2. Intentionalism and Phenomenalism

In discussing Kant's idealism about appearances, I use the term ‘phenomenalism’ broadly, to refer to all interpretations which mentalize Kantian appearances by seeing their existence as involving no ontological commitments over and above the existence of mental states, whether or not this involves positing distinct mental objects. This means that it includes interpretations like those of Van Cleve (Reference Van Cleve1999) and Guyer (Reference Guyer1987), who deny that their interpretations involve positing mental objects (and therefore also deny that they have traditional ‘two world’ interpretations). Similarly, it seems to me that at least some intentional objectFootnote 3 (and adverbialFootnote 4) interpretations of Kantian appearances keep appearances in the mind without making them into mental entities. I do not argue that this is necessarily the case, because intentional theories of perception differ considerably and are controversial, as is the status of intentional objects.Footnote 5 This means that we cannot use this notion to explain Kant's position without a detailed explanation and defence of the understanding of such objects. I will say something briefly about why I think the two main ways in which the notion of intentional objects of perception may be appealed to to explain the status of Kantian appearances will not help us to have an account which retains some kind of mind-dependence, yet does not involve appearances having a phenomenalistic existence, literally in the mind.

The key feature of intentional objects, as they are standardly understood, as objects of thought, is that they may not exist: I can think about things which do not exist.Footnote 6 This has been thought by some to enable an intentional account of perception to respond to the traditional arguments which appeal to hallucination and non-veridical perception to reject direct realism: they point out that, while something which does not exist can be the object of my thought, this does not mean that in the case in which I think about something which does exist, the object of my thought is anything less than the object itself. Similarly, it is suggested that the fact that I can, for example when hallucinating, represent things which do not exist, does not show that, in the case of veridical perception, the direct object of perception is anything other than the physical object perceived. When you are thinking about something which does not exist, the object of thought is merely represented; similarly, it may be argued that we can account for the occasional illusory character of perceptual consciousness in terms of representational mental states or intentional objects. The idea is that hallucination involves objects which are merely intentional,Footnote 7 whereas in cases of actual perception the intentional objects of perception are the external object themselves. Whether or not this is a way in which a direct realist about perception can deal with the argument from hallucination,Footnote 8 it is not clear that it can help us to explain Kant's idealism about appearances in a non-phenomenalist way. This is because there are two ways in which the analogy between Kantian appearances and intentional objects or intentional content can be understood, one of which will lead to seeing Kant as a phenomenalist, and the other seeing him as a straightforward realist.

On the one hand, Kantian appearances could be understood in terms of an analogy with the merely intentional objects in terms of which intentional theories of perception characterize hallucination. On this reading, Kantian appearances are compared with representational mental states which are in the mind. As we have seen, in the theory of perception at least part of the point of introducing intentional objects is supposed to be to enable us to characterize hallucinations by appealing to representational mental states rather than mental entities, while asserting that in the case of non-hallucinatory perceptual experience the intentional objects of the experience are the external objects themselves. This means that to compare Kantian appearances with merely intentional objects is to keep appearances in the mind, resulting in a more sophisticated version of a phenomenalist interpretation. An example of this kind of view is provided by Aquila's interpretation. He says: ‘To exist as an appearance is to exist in what I shall call a “phenomenalistic” sense; it is to exist, in a certain sense, merely “intentionally” ’, adding that it is ‘not strictly speaking possible to suppose that the very same thing might exist in both of the senses in question’ (Aquila Reference Aquila1983: 89–90). This kind of intentionalist interpretation can be grouped with the phenomenalist interpretations.

On the other hand, the point of the analogy with intentional theories of perception could be to compare Kantian appearances with the mind-independent objects which the theory holds to be the intentional (direct) objects of perception in cases of veridical perception.Footnote 9 On this way of reading the analogy, the intentional objects in terms of which Kantian appearances are explained are supposed to be that to which the mind is actually directed – the actual objects of experience. It is in this sense that Adams attributes phenomenalism to both Leibniz and Kant.Footnote 10 Adams denies that phenomena (his primary concern is with Leibnizian phenomena but he thinks it applies to Kant as well) are mental images, and says they should instead be understood in terms of intentional objects, by which he means things which are objects for minds or objects of perception, or the ‘representational contents of our perceptions – what appears to us’.Footnote 11 But if Kantian appearances are compared to the independent things which are the objects of veridical perception, then the appeal to intentional objects does not explain the sense in which appearances are mind-dependent. In the sense in which intentional objects are not intentionally inexistent objects of thought, but rather ‘what appears to us’, they can, and do, feature in straightforwardly realist accounts of perception. A realist thinks that external objects can be and often are objects for minds, or objects of perception, and that the ‘representational contents of our perceptions’ are features of the way things mind-independently are.Footnote 12

