KIYOSHI CHIBA | Kants Ontologie der raumzeitlichen Wirklichkeit | Walter de Gruyter 2012


By Kiyoshi Chiba

Chris Onof’s detailed and penetrating objections (here and here) have clarified which parts of my book are unclear, misleading or hard to follow. I am pleased to have the opportunity to make my view and intentions clearer by responding to his objections. Onof’s objections concern various of my claims, explanations and arguments. In this short reply, I cannot examine them all, so I would like to concentrate especially on the following topics:

(1) The distinction between Cognition-Independence (CI) and Cognition-Transcendence (CT), and Onof’s favoured combination “CI and ¬CT”, (2) Onof’s objection to my interpretation of Kant’s solution of the antinomies, (3) the distinction between regress in infinitum and regress in indefinitum and … (4) the charge of subjective idealism (Onof’s “Worry W”) and the Fourth Paralogism of the A-version of the Critique of Pure Reason (CPR). (I use hereafter also the following abbreviations: TR for Transcendental Realism, TI for Transcendental Idealism, KORW for my book and OC for Onof’s critique of my book.)

1. Cognition-Independence (CI) and Cognition-Transcendence (CT)

In KORW, I defined realism as a position which claims: The truth of statements of the class in question is independent of our verification/verifiability; ontologically paraphrased: Existence of objects of the class in question is independent of our cognition/cognizability. Furthermore, I distinguished the notion ‘Cognition-Independence’, which characterizes realism, from the related notion ‘Cognition-Transcendence’, which says that it is possible that some objects of the class in question are absolutely unknowable. Following Onof’s abbreviation, hereafter I will call these notions respectively ‘CI’ and ‘CT’, and ‘Cognitive-Dependence’ and ‘Cognitive-Immanence’ respectively ‘¬CI’ and ‘¬CT’. According to these abbreviations, my anti-realist interpretation claims that TI assumes ¬CI regarding spatiotemporal objects.

Before discussing Onof’s view, I would like to answer his question. In Section 7, he criticizes me saying that it is not clear what I meant by “verifications”, which should function as truth-makers in the anti-realist framework, especially, whether I meant by it (a) possible or solely actual knowledge and (b) general or solely perceptual knowledge. Well, in my definition, the term ‘anti-realism’ does not denote a particular non-realist position such as Putnam’s ‘internal realism’, but rather a genus which embraces every counter-position to realism, so that anti-realism has various possible versions. Therefore, some versions admit only actual and perceptual knowledge—or even merely knowledge available to Me (solipsism)—as truth-maker, and some versions accept possible and general knowledge as well.

In Section 7 of OC, Onof complains that it is confusing that I classified Van Cleve’s ‘traditional idealism’ as a form of anti-realism. But there is no problem here according to my definition. Every position which assumes ¬CI is a version of anti-realism, and it is obvious that Van Cleve’s ‘traditional idealism’ assumes a kind of ¬CI.[1]

However, it is another question whether the version of anti-realism which I ascribed to Kant’s TI admits (a) possible or solely actual knowledge and (b) general or solely perceptual knowledge as truth-maker. As for question (a), it is clear (and Onof himself acknowledges it in Section 7 of OC) that my favoured version of anti-realism admits possible knowledge as truth-maker. As for question (b), my favoured version admits also general knowledge as truth-maker. I confess that I did not explain the notion of ‘verification’ in KORW. I took ‘verification’ as it is understood in standard philosophical discourse, namely as a sum of knowledge which epistemically justify a particular claim to knowledge. It involves perceptual knowledge, naturally, but also general knowledge, for example that of logic, mathematics and empirical sciences, as well as epistemological principles such as “Our perceptual knowledge is mostly reliable”. If one tries to characterize the notion of ‘verification’ more precisely, one will immediately encounter lots of problems, but I think the above-mentioned simple characterization is enough at least in order to understand the arguments in KORW.

Now, let’s examine Onof’s objection. He claims that TI should not be understood as ¬CI, but as the combination of CI and ¬CT (regarding spatiotemporal objects). This is the main thesis of OC as a whole, which is introduced in Section 5 (with respect to what Onof calls “Argument A”), and repeatedly argued for throughout OC. It is, however, not clear what Onof means by this combination exactly. My conclusion is that his position amounts either to the position akin to the metaphysical two-aspect view à la Allais/Rosefeldt or to the ¬CI-anti-realism that I defend.

