1. Transcendental Idealism and Ontology

In Kant's lecture courses on metaphysics we find clear and succinct expressions of his conception of ontology. There, ontology is defined as ‘the science … which is concerned with the more general properties of all things’ (Kant Reference Kant2002: 295; TP1 Footnote 1 2.309) and ‘the science of the properties of all things in general’ (Kant Reference Kant1997: 140; LM Footnote 2 29.784), and it is said that ‘Ontology thus deals with things in general, it abstracts from everything particular’ (Kant Reference Kant1997: 307; 28.541) and ‘Ontology … contains the summation of all our pure concepts that we can have a priori of things’ (Kant Reference Kant1997: 308; 28.541–2). Similarly, Baumgarten's Metaphysics, which Kant often used as the textbook for his metaphysics courses, defines ontology as ‘the science of the general predicates of a thing’ (Baumgarten Reference Baumgarten2009: 89), while the second chapter of Wolff's Rational Thoughts on God, the World and the Soul of Human Beings, Also All Things in General, in which Wolff sets out the basic principles of his ontology, is titled ‘On the First Principles of Our Cognition and of All Things in General’ (Wolff Reference Wolff2009: 9).Footnote 3 In each of these characterizations of ontology we are given that discipline's distinctive object of concern: things. More specifically: that which pertains to things in the most general fashion, and thus ‘things in general’. This already enables us to make sense of Kant's famous claim to have displaced ontology by ‘humbling’ the categories in the Critique of Pure Reason. No longer describing the general character of things and allowing cognitive access to them (i.e. no longer functioning as ontological concepts), the categories become mere functions of synthesis in the human understanding's comprehension of the intuitional manifold, and so the Transcendental Analytic's ‘principles are merely principles of the exposition of appearances, and the proud name of an ontology, which presumes to offer synthetic a priori cognitions of things in general in a systematic doctrine … must give way to the modest one of a mere analytic of the pure understanding’ (Kant Reference Kant1998: 345; CPR Footnote 4 A247/B303). On this interpretation, then, the crux of Kant's rejection of ontology is not so much a rejection of the notion that we can have cognitive access to a certain kind of thing—for example, something like a Platonic essence or Leibnizian monad—although that is of course an important facet of his critique of rationalism. Instead, it would be more precise to say that Kant's rejection of ontology is a rejection of the notion that the a priori conditions of cognition afford access to things as such.Footnote 5

Furthermore, if the a priori conditions of cognition do not afford access to things as such, then Kant must have some way of explaining how that to which these conditions do afford access—i.e. appearances—are not things and thus do not entail a commitment to any ontology. Appearances must in some sense be ontologically insignificant. They must have some sort of non-ontological status. It is here that the appearance/thing in itself distinction becomes crucial. My contention is that transcendental idealism can only guarantee that appearances have non-ontological status by emphasizing their epistemic significance and distinguishing them from things in themselves, which, as things,Footnote 6 have ontological status. That is, appearances (regardless of whether these be understood ‘substantively’ or ‘adverbially’Footnote 7) are just how things appear, which means they are not things in their own right but rather just a part of how things are known, whereas things in themselves are simply things as such (and ‘in themselves’ just signifies their independence from the subjective conditions of cognitionFootnote 8). In other words, the appearance/thing in itself distinction is a distinction between things as they are known and things as they are. Hence it could be said that the significance of Kant's arguments for transcendental idealism in this respect is that their affirmation of knowledge of appearances and denial of knowledge of things in themselves differentiates transcendental idealism (a theory of knowledge) from any ontology (a theory of things in general). More specifically, Kant's restriction of the objective validity of the categories, space, and time to appearances separates our cognitive relation to objects from the being of things, so that corresponding to this divorce of epistemology from ontology is the distinction between an object Footnote 9 (that which is known) and a thing Footnote 10 (that which is). Thus while the intuitional manifold of sensibility alone provides human cognition with content, it only presents the way in which we are affected rather than things as they are in themselves since sensations are received in a spatiotemporal form that is the contribution of the constitution of human sensibility instead of that which affects us. Accordingly, when the understanding constitutes objects of knowledge by categorially determining the intuitional manifold, it is synthesizing how things appear under the subjective conditions of intuition instead of representing things as they are. Hence Kant (Reference Kant1998: 511; CPR A492/B520) claims that ‘Space itself, … together with time, and, with both, all appearances, are not things, but rather nothing but representations’.Footnote 11