In summary, on the one hand, appearances could be compared with the merely intentional objects, or representational mental states, involved in hallucinatory perceptual events. In this case, appearances are once again compared with something which is characterized entirely mentally, which will result in a (sophisticated) phenomenalist reading of transcendental idealism. On the other hand, appearances could be compared with the actual objects which are the intentional objects of perception in normal veridical perception, but now the comparison with intentional objects fails to capture any sense in which Kant's position is idealist, or any sense in which empirically real objects have a necessary connection with possible perception. Clearly I have not said enough here to exclude the possibility of alternative intentional object interpretations; my aim is simply to show that at least some intentionalist interpretations can be classed with phenomenalist readings, and will be vulnerable to objections to the latter.

3. Is Kant a Disjunctivist?

I appeal to a particular theory of perception to explain the account of secondary qualities we need to explain a kind of mind-dependence which does not involve existence in the mind. I have followed John Campbell's terminology in calling this the relational view, where the term ‘relational’ marks the fact that ‘the object perceived is a constituent of the conscious experience itself’ (Campbell Reference Campbell2002: 117). This contrasts with the so-called Cartesian view of perception, which understands perception as involving mental states with the right kind of (usually causal) relation to objects outside the mind, where a mental state with the same content could exist without the objects outside the mind. For the relational view, the presence to consciousness of the object is part of what makes the mental state the state that it is.

Roche says that it is not clear whether I am merely appealing to a relational account of perception to explain the account of secondary qualities I want to draw on to explain Kant, or whether I am attributing to him a relational account of perception. He argues that the latter would be implausible, because it involves, in his view implausibly, seeing Kant as a disjunctivist. I certainly think that the relational view enables us to present the account of colour we need to make sense of Kant's suggested secondary quality analogy, but in addition, like a number of other philosophers, it seems to me that a central part of the innovation of Kant's system is his rejection of the Cartesian conception of experience.Footnote 13 However, ‘disjunctivism’ and ‘the relational account’ of perception are contemporary names for contemporary theories. There is no systematically developed ‘theory of perception’ in our contemporary sense in Kant, and almost no discussion of whether and how illusion and hallucination pose problems for a direct theory of perception. This is simply not one of Kant's concerns. This means that there is a clear sense in which it is problematic to argue either that Kant did or did not hold one of these theories. Still, this does not rule out finding in Kant some of the ideas which these theories of perception are trying to capture. In particular, it seems to me that Kant thinks that perception of empirically real objects involves those objects being directly present to consciousness, without mental intermediaries: it involves being in representational mental states which essentially involve the (empirically real) object as constituents.

Roche thinks that a relational view is committed to disjunctivism, as the only way in which it can deal with hallucination and dreaming, and that Kant cannot be seen as a disjunctivist. One of Roche's reasons for thinking that Kant cannot be committed to something like disjunctivism is that although, as he acknowledges, Kant consistently holds that hallucinations and dreaming are not experiences, he argues that Kant does think that veridical experiences, hallucinations, and illusions form a common kind. This issue is complicated, since there is of course much dispute about what is meant by ‘disjunctivism’Footnote 14 and the sense in which disjunctivists deny a ‘common kind’ between veridical experiences, hallucinations, and illusions. Many disjunctivists may allow that there is a common kind in some sense, for example, a phenomenologically indistinguishable mental state, or common cognitive processing (Campbell Reference Campbell2002: 118). In addition, there is debate about whether the disjunctivist move works only for hallucination and not for illusion,Footnote 15 and the position I draw on to illustrate Kant's idealism is one which does not make a disjunctivist-type move for illusion. Since Kant is not in the business of situating his position in relation to this contemporary debate, we are unlikely to find a clear explication of the sense in which he does and does not think there is a ‘common factor’, in a way which would determine how to position his view in the contemporary debate. And even if it is true that a relational view must make something like a disjunctivist move to deal with hallucination (and possibly also illusion), this would not show that if Kant had a relational view we will therefore find a disjunctivist move in his writing. He may simply not have developed the consequences which his account of perception has for how to think about illusion and hallucination, since dealing with these is not his concern. He may not have realized that a relational view requires something like the disjunctivist move (assuming it does). Thus, weighing the fact that Kant seems to think that hallucination and dreaming have something in common with veridical perceptionFootnote 16 against his frequently saying that hallucinations and dreams are not experiences does not seem to me to give us decisive grounds for situating his account of perception either way, in terms of the contemporary debate. However, while I do not think that Kant says enough about the ‘common factor’ to ground a particular interpretation, I think there are a number of other considerations which support thinking that Kant's view does involve a rejection of the Cartesian conception of experience.

