On A. D. Smith’s constancy based defence of direct realism

This paper presents an argument against A D Smith’s Direct Realist theory of perception, which attempts to defend Direct Realism against the argument from illusion by appealing to conscious perceptual states that are structured by the perceptual constancies. Smith’s contention is that the immediate objects of perceptual awareness are characterised by these constancies, which removes any difficulty there may be in identifying them with the external, or normal, objects of awareness. It is here argued that Smith’s theory does not provide an adequate defence of Direct Realism because it does not adequately deal with the difficulties posed by the possibility of perceptual illusion. It is argued that there remain possible illusory experiences where the immediate objects of awareness, which in Smith’s account are those characterised by perceptual constancies, cannot be identified with the external objects of awareness, contrary to Direct Realism. A further argument is offered to extend this conclusion to all non-illusory cases, by adapting an argument of Smith’s own for the generalising step of the Argument from Illusion. The result is that Smith’s theory does not provide an adequate Direct Realist account of the possibility of perceptual illusion.

theory in response to a number of criticisms levelled against it by Susanna Siegel (2006, pp. 378-410). In particular, Siegel has argued that many of Smith's claims about the nature of perceptual consciousness, claims which feature prominently in his answers to the arguments from illusion and from hallucination, are compatible with Indirect Realism. Smith responded to these criticisms in two ways: partly by articulating his views more fully and partly by arguing that his own theory of perception is not meant to show Indirect Realism to be false, but to show that Direct Realism can be true even in the face of the arguments from illusion and from hallucination. I find Smith's responses to be generally adequate to Siegel's criticisms. In this paper, though, I argue against Smith's theory on grounds distinct from those proposed by Siegel. I will argue that Smith's theory does not provide an adequate defence of Direct Realism because it does not adequately deal with the difficulties posed by the possibility of perceptual illusion. This is because, for all Smith says to the contrary, it still remains that in special kinds of illusory cases the immediate objects of sensory awareness, even when these are construed as the deliverances of perceptual constancies in the manner Smith requires, cannot be identified with the ordinary, external objects of awareness. In short, Smith's theory does not provide an adequate answer to the argument from the possibility of perceptual illusion.
I begin my discussion in the next section with a brief exposition of Smith's theory and the way it is intended to provide a defence of Direct Realism in the face of the argument from illusion. In Sect. 3 I argue that in certain special cases of illusion the sensory state must, even according to Smith's general analysis of sensory and perceptual states, be counted as an instance of indirect perceptual awareness. I then provide a further argument, a modification of one of Smith's own, to show that if we treat the special cases as instances of indirect awareness we should also treat the normal cases in this way. The upshot of these arguments is that the possibility of perceptual illusion cannot be made compatible with Direct Realism simply by an appeal to the perceptual constancies, in the way advocated by Smith.

Smith's defence of direct realism
Smith presents his theory of perception as a defence of Direct Realism, in the face of the arguments from illusion and hallucination. Here 'Direct Realism' is to be understood as a kind of theory of perception which does not make a certain kind of existential commitment which is made by its realist alternative, Indirect Realism. Indirect Realism postulates, in addition to all the objects of awareness acknowledged by common sense and empirical discovery (referred to by Smith as 'normal' objects of awareness), a distinctive category of object: those in virtue of which we enjoy sensory awareness of the others. The Indirect Realist postulates two categories of objects of awareness and conceives of perceptual awareness as mediate; the Direct Realist, on the other hand, postulates only one category of object of awareness and perceptual awareness is not construed as mediate in the way articulated by the Indirect Realist.