In short, whereas Kant understands ontology to be concerned with the possibility of things, the distinguishing feature of his transcendental idealism is its more modest concern with the possibility of cognitive experience. Thus even concepts like existence or actuality, which would otherwise seem to be the most obviously ontologically significant, are recast in terms of cognitive experience, as the second Postulate of Empirical Thinking in General makes clear: ‘That which is connected with the material conditions of experience (of sensation) is actual’ (Kant Reference Kant1998: 321; CPR A218/B266). Accordingly, for Kant, claims made about the ‘actuality’ of ‘objects’ are ultimately claims about cognitive experience (whose conditions of possibility, let us not forget, are ‘at the same time conditions of the possibility of the objects of experience’: Kant Reference Kant1998: 283; A158/B197) rather than ontological claims, even if they superficially look like the latter. The same goes for the rest of the categories, which Kant has critically appropriated from ontology in exactly this manner. Therefore, on this interpretation, transcendental idealism, as a doctrine exclusively about how things are known (i.e. as they appear), and justified by arguments which refuse such knowledge any traction on things in themselves, is free from any ontological commitments since its denial of knowledge of things in themselves amounts to an agnosticism about how things are. In other words, appearances have a purely epistemic, non-ontological status Footnote 12 since they reflect transcendental idealism's disavowal of any claims concerning things as they are. This is what constitutes transcendental idealism's ontological agnosticism. Importantly, this is also what allows transcendental idealism to be distinguished from something like Berkeley's subjective idealism, for the fact that transcendental idealism restricts knowledge to appearances does not mean that it restricts all things to appearances:Footnote 13

2. Ontological Agnosticism and the Problem of the Thing in itself

It is only in light of the foregoing interpretation of the appearance/thing in itself distinction, I submit, that transcendental idealism can explain the non-ontological status of appearances and justify Kant's rejection of ontology. However, the primary contention of this section is that this achievement is in vain. For this ontologically agnostic interpretation of transcendental idealism requires things in themselves to be the bearers of being so that appearances are not, and now there is a much more serious problem: the non-ontological status of appearances depends on the actual existence of things in themselves. This is because one cannot claim that appearances are just how things appear without presupposing that there is more to things than how they appear. If that presupposition is not justified, the additional possibilities are left open that appearances are all that there is or that appearances are just not the appearances of any other things (which may or may not exist), in which cases appearances cannot be contrasted with things in themselves in a way that secures the non-ontological status of the former.

It is true that the notion that there is nothing more to things than their perceptual appearance is central to Berkeleyan idealism and, as we saw in the preceding section, transcendental idealism distinguishes itself from Berkeleyan idealism insofar as the latter mistakes a condition of cognitive experience for a condition of things in general and dogmatically denies the existence of things beyond perception.Footnote 15 However, this point by itself just reproaches Berkeley for hastily dismissing the possible existence of things in themselves, for the denial that the conditions of cognitive experience are conditions of things in general is equally consistent with the possible nonexistence of things in themselves (or their possible separate existence from appearances). In other words, the problem here is that, although Kant's distinction between the conditions of cognitive experience and the conditions of things in general may distinguish him from Berkeley, for whom the objects of sense-experience definitely have ontological status, it is not enough to validate the claim that Kantian appearances definitely do not have ontological status. Kant could not, for instance, flatly claim, and without reference to things in themselves, that in transcendental idealism appearances have non-ontological status because they are only considered in terms of their epistemic function, for in this respect that would merely mark a nominal difference between Kantian appearances and Berkeleyan ideas (and let us not forget that the latter have an epistemic function too). For if it happened to be the case that things in themselves did not exist, then something like Berkeley's ontology would be true (though unverifiable), and Kantian appearances would be things in general. Only the actual existence of things in themselves could rule this out, and only appearances being how these things appear, I contend, could justify the claim that appearances have non-ontological status.Footnote 16

Clearly, then, Kant's conception of the ‘negative’ noumenon as a limiting concept will not suffice to secure the non-ontological status of appearances either.Footnote 17 This is because this concept is only necessary to remind us that we are not justified in assuming our sensibility extends to all things there are, so that we cannot deny the possibility of things beyond human sensibility or faculties of intuition different from our own.Footnote 18 That is to say, it is simply a reflection of the difference between transcendental idealism's conditions of possible experience and ontology's conditions of things in general:Footnote 19

Therefore, since the necessity of thinking noumena is not intended to affirm the existence or nonexistence of things beyond human sensibility, it does not conflict with the (unknowable) possibility of either. For the same reason, it cannot guarantee that appearances have no ontological status.