Roche's other reason for thinking that Kant does not have a relational view of perception is that he thinks this will not make sense of the Kantian notion of intuition. My view is exactly the reverse: I think we cannot make sense of the Kantian notion of intuition without seeing that Kant rejects the Cartesian account of experience. Before discussing this, I mention another reason I have for thinking Kant is committed to the idea that empirically real objects in space are directly presented to us in perception, without mental intermediaries: the way Kant presents the positions he rejects in the Refutation of Idealism. This also, it seems to me, provides one of the strongest reasons for rejecting a phenomenalist interpretation of appearances. In the Refutation of Idealism Kant argues that outer experience is primary, and that we could not be aware of the temporal order of our inner mental life if we did not have experience of objects outside us in space. He characterizes the idealism he rejects as a view which sees the immediate objects of perception as mental items and argues, against this, that external objects in space are the immediate objects of perception (A367, B276–9).

Kant divides his idealist opponents into those who, dogmatically, deny the existence of matter or external objects (such as Berkeley) and those who, sceptically, doubt the existence of external matter (such as Descartes). He identifies the external objects Descartes doubts with the material objects that Berkeley denies, and identifies the mental access to ideas Descartes allows as beyond doubt with the access Berkeley gives to his empirical objects. Kant says that he wants to prove, against both these views, that the immediate objects of perception are external objects in space. Clearly, Berkelean empirical objects are immediately perceived, but Kant takes it that they are not external objects in space, but rather something mental, and wants to show on the contrary that the immediate objects of perception are external, material objects in space. Descartes's view introduces mental intermediaries through which we perceive physical objects in space, and Kant rejects this part of his position. While the details of the argument of the refutation of idealism are somewhat obscure and remain controversial, Kant's presentation of his aims, and of what he is arguing against, is clear. He says that he wants to establish the reality of the very outer objects whose existence (he thinks) Berkeley denies and Descartes renders doubtful, he wants to reject the view that our experience of our own minds is more immediate than our experience of outer objects, and he wants to establish, on the contrary, that experience of objects in space is direct and immediate. The external objects whose existence Descartes renders doubtful are neither mental items nor constructions from mental items, and Kant does not suppose they are. Kant's assertion of the existence of the very external objects which Berkeley denies and Descartes doubts is not compatible with thinking that these external objects exist as constructions out of the mental entities or states which Berkeley and Descartes allow. And Kant's statement of what he is trying to achieve is that he wants to reject the idea that our experience of objects in space goes via mental intermediaries. Thus, although he does not explictly situate his view in terms of the theories of perception we discuss today, he clearly commits himself to a key part of the direct realist or relational position.

Roche argues that a relational view of perception cannot make sense of the Kantian notion of intuition, because ‘It is natural to construe intuitions as the direct objects of perception (a sense-datum reading) or as the representational vehicles – not themselves perceived – to objects appearing to be in space and time (an intentional object reading). Both construals are incompatible with a relational view, which denies any sort of perceptual intermediary’. My view is that intuitions are (or rather essentially involve) the direct objects of perception, but this does not involve a sense-datum reading, and is quite compatible with a relational view. When we have an intuition, this involves the object the intuition represents being present to consciousness.