Our common sense view of perception is most naturally expressed as a variant of Direct Realism, because prior to serious reflection on perception we do not acknowledge a distinctive category of mediating objects of awareness. One motivation for making such an acknowledgment comes from reflection on the possibility of perceptual error: commonly, perceptual illusion or total hallucination. Each of these forms of perceptual error can be exploited to generate an argument in favour of Indirect Realism, known, respectively as the argument from illusion and the argument from hallucination. The idea behind these arguments is that perception involves a conscious mental state, referred to as a sensation, which is characterised by a range of properties which do not merely specify the content of the state but are genuine properties of the vehicle of that content, such as colour, pitch, texture, etc. Smith gives these properties the label 'sensory qualities'. Now, in cases of perceptual error there is no object characterised by the relevant properties, either because there are no objects with these particular properties (illusion) or because there are no objects of the relevant kind whatever (hallucination). These are possible situations and these possibilities demand some explanation. As my own objection to Smith's theory concerns its capacity to deal adequately with perceptual illusion, in what follows I will focus on the argument from illusion.
The Indirect Realist has available an explanation of perceptual illusion: in such situations there are mediating objects of awareness characterised by these properties. Moreover, these mediating objects are identified with sensations: they just are mental states characterised by sensory properties. As Smith conceives it, the argument from illusion poses the following challenge to someone wishing to defend a Direct Realist theory of perception: sensations, defined as above, must be acknowledged in the theory, but without these mental states being construed as objects of awareness. Given the sheer possibility of perceptual illusion and the thought that any account of cases of illusion must apply equally to non-illusory cases, failure to do so will result in the conclusion that sensations are not just occasionally objects of awareness but always immediate objects of awareness. This will just be to concede the argument to the Indirect Realist, who holds that in perception we are aware of external, or what Smith calls 'normal', objects always only in virtue of an awareness of some other object which we are aware of immediately.
The Direct Realist must offer an explanation at least as plausible, but preferably better, than the Indirect Realist's. Smith argues at length that various forms of Direct Realism which divide perceptual states into a conceptual mental state and a non-conceptual sensation fail to provide a plausible construal of sensations that does not make them objects of awareness. 1 Moreover, Smith argues that simply rejecting the act/object analysis of perception, which the sense-datum theorists of the early 20th century appealed to in their arguments for Indirect Realism, will be insufficient to meet the challenge presented by the argument from illusion: For even if sensory qualities are inherent qualities of sense experience itself, it is far from clear how we can avoid the conclusion that we are aware of them as objects whenever we are perceptually conscious, or that we immediately aware of the experience itself that exhibits such qualities. On such a view perceptual experience would be self-presenting, and the upshot of the Argument would be that we are only ever aware of our own experiences, with such experiences themselves constituting the veil of perception…. Perception is sensuous in a way that mere thought is not because it involves perceptual sensation. Our problem is, therefore, to see how such sensations can fail to be our immediate objects of awareness whenever we perceive. Smith (2002, pp. 60-61).
Smith's own theory of perception aims to address this problem: it acknowledges sensations, in the sense specified earlier: conscious mental states characterised by sensible qualities, which do not merely specify the content of the state but are genuine properties of the vehicle of that content. However, in Smith's theory sensations are not construed as objects of awareness at all. To see why Smith thinks he can claim this we need to introduce two other states of mind that Smith's theory postulates 2 : 1. Phenomenally perceptual conscious states 2. Perceptions 'Phenomenally perceptual conscious states' are states in which sensations are structured in a distinctive manner: the phenomenal character of these experiences is structured by at least one of three non-sensory features: i. Phenomenal three-dimensionality ii. Kinetic structure or position constancy iii. The Anstoss, or a non-sensory awareness of one's own agency 'Perceptions', on the other hand, are states where these features do not actually structure the phenomenal character of the experience, but the perceptual systems of the subject are such that movement would result in 'phenomenally perceptual conscious states'. 'Mere' sensations are states where movement would not result in such phenomenally perceptual conscious states, such as in the case of after-images. Now, the reason that Smith's theory can offer a response to the argument from illusion is because, on his account, objects of awareness only begin to enter the story when we consider phenomenally perceptual conscious states. The proper specification of the properties possessed by the objects of perceptual awareness are, Smith claims, to be found by considering the non-sensory features distinctive of these phenomenally perceptual conscious states. A correct list of the properties possessed by the objects of awareness is not to be found by considering the sensory qualities that 'mere' sensations possess, or are 'characterized by': these latter properties are not, according to Smith, properties of the object of awareness at all. For Smith, what counts as the immediate object of awareness is 'characterized' by these non-sensory features, which he calls 'phenomenological constancies '. 3 Such constancy is not the function of any ''judgment'' that supplements the deliverances of the sense; it characterizes what we are aware of in the most ''basic'' and ''immediate'' sense. Smith (2002, p. 178) Smith's idea seems to be that, whatever the immediate object of awareness turns out to be, the actual or potential operation of the constancies mean that the object of immediate awareness will possess the properties of being spatially related to my sense organs, being in one unchanging position as I move around it, or as resisting the imposition of my agency in specific ways as I push against it. None of this is true of 'mere' sensation, so sensations are not, on this account, to be construed as objects of awareness, contrary to Indirect Realism.
