM is the only possible method for proving the possibility of a priori cogni-tion of objects, these imagined creatures could never be justified in applying their representations of space and time, even if “problematically,” to objects of our intui-tion, for the simple reason that (1**) would not hold in their case either.

29

Let us say that an a priori representation in a subject has “intra-specific” applica-tion if it applies (or relates) to objects of intuition with which the subject’s own species is endowed, and that it has “inter-specific” application if it applies also to objects of intuition with which a species other than the subject’s own is endowed. Then the disparity between the categories and representations of space and time comes down to this: On the one hand, given the falsity of (1**), representations of space and time can never have inter-specific application even if they must have intra-specific application for any non-human rational creatures with the same kind of intuition as ours. On the other hand, given (1*), not only must the categories have intra-specific application, but they could have inter-specific application as well if rational creatures of some non-human species were known to exist. Whether or not representations of space and time, or the categories, can have inter-specific applica-tion is determined here according to whether or not such application can be sanc-tioned by the TM; and if it cannot be sanctioned by the method, it can never be sanctioned at all, since the method is supposed to be the only possible one for proving the relation of a priori representations to an object. To the extent that representations of space and time can never have inter-specific application because of the falsity of (1**), let us say that it is “TM-impossible” for them to have inter-specific application. Similarly, let us say that inter-specific application is not TM-impossible for the categories. This difference provides more perspicuous grounds than do the truth of (1*) and the falsity of (1**) for permitting the categories but not representations of space and time a “further extension…beyond our sensible intuition…to objects of intuition in general, whether the latter be similar to our own or not, as long as it is sensible and not intellectual.”

This “further extension” is held to be not only admissible but also necessary or un-avoidable: “necessary in order not to extend sensible intuition to things in themselves, and thus to limit the objective validity of sensible cognition” (A254/B310). In their privileged though “problematic” extension apart from representations of space and time, the categories become, as mentioned earlier, the concept of a noumenon in the negative sense. This concept is now assigned a strictly “negative” use, to function as “a boundary concept…to limit the pretension of sensibility” (A255/B310-1). We need this “boundary concept” at least to curb our misguided tendency to regard our sensibility as capable of going hand in hand inter-specifically with the understanding: the “boundary concept” may be said to serve as a reminder, instituted by a “critical” thinker but placed outside the boundaries of her cognition, that she should not be deluded by a mere logical possibility of non-human sensible intuition sharing the same form as ours into believing that our representations of space and time might have inter-specific application to noumena in the negative sense. As Kant would say here, “Now in this way our understanding acquires a negative expansion, i.e., it is not limited by sensibility, but rather limits it by calling things in themselves (not considered as appearances) noumena” (A256/B312; original emphasis).

It is important to note that, for Kant, we must, in order to entertain the logical possibility of things in themselves being in space and time, conceive of them as noumena in the negative sense, i.e., as objects of non-human sensible intuition. It follows that in holding that such a logical possibility may as a matter of fact be realized, we are committed to the TM-possibility of our representations of space and time applying inter-specifically to objects of intuition possessed by non-human rational creatures. But if Kant is right, such application is TM-impossible. Although it is not, logically speaking, self-contradictory to entertain the thought that noumena in

30

the negative sense may be in space and time, this is always an “empty” thought, a “mere play of representations.” Given that inter-specific application is TM-possible for the categories and yet TM-impossible for representations of space and time, it can be inferred within Kant’s framework that, although it is TM-possible for things in themselves (qua noumena in the negative sense) to be “thought” by us, it is TM-im-possible for them to conform to our representations of space and time; so is it TM-impossible for them to be in space and time as forms of our intuition. In this way, the ideality thesis is derived from the further extension of the categories unaccom-panied by representations of space and time beyond appearances. So derived, the ideality thesis does not conflict with the limitation thesis insofar as the intra-specific application of an a priori representation in us is a matter independent of whether or not its inter-specific application can ever be justified: whereas the ideality thesis concerns the TM-impossibility of inter-specific application of representations of space and time to things in themselves, the limitation thesis concerns human cognition of objects, and it requires that such cognition be cognition of appearances, not of things in themselves, simply because the categories and representations of space and time can be applied intra-specifically to appearances alone.

Finally, we may ask if the series of arguments that culminate in the ideality thesis can justify a bifurcation of “things in themselves” and “appearances” into two distinct worlds of objects, the latter being merely “representations” in us. At the beginning of §3 we saw Kant announce the limitation thesis, that “we can have cognition of no object as a thing in itself” (Bxxvi), in the Preface to the second edition of the Critique. Immediately after the announcement, Kant continues, “Yet the reservation must also be well noted, that even if we cannot cognize these same objects as things in themselves, we at least must be able to think them as things in themselves” (ibid.; italics added). The identity construal is still in use after Kant thinks he has reached the limitation thesis. And I think this is how it should be even after he has reached the ideality thesis. For one thing, Kant’s Copernican “experiment” is an attempt to prove the possibility of a priori cognitions of objects by assuming that “the objects must conform to our cognition” (Bxvi)—or, I take it, by assuming that the objects (in Strawson’s weighty sense) must conform to the forms of “intuiting” and “thinking” involved in our sensible cognition of them. The kind of object that must conform to these forms is what Kant later in the Preface says “should be taken in a twofold meaning, namely as appearance or as thing in itself” (Bxxvii). It would be contrary to Kant’s intention if it should turn out that “the objects [that] must conform to our cognition” must be considered only as “appearances” that are, one and all, inner and outer alike, mental entities in us. The transcendental ideality of space and time shows, not that “appearances” are, literally, in us, but only that, for a subject of human cognition, there is an ineliminable “subjective” aspect to the way in which we represent objects to ourselves through sensibility, given the TM-impossibility of inter-specific objective application of representations of space and time, and given the TM as the only possible method for proving the possibility of a priori cognitions of objects. If “subjectivity” is a suitable term here, it ought to attach not to objects considered as “appearances,” but to the representing of them by a human subject.