Roche argues that the passages suggesting that intuitions involve the presence of an object can be taken to emphasize the passivity of intuition – that they involve something being given to our senses. But this is not the only significant feature of the way Kant talks about intuition; in addition, there is the way he defines them, and there is their role. The essential defining features of Kantian intuition are that intuitions are singular and immediate,Footnote 17 and the role of intuition is that of giving us the objects about which we think. I argue that a clear and straightforward way of understanding the idea that intuitions are singular and immediate is to see intuitions as representations which essentially involve the presence to consciousness of the particular things they represent.Footnote 18 In other words, intuitions represent objects immediately because they present the object itself, as opposed to referring to an object through the mediation of further representations (which enable us to think about the object whether it is present or not). Immediacy says that an intuition is not simply a representation which is caused by a particular thing, but that it is in fact a presentation to consciousness of that thing. By way of contrast to the immediacy of intuition, Kant explains the mediacy of concepts by saying ‘since no representation pertains to the object immediately except intuition alone, a concept is thus never immediately related to an object, but is always related to some other representation of it (whether that be an intuition or itself already a concept). Judgment is therefore the mediate cognition of an object, hence the representation of a representation of it’ (A68/B93; A19/B33; A239/B298; Proleg. 4: 282). In saying that a concept is always related to an object through a representation of it, whereas an intuition relates to an object immediately, he rejects the idea that intuitions involve mental intermediaries. I therefore think that when Kant talks about intuitions as involving the presence of objects this goes further than merely emphasizing their passivity, which, in addition, would not distinguish intuitions from sensations. Clearly, there is much more to be said about the contested notion of Kantian intuition, but, with these considerations in mind, I do not think Roche has said enough to show that Kant's notion of intuition is incompatible with his having something like a relational view of perception.

4. Mere Representations in Us

My view tries to explain a sense in which appearances are mind-dependent without making them exist literally in the mind. But Roche argues that Kant describes his view in a way which makes it more strongly idealist than my interpretation can account for. Kant says that appearances exist in minds,Footnote 19 that they are mere representations,Footnote 20 and that their existence has an essential connection to possible perception.Footnote 21 I agree that these features of the way Kant describes appearances are important, that they are presumably the reason phenomenalist-type interpretations have been so common, and that they must be explained by an adequate interpretation of transcendental idealism. Further, ideally, we need to explain these features of the way Kant describes his position at the same time as doing justice to his empirical realism, and his rejection of Berkelean idealism and of the Cartesian conception of experience.

I argue that Kant's claim that appearances are ‘in us’ does not commit us to a phenomenalist reading of appearances. Even more strongly, some interpreters have argued that it need not be understood as involving any kind of idealism. Allison (Reference Allison2004), for example, argues that we must read such passages bearing in mind Kant's disambiguation between the empirical and transcendental senses of ‘in uns’ and ‘ausser uns’ (A373). Allison says that understood empirically, the terms ‘in us’ and ‘outside us’ mark a distinction beween objects of inner and outer sense respectively, but understood transcendentally, they mark a distinction between ‘two manners in which objects can be considered in relation to the conditions of human sensibility’ (Allison Reference Allison2004: 24). Allison is clearly right to point out the significance of the disambiguation between the empirical and transcendental senses of ‘in us’, but of course this does not settle the interpretative question: establishing that, for Kant, there are distinct empirical and transcendental senses in which objects can be ‘in’ our minds does not establish any particular account of the latter.Footnote 22 Clearly, we need to pay attention to the passage in which Kant describes his disambiguation. Kant says that:

Kant says that what is outside us in the empirical sense are things which exist in space, whereas what is transcendentally external is ‘something that, as a thing in itself, exists distinct from us’ (A373; my underlining). The empirical contrast between ‘in us’ and ‘outside us’ is between things which are in our minds and things that are in space, while the transcendental contrast is between things whose existence is distinct from us and things whose existence is not distinct from us. In the latter sense, to say that things are in us is to say that their existence is not distinct from us. At the very least, Kant's point here is to insist that the question of whether objects are outside us in space is a separate question from whether they are independent of us. This, I think, supports thinking that transcendental idealism involves a kind of mind-dependence which does not involve existence in the mind, and that Kant's saying that appearances exist ‘in us’ does not count against this.Footnote 23

Roche is worried that my reading cannot do justice to Kant's use of the term ‘representations’. This term is seen by many commentators as support for a phenomenalist interpretation (see, for example, Van Cleve Reference Van Cleve1999: 123; also Barker Reference Barker1969; Wilkerson Reference Wilkerson1976: 186). It is repeated many times, throughout the Critique, and not only does Kant call appearances representations, he says that they are mere representations, and he frequently links calling appearances representations to asserting their mind-dependence. He says that appearances are mere representations which do not exist apart from a connection to possible perception (A225/B272; A376; A493/B521; A493–4/B521–2).