In short, it seems as though the non-ontological status of appearances can only be secured if ontological status is located elsewhere. However, as transcendental idealism's earliest critics famously protested,Footnote 21 it is exactly this location of an ‘elsewhere’ in things in themselves that Kant deprives himself of the right to identify in virtue of his restriction of the objective validity of the categories to appearances. For example, claiming things in themselves actually exist uses the category of existence beyond its domain of legitimate application, while inferring their existence as causes of appearances does the same with the category of causality as well (and Kant does bothFootnote 22). In the A-edition Fourth Paralogism Kant himself casts suspicion on the latter inference by conceding the point to scepticism,Footnote 23 and Hume had already argued that the confinement of knowable objects to sense-experience leaves one with little room to manœuvre in this dilemma.Footnote 24 Is there, then, a way for transcendental idealism to establish the existence of things in themselves in a manner that secures its ontological agnosticism without generating problems of this magnitude? In the remainder of this section I will argue that there is not.

At many points Kant seems to insist that the actual existence of things in themselves is already established since it is entailed by the very notion of appearances. The basic argument is apparently that our sensory representations, as ‘appearances’, presuppose things in themselves as their thinkable, if not directly cognizable, ground, because ‘otherwise there would follow the absurd proposition that there is an appearance without anything that appears’ (Kant Reference Kant1998: 115; CPR Bxxvi–xxvii). The Prolegomena contains passages along these lines, such as the following: ‘appearances actually do relate to something distinct from them (and so entirely heterogeneous), in that appearances always presuppose a thing in itself, and so provide notice of such a thing, whether or not it can be cognized more closely’ (Kant Reference Kant2004: 106; P 4: 355) and

However, it is rather obvious that Kant cannot claim that things in themselves exist on the dual basis of our acquaintance with appearances and an analysis of the meaning of the word ‘appearance’ without begging the question. For if things in themselves are indeed logically implied by the concept of appearance, then Kant must first explain why ‘appearance’ is the appropriate term for the sensory representations given in sensibility. To do that, he must already know that these sensory representations are the appearances of things in themselves, but that is precisely what was supposed to be concluded by these means. If it is not already established that sensibility gives appearances of things in themselves, why call these representations ‘appearances’ if this term logically implies things in themselves as their ground? If there is no such logical implication, then we are back to square one. Lorne Falkenstein (Reference Falkenstein1995: 325) apparently endorses this kind of argument when, distinguishing Kant's idealism from Berkeley's, he claims that for Kant ‘the very notion of an appearance carries with it the thought of something of which it appears, so that things in themselves can at least be thought (indeed, known to actually exist), though nothing more can be known of them’.Footnote 25

However, it is perhaps Kant's own interpretation of the consequences of his arguments for transcendental idealism that is the origin of such problematic reasoning insofar as he presents an unwarranted assumption of the existence of things in themselves in the guise of a restriction of knowledge to ‘things as they appear’ (i.e. appearances). In other words, Kant's articulations of the results of his arguments for transcendental idealism insinuate the actual existence of things in themselves by concluding that we can only know ‘things as they appear’ rather than ‘things as they are in themselves’ (much as the interpretation of transcendental idealism given in the preceding section does). This gives the misleading impression that the notion of ‘appearance’ sanctioned by these arguments legitimizes the affirmation of existent things in themselves as the necessary correlate of appearances. For example: ‘We have therefore wanted to say that all our intuition is nothing but the representation of appearance; that the things that we intuit are not in themselves what we intuit them to be, nor are their relations so constituted in themselves as they appear to us’ (Kant Reference Kant1998: 168; CPR A42/B59) and

Unfortunately, this simply does not follow from Kant's arguments for transcendental idealism. In such passages there seems to be an equivocation between two possible meanings of ‘appearance’: (1) sensory representation caused by, or somehow related to, a thing in itself and (2) sensory representation full-stop. Kant's prioritization of the first sense of ‘appearance’ here to affirm the existence of things in themselves is unjustified. This is because, strictly speaking, the arguments for transcendental idealism only conclude that space, time, and the categories are objectively valid with respect to sensory representations (‘appearance’ in the second sense)—such that objects of knowledge are constituted by the human cognitive apparatus—and invalid with respect to things that exist independently of these representations (things in themselves)—such that those things may exist though we can know nothing more about them. In other words, they conclude that we cannot know anything beyond sense-experience, and thus we cannot know things in themselves. They do not license the conclusion that sensory representations must be the appearances of things in themselves (‘appearance’ in the first sense). Thus they cannot legitimize ‘appearance’ as the appropriate term for these representations if the concept of appearance logically implies things in themselves. In short, while sensory representations being the appearances of things in themselves may be logically consistent with the arguments for transcendental idealism, it is not established by them, and the latter is what is needed.Footnote 27 (For the same reason, the arguments for transcendental idealism cannot be taken as the grounds for Kant's rejection of ontology as it is interpreted above, for if they do not establish that appearances are just how things appear then they do not establish that appearances have no ontological status.)

On the other hand, if Kant's argument about appearances presupposing things in themselves is about ontological dependence rather than logical implication, then it would seem to be a non-sequitur. That is, if it is argued that appearances cannot exist on their own because they are sensory representations, and so must have things in themselves as their ground, then the possibility that the transcendental subject instead functions as that ground would have to be eliminated first (which would require ruling out the possibility of unconscious subjective sources of appearancesFootnote 28). Rae Langton avoids this difficulty by contending that for Kant objects of sensibility are not mere sensory representations but in fact the transcendentally real relational properties of things in themselves,Footnote 29 while the latter are metaphysical substances which also possess intrinsic properties.Footnote 30 Accordingly, given that outer sense represents nothing but external relations, and external relations allegedly presuppose relata to bear themFootnote 31 but do not supervene on their intrinsic properties,Footnote 32 it is possible for Kant to legitimately claim that there are things in themselves while denying knowledge of what they are (i.e. their intrinsic properties).Footnote 33 Furthermore, this relation of ontological dependence would explain why Kant thinks appearances presuppose the existence of things in themselves.Footnote 34 However, Langton's methodology provides grounds for doubting whether her solution to the problem of the thing in itself is of any help to transcendental idealism in the end.Footnote 35 Moreover, even if the argument Langton proposes for the existence of things in themselves works, it does not establish the non-ontological status of appearances (and to be fair, that was never her concern): as relational properties of substance, the appearances of Langton's Kant are explicitly ontological.

Even though Langton as well as Kenneth Westphal endorse metaphysical conceptions of appearances, their respective treatments of the problem of the thing in itself have one thing in common that could be used to argue for the existence of things in themselves in order to justify a non-ontological interpretation of appearances. That is an attempt to defend Kant's affirmation of the existence of things in themselves and their correlation with appearances on the basis of an interpretation of the legitimate use of certain unschematized categories.Footnote 36 This involves emphasizing the cognitive indeterminacy of the affirmation of the mere existence of things in themselves, in contrast with the cognitive determinacy that the schematized categories provideFootnote 37 or knowledge of things as they are in themselves would afford.Footnote 38 It is then apparently concluded that since affirming the existence of things in themselves—whether as the bearers of intrinsic properties (Langton) or the sources of sensory affection (Westphal)—does not involve the determinacy of empirical cognition or knowledge of intrinsic properties, it is permitted within use of the unschematized categories.Footnote 39 However, I am not quite convinced that this conclusion follows, for the ‘indeterminacy’ at issue here seems equivocal, and neglect of the category of existence may be to blame.Footnote 40 That is, while affirming that there are things in themselves without specifying what they are would be making a largely indeterminate claim about things, it is still more determinate than legitimate use of the unschematized categories would allow since it is still a claim about actually existent things instead of merely thinkable ones. In other words, it may not have the specificity that empirical cognition or ascription of determinate intrinsic properties has, but it is still a positive existence claim about some particular things, even if nothing more is said about them. This is clearly more than merely thinking the possible existence of things in themselves, or their possible characteristics via the logical content of certain categories. In short, I think it is determinate enough to be closer to what Westphal (Reference Westphal2004: 52) calls the ‘transphenomenal “application” of concepts Kant proscribes’ (i.e. ‘the purported subsumption of unsensed particulars under nonschematized concepts in determinate, theoretically cognitive judgments’) than one of the ‘legitimate ways of identifying particulars’ other than empirical cognition. On the other hand, Westphal proposes a sophisticated account of transcendental and ‘epistemic’ reflection as one of these ‘other legitimate ways of identifying particulars’ and argues that Kant can legitimately speak of noumenal causes of sensory affection since such ‘causes’ are postulated at the transcendental level of discourse about the conditions of cognition (as opposed to the empirical level of cognition itself).Footnote 41 Nevertheless, I am not yet convinced that this is fully consistent with legitimate use of the unschematized category of existence or, even if it is, whether it can eliminate the possibility of unconscious subjective sources of sensory representations.Footnote 42

Finally, perhaps it could be argued that Kant's practical philosophy resolves the theoretical philosophy's difficulties with things in themselves, and that it could be mobilized to guarantee the non-ontological status of appearances as the appearances of things in themselves. Now whether Kant's conception of noumena in the practical philosophy should be interpreted ontologically is certainly debatable, but for the sake of the argument let us assume it should be. Suppose we are justified in holding that things in themselves (or at least some of them) are free noumenal souls. I think this would still fall short of establishing that appearances are the appearances of things in themselves, because the latter requires establishing that every appearance we cognize empirically is the appearance of a thing in itself (or multiple things in themselves, as the case may be). In this case, that would mean guaranteeing that every appearance is the appearance of a noumenal soul, such that appearances always correspond to such souls (though the reverse correspondence need not hold). What resources, though, does Kantianism possess that could ever assure this correspondence? How could transcendental idealism ever be justified in holding that every appearance—i.e. not just human actions but inanimate, non-human objects and events as well—is grounded in a noumenal soul? I do not see how it could, and for that reason I do not think the practical philosophy can secure the non-ontological status of appearances either. The same difficulties would arise, I think, if we considered the possibility that any kind of ‘subject in itself’ is the ground of appearances.Footnote 43

To summarize thus far, the actual existence of things in themselves, with appearances interpreted as how these things appear, could secure the non-ontological status of appearances and substantiate Kant's rejection of ontology. Transcendental idealism, though, does not have the resources to establish the actual existence of things in themselves. Furthermore, to the extent that transcendental idealism deprives itself of the resources to do this by restricting the objective validity of the category of existence to appearances, the ontologically agnostic version of transcendental idealism outlined above culminates in self-refutation: appearances have non-ontological status only if transcendental idealism presupposes an ontological claim (concerning the actual existence of things in themselves) whose legitimacy should have been foreclosed by the very rejection of ontology that transcendental idealism's ontological agnosticism was supposed to have exemplified in the first place. If this is correct, then it seems as though transcendental idealism must grant the possibility that appearances have ontological status and relinquish the ontological agnosticism described above if it is to avoid self-refutation.

3. The Problem of Allison's Two-Aspect View

Nevertheless, since the foregoing interpretation and problematization of transcendental idealism's ontological agnosticism depend on an ontological conception of things in themselves, it is worth considering whether Henry Allison's two-aspect view of transcendental idealism, which proposes a non-ontological conception of appearances and things in themselves, has the resources to provide a less problematic account of transcendental idealism's ontological agnosticism. Indeed, Allison (Reference Allison2004: p. xv) contends that Kant's anthropocentric reformulation of the standards of objective knowledge entails that ‘transcendental idealism is grounded in a reflection on the a priori conditions of human cognition … rather than, as in other forms of idealism (for example, Berkeley's), on the ontological status of what is known’. At first glance, at least, it seems as though the two-aspect view could distinguish transcendental idealism from subjective idealism, secure the non-ontological status of appearances, and substantiate Kant's rejection of ontology without risking self-refutation. I will argue here, however, that this is not the case.

For Allison, the crucial mistake of previous interpretations of transcendental idealism is to take the distinction between appearances and things in themselves as an ontological distinction between two kinds or classes of entities:

As an alternative, Allison (Reference Allison2004: 16) proposes that ‘appearances and things in themselves be understood as holding between two ways of considering things (as they appear and as they are in themselves)’. That is, this non-ontological reading of the appearance/thing in itself distinction consists in construing the latter as two ‘ways of considering’ one class of things rather than indicators of the respective natures of two classes of things. Naturally, considering things ‘as they appear’ means considering them ‘as they are in relation to the subjective conditions of human cognition’, while considering the same things ‘as they are in themselves’ means considering them ‘independently of these conditions’ (Allison Reference Allison1996: 3).

It could be said that Allison's two-aspect view is something of a generalization of Kant's account of transcendental reflection in On the Ground of the Distinction of All Objects in General into Phenomena and Noumena, for things in themselves so conceived, much like noumena in the negative sense, only function conceptually ‘at the metalevel of philosophical reflection’ (Allison Reference Allison1996: 3) rather than ontologically. Because of this, Allison argues that the two-aspect view does not entail any inconsistency for transcendental idealism. Considering things independently of the subjective conditions of human cognition is not knowledge of them, nor does it require positing some additional class of things beyond what we know to exist via such subjective conditions. Consequently, ‘the temptation to worry about the existence of things in themselves disappears once it is recognized that Kant is not primarily concerned with a separate class of entities, which, unlike appearances, would supposedly “be there” even if there were no finite cognizers’ (Allison Reference Allison2004: 51).Footnote 44

However, unlike negative noumena, the two-aspect view does more than simply countenance the mere possibility of modes of intuition different from human sensibility or the possible existence of things beyond human sensibility. This is because Allison's claim that the appearance/thing in itself distinction demarcates two ways of considering things rather than two classes of things concentrates the concepts of ‘appearance’ and ‘thing in itself’ on one class of things—i.e. those that are considered under the two aspects. The crucial question, though, is this: given an object known via the subjective conditions of human cognition, what are Allison's grounds for considering this very same object independently of those conditions? If such an object cannot be reduced to ‘mere representations’,Footnote 45 surely it makes sense to consider it independently of sensibility only if there is something of that object that actually is independent of sensibility. Otherwise, what is it that is being considered ‘in itself’? To his credit, Allison is keenly aware of this question, but unfortunately his answer to it is inadequate. As he candidly summarizes the matter:

To this question Allison offers the following response:

Now, this much is clear: there must be something more to things than how they appear for the two-aspect view to even get off the ground. For example, it is said that under the two-aspect view of transcendental idealism ‘we know real, mind independent objects (although not considered as they are in themselves)’ and so ‘the position is not phenomenalistic’ (Allison Reference Allison1996: 3), and that like the transcendental realist ‘Kant likewise assumes’ that ‘things exist independently of their relation to the condition of human sensibility’ (Allison Reference Allison2004: 24–5). Without these things that exist independently of the subjective conditions of cognition Allison has no basis to secure the non-ontological status of appearances, for it is now evident that the latter are only without ontological status if they are actually these ‘subject-independent things’Footnote 46 considered in a certain way. In other words, ‘appearance’ and ‘thing in itself’ can be interpreted ‘adverbially’ as two aspects only if there is something ‘substantive’ to consider under those two aspects.Footnote 47 This brings to light the fact that the two-aspect view implicitly presupposes an ontological thesis concerning the actual existence of these subject-independent things, for without such things there is nothing to consider ‘as it is in itself’. Whereas transcendental idealism as presented in the first section attempts to secure the non-ontological status of appearances by emphasizing the ontological status of things in themselves, Allison attempts to secure the non-ontological status of appearances and things in themselves by tacitly shifting the burden of ontological status onto these subject-independent things. Consequently, these subject-independent things are just Allison's proxies for the things in themselves of other interpretations of transcendental idealism, including the one presented above.

The real problem, however, is that Allison simply assumes the actual existence of these subject-independent things instead of arguing for it. He claims objects of knowledge exist independently from the subjective conditions of cognition but fails to explain how this is justified.Footnote 48 Therefore, by assuming the actual existence of subject-independent things and restricting their knowability to ‘how they appear’, Allison simply regenerates the problem of the thing in itself (even though he has changed the meaning of the term ‘thing in itself’), which plagued the very metaphysical interpretations of transcendental idealism his two-aspect view was supposed to correct and replace. Furthermore, as a non-ontological interpretation of transcendental idealism, the two-aspect view suffers from a problem of self-refutation very similar to the one discussed at the end of my second section: appearances and things in themselves have non-ontological status only if the two-aspect view presupposes an ontological thesis (concerning the actual existence of subject-independent things) that should have been eliminated by the two-aspect view's non-ontological conception of things in themselves. In other words, Allison's version of transcendental idealism is dependent upon that which it claims to overcome, because it has to presuppose an ontological thesis in order to misunderstand itself as an ‘alternative to ontology’ (cf. Allison Reference Allison2004: 98). In sum, whether things in themselves are conceived non-ontologically à la Allison or ontologically as above, there is in both cases an unjustified ontological thesis concerning things that exist independently from the subjective conditions of cognition that is used to explain the non-ontological status of appearances.