At first glance, it seems obvious that Onof’s position, pace his denial, results in something like the metaphysical two-aspect interpretation à la Allais/Rosefeldt,[2] insofar as it assumes CI regarding spatiotemporal objects. For this is just what my argument which Onof calls “Argument A” has proved. Let me summarize this argument here:

Argument A begins with confirming the following: The Kantian appearance, whether understood as a distinct thing or as an aspect of a thing, must be subject-dependent in some sense;[3] interpretations which do not meet this condition, such as the outright realist interpretation of Rae Langton (1999), are to be rejected from the outset. On this premise, Argument A goes as follows: Insofar as an interpretation assumes CI regarding spatiotemporal objects, it must admit that spatiotemporal objects have not only a subject-dependent dimension (no matter what this may mean) but also a subject-independent dimension, just because of its CI-assumption. And in this case, it is bound to regard the former dimension as the appearance-aspect and the latter as the ‘in itself’-aspect of what we cognize as spatiotemporal objects, which is nothing other than what the metaphysical two-aspect interpretation maintains.

Onof does not attack particular steps of Argument A. His objection concerns rather the conclusion of this. He argues namely: Argument A does not exclude a version of the realist two-world interpretation which regards both spatiotemporal objects (appearances) and things in themselves as CI but the former as ¬CT and only the latter as CT.

This objection fails, I think, because Onof assumes here that CT—better: unknowability[4]—is a defining feature of things in themselves. This assumption is untenable. As Langton (1999:10ff.) rightly stresses, Kant’s claim that things in themselves are unknowable must not be understood as an analytical statement from the meaning of the term ‘things in themselves’. Just because of the non-analyticity of this claim, Kant had to advance arguments for the unknowability of things in themselves. If an interpretation cannot appreciate this point, this is not an ignorable minor error but a fatal mistake. Onof makes just this kind of mistake in his objection to Argument A.

As far as I understand his objection to Argument A, I have found no other argument. So it appears that we can conclude that Onof’s interpretive option must, just because of its CI-commitment, result in the metaphysical two-aspect interpretation.

In Section 10, however, Onof adds an essential feature of his position which he characterised as the CI/¬CT-combination before. It acknowledges ¬CI for determinations—I would rather say: properties—of spatiotemporal objects and assumes CI only for their existence. But then, which difference is there between us exactly, if any?

It depends on how Onof conceives of the individuation of spatiotemporal objects (in other words, of the particularness of particular spatiotemporal objects as such). Does he think that the individuation is CI or ¬CI? The answer seems obvious at first glance. It would be absurd to say that something is individuated independently of its properties; so, if all properties of something are ¬CI, so is its individuation. In Section 8 of OC, Onof really says that “there is no such individuated object in the realm of things in themselves”. However, there are some passages and expressions as well which arouse the suspicion that Onof considers the individuation rather as CI. The clearest case is found in footnote 10. Onof argues there that the principle of bivalence is valid for statements about the existence of particular spatiotemporal objects, but not for those about their properties. This difference comes from the circumstance that spatiotemporal objects are CI regarding their existence but ¬CI regarding their properties. This means, however, that Onof claims, at least in this footnote, that the state of affairs that particular spatiotemporal objects exist is CI. This means that Onof assumes that the individuation of spatiotemporal objects is CI. If he did not mean so, the difference between sentences about the existence of particular objects and those about their properties would make no difference regarding logic.

However, why is it important at all whether Onof regards individuation of spatiotemporal objects as CI or ¬CI? The reason is that, if Onof regards it as CI, his position is no different than the metaphysical two-aspect interpretation. And if he regards it as ¬CI, his position turns out to coincide with my ¬CI-anti-realism. I think the former disjunct is clear enough for readers of KORW, so I will discuss here only the latter disjunct.

The explanation is simple: If the individuation is ¬CI, then there can be no particular, that is, individuated, objects that exist independently of our cognition, so we cannot meaningfully assert that particular objects exist independently of our cognition. This point would be made more clearly if we talk of the truth of statements instead of the existence of objects. On the assumption of ¬CI regarding individuation, existential statements such as “There are two things in front of me”, cannot be true independently of our cognition, because the distinctness of particular spatiotemporal things is not possible independently of our cognition. Onof’s suggestion about the peculiarity of the category of existence (Section 11) does not change this point.