As Roche points out, I suggest that my reading can accommodate this, since, as I see it, appearances are representations in the sense that they present objects to us: they exist in relation to conscious minds. However, he says that this cannot be what Kant means by representations, since Kant includes under the term ‘intuitions, concepts (conceivings), judgements (judgings), cognitions, perceptions, and sensations (for example, A320/B376-77)’. I take it that ‘representations’ is Kant's most generic term for mental contents, which includes both sensations and experiences. It is clearly not the case that Kant thinks that all representations have the same status, since while appearances are representations not all representations are appearances. Someone who has a relational view of perception may say that perceptual states have representational mental content which essentially involves the objects perceived, without thinking that all representational mental states have this status; thus, they could think that some representations exist merely in the mind, while others involve the presence of the object represented. Thinking that all representations involve the mind, and that some representations are literally in the mind, does not commit us to thinking that all representations are literally in the mind. On my view, Kant's calling appearances mere representations is meant to mark the idea that they exist only in relation to the possibility of their being perceived by us. In other words, not only are they presented to consciousness, they are essentially things which can be presented to consciousness. Below, I say a bit more to clarify this; for now, I comment simply that it does not seem to me that the fact that Kant calls appearances representations, and also calls intuitions, concepts, and perceptions representations, is sufficient to support Roche's claim that my interpretation is not idealist enough.

Roche also argues that my view cannot give a satisfactory account of the long tradition of holding that in some sense we impose spatio-temporality onto experiences. His point here is not just an appeal to the interpretative tradition (as oppposed to the text, where we do not actually find talk of imposition) but also to Kant's idea that our representations of space and time are a priori intuitions, and that a priori intuitions are not passively received, but rather what we bring to, or ‘put into’, the objects of experience. He claims that my reading accounts for the ideality of appearance properties, but not the ideality of space and time. This is an important issue. In Kant's account, space and time are actually more mind-dependent than the empirical properties we experience spatio-temporal objects as having; they are merely the forms of our intuition, and represent ‘no property at all of any things in themselves nor any relation of them to each other’ (A26). However, I think this is something my account can accommodate. Below, I sketch my use of the secondary quality analogy in a way which, I hope, explains how it can accommodate the ideality of space and time, as well as the question of whether or not my account identifies intrinsic properties and mind-dependent relational properties.

5. The Secondary Quality Analogy

In the paper in which I explore the secondary quality analogy (Allais Reference Allais2007), I draw on a metaphor John Campbell uses to give us a grip on the idea that what we know about cognitive processing need not drive us to a representational view of perception. He asks us to compare perception to viewing a scene through a pane of glass: in the metaphor, a representative account of perception involves seeing images on the glass, while for the direct realist, we see straight through the glass, to the objects beyond it (Campbell Reference Campbell2002: 119). He then asks us to imagine that the glass is volatile, and that keeping it transparent requires constant recalibration and adjustment. The upshot of the constant activity is to keep the world in view, not to create images on the glass. The analogy cannot be taken too far (it could be thought to suggest a homunculus looking through the eyes), but what it is meant to give us a grip on is the idea that cognitive processing does not threaten the relational account of perception.

