¹ Recently, Allen Wood has termed these the ‘causality’ and ‘identity’ interpretations respectively (Kant (Oxford: Blackwell Publishing, 2005), pp. 63–76)Google Scholar . I agree with Wood that the label ‘two-world’ is a misnomer, since it is applicable to both versions; but, for reasons that I cannot get into here, I am not convinced that his proposal is more useful than the standard terminology for characterizing the contrasting interpretations of transcendental idealism. Thus, I shall, for the most part, continue to use the latter.

² In arguing for a non-metaphysical interpretation of transcendental idealism, I do not intend to deny that this idealism has important ontological or, more broadly, metaphysical implications. Clearly, the arguments of the Aesthetic, Analytic and Dialectic, all of which are intimately connected with transcendental idealism, have such implications and were intended by Kant to have them. Thus, if anyone wishes to preserve the term ‘metaphysical’ for Kant's central claims I have no objection. In fact, there would be ample Kantian support for doing so. As will become clear in due course, what I wish to insist upon here is simply that transcendental idealism is not it self to be understood as a metaphysical theory that affirms that the phenomenal has a lesser degree or kind of reality than the noumenal.

³ This approach is compatible with, but distinct from, my previous treatments of the topic, the most recent and comprehensive of which is to be found in Kant's Transcendental Idealism, revised and enlarged edition (New Haven: Yale University Press, 2004), especially pp. 27–34Google Scholar .

⁴ The view I am here attributing to Kant has obvious affinities with the position which Hilary Putnam terms ‘internal realism’, and which he regards as Kantian. I am not sure, however, to what extent Putnam would be willing to accept my reading of Kant as an account of what Kant actually held as opposed to what he should have held. For a Useful discussion of Putnam's ‘internal realism’ and its relation to Kant see Moran, Dermot, ‘Hilary Putnam and Immanuel Kant: two “internal realists”?’, Synthese 123 (2000), pp. 65–104CrossRefGoogle Scholar .

⁵ All references to the Critique of Pure Reason are to the standard A/B pagination of the first and second editions and cite the translation of Guyer, Paul and Wood, Allen, The Cambridge Edition of the Works of Immanuel Kant (Cambridge: Cambridge University Press, 1998)Google Scholar . References to other works of Kant are to the volume and page of Kants gesammelte Schriften, herausgegeben von der Deutschen (formerly Königlichen Preussischen) Akademie der Wissenschaften, 29 volumes (Berlin: Walter de Gruyter (and predecessors), 1902 ff). Citations from the translation of Kant's Inaugural Dissertation (abbreviated as ID) are to the translation by Walford, David, The Cambridge Edition of the Works of Immanuel Kant, Theoretical Philosophy 1755-1770, translated and edited by Walford, David in collaboration with Ralf Meerbote (Cambridge: Cambridge University Press, 1992)Google Scholar ; from the Prolegomena to any Future Metaphysics (abbreviated as Pro) are to the Hatfield, Gary translation, The Cambridge Edition of the Works of Immanuel Kant, Theoretical Philosophy after 1781, edited by Allison, Henry and Heath, Peter (Cambridge: Cambridge University Press, 2001)Google Scholar .

⁶ This seems to be denied by Ameriks, who at least at one point characterizes transcendental realism as a ‘particular metaphysical position’, albeit without further identifying the position in question. See Ameriks, Karl, Kant and the Fate of Autonomy (Cambridge: Cambridge University Press, 2000), p. 290CrossRefGoogle Scholar .

⁷ Interpreters who take this view include Turbayne, Colin, ‘Kant's refutation of dogmatic idealism’, Philosophical Quarterly 5 (1955), 228CrossRefGoogle Scholar , and Al-Azm, Sadik J., The Origins of Kant's Argument in the Antinomies (Oxford: Oxford University Press, 1972), p. 148Google Scholar .