In fact, Onof’s position seems to amount to the interpretive option which I considered on p. 52 (footnote 76) and p. 85ff. (especially concerning Robinson 1994) of KORW. This option refrains from maintaining that individual cognition-independent things appear to us as particular spatiotemporal objects, and claims instead that the cognition-independent noumenal world (or reality-in-itself) appears to us as the spatiotemporal world, while regarding every individuation of particular spatiotemporal objects as cognition-dependent. And I classified this interpretive option as a version of the anti-realist interpretation (in the sense of my ¬CI-anti-realism). Additionally, it is noteworthy that this option was just the version of anti-realism I finally ascribed to Kant towards the end of Chapter 8 of KORW (p. 398ff.), which allows us to say that the phenomenal world is the appearance of the world of things in themselves.

So I would like to conclude this section by saying that, if Onof does not assume CI regarding individuation, his position is not substantially distinct from this interpretive option; especially, it cannot be said that his position is distinct from my ¬CI-anti-realist interpretation in that it assumes CI regarding the existence of spatiotemporal objects.

This suggests that there is no substantial difference between Onof and me, at least in the essential feature of anti-realism what we each will defend. However, this naturally does not exclude that there can still be some disagreements between us on particular interpretive points. I will discuss such points in the following.

2. Kant’s Solution of the Antinomies (re Section 6 and 8 of OC)

Onof’s objection regarding this topic concerns various points, but his main objection is that my interpretation makes Kant’s solution of the antinomies non-transcendental, that is, empirical. Before replying to this objection, I would like to comment on Onof’s following summary of my argument which he presents in Section 6 of OC:

(a1) The CI non-existence of totalities of series of conditions implies the CI-non-existence of spatiotemporal objects, which negates TR and thereby affirms TI.

(a2) TI additionally has an account of why for ¬CI existing objects, totalities of series of conditions cannot (¬CI) exist.

Since this is different from how I understand my own argument, I am not sure how this summary relates to my concrete argument. It may not be false (because it may be compatible with my argument on some understandings of it), but in order to prevent readers from possible misunderstanding, I modify this summary as I myself understand the argument in question:

(a1) TR is negated and TI is affirmed instead, because TR assumes the CI-existence of spatiotemporal objects, which inevitably leads to the existence of every member of series of conditions, which further amounts to the existence of totalities of series of conditions, which finally leads to the antinomies.[5]

(a2) TI’s assumption of the ¬CI-existence of spatiotemporal objects additionally has an account of why for spatiotemporal objects, totalities of series of conditions cannot exist.

Now, let’s examine Onof’s objection. I will argue two points: (1) Onof’s evaluation that according to my interpretation, not only infinite totalities but also very large finite objects turn out to be impossible and (2) his objection to my appeal to non-formal factors in explaining Kant’s solution of the antinomies.

(1) Onof criticizes my interpretation that it takes Kant’s rejection of absolute totalities to be based on “the limitations of our cognitive faculties understood in their empirical employment” (Section 6). There are indeed some passages in KORW which give such an impression. Onof refers to p. 135 (See OC, footnote 7[6]), but a clearer passage would be the following one: “The series of conditions is so large that we cannot run through its elements completely through regress; therefore, [according to anti-realism] its elements cannot be given completely” (KORW p. 133). In this passage, I did say that on the assumption of ¬CI, absolute totalities are impossible because they are too large for us to cognize. However, I added immediately after this passage that such an explanation is in fact problematic (see ibid.), and I discussed on p. 155 why this explanation is problematic. The reason was that even on the assumption of ¬CI, there should be no problem in cognising absolute totalities if they are finite, no matter how large they are. Moreover, if one further observes Chapter 7, it becomes perfectly clear, I hope, that I did not maintain that we cannot cognise absolute totalities just because of the empirical limitations of our cognitive faculty.[7] What is excluded on the assumption of ¬CI is only the possibility of infinite absolute totalities, because of the a priori impossibility of cognising them, just as Onof concludes.