I suggest that we can illustrate a possible account of non-veridical perception by introducing a variation on Campbell's analogy of the glass-like transparent medium of perception. Imagine that a malfunction in the adjustment mechanism causes the medium to fail to be completely transparent, and instead distorts the shapes of things to some extent, as, for example, a fishbowl lens might. According to the relational view, just as when the medium is transparent, it is not the case that the distorted things are seen through seeing an image on the glass—it is the things themselves, beyond the glass, which we directly see. If the fully transparent glass allows us to be directly in touch with the objects themselves, we can also be directly presented with the objects themselves when they are seen through a distorted glass. The curved appearance of the distortedly appearing objects is not a sensation objects produce in us; it is a feature of the objects which are seen, but which are seen in a way which differs, to some extent, from the way they are apart from their being seen. The argument against direct realism from non-veridical perception tries to force a non-identity between the physical object and what is directly perceived by claiming that the latter has properties that the former does not have. In response, the view I am describing denies that there is something of which the subject is aware which possesses sensible qualities that the physical object lacks: the physical object has the property of appearing the way it does, despite the fact that the way it appears differs, to some extent, from the way it is apart from its appearing. We must move slowly here: this is not meant to be an account of the status of Kantian appearances (it is an account of empirical illusion), but merely of one way of thinking about non-verdical perception, which allows that although seeing is relational in the sense that the object is essentially present to consciousness, it does not follow from this that the way the object is seen as being is entirely the way it is apart from its being seen, or that seeing is entirely transparent.

In this context, there is a natural use for talk of the way things are in themselves (of course, we are some way from Kant's use of this notion). We can now consider the shapes of things as they are apart from their being seen, or as they are apart from their visually appearing, and the way the shapes of things appear. We can use the term ‘as they are in themselves’ to refer to the way they are apart from their visual appearing. The view we are considering allows it to be a possibility that the way things visually appear may differ from the way they are in themselves, even though the visual appearing is something which essentially involves the objects, and not something which exists merely mentally. If this is possible, then it seems also to be possible that there are properties which belong to objects only as they perceptually appear. We can call properties of objects which we experience in perception manifest properties. We can then introduce the idea of essentially manifest properties: properties which objects have only in their perceptually appearing to subjects. The account of colour we need for the secondary quality analogy is that colour is an essentially manifest property.

To say that colour is an essentially manifest property is not to say that colour is a property of a mental state. Rather, on this view, colour is a relational property of objects, which they have as they appear to or are experienced by (certain kinds of) conscious subjects. Philosophical accounts of colour are subject to pulls in opposite directions by two features of colour. One pull is to preserve the idea that colour is, as it phenomenologically appears to be, a property of objects. The other pull is created by the idea that colour has an essential relation to visual experience. The idea that colours are essentially manifest properties attempts to accommodate both pulls. According to this view, there is an essential relation between colour and visual experience; colours are essentially features of the way objects look to conscious subjects. Colour is neither a mind-independent power to produce a purely mental effect, nor a purely mental effect produced by such a power, but rather is the mind-dependent way something mind-independent appears.

The view can be further explained by looking at one traditional argument for the idea that colour is mind-dependent, the idea that colour can be accessed only by sight (unlike shape and size, which can be accessed by more than one sensory modality). David Smith argues that this is the strongest argument for the secondary quality status of colour (although he characterizes this status a bit differently), saying that ‘it is not a contingent peculiarity of color qualia that they are not enjoyed through the exercise of smell or touch: for they define the sense modality of sight. When an object is perceived through a sense modality, the intrinsic nature of that sense determines how that object sensorily appears to the subject’ (Smith Reference Smith1990: 239). The idea is that we cannot make sense of the idea of accessing colour through any modality other than vision, because colour has an essential relation to visual experience; it is essentially a feature of the way objects look.Footnote 24

I suggest that we understand the sense in which appearances are mind-dependent along these lines: we should understand Kant's idealism about things as they appear to us as saying that what appears to us (what we experience) are essentially manifest features of reality, and that we can have knowledge only of essentially manifest features of reality. Appearances do not exist merely in mental states, but they do not exist in things as they are apart from their perceptually appearing to us. Appearances are not just perceptible things, things which manifest themselves to us, they are essentially perceptible, or essentially manifest. As Kant puts it, ‘What is not to be encountered in the object in itself at all, but is always to be encountered in its relation to the subject and is inseparable from the representation of the object, is appearance’ (B70n.). Immediately before presenting the secondary quality analogy Kant defines idealism as the view that the ‘things that we believe we perceive in intuition are only representations in thinking beings’ and explicitly rejects this view. However, he immediately goes on to say that appearances are representations (Proleg. 4: 289). In the same passage, Kant seems both to assert and to deny that the things we perceive in intuition are representations in us. Using the idea of essentially manifest properties, this can be explained. On the one hand, appearances are not representations in thinking beings in the sense of things which exist only in our minds. On the other hand, appearances are mere representations in the sense that they belong to objects only as those objects are for the consciousness of subjects like us. They are objects only for conscious subjects, not things that exist merely in consciousness.Footnote 25