Smith argues that the many philosophers who have been tempted by Indirect Realism have gone wrong in two places: firstly, in thinking that the immediate objects of awareness should be construed as being characterised by any old sensible qualities which a sensation is characterised by, such as the increase in extensity that is a feature of objects getting closer. 4 Secondly, they have gone wrong in failing to recognise that Smith's non-sensory features are just as much 'original' features of perceptual phenomenology than are those changes in extensity just mentioned. The difficulties for Direct Realism in providing an adequate response to the argument from illusion if these views are adopted would be 'overwhelming', as Smith believes. The only reason Smith can see for adopting these views, however, arises in the visual case and amounts to the claim that visual experience is not phenomenally three-dimensional in character. Smith thinks that there is 'nothing whatever' to be said in favour of this, 5 so we have no reason to adopt these views that would present overwhelming difficulties for Direct Realism.
In response to Smith's account of phenomenally perceptual conscious states and the significance of the constancies for the argument from illusion, Siegel has observed that, although there is much to be said in its favour, nothing said by Smith rules out postulating mediate objects of awareness which possess these constancy features. This is to say that nothing Smith has to say rules out Indirect Realism. At this point Smith could have availed himself of an argument he gives briefly at the outset of The Problem of Perception, that, Indirect Realism fails not simply because it offers a poor account of perceptual illusion, but because it is incoherent: The point about Indirect Realism is not that it is epistemologically suspect, but that it is incoherent…. We need, first, to be clear precisely what it is that this position invites us to be Realists about. The domain in question is that of the physical components of the empirical world. An empirical world is an environment of which we are cognizant through perception-one that contains entities we can and do perceive…. Any hypothesized realm of concrete, nonconscious elements… with which we could have no perceptual dealings whatever, which contained no entity that we could possibly perceive would not be our physical world…. A realm is empirical, is our world, only in virtue of elements in it being perceptible; but for the Indirect Realist this can only be a matter of indirect perception. So this realm is supposed to be both empirical and yet irremediably behind a ''veil of perception'' (Smith 2002, pp. 13-14).
This brief argument is, I think, not particularly compelling because it tacitly supposes that the possibility of perception which makes an empirical world our world must be the possibility of unmediated perception, which is not obviously plausible. However, the central thrust of Smith's attempt to defend Direct Realism does not rely on the success of this argument. Smith does not intend this argument against Indirect Realism to conclusively demonstrate the truth of Direct Realism, because Smith considers Idealism to be a legitimate alternative, competing explanation of the facts about perception. 6 Moreover, the response Smith gives to Siegel's criticism does not appeal to this argument. The reply he offers to Siegel has two features: first, where Siegel suggests that various versions of the sense-datum theory can incorporate the constancy features Smith appeals to, Smith offers a series of detailed objections. Secondly, Smith observes that even if there were a version of Indirect Realism that could do so, this would be irrelevant to the limited, 'essentiality defensive task' of defending Direct Realism from the challenge raised by the possibilities of illusion and hallucination: …all I need to do is to give a plausible account of a way in which sensations can fail to be objects of awareness: to offer an unrefuted and at least equally plausible alternative account to the sense-datum theory. (Smith 2006, