1

Lucy Allais’s bookFootnote ¹ is an extremely rich and wide-ranging work, complex and creative yet clear and compelling. It sheds new light on a host of traditional topics and will no doubt become an essential point of reference in future debate. One especially interesting and distinctive feature of the book is how it integrates familiar issues in Kant’s metaphysics and epistemology with novel and emerging ideas about his philosophy of perception. Another is how it draws inspiration in this from a wide variety of views in contemporary philosophy. Two in particular are central to the book and they will be my focus here.

First, in her account of intuition, Allais draws on a form of relationalism about perception (RP). Second, in her account of the properties of the objects of possible cognition, she draws on a form of relationalism about colour (RC). Both are central to the book because both are central to its account of transcendental idealism. The first provides resources for Allais’s account of Kant’s master argument for transcendental idealism. The second provides the structural core of her account of the nature of that idealism.

In light of the nomenclature I have just employed, we might expect these two views to fit naturally together. Allais certainly does. Indeed, she says RC ‘requires’ RP (p. 121; cf. pp. 16, 103, 117). I will argue, however, that not only do the views not fit naturally together or entail one another – they are fundamentally incompatible. One salient way to characterize the problem is that RP requires a more robust form of realism about the properties of the objects of perception than can be accommodated in the (at least partially) idealistic framework of RC. Though ingenious, Allais’s new interpretation of Kant’s notoriously difficult balancing act leaves it unstable.

In §2 I outline RP with a view to honing in on the tension with RC. In §3 I outline RC and explain my worry. Up to this point I focus exclusively on philosophical issues. In §4 I return to Kant and briefly explain why it will not help to appeal to differences between the Kantian and the contemporary contexts. I end with a proposal and a question.

2

What is relationalism about perception, or RP? One core aspect of the view is the claim that perception essentially involves the subjects of perception standing in primitive, non-representational relations of conscious acquaintance to the objects of perception. The metaphysical nature of the relations in question is such that their obtaining – and therefore the occurrence of the perceptions that they (at least partially) constitute – necessarily implies the concurrent existence and presence to the subject of the objects of perception. Note, however, that nothing has yet been said about the nature of these object-relata. So far, they are just whatever is perceived – those things, whatever they may be, to which the subject stands in the conscious acquaintance relation. More specifically, then, the form of relationalism that Allais draws on in her reading of Kantian intuition is one that takes the object-relata in such relations to include the kind of external, physical entities that populate our environment – things like apples and the trees they grow on.

In this the view contrasts with an alternative which can also be classified as a form of relationalism because it likewise analyses perception in terms of primitive, non-representational relations of conscious acquaintance. The difference is that this alternative view then takes the object-relata in such relations always to be non-physical ‘sense-data’ – somewhat mysterious entities that are, at best, merely caused by external, physical things like apples and trees. It is thereby sometimes characterized as an indirect realist theory of perception in that it yields a distinction between the sense-data that are the only direct objects of perception – perception being essentially the subject’s conscious acquaintance with them – and the real-world objects that might be counted as the indirect objects of perception in virtue of standing in the right kind of causal relations to sense-data. These causal relations might in turn be taken to ground representation relations and at this stage the view can also be called ‘representationalist’ in the rather restricted, ‘Cartesian’ sense in which Allais tends to employ that moniker, where the vehicle of representation is also the object of consciousness. But it is worth noting the initial, core similarity between these two views, in part because Allais does not, in part because it will be relevant below and in part because she often proceeds as though they are the only options, for instance inferring the plausibility of one from the inadequacy of the other (pp. 11–12, 106, 114, 154, 157). Non-relationalist, so-called ‘intentionalist’ versions of representationalism on which the vehicle of representation is not itself the object of consciousness get rather short shrift in the book.