One of Roche's worries is how space and time fit into this picture. Kant thinks that we could not experience individual objects without experiencing them as located in a spatio-temporal framework, and that the a priori nature of this framework means that it cannot be taken as reflecting mind-independent reality, but rather must be something brought to the world by our minds.Footnote 26 Let us return to the metaphor of the glass. We were imagining that the glass was doing something analogous to cognitive processing: taking sensory input and binding and organizing it in order to enable us to be in direct cognitive contact with objects beyond the glass. The cognitive processing is not producing an image which it makes available to consciousness,Footnote 27 but rather is making it possible for us to perceive the world. Now imagine that the glass has a built-in way of structuring, binding, and organising which does not reflect any aspect of the world beyond it, even though it is still input from the world beyond it which is being organized using this structure. The structure is a feature of the glass, and not of what is seen through it. It is still the case that we are seeing through the glass, to the world outside our minds, but if the structure which enables us to experience individual things does not present the world as it is in itself, the things that we experience through the structure will not be experienced as they are in themselves. On this view, the idealism about appearances in general would therefore follow, as Kant takes it to, from the ideality of our representations of space and time. Thus, it seems to me that my view can accommodate the idea that space and time are only features of objects as they appear, and do not represent any feature of things in themselves.

The view of colour as essentially manifest involves thinking that things have mind-independent natures which appear to us as coloured, and which our visual experience of colour does not give us insight into, as they are in themselves (as they are apart from their visual appearing). Kant's view says that all properties that we experience are like this: there is some mind-independent reality which grounds the way we experience things as being, but our experience gives us no insight into the nature of this mind-independent reality as it is in itself. This raises the question of the relation between mind-independent properties and appearance properties, another point on which Roche has worries about my account. He argues that my position has a difficulty which Kant is often said to face:

This is an extremely tricky point, one on which I have not been sufficiently clear, and which I still struggle to express precisely. I think that once again the secondary quality analogy is helpful here. The account of colour on which I have been drawing says that colour is a property of physical objects, but not one they have as they are in themselves, apart from the way they are perceived by creatures like us. As an essentially visual (essentially manifest) property, its existence is not separate from the possibility of its being perceived by creatures like us. Objects have mind-independent natures which we experience as coloured, but as it turns out, there is no simple correlation between a particular colour and a particular mind-independent surface property. Instead, the physical bases of colour are vastly disjunctive, and it may be that the only thing that unites the various properties we experience as red is our experiencing them as red. This would mean that redness is a property which can be picked out only at the level of visual experience. Despite this, it seems that there is a sense in which it is the mind-independent features of reality which are appearing to us as, for example, red. On this account, redness, as an essentially manifest property, belongs to objects only in relation to the possibility of their being perceived, and does not present the intrinsic nature of the features of objects which appear red. But redness is also the way in which something mind-independent appears. An important point to note about this is that since redness is a feature of objects which exists only at the level of visual experience, although it does not present the intrinsic nature of the many physical properties which appear red, it would not be right to say that it misrepresents them.

Of course, the secondary quality analogy is limited in a number of ways, and a crucial one is that, in this case, we ordinarily take it that we do know something about the mind-independent properties of objects which we experience as red—this is what enables us to know that they are vastly disjunctive. Still, the analogy is helpful in presenting two ideas: one is that if we think of colour as an essentially manifest quality we can see it as mind-dependent yet not a systematic illusion or misrepresentation, and the other is that we can see a sense in which colour cannot be identified with any mind-independent property, but can also see the point of saying that there is a sense in which there are certain mind-independent features of things which are the features we experience as coloured. Both these points are important when we extend the analogy, as Kant does, to all manifest properties of objects. As we know, Kant frequently speaks of things as they appear to us and these same things as they are in themselves, and this has frequently been taken to support non-idealist, deflationary interpretations of transcendental idealism. On the other hand, he says that appearances are mere representations, which exist in us, and whose existence is not separable from a link to possible perception. There is a sense in which it is things themselves which appear to us, but another sense in which we cannot identify things in themselves and appearances, since the latter are mind-dependent.