One may wonder here: For what reason was such an inadequate, misleading explanation advanced on pp. 133–5 of KORW? The answer is: In that part, I dealt with the passage A498ff./B527ff., and Kant argues there simply that absolute totalities of series of conditions cannot be given because they go beyond our capacity of synthesis, without considering the possibility that they are finite.[8] In fact, Kant excludes this possibility in other passages not with an anti-realist argument (that which is uncognisable cannot be actual), but with metaphysical arguments I elucidated in Chapter 7 of KORW (pp. 256–9). But these metaphysical arguments are not mentioned in A498ff./B527ff. and Kant argues there rather as if the absolute totalities were rejected solely in response to the anti-realist argument, no matter whether they are infinite or finite. So I followed Kant in this point at least on pp. 133–5, in order to analyse Kant’s argument in A498ff./B527ff. This is why I intentionally provided there what I myself took to be an inadequate explanation. But, let me excuse myself, I did not forget to warn about its inadequacy there!

(2) In Chapter 7, I established that Kantian anti-realism should admit not only actual, but also possible verifications as truth-makers of statements about spatiotemporal objects. Now, Onof is right in pointing out that the notion of possibility I elaborated in Chapter 7 “clearly derived from [the notion] of actuality” (OC, Section 8). He thinks that this is further evidence for the non-transcendental character of my interpretation. His objection is directed to the point that I appealed to non-formal factors in my interpretation of Kant’s solutions of the antinomies. He thinks that Kant’s solution of the antinomies must be explained solely in appeal to formal factors such as “the limit on possible experience outlined in CPR” or “a truth of transcendental logic, namely that an endless temporal series of representations cannot be grasped under the transcendental unity of apperception” (OC, ibid.).

To reply to this objection, I would like to point out first: It is not false to say that both Chapter 4 and Chapter 7 of KORW concern Kant’s solution of the antinomies. There is, however, a significant difference in the tasks of both chapters. The task of Chapter 4 was to clarify why TR inevitably falls into antinomies. The answer was: because of its commitment to realism of spatiotemporal objects, and it followed from this that in order to avoid antinomies, TI must reject realism and assume some version of anti-realism. I suppose that it is especially this task that Onof calls “resolution of antinomies” in OC. And I would like to stress: As far as this task, above all, to explain why TI can reject absolute totalities of series of conditions, is concerned, we have only to appeal to the a priori impossibility of cognizing such totalities, as Onof does.

In contrast, the task of Chapter 7 is to explain how spatiotemporal reality is understood in Kant’s anti-realist ontology, in other words, to elucidate the meaning of ‘actuality’ with respect to spatiotemporal objects. True, this task partially relates to the solution of the antinomies, for not every version of anti-realism of spatiotemporal objects can solve the antinomies in the Kantian way, i.e., to conclude that both thesis and antithesis are false (see pp. 259–61 and 265–71), so I tried to elaborate versions of anti-realism which make such a solution possible. However, the main objective of Chapter 7 is to elaborate the concept of ‘actuality’ regarding spatiotemporal objects. For this purpose, the appeal to formal factors such as “the limits on possible experience outlined in CPR” or “a truth of transcendental logic” is obviously insufficient. Even Kant himself has to resort to material factors such as sensation, in order to define ‘actuality’ in the Postulates of Empirical Thinking in General. It was just for this reason, and not because I understood Kant’s solution of the antinomies as empirical, that I appealed to non-formal factors in Chapter 7.

3. Regress in infinitum/in indefinitum (re Section 9 of OC)

As for this topic, Onof advances two objections: My interpretation does not comport well with A519/B547, and more importantly, it misses the constitutive/regulative distinction between mathematical and dynamical principles of experience.

The first objection would be a good starting point. I confess that especially as to the topic of regress in infinitum/in indefinitum, there are some passages which seem to contradict my interpretation. As I pointed out at the beginning of Section 7.2.3 of KORW (p. 278), Kant’s account on this topic is not always consistent, and this suggests that Kant himself was not totally clear about this theme.[9] So I did not intend all along to advance an interpretation which explains Kant’s every statement about this topic consistently. I think we have to be satisfied if we can get an interpretation which makes Kant’s statements as a whole understood relatively well.