In any case, the particular form of relationalism that Allais draws on – what I am simply calling RP – is sometimes called a naïve realist theory of perception, and one way in which this name is fitting is that the view says that the direct objects of perception are what they naïvely seem to be – the kind of external, physical entities that populate our environment. According to Allais, ‘intuitions involve the presence to consciousness of their objects’ (p. 153), where ‘being perceptually presented with a particular involves a subject’s being aware of something outside of and other than themself, as opposed to being aware of a sensation, such as an itch, which is simply awareness of a state of the self’ (p. 154).

Another way to characterize RP, and one which Allais makes much of (pp. 12, 105ff., 134, 197), is as the view that external, physical things can be literal constituents of perception. Or as she sometimes puts it:

This does not mean that such objects are ‘in the mind’ in anything like the way that sensations or dreams might be said to be. That way lies the sense-data view. According to RP, worldly objects can only be said to be in the mind to the extent that the mind can likewise be said to be in the world. Metaphors abound that run in both directions. In perception, the mind spreads itself on the world. In perception, the world is digested by the mind. How, then, more precisely, are we to understand RP’s particular constitution claim, the idea that external, physical objects can be in consciousness?

Note that it would not suffice to appeal to the mere fact that RP analyses perception as a kind of existence-and-presence-entailing relation. Other views distinct from RP do this. One might, for instance, analyse perception in terms of suitable causal connections holding between subject and object, where ‘suitable’ demands existence and presence. But we are not tempted to characterize such views as ones on which the object of perception is a literal constituent of the state of perceiving it – object-dependent does not entail object-involving. And even if we were so tempted, the constitution claim would then not distinguish RP.Footnote ² RP claims precisely that there are extra-causal conditions on perception. Instead, then, the answer to how we are to understand RP’s talk of constitution lies in its account of what in contemporary parlance gets called phenomenal character.

The phenomenal character of a state of perceiving consists of its phenomenal properties, roughly those properties that determine ‘what it is like’ or ‘how it feels’ for a subject to be in that state. RP says that (at least some of) these phenomenal properties just are properties of the perceived object. As Allais puts it, ‘the qualitative features of perceptual experience are features of the objects perceived’ (p. 107). Thus the redness in a visual perceptual appearance, for example, is the redness in the object perceived. It is in this way that the object counts as a literal constituent of the perception of it, as being in consciousness.

The idea here, and one of the core motivations for contemporary versions of RP, is that this picture provides the best explanation of phenomenal character. And it is a reductive, naturalistic explanation. We do not need to appeal to shadowy mental paint, mysterious private objects, or raw subjective feels in order to explain perception and the character of phenomenal consciousness. We simply appeal to the familiar, everyday objects that we perceive and with which we are consciously acquainted. The heart of RP is an appeal to the properties of such objects as an explanation of perceptual appearances.

And it is here that we come to the fundamental tension between RP and RC, between relationalism about perception and relationalism about colour. For in RC, the order of explanation is precisely the reverse. Where RP explains the phenomenal character of perception (at least partially) in terms of the properties of objects, RC explains what it is for objects to have certain properties (at least partially) in terms of the phenomenal character of perception. Combining RP and RC, were it possible at all, would lead to an explanatory circle that ultimately leaves us with no account of either phenomenon.

3

What is relationalism about colour, or RC? One core aspect of the view is the claim that colours are relational properties of objects, properties ‘which are partly dependent on how coloured objects are in themselves and partly dependent on subjects (and also environmental context)’ (p. 122). As with RP, however, this core claim leaves a lot undetermined. In particular, nothing has yet been said about what in the subject acts as the relevant co-determiner. More specifically, then, the particular form of relationalism about colour that Allais draws on to model the properties of cognizable objects – what I am simply calling RC – locates that co-determiner in the phenomenal character of the subject’s perceptual episodes. Colours, according to RC, are ‘qualities objects have only in their perceptual appearing to subjects like us’ (p. 121) or ‘qualities which belong only to things’ perceptual appearing’ (p. 122) – they are what Allais calls ‘essentially manifest qualities’ (pp. 117ff.):

RC is thereby supposed to accommodate two intuitive features of colour that can otherwise seem to pull in opposite directions. On the one hand, colour is presented to us in visual perception as a property of objects. It seems to us as though it is the apple that is red. Colour is more like shape than blurriness in this regard. On the other hand, it seems essential to colour that it is something presented in visual experience. Colour just is, so the claim goes, a feature of the way things appear, and thus dependent on or a feature of visual perception. By analysing colours as relational, as properties of objects-as-they-appear-in-visual-perception, RC respects both ideas – it captures both the objective and subjective character of colour.