⁸ I say stipulative definitions, since Kant offers a significantly different one in each edition, a point which is often overlooked because of their partial overlap. In the first edition, Kant writes: ‘I call all cognition transcendental that is occupied not so much with objects but rather with our a priori concepts of objects in general’ (A12 ). In the second, transcendental cognition is defined as that which ‘is occupied not so much with objects but rather with our mode of cognition of objects insofar as this is to be possible a priori’ (B25). Although these definitions have been frequently discussed in the German, philologically oriented literature, the most thorough treatment of the subject is by Pinder, Tilmann, ‘Kant's Begriff der transzendentalen Erkenntnis’, Kant-Studien 77 (1986), 1–40CrossRefGoogle Scholar . According to Pinder, in the A version Kant is trying to indicate that the central focus of transcendental cognition and, therefore, of the Critique itself will be on our a priori concepts of objects rather than on objects (or things) themselves, which would characterize the ontological approach. Since a concern with such concepts involves also one with the objects (if any) supposedly falling under them, it will be concerned (albeit indirectly) with the latter as well. Thus, Kant's use of the ‘not so much … but rather’ [‘nicht sowohl… sondern’] locution. By contrast, in the B version, Pinder thinks that Kant's focus has shifted to a more narrow concern with the possibility of synthetic a priori judgments, which reflects the central concern of the Prolegomena Since the details of this shift, as important as they may be for an understanding of the development of Kant's thought, are not directly relevant to the concern of this article, I have attempted to provide a characterization of Kant's definitions that covers both versions.

⁹ Wolff describes ontology, which he equates with first philosophy, as ‘that part of philosophy which treats of being in general and of the general affections of being’. And he thereby defines it as ‘the science of being in general, or insofar as it is being’. ( Wolff, Christian, Preliminary Discourse on Philosophy in General, 72, translated by Blackwell, Richard J. (Indianapolis and New York: The Bobbs-Merrill Company Inc., 1983), p. 39)Google Scholar . Similar formulations are to be found elsewhere in Wolff and in Baumgarten.

¹⁰ This conception of the transcendental is obviously at work in Kant's dismissive treatment of the transcendentalia of scholastic metaphysics (Bl 12 -16 ).

¹¹ That Kant lists four possibilities, rather than merely the three that I suggested in the first edition of Kant's Transcendental Idealism, has been noted by Falkenstein, Lome, Kant's lntuitionism: A Commentary on the Transcendental Aesthetic (Toronto: University of Toronto Press, 1995), p. 147Google Scholar . As he correctly notes, this was already pointed out by Vaihinger, both Hans, Commentar zu Kants Kritik der reinen Vernunft 2 (Stuttgart: W. Spemann, 1881-1892), pp. 131–4Google Scholar , and Martin, Gottfried, Kant's Metaphysics and Theory of Science, trans. Lucas, Peter (Manchester: Manchester University Press, 1955), pp. 11–12Google Scholar . Moreover, there are several other texts in which Kant clearly distinguishes between these possibilities, including ID 2: 400 and 403, Reflexion 5298: 18, 146-7, and Reflexion 5404: 18, 174. Nevertheless, at least from the time of the Dissertation, Kant effectively assumed that the only two alternatives worthy of serious consideration were the Newtonian and the Leibnizian positions.

¹² It might be wondered why Kant should claim that the Leibnizians ‘ontol-ogized’ space and time in this sense, since, like Kant, Leibniz held that they were ‘ideal’ in the sense that they pertain only to phenomena. Quite apart from the question of the adequacy of his interpretation of Leibniz, however, it is clear that Kant's Leibniz is committed to the thesis that spatiotemporal predicates (properly construed) are applicable to things in general. As Kant saw it, this is because ‘Leibniz intellectualized the appearances’ (A271/B327 ), by which Kant meant that for Leibniz the difference between what ‘appears’ or is sensibly represented and what is grasped intellectually or conceptually is a matter of degree of clarity and distinctness rather than of kind. Thus, although Kant was well aware that Leibnizian monads are not in space and time, he also insisted that for Leibniz the spatiotemporal relations holding between the ‘phenomena bene fundata’ are reducible in principle (though not for us) to the purely conceptual relations supposedly holding at the monadological level. Moreover, the latter relations are clearly thought by the Leibnizians to apply to things in general. I discuss this issue in Kant's Transcendental Idealism (2004 edn), pp. 29-31 and at greater length in ‘Kant and the two dogmas of rationalism’ (to appear in Blackwell Companion to the Rationalists, edited by Alan Nelson).

¹³ I emphasize the Aesthetic because it is here that the ontological reading of Kant's ideality thesis seems most compelling. Thus, if this reading can be challenged here, the stage is set for a comprehensive non-metaphysical interpretation of transcendental idealism.

¹⁴ Once again, if anyone wishes to insist that this remains a move within ontology because it involves a global rejection of the generally accepted ontological alternatives, I have no objection. I would point out, however, that it fundamentally changes the nature of the game by transforming what were formerly regarded as ontological into epistemic conditions.

¹⁵ Kant holds open the (logical) possibility of both finite cognizers with forms of sensibility other than space and time and of a non-sensible (intellectual) mode of intuition.

¹⁶ I analyze and attempt to defend , Kant's argument in Kant's Transcendental Idealism (2000 edn), esp.pp. 122–32Google Scholar . Here, I wish merely to point out that a direct and important corollary of this reading is the assignment of a central place to the intuition arguments of the Aesthetic, since it is through these alone that Kant attempts to link the representations of space and time with human sensibility. Assuming their apriority, if space and time were assigned to the understanding rather than to sensibility, Kant would have had to conclude that they are predicable of things in general. Accordingly, it is of no little significance to the Critical project, for Kant to be able to show that, ‘Space is not a discursive or, as is said, general concept of relations of things in general, but a pure intuition’ (A24-5/B39).

¹⁷ In addition to the passages cited below, Kant refers to a putative transcendental use of the pure concepts and/or their associated principles at A139/B178, A19/B266, A242, A246/B303, A247/B304, A296/B352-3, A402-3, A515/B544. In most of these places it is contrasted with a legitimate empirical use. In the Transcendental Deduction, however, Kant views this distinction in a quite different way with respect to the faculties of sense, imagination and apperception, each of which is claimed to have a legitimate transcendental use as well as an empirical one (A94/B127 ). But here ‘transcendental’ refers to their function as conditions of the possibility of experience, which obviously does not involve any reference to things in general.

¹⁸ Although Kant explicitly denies that the pure categories can be defined at A245, he there also states that they are ‘nothing other than the representations of things in general, insofar as the manifold of their intuition must be thought through one or another of these logical functions’. For Kant's definitions, see A93/B126, A248/B305, A253 and A290/B346. Thus, following Lewis White Beck, I think it best to regard Kant as providing a nominal definition of the categories. See , Beck, ‘Kant's Theory of Definition’, Studies in the Philosophy of Kant (Indianapolis, New York, Kansas City: The Bobbs-Merrill Company Inc., 1965), pp. 61–73Google Scholar .

¹⁹ See, for example, Bxxvii, where Kant claims that if the distinction between things as objects of experience and the very same things as things in themselves were not drawn, then the principle of causality would be valid of things in general as efficient causes; and B410, where Kant suggests that if the rational psychologists were right, synthetic propositions ‘could reach as far as things in general and in themselves’.

²⁰ See A324-5/B38 O-82.

²¹ , Guyer, Kant and the Claims of Knowledge (Cambridge: Cambridge University Press, 1987), p. 333CrossRefGoogle Scholar .

²² Ibid., p. 336.

²³ Kant himself explicitly says as much when he remarks in response to the Garve-Feder Review: ‘The principle that governs and determines my idealism throughout is … All cognition of things out of mere pure understanding or pure reason is nothing but sheer illusion, and there is truth only in experience’ (Pro 4: 374).

²⁴ By a ‘pure understanding’ Kant here means one which, unlike ours, operates independently of the conditions of sensibility, that is, one which purports to cognize objects through the pure or unschematized categories. This locution is especially prominent in the Phenomena and Noumena chapter.

²⁵ A related but somewhat different version of the triviality objection has been voiced recently by Langton, Rae, Kantian Humility, Our Ignorance of Things in Themselves (Oxford: Oxford University Press, 1998), pp. 8–12Google Scholar . Focusing on Kant's denial that we can have knowledge of things in themselves (which is what she understands by ‘Kantian humility’) rather than on their non-spatiotemporality, Langton argues that, on my reading, this supposedly momentous discovery on Kant's part reduces to the trivial analytic claim that we cannot know things in themselves because doing so would (by definition) mean knowing them in abstraction from the conditions of our knowledge. My response is that Kant's revolutionary and certainly non-trivial claim is that our cognition is governed by sensible conditions. Granted, given this, together with Kant's account of what knowledge of things in themselves (or as they are in themselves) would require, the unknowability thesis follows. But this hardly makes the latter claim trivial, particularly since the transcendental realists whom Kant was attacking did not acknowledge that human cognition is subject to sensible conditions in anything like the sense insisted upon by Kant and, as a result, they assumed that we could cognize things as they are in themselves.

²⁶ For a recent statement of this line of objection, see Howell, Robert, ‘The conundrum of the object and other problems from Kant’, Kantian Review 8 (2004), p. 120CrossRefGoogle Scholar .

²⁷ I have treated this topic in some detail in ‘The non-spatiality of things in themselves for Kant’, Journal of the History of Philosophy 14 (1976), pp. 313–21CrossRefGoogle Scholar ; Kant's Transcendental Idealism (1983 edn), pp. 111–14Google Scholar ; Idealism and Freedom (Cambridge: Cambridge University Press, 1997), pp. 8–11Google Scholar ; and in Kant's Transcendental Idealism (2004 edn), pp. 128–32Google Scholar . Recently, my analysis of this issue has been challenged by Falkenstein, Lome, Kant's Intuitionism: A Commentary on the Transcendental Aesthetic (Toronto: University of Toronto Press, 1995), esp. pp. 301-5Google Scholar . According to Falkenstein, I ignore the possibility that a transcendentally real space might be just like the space of human sensibility except for its dependence on the latter. For my response to this criticism, see Kant's Transcendental Idealism (2004 edn), pp. 130–2Google Scholar .

²⁹ The point here is the same as in the distinction drawn above between the ways in which the concepts of things in general and of things as they are i n themselves involve an independence from the conditions of sensibility. The critic is, in effect, treating Kant's claim about the latter (the non-spatiotemporality of things as they are in themselves) as if it were about the former.

³⁰ I initially appealed to this analogy in Kant's Transcendental Idealism (1983 edn), pp. 241–2Google Scholar . It has been criticized by Cleve, James Van, Problems from Kant (Oxford: Oxford University Press, 1999), pp. 147–9Google Scholar . I defend my use of this analogy against Cleve's, Van criticisms in Kant's Transcendental Idealism (2004 edn), pp. 42–5Google Scholar .

³¹ P. F. Strawson expresses this view with admirable succinctness when he defines transcendental idealism as the doctrine that ‘reality is supersensible and that we can have no knowledge of it’ (The Bounds of Sense (London: Methuen, 1966), p. 38)Google Scholar .

³² The classical formulation of this dilemma is by Prichard, H. A., Kant's Theory of Knowledge (Oxford: Clarendon Press, 1909), esp. pp. 71–100Google Scholar . Kant clearly was in the grip of this picture in the Dissertation, when he claims that ‘things which are thought sensitively are representations of things as they appear, while things which are intellectual are representations of things as they are’ (ID 2: 292 ).

³⁴ Kant underscores this point in the introductory portion of the Transcendental Deduction common to both editions, when he notes that the seemingly unrestricted scope of the categories ‘not only arouses suspicion about the objective validity and limits of their use but also makes the concept of space ambiguous by inclining us to use it beyond the conditions of sensible intuition, on which account a transcendental deduction of it was also needed above’ (A88/B12 O-1 ).

³⁵ This is the noumenon in the negative sense, which is just the concept of an object insofar as it is not the object of a sensible intuition. By contrast, a noumenon in the positive sense would be an actual object of a non-sensible intuition. The latter is a problematic concept for Kant in the sense that we cannot determine whether such an entity (or mode of intuition) is really possible. Kant indicates the connection between what I have termed the two-steps of his analysis when he remarks that ‘the doctrine of sensibility is at the same time the doctrine of the noumenon in the negative sense’ (B307 ). I discuss the different senses of the noumenon and their relation to the transcendental object in Kant's Transcendental Idealism (2004 edn), pp. 57-64.

³⁶ The concept of the noumenon serves to limit the ‘pretension’ of the understanding as well, albeit in an indirect manner, by way of the dependence of the understanding on sensibility. Since cognition through the understanding (discursive cognition or judgement) requires that its object be given in sensible intuition, and limitation on the scope of the latter will limit that of the former as well.

³⁷ See Pro 4: 341, where Kant poses the issue in this logical form. It is also noteworthy that Arthur Collier, with whose work Kant was probably familiar, used virtually the same antinomial argument in an attempt to prove that ‘an external world, whose extension is absolute, that is, not relatively depending on any faculty of perception’, is self-contradictory. (Clavis Universalis, in Metaphysical Tracts by English Philosophers of the Eighteenth Century, ed. Parr, Samuel (London: Edward Lumley, 1837), pp. 46–50)Google Scholar .

³⁸ Although this assumption does not enter as a premise into either the thesis or antithesis argument of any of the antinomies, it underlies the cosmological debate as a whole. In particular, it makes it possible for each party to argue apagogically from the falsity of the alternative to the truth of its own claim. This also enables Kant to vouch for the soundness of each of the proofs, while at the same time claiming that the whole dispute is based on a deep misunderstanding. Admittedly, these proofs remain highly controversial, but I have endeavoured to defend those of the first and third antinomies against the standard objections in Kant's Transcendental Idealism (2004 edn), pp. 366-84.

³⁹ For my analysis of this argument, see Kant's Transcendental Idealism (2004 edn), pp. 388-95.