An objection to this might be that Kant denies that we have any knowledge of things in themselves, and further, in opposing Leibniz, he denies that phenomena should be thought of as unclear or confused representations of things in themselves, saying, on the contrary, that they do not represent things in themselves at all. But if we say that things have a way they are in themselves which grounds the way they appear to us, this might be thought to imply, against Kant, that our experience is some kind of apprehension of things in themselves, and therefore that things in themselves do appear to us.Footnote 28 Kant says that it would be wrong to think that we have confused representations of things in themselves, rather ‘We do not apprehend them in any fashion whatsoever’ (A44/B62). He argues that, for Leibniz, ‘appearance was the representation of the thing in itself’ (A270/B327), and he rejects this view, saying that appearances ‘do not represent things in themselves’ (A276/B332).

In response to this, it is important to see that we can deny that we have representations of things as they are in themselves without denying that there is a sense in which we have representations of things themselves,Footnote 29 and, in fact, this is something we find Kant doing. While he denies that we have representations of things as they are in themselves, including confused ones, he also sometimes says that appearances represent things themselves:

Here he says that appearances are the immediate and sensible representation of something which is an object independent of our sensibility. The passages in which Kant denies that we represent things in themselves in any way occur in places where his aim is to reject Leibniz's account of confused representations of things as they are in themselves (A44/B61–2; A271/B327; A276–B327). For Leibniz, the understanding is ‘immediately related to things in themselves’ (A271/B327), whereas Kant thinks that neither sensibility nor understanding give us knowledge of things as they are in themselves. So we can read Kant as follows: where he rejects Leibniz's view that appearance is the representation of the thing in itself, this is not because he is rejecting the idea that appearances represent things (themselves), which, as we have seen, he sometimes asserts, but rather because he rejects the idea that appearances represent things as they are in themselves. Immediately after criticizing Leibniz's view, Kant says that ‘we do not understand through pure reason what the things that appear to us might be in themselves’ (A276/B332; my emphasis), and here it is clearly things which appear to us which have a nature in themselves that we cannot know. Appearances are appearances of things themselves, and this means that there is a sense in which they represent things themselves and a sense in which they do not – they do not represent things as they are in themselves. Appearances are not confused or illusory representations of things as they are in themselves, but are representations of things which have a way they are in themselves.

There is much more to be said about the metaphysical status of essentially manifest properties, and the sense in which they are mind-dependent; here I will make one brief final comment about the direction in which I think we should go. As we have seen, one of the features of the text which seems to support a phenomenalist reading is the fact that Kant repeatedly says that outer objects exist or are real in perception only (wirklich nur in der Wahrnehmung), and that the actuality of a thing requires ‘its connection with some actual perception in accordance with the analogies of experience’ (A225/B272, my emphasis; A376; A493/B521; A493–4/B521–2). Kant clearly expresses a necessary connection between the existence of empirical objects and our being able to experience them, and this must be accommodated and explained by any convincing interpretation of transcendental idealism.Footnote 30 Kant does not say that objects exist only in our actual experience of them (esse est percipi) but that their existence is not separable from our possible perception of them. He allows as part of what is empirically real things which are so small or so far away that we cannot perceive them, or for which we do not have the requisite sense organs (A225–6/B273; A522/B550). Once again, this creates interpretative pulls in different directions: we need a kind of mind-dependence which involves an essential relation to perception, while allowing a realism which accommodates the existence of things too far away for us to perceive. Although I cannot argue for this here, I suggest that, despite significant differences between the positions, it is helpful to compare Kant's idealism with a kind of anti-realism which rejects verification-transcendence.Footnote 31 The idea is that essentially manifest properties are not experience-transcendent. Kant does not think that objects depend for their existence on particular acts of apprehension by particular subjects, nor does he think that objects exist only as constructions out of merely mental states, but he thinks that the fact that an object exists is not separable from the fact that we can, in principle, have experience of it: in other words, Kant rejects experience-transcendence for appearances. Clearly, there is much more to be said on all of these points; I am grateful for the opportunity to have said a bit more to explain my view, and to Andrew Roche, whose stimulating paper provided the opportunity to do this.