Because of these textual circumstances, I am not so confident of the rightness of my interpretation of this topic as of that of the others. I am willing to abandon my interpretation if a more convincing interpretation is proposed. It is a matter of comparison. And as far as Onof’s proposal in particular is concerned, I do not think that it is better than my interpretation. I will explain why. There are many interesting points to discuss, but let me choose here the following three points:

(1) The core idea of Onof’s distinction between regress in infinitum and regress in indefinitum is that the difference between them does not lie in whether every member of the series has a successor member or not (as I interpreted it), but in whether there are rules identifying the successor for any member of the series. I would like to point out first: I think it is too much to say with regard to regress in indefinitum that there is no rule identifying successors; we have no a priori ground to assert this positively. The possibility is not, and also need not be, excluded that there is such a rule but that it is not known to us. What Onof needs for his argument—and especially for explaining the constitutive/regulative distinction—is, as far as I understand his argument, only this kind of epistemic difference between two sorts of regress, that is, whether we know such rules or not.

But this point turns out to be a disadvantage of Onof’s proposal. This epistemic difference cannot affect the ontological difference of the sorts of series of conditions which are to be cognised through those two types of regress. For once it is guaranteed that every member of a series of conditions has successor members, we do know from this alone that the series extends itself in infinitum, no matter whether we can concretely identify its members or not. This means that Onof’s proposal misses the ontological import of the distinction between regress in infinitum and regress in indefinitum.

(2) Onof assumes that in case of a regress in indefinitum, namely (R′), we know that every member of a series of conditions has successors but do not know how to identify them. There is an unclear point here: Does he allow the possibility in case of regress in indefinitum that some successors cannot be identified, namely, cognised? If he does, he must allow that there can be uncognisable spatiotemporal objects, which contradicts ¬CT which he advocates. If he does not, he must admit that even in case of regress in indefinitum, every member of the series of conditions can be identified somehow, even though we do not know law-like relations between the members. In this case, however, any substantial difference between Onof’s (Q′) and (R′) vanishes. This arouses the suspicion that Onof’s proposal does not really enable him to explain the constitutive/regulative distinction he means to explain.

(3) There are some problems—or at least insufficiencies—in Onof’s treatment of the issue of logic. For example, he assumes simply, without any argument, that in his ¬CT-anti-realism, the principle of bivalence (which says that every statement is determinately either true or false) is retained unproblematically (see footnote 8 of OC). However, this is in fact highly problematic (if not obviously absurd), for, on the assumption of ¬CT, one can easily construct the following argument against the principle of bivalence:

According to ¬CT, a statement is verifiable if it is true, and it is falsifiable if it is false. As a result, the principle of bivalence means that every statement is either verifiable or falsifiable. However, we do not know how to justify such a general statement; in any case, the principle of bivalence cannot be affirmed on the assumption of ¬CT, until that general statement is established.

In our personal correspondence, Onof replied to this objection that it is assumed in his ¬CT-anti-realism that every spatiotemporal object is knowable, so that every statement about spatiotemporal objects is either verifiable or falsifiable. (I understand that by “every spatiotemporal object” he meant individual spatiotemporal objects; I think this does not distort his intention.)

However, this is not yet enough to defend the principle of bivalence in his system. Even if every individual object should be knowable, it does not mean that everything about spatiotemporal reality is knowable; for example, general statements such as empirical laws of nature may remain unknown. To explain this circumstance, it is instructive to consider an analogous case in mathematics: Let’s think about Goldbach’s Conjecture (“Every even number is a sum of two prime numbers”). We know that for every individual even number (however large), it is decidable whether it meets Goldbach’s Conjecture or not. However, this does not mean that Goldbach’s Conjecture itself is decidable (and it indeed has been undecided up until now, 2013). The lesson we should learn from this case is that the knowability of every individual object alone does not guarantee what is required for affirming the principle of bivalence in ¬CT, namely, that everything about the spatiotemporal reality is knowable. This consideration suggests that it is, at least, highly questionable that the principle of bivalence can be affirmed on the assumption of ¬CT.[10][11]

I would like to stress: What I am now considering is not just a superficial, technical issue. The question which logic should be assumed affects how Onof’s proposal itself should be understood.[12] He says nothing about which logic he assumes in his formalization of (Q′) and (R′), so let’s examine thinkable alternatives.