Allais usefully contrasts RC with a variety of other views (pp. 118ff.). ‘Scientific objectivism’, for example, identifies colours with object-properties like microphysical surface structure or causal power. It thereby respects the objective character of colour, but since such properties do not themselves ‘manifest’ in perception – we are not perceptually acquainted with such properties – the view fails entirely to capture colour’s subjective character. At the other extreme there are ‘idealist’ views, which identify colours with purely mental entities like ideas or sensations. These views respect the subjective character of colour – its essential connection to perceptual appearances – but only at the expense of making it a property of subjects and their states rather than of objects.

These and other contrasts that Allais draws are correct and apt. But what is relevant here is what RC shares with idealist views. In their respective, otherwise quite different accounts of the nature of colour – in their respective explanations of what colour is – both kinds of view make ineliminable appeal to phenomenal character. And the problem is that RP, as an account of the nature of phenomenal character in terms of the nature of objects and their properties, can have no truck with any appeal to phenomenal character in an account of the properties of objects. The point can be put in terms of grounding rather than explanation. According to RP, facts about the properties of objects ground facts about phenomenal character. According to RC, facts about phenomenal character ground facts about the properties of objects. And we cannot have it both ways because grounding is an asymmetrical relation.

Let me elaborate. Suppose we ask what it is for an object to be red. RC, according to Allais, says that an object is red ‘only if there is a way it would appear to subjects who are suitably situated and suitably receptive’ (pp. 123–4).Footnote ⁴ But now suppose we ask: in what way? In what particular way would a red object appear? Well, red objects appear red. So the account looks in danger of empty circularity. It looks like it says merely that being red is a matter of appearing red. Now, theories like RC have a standard, two-step response to this worry. First, they propose an equivocation in the two uses of the term ‘red’ in such statements. While the first occurrence names a (relational) property of the object, the second names a property of the subject, namely her being in a certain sensory state, a state with a certain phenomenal character. Second, they provide an internal, non-relational characterization of that state. What is crucial in this second step is that the account of the state of the subject make no appeal to the property of the object that brought it about or determined it, for we are precisely trying to give an account of that object-property – we are looking for the determining relata that together would individuate the object’s redness. But RP rejects such purely internal characterizations of sensory states because it reduces their phenomenal character properties to the properties of objects.

Consider the following toy case. In answer to our second question above, a subject perceiving a red object demonstratively picks out in introspection the relevant sensory state, an aspect of the phenomenal character of her perception of a red object – ‘the red object is appearing that way, which way I hereby dub “red*”’, she says. Thus the equivocation is explicit and circularity avoided – RC says that to be red is to appear in a red* way. But it is precisely this kind of move that RP blocks. For according to RP, what such a subject is ‘pointing to’ in our toy case is nothing but the property of the object – RP is the view that red*-ness is redness.

In fact this is all a little too rough. Theories like RP tend to be presented as general theories, supposed to hold for all perception. This is how I presented it above, following Allais. This need not be the case, however, and localized versions of theories like RP and RC would be compatible so long as they kept fully separate, on the one hand, the properties of phenomenal character for which one gives a relationalist analysis in terms of the properties of objects, and on the other hand, the properties of objects for which one gives a relationalist analysis (partially) in terms of the properties of phenomenal character. In principle there is no reason why one could not simultaneously maintain something like RP with regard to P-properties and something like RC with regard to Q-properties, if, and only if, the set of P-properties does not overlap with the set of Q-properties.