⁴⁰ For the definitive account of transcendental illusion, see Grier, Michelle, Kant's Doctrine of Transcendental Illusion (Cambridge: Cambridge University Press, 2001)CrossRefGoogle Scholar . My own systematic discussion of the topic, which is greatly indebted to Grier's but differs on some points, is to be found in Kant's Transcendental Idealism (2004 edn), pp. 322-32.

⁴¹ Since by such a synthesis Kant understands one that makes use merely of the pure or unschematized categories, in stating that it represents things as they are he is clearly not suggesting that it provides cognition of things as they are in themselves. His point is rather that it regards the items synthesized (the conditioned and its conditions) as a collection of objects whose nature is fixed apart from any sensible conditions that may be necessary for us to access them, that is, as a collection of things considered as existing in themselves. It should also be kept in mind that one of the ways in which Kant characterizes the thought of things as they are in themselves is as objects of a ‘pure understanding’. See note 24.

⁴² The essential point here, which has been developed at length by Grier in Kant's Doctrine of Transcendental Illusion, is the distinction between the illusion and the metaphysical fallacies it generates (which include those committed by the participants in the antinomial conflict). Whereas the former is unavoidable, the latter are not, even though they are based on this illusion. As Grier shows, failure to keep this distinction in mind underlies much of the confusion regarding Kant's critique of metaphysics in the Dialectic.

⁴³ A typical representative of this approach is Guyer. See his Kant and the Claims of Knowledge, pp. 385-415.

⁴⁴ Setting aside the question of philosophical adequacy, it seems clear that the second alternative comes closer to capturing Kant's actual views on the matter. See, for example, Bxxvii-xxviii and R5642: 18,401. At issue is only whether the two-aspect formulation is to be taken metaphysically.

⁴⁵ Beck, Lewis White, ’Five concepts of freedom in Kant’, in Srzednick, J. T. J. (ed.), Philosophical Analysis and Reconstruction, a Festschrift to Stephan Korner (Dordrecht: Reidel, 1972), pp. 42-3Google Scholar .

⁴⁶ See Irwin, Terence, ‘Morality and personality: Kant and Green’, in Wood, Allen (ed.), Self and Nature in Kant's Thought (Ithaca and London: Cornell University Press, 1984), p. 38Google Scholar .

⁴⁷ Following the language of Dummett, Putnam and others, I characterize this as a doctrine of ‘warranted assertibility from a point of view’. See Kant's Transcendental Idealism (2004 edn), p. 48.

⁴⁸ I here find myself in fundamental disagreement with Karl Ameriks, who has criticized non-metaphysical interpretations of transcendental idealism such as mine on the grounds that they give ‘no reason to think that the non-ideal has a greater ontological status than the ideal’, which he sees as incompatible with Kant's deepest philosophical commitments concerning ‘the absolute reality of things in themselves with substantive non-s patio-temporal characteristics’ ( ‘Kantian idealism today’, History of Philosophy Quarterly 9 (1992), p. 334)Google Scholar . I believe that Ameriks is correct in pointing out that on such readings the non-ideal has no greater ontolog-ical import than the ideal; but I question his further claim that this is incompatible with Kant's deepest philosophical commitments. These commitments, I suggest, are to a robust empirical realism, on the one hand, and to a conception of freedom capable of supporting the autonomy of the will required by his moral theory, on the other. Although I assume that Ameriks would concur on both these points, I have tried to show that both are threatened rather than preserved by an ontological reading of transcendental idealism of the sort that he evidently favours.

⁴⁹ This is especially true of Guyer, who views the Refutation or, more precisely, a version of it contained in Kant's late Reflexionen, as the culmination of Kant's transcendental theory of experience. See ‘Kant's intentions in the Refutation of Idealism’, The Philosophical Review, 92 (1983), 329–83CrossRefGoogle Scholar ; and Kant and the Claims of Knowledge, pp. 279-329.

⁵⁰ For my analysis of the Refutation of Idealism see Kant's Transcendental Idealism (2004 edn), pp. 285-303. I there argue that rather than being incompatible with transcendental idealism, Kant's argument is dependent on it.

⁵¹ At least with regard to the direction of Kant's argument in the Analytic I am in agreement with Ameriks. See his ‘Kant's Transcendental Deduction as a regressive argument’, Kant-Studien 69 (1978), 273–87Google Scholar ; and Kant and the Fate of Autonomy, pp. 55-63 and passim.

⁵² I wish to thank the audiences at the meeting of the Pacific Study Group of the North American Kant Society and the colloquium of the Stanford Philosophy Department, before whom I presented earlier versions of this article, for their invaluable comments and criticisms. I have endeavoured to address many of these in this greatly revised version.