§9. Conclusion: A Copernican World-View for Rational Creatures

We have found that, at a stage after the completion of the Deduction and a parallel deduction of space and time, a disparity between the categories and representations of

31

space and time becomes visible regarding whether or not they are (a priori) condi-tions of the possibility of any sensible cognition of objects. This disparity leads to Kant’s thesis about the further extension of the categories, a thesis allowing them to leave behind representations of space and time and to carry us a priori though “pro-blematically” and in a strictly “negative” way to a non-spatiotemporal realm of things in themselves. On the basis of the interpretation suggested in the last section, it may be said that the resulting thesis of the transcendental ideality of space and time emerges after human cognition is envisaged as belonging only to one among a plurality of logically possible kinds of objective cognition attainable by rational creatures. Insofar as it is underpinned by the externally unique Deduction, the limitation thesis demands a kind of epistemological modesty that may be said to be a transcendental stance taken solely from within the cognitive boundaries for human beings. Without the additional thesis about the further extension of the categories, we remain subjects of human cognition, situated in a world of nature that we must regard as spatiotemporal. Epistemological modesty at this human level is compatible with an agnostic attitude toward the question whether or not things that affect our senses are in themselves in space and time. But once the further extension of the categories is brought out and carries us beyond the boundaries of human cognition, we are to take a transcendental stance as (finite) rational beings whereby we are compelled on a priori grounds to “think” objects through the pure understanding in abstraction from their relation to our sensibility, and consequently to let our formerly agnostic modesty resolve into a stronger one, which recognizes that no sensibility of any logically possible kind can ever contribute to a Copernican world-view of rational beings as such. Kant once remarked, “[T]he danger is not that I will be refuted, but that I will not be understood” (Bxliii). So, perhaps we should at least come to this conclusion, that whether or not it is refutable, it is the stronger modesty, which cannot be captured by the limitation thesis unaided by the ideality theses, that is truly and coherently Kantian.

Obviously, Kant holds dear the room, or “space” (A289/B345), allowed for the further extension of the categories even though the room must be left “empty” (because the categories acting in their “extended” capacity must nevertheless be treated as empty concepts, whose grounds for non-emptiness are only half-available, as explained in the last section): “we remain at liberty, indeed we are called upon by reason to fill [the empty room] if we can through practical data of reason” (Bxxi-ii). Kant is thinking here of a room for “cognitions a priori that are possible, but only from a practical standpoint” (Bxxi). But whether or not a non-spatiotemporal room is indeed left open and “empty,” waiting to be filled up “through practical data of reason,” depends on Kant’s thesis about the further extension of the categories apart from representations of space and time; and the latter depends in turn on the kind of disparity described in the last section between these two sets of a priori representa-tions. It is therefore lamentable, I must say, that Kant should argue for the trans-cendental ideality of space and time as early as in the Aesthetic, prior to the Deduction, rather than at a stage posterior to both.

32

References

Allison, Henry, 1983, Kant’s Transcendental Idealism, Yale University Press.

---------------, 1996, Idealism and Freedom, Cambridge University Press.

Bennett, Jonathan, 1966, Kant’s Analytic, Cambridge University Press.

Guyer, Paul, 1987, Kant and the Claims of Knowledge, Cambridge University Press.

Henrich, Dieter, 1976, Identität und Objektivität, Heidelberg: Karl Winter Uni-versitätsverlag.

Hume, David, A Treatise of Human Nature, ed. L. A. Selby-Bigge, Oxford University Press, 1978.

Kant, Immanuel, What Real Progress Has Metaphysics Made in Germany since the Time of Leibniz and Wolff, trans. Ted Humphrey, N.Y.: Abaris Books, 1983.

---------------, Critique of Pure Reason, trans. Paul Guyer and Allen W. Wood, Cambridge University Press, 1998.

---------------, Correspondence, trans. Arnulf Zweig, Cambridge University Press, 1999.

Kemp Smith, Norman, 1962, A Commentary to Kant’s Critique of Pure Reason, 2nd ed., New York: Humanities Press.

Kitcher, Patricia, 1982, “Kant on Self-Identity,” Philosophical Review 91: 41-72.

Körner, Stephan, 1955, Kant, London: Penguin Books.

Langton, Rae, 1998, Kantian Humility: Our Ignorance of Things in Themselves, Oxford University Press.

Strawson, Peter F., 1966, The Bounds of Sense, London: Methuen and Co.

Wolff, Robert Paul, 1963, Kant’s Theory of Mental Activity, Harvard University Press.