More carefully, then, the problem for Allais’s account is that it generalizes RC to provide the model for all possible properties of cognizable objects. According to Allais, the possible properties of such objects – Kantian ‘appearances’ – are exactly those properties possibly manifest in perception. Hence the ‘perceptuals’ in my title. Allais’s concern is of course not only with colour but with any similarly perceptual property. This leaves no room for RP. Or coming from the other direction, Allais’s account mobilizes a general version of RP, leaving no room for RC, never mind a generalization of it.

4

Where does this leave us? One response might be that it is far from clear to what extent, if at all, Kant was concerned with anything like the notion of phenomenal character that has been my focus here. It certainly does not follow merely from his use of related vocabulary. In particular, the reductive account in terms of which I have explained RP does not look especially Kantian, so surely it does not matter if this cannot be retained when the view is combined with RC. Such worries about Kantian provenance are well founded. Indeed, they are part of my point. But this response mislocates the force of my objection. I have argued that combining RP and RC is philosophically inconsistent. The problem is not just that one or other of the views loses some theoretical benefit when combined, but rather that it is far from clear that their combination even makes sense. And I do not think it would help to attempt to somehow divorce either of the views from considerations of phenomenal character.

Take RC. One way of resolving the tension with RP would be to offer the required, fully internal characterization of the subject-relata in bodily terms, as, for instance, and very roughly, features of the retina. Thus being red is a relational property of an object not insofar as the object phenomenally appears a certain way, but rather insofar as it produces a certain retinal effect. Now this is simply a different view to RC, albeit one that might still be classified as relationalist. More importantly, however, it could no longer serve as a model for transcendental idealism. First, body-dependence is not mind-dependence, for Kant. Second, for Kant, bodily states are just as ‘empirical’ as red apples, and thus would themselves be in need of the ‘transcendental’, relationalist analysis.

In any case, I think RC provides a brilliant and illuminating way of modelling Kant’s signature doctrine. The real culprit, in my view, is RP. I have argued elsewhere that the way RP would have us understand the ‘immediacy’ of intuition, for example, cannot be the way Kant understands it.Footnote ⁵ My point here can be seen as reinforcing and expanding that claim. Given her employment of RC, it cannot be the way Allais understands the immediacy of intuition either.

It is again far from clear whether it is so much as possible to separate RP from its reductive account of phenomenal character, without, that is, fundamentally changing or obscuring the nature of the view. We have seen that this account is how RP cashes out its talk of external objects being literal constituents of the state of perceiving them. Otherwise put – and I think this is the real heart of the matter – how are we to understand RP’s special conscious acquaintance relation if not in terms of phenomenal character and specifically the claim that its properties just are those of the object? It was precisely this claim that gave us an idea of the way in which the relation was not only existence-and-presence-entailing but non-representational and non-causal, or at least not merely causal; it was precisely this that unpacked the notion of our immediate perceptual acquaintance with the world, this that constituted the way in which reality is manifest to us in perception. Yet it is precisely this that is stripped from the view when we try to combine it with RC.

In effect, then, I am recommending that Allais jettison RP. And so finally, it is useful to ask: what would be lost if she did? Significantly, I think the answer is: not much. The central task of any account of intuition is to explain how it is able to play its role of ‘giving’ us objects for cognition, which is to say the role in virtue of fulfilling which it forms a necessary condition on cognition. After all, that intuition forms such a condition could well be described as the single most important commitment of the entire Critical philosophy. Allais understands cognition as a kind of singular thought, so we can be more specific with our question. What does it contribute to explaining its role in singular thought to construe intuition in terms of some special RP-style constitution relation, over and above some suitable causal or otherwise object-dependent relation? The widespread popularity of causal and otherwise non-RP accounts of singular thought is testament to the difficulty of answering this question. And without an answer – an answer which is only more elusive in the presence of RC – I cannot see what real theoretical work RP is doing for Allais’s Kant.Footnote ⁶