Let’s consider that, as was assumed by Onof himself, that his ¬CT-anti-realism adopts a bivalent logic. In this case, it is very natural to think that this logic is the standard logic which is taught in normal textbooks of predicate logic, that is, the classical logic with the classical bivalent semantics. I would like to ask readers: How many of you have even tried to think what comes about if Onof’s formalization is understood in other than the standard textbook-logic? On the assumption of classical bivalent semantics, the domain of quantification is regarded as a set in the classical sense, that is, as a set all members of which are given completely. However, affirming the first half of Onof’s (Q′) and (R′), i.e., “Every member [of the series] has successor members [xy (Sx ⟶ (Sy and y is successor of x))]”, on this assumption, amounts to saying that the series makes up a non-finite totality, that is, it is actual-infinite. If Onof’s proposal is understood in this way, we have to say that it fails to explain not only the difference between regress in infinitum and regress in indefinitum, but just Kant’s solution of the antinomies itself.

Let’s try then the other possibility that Onof’s ¬CT-anti-realism must adopt a non-bivalent logic, as I argued above. The candidate which is most easily thought of as such a logic based on ¬CT would be intuitionistic logic. Do you think, dear readers, that this step of my argument sounds dogmatic? If you think so, you should consider how difficult it is to propose another candidate concretely! But, if Onof’s formalization is understood intuitionistically, the first half of (Q’) und (R’) amounts this time to saying that for every member of the series, we are able to identify its successor members.[13] So the difference between (Q′) and (R′) vanishes again, this time only by a consideration of the adequate logic for understanding Onof’s formalization.

I confess that neither of the above-mentioned arguments is conclusive. It may be possible to find a bivalent logic without assuming domains of quantification as classical sets or to construct a non-bivalent anti-realist logic which is essentially different from an intuitionistic one. But it is not obvious that such logics are possible at all, still less that such logics support Onof’s proposal. One thing is clear: Until he fixes an adequate logic for his formalization, we cannot even understand what his proposal really amounts to.

For the reasons I have discussed so far, I do not think that Onof’s proposal, at least as it stands in OC, is a better one than mine for explaining the difference between regress in infinitum and regress in indefinitum. However, I do not intend to deny the possibility that Onof’s idea may be further elaborated into an interesting and attractive system which is essentially different from my interpretation.[14] It is naturally better, also for me, that many candidates for anti-realist interpretations are available.

4. The Charge of Subjective Idealism and the Fourth Paralogism (re Section 4 and 10 of OC)

Onof expresses a suspicion (which he calls “Worry W”) that the CI-anti-realism I defended is a kind of subjective idealism. He means by this the position that maintains that “appearances are nothing but representations” (Section 4), namely, the position which simply identifies spatiotemporal objects with representations.[15] Although this position is clearly a version of what I call “anti-realism”, it is not the version I ascribed to Kant’s TI. I explicitly pointed out on p. 171 of KORW that Kant’s expressions which simply identify objects with representations should not be taken literally (and explained the reason why as well).

In Section 10 of OC, Onof also admits this. So the point of his suspicion would be, to be exact, not that my CI-anti-realism is a kind of subjective idealism in this sense, but rather that it was confusing that I brought up such passages as textual evidence for an anti-realist interpretation even though I thought that they were inadequate for Kantian anti-realism. Onof says that “an account is needed” for this way of arguing (OC, Section 8). I agree, and here I would like to take the opportunity of explaining this point.

I would like to stress: If certain passages in Kant’s works turn out to be inexact or inadequate for his considered view, this does not mean that such passages cannot be used as textual evidence for any kind of interpretation. An inadequate expression is an inadequate expression of something. Mostly, this “something” is discernable. Do you think, for example, that the passages which identify spatiotemporal objects with representations are inadequate expressions of a realist ontology? This would be a highly implausible, even unreasonable, reading. It would be much more plausible to think that such expressions indicate at least some anti-realist ontology, even if in an inadequately exaggerated, misleading way. For this reason, I brought up the kinds of passage in question (and also some passages which suggest excessively subjectivist versions of anti-realism, even solipsism) as textual evidence for my interpretation. I think that such an interpretive procedure is not illegitimate; otherwise we would have to concede that almost no passage in Kant’s works can be used as textual evidence for any interpretation.