4. Ontological Agnosticism without Things in Themselves

In light of the difficulties with transcendental idealism's ontological agnosticism discussed so far, it is worth considering the possibility that there is no need for the actual existence of things in themselves (or Allison's subject-independent things) to explain the non-ontological status of appearances or to justify Kant's rejection of ontology. For a start, perhaps one could simply appeal to the infinite divisibility of appearances without having to refer to anything beyond them. The argument could be formulated as follows. Since space and time are infinitely divisible, and appearances are infinitely divisible since they are in space and time, appearances have no simple parts from which they could be composed. Hence as purely external relations they are ultimately ‘nothing’ at all: there is nothing ontologically substantial about them, and thus there is nothing in them that could constitute their ‘being’.Footnote 49 This proposal has the benefit of being agnostic about the existence of things in themselves: they may exist, in which case they may or may not be the causes of appearances, and appearances may or may not be like the relational properties of substances Langton seizes on (though we could never know); or they may not exist, in which case some sort of ontological nihilism might be true (though unverifiable), for ‘all that there is’ (excepting, perhaps, transcendental subjectivity) would ultimately be nothing, with the notion of ‘something’ rendered illusory if convenient, much like perceived composites are in mereological nihilism. It is probably true that even the mere possibility of some kind of ontological nihilism being true would have been hard for Kant to accept,Footnote 50 but that does not mean it should not be accepted. Perhaps this would also require revision of the division of the concept of nothing in Kant (Reference Kant1998: 382–3; CPR A290–2/B346–9) as well as the second Antinomy. Nevertheless, these grounds for establishing the non-ontological status of appearances are clearly not epistemic grounds but much more like metaphysical ones. In this case, then, the non-ontological status of appearances would not be a consequence of their purely epistemic status, and Kant's rejection of ontology would be no closer to validation. Furthermore, the argument depends on Kant's premise that whatever is in space and time must be infinitely divisible since space and time are infinitely divisible—a supposition that, as I think Falkenstein (Reference Falkenstein1995) has convincingly argued, is undermined by its failure to exclude the extensionless physical monads of Kant's own pre-Critical Physical Monadology.Footnote 51

There remains another, perhaps more obvious possibility for explaining the ontological agnosticism of transcendental idealism without reference to actually existent things in themselves. At the beginning of my second section I pointed out that taking transcendental idealism as a doctrine regarding the conditions of cognitive experience rather than the conditions of things in general does not by itself entail a commitment to the existence or nonexistence of things in themselves since it is equally compatible with both, and that if things in themselves did not exist appearances would be things in general (though unverifiably so). Accordingly, at the end of that section I suggested that transcendental idealism might have to accept the possibility that appearances have ontological status if it is to avoid self-refutation. In the remainder of this section, however, I would like to consider the possibility that the ontological agnosticism of transcendental idealism could actually be conceived along these lines.

Let us assume that the transcendental idealist concedes the point that the non-ontological status of appearances cannot be guaranteed but maintains that the merely possible existence of things in themselves is enough to ground genuine ontological agnosticism. In this case, the transcendental idealist would have to admit that we might have cognitive access to things in general, but no more than s/he affirms that we might not: maybe appearances are all that there is, and thus things in general, and maybe appearances are just how things are known. Transcendental idealism may not be able to establish that appearances are not things in general, but it is equally unable to establish that they are. Now, perhaps it could be said that it is precisely this uncertainty about things in general—i.e. this uncertainty as to whether they are appearances or things in themselves (or even both or neither)—that constitutes transcendental idealism's ontological agnosticism.

Let us see whether this revised version of transcendental idealism's ontological agnosticism can withstand scrutiny. Under this revision, transcendental idealism abandons any commitments to the existence or nonexistence of things in themselves as well as the ontological or non-ontological status of appearances. Presumably, everything else (e.g. the Transcendental Aesthetic's subjectivization of space and time as a priori forms of intuition, the Transcendental Analytic's account of synthetic objectivity, the Transcendental Dialectic's critique of rationalist metaphysics, etc.) would remain intact, so let us return our attention to the status of appearances. Transcendental idealism (with or without the above revision) is committed to the existence of appearances insofar as it holds that there are appearances. More exactly, it is committed to the existence of sensations that are synthesized as intensive magnitudes,Footnote 52 spatiotemporal forms that are synthesized as extensive magnitudes,Footnote 53 permanent perceptions that are synthesized as substances, sequential perceptions that are synthesized as causes and effects, and simultaneous perceptions that are synthesized as mutually interacting substances.Footnote 54 In light of this fact, let us re-examine the basis for transcendental idealism's revised ontological agnosticism, its uncertainty about things in general. As we saw just above, this agnosticism and uncertainty result from transcendental idealism's inability to determine whether or not appearances are things in general. Transcendental idealism is unable to make this determination because it does not have the resources to establish that appearances are just how things are known (in which case appearances would have non-ontological status, as in the first section above), just as it does not have the resources to establish that appearances are things in general (in which case appearances would have ontological status, as in subjective idealism). It does not have the resources to do the former because of the problem of the thing in itself, and it does not have the resources to do the latter because it cannot rule out the possibility that there are things in themselves. Indeed, it could be said that the reasons it does not have the resources to do either stem from the same source: because the category of existence—which, to be exact, is the category of ‘existence-nonexistence’Footnote 55—is only objectively valid with respect to appearances, transcendental idealism cannot say whether there are or are not things beyond appearances.

The fact that transcendental idealism cannot deny the existence of things in themselves is most important for this revised ontological agnosticism, though, because this is what permits the transcendental idealist to refrain from conceding that appearances actually are things in general, despite not being able to deny that they actually are. Again, s/he avoids conceding that appearances are things in general because s/he cannot eliminate the possibility that things in themselves exist. S/he cannot find any reason for the existence of things in themselves to be impossible. Furthermore, since the category of existence is only objectively valid with respect to appearances, I presume that the transcendental idealist could make these claims about the possible existence of things in themselves only by using the unschematized category of existence (and perhaps the other unschematized categories of modality as well). If this is the case, then the kind of possibility at issue here should be logical possibility, for ontological possibility would entail illegitimate use of the categories and cognitive possibility would concern appearances. Therefore, we could say: the transcendental idealist cannot deny the existence of things in themselves because there is no contradiction in the notion that things in themselves exist.Footnote 56 By the same token, s/he does not have to concede that appearances are things in general because there is no contradiction in the notion that appearances have non-ontological status as how things in themselves appear. In short, the ultimate basis of this revised version of ontological agnosticism is that the transcendental idealist can refrain from conceding that appearances are things in general because there remains the logical possibility that things in general could be other than appearances (i.e. things in themselves).

Now we have a fuller understanding of this revised version of transcendental idealism's ontological agnosticism. Despite transcendental idealism's commitment to the existence of appearances, it is the lingering logical possibility, the mere non-contradictoriness of the notion that things in general are other than appearances, and in such a way that appearances are just how these things are known, that grounds transcendental idealism's uncertainty about things in general and vindicates Kant's rejection of ontology. However, insofar as this clarifies transcendental idealism's ontological agnosticism, it also makes more explicit the general criterion of ontological agnosticism that is being employed in this particular case. This criterion, I think, would be something to the effect that a philosophical doctrine is ontologically agnostic if it cannot eliminate the merely logical possibility that there are things whose existence would entail the non-ontological status of some item or items that are integral to that philosophical doctrine.Footnote 57 If this criterion were to be employed beyond the particular case of transcendental idealism, then any philosophical doctrine that cannot eliminate this logical possibility would be as ontologically agnostic as transcendental idealism (so interpreted). Moreover, a general criterion of ontological agnosticism must presuppose a conception of ontology—or a ‘general criterion of ontology’, if you will—for it cannot explain how ontology can be avoided (or displaced) without a conception of what ontology is. This is indeed the case here, but the problem is that the conception of ontology presupposed by this criterion of ontological agnosticism is not the conception of ontology that the present essay has been working with so far, i.e. Kant's conception of ontology. That is, this criterion of ontological agnosticism does not presuppose that simply any philosophical doctrine regarding things in general is an ontology. Instead, it adds the additional condition that only a philosophical doctrine regarding things in general that can eliminate the logical possibility mentioned in this criterion is an ontology. To be precise, according to this presupposed conception of ontology, only a philosophical doctrine regarding things in general that can eliminate the merely logical possibility that there are things whose existence would entail the non-ontological status of what that philosophical doctrine regards as things in general is an ontology. More simply: if ontological agnosticism is grounded in the inability to eliminate the logical possibility mentioned in the criterion for ontological agnosticism, then an ontology must be grounded in the ability to eliminate that logical possibility.Footnote 58

I think this is an unreasonably austere requirement to place on any would-be ontology and an excessively high price to pay to bear the name of ontology, not to mention the fact that it is hardly uncontroversial to suppose that a merely logical possibility has this kind of ‘meta-ontological’ authority.Footnote 59 In other words, this criterion of ontology is so austere—which is by itself contentious—because it presupposes a contentious meta-ontological premise. Perhaps it could be argued that this is reason enough to reject this conception of ontology and the criterion of ontological agnosticism that presupposes it. However, the most important problem here is that the criterion of ontological agnosticism under consideration is incapable of supporting Kant's rejection of ontology, precisely because it presupposes a conception of ontology that is different from Kant's. Consequently, I do not think this revised version of transcendental idealism's ontological agnosticism employs a defensible criterion of ontological agnosticism or that it can vindicate Kant's rejection of ontology.Footnote 60