To bring this reply to a close, I will discuss one topic regarding the Fourth Paralogism. At the beginning of Section 10 of OC, Onof describes my argument in Chapter 5 of KORW (more concretely, Section 5.2.2) as follows: “Although for Chiba the realist can provide a plausible interpretation of much of the text before A372, he points out a key passage in support of his interpretation to be Kant’s claim […] that ‘their reality [sc. the reality of spatiotemporal things], just as much as that of my own thoughts, rests on immediate consciousness’ (A372).” This description suggests that my argument for an anti-realist interpretation of the Fourth Paralogism depended solely, or at least mainly, on the individual passage in A372.

I have to say that this is a serious misrepresentation of my argument. I am generally opposed to the interpretive strategy of justifying (or criticizing) a certain interpretation only with quoting individual passages; instead, I repeatedly stressed the importance to consider the context of Kant’s arguments (see for example, p. 41, 90ff., 94, etc. of KORW). My interpretation of the Fourth Paralogism was no exception. That is to say, I did not argue that, as Onof presents it, Kant’s argument other than the passage in question in A372 can be understood well in accordance with a realist interpretation, but rather that the realist interpretation cannot make the whole argument in the Fourth Paralogism understandable, even though it may succeed in pretending as if some individual passages, isolated from the context of the argument, could allow for a realist reading. I made this point explicit on p. 176ff. of KORW.

There still remain many points to discuss, but because of the limit of space and time, I have to end my reply here. I am very glad that I could make many points clearer in responding to Onof’s objections and get many new insights as well. I thank Christian Onof for his careful examination of my book and stimulating objections.



Robinson, H. (1996) ‘Kantian Appearances and Intentional Objects’, Kant-Studien 87: 448–54.

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

Van Cleve, James (1999) Problems from Kant, Oxford: Oxford University Press.

[1] Like many Kant-scholars, Onof makes a typical mistake when he says: “[. . .] actual perceptual knowledge of objects is required for objects to exist, which is a form of dogmatic idealism à la Berkeley” (OC, Section 7, italics mine). As I clarified in Section 5.2.3 of KORW (p. 180ff.), the definitional feature of the Dogmatic Idealism is only to deny the existence of spatial things. Furthermore, I think Onof makes a double mistake when he says that Van Cleve’s “traditional idealism” “certainly includes Kant’s dogmatic idealism” (ibid.). First, “traditional idealism” does not deny the existence of special things, and secondly, even if the first point is ignored, “traditional idealism” need not restrict itself within actual perceptual knowledge; according to the characterisation of Van Cleve (1999:12-14), “traditional idealism” is traditional just in comparison with “contemporary anti-realism” à la Dummett (ibid.; anyway, in Van Cleve’s sense and not in mine), and it embraces not only actuality-based phenomenalism but also a possibility-based one, furthermore, not only “ontological phenomenalism” but also “analytical phenomenalism” (see Van Cleve 1999: 71). Unfortunately, the subtlety of Van Cleve’s characterisation of idealism/phenomenalism has been overlooked by many critics of his interpretation (typically by Ameriks 2005; see footnote 46 (p. 34) of KORW).

[2] Or still worse, to the “two-aspect-plus-another-world” view which I described on p. 88 of KORW, namely, the interpretive option which assumes an appearance-aspect and an ‘in itself’-aspect of what we experience as spatiotemporal objects, and additionally distinctive things in itself that do not appear to us. However, I think it is obvious that such a position is not substantially distinguishable from the metaphysical two-aspect interpretation, so I will ignore this option hereafter.

[3] I did not stipulate in what sense; this is the task of advocates of each interpretation. It is noteworthy that Argument A works without stipulating this point.

[4] Pace Onof, cognition-transcendence and unknowability are not identical. The former, in my definition (see p. 19 of KORW), allows that some objects of the class in question are knowable. In contrast, things in themselves are totally unknowable. However, I ignore this point in this reply.

[5] I would like to stress especially that it is not the reason for rejecting TR that on TR’s assumption of the CI-existence of spatiotemporal objects, non-existence of totalities of series of conditions implies non-existence of spatiotemporal objects.

[6] In this footnote, by the way, Onof regards even my consideration of super-tasks (“the possibility of the completion of an infinite number of tasks in a finite time”) as an example of empirical matters. I think Onof’s goes too far here. If he really thought that the impossibility of super-tasks is only an empirical matter, he could not claim, as he actually does in Section 8 of OC, that it is an a priori impossibility that we cannot go through the infinite totality (because for claiming this, Onof would have to deny the possibility of super-tasks, which is, however, ex hypothesi an empirical matter).

[7] For example, see the following points: (1) My characterisation of a “possible verification-route [mögliche Verifikationsroute]” (p. 293), which allows the factually impossible possibility that a cognitive subject will live endlessly. (2) Footnote 430 (p. 300ff.), which says that the assumption of possible verification-routes admits finite idealisation of human cognitive faculty. (3) The system ZN+ (p. 320) acknowledges as truth-makers not only possible verifications of actual cognitive subjects, but also those of possible cognitive subjects (which I called “possible co-subjects [mögliche Mitsubjekte]”).

[8] Interestingly, this is the case not only with Kant in A498ff./B527ff., but also with Onof in his entire argument in Section 8 of OC.

[9] This circumstance is well understandable in my interpretation, because Kant did not have any well-constructed system of non-bivalent logic.

[10] In Section 7.4.1 of KORW, I examined and rejected a strategy which purports to justify the principle of bivalence within the framework of anti-realism by appealing to the determinism of the phenomenal world (which seems to have the consequence that every individual spatiotemporal object is knowable). I still do not think that that examination was totally redundant, but I have noticed this time that the principle of bivalence can be rejected generally with the argument I advanced above, without any appeal to the examination of the problematic of the determinism. I am thankful to Onof for giving a clue to this idea.

[11] Afterthought on 12 December 2013: In our latest correspondence, Onof showed me a sketch of an argument for defending the principle of bivalence in his system against my discussion here, also for empirical laws. I expect that he will develop this argument at some point in the future.

[12] Here, I would like to correct one point of Onof’s description of my argument: At the beginning of Section 9 of OC, he depicts (Q) and (R) and says “If this [contrast of (Q) and (R)] is indeed the contrast between the two types of regress, then this provides a powerful argument against the adoption of a bivalent logic to interpret Kant”. However, I did not intend to present this contrast as an argument against the principle of bivalence. When I discussed this topic on p. 304ff., the principle of bivalence had already been rejected on p. 302 in the system of ZN (the time-neutral version of anti-realism). Therefore, in order to advance (Q) and (R), it was not necessary to justify the rejection of a bivalent logic in addition. If this circumstance is taken into consideration, it would appear more natural to interpret the difference between regress in infinitum and regress in indefinitum in the way I explained with (Q) and (R).

[13] In a more precisely intuitionistic way of saying: For every member of the series, we have effective procedures to discover its successor members.

[14] Afterthought on 12 December 2013: In our latest correspondence, Onof actually suggested his idea for constructing such a logic, based on Kant’s consideration about a distinction between general and transcendental logic. I expect that Onof will develop this idea at some point.

[15] In Section 10, he describes subjective idealism a little differently: “[T]he claim that the ‘reality’ of appearances, ‘just as much as that of my own thoughts, rests on immediate consciousness’ (A372), could be read in a straightforward subjective idealist fashion […]”. This description suggests that what Onof regards as subjective idealism is rather the version of anti-realism which admits only actual verifications (“immediate consciousness”) as truth-makers. (Note that this position as well as the quotation from A372 do not necessarily commit one to the simple identification of spatiotemporal objects with representations.) As was already discussed, what I ascribed to Kant’s TI is not subjective idealism in this sense either.

Acknowledgements: I am very thankful to Christian Onof for thoughtful comments on this reply as well as advices for improving my text.

© 2013, Kiyoshi Chiba

Kiyoshi Chiba is Associate Professor of Philosophy at the University of Yamagata, Japan. Chiba obtained his Ph.D. in Philosophy in 2010 from the University of Bonn, Germany. His dissertation “Kants Ontologie der raumzeitlichen Wirklichkeit. Versuch einer anti-realistischen Interpretation der Kritik der reinen Vernunft”, the book under discussion here, was published by Walter de Gruyter in 2012 in the series Kant-Studien Ergänzungshefte and can be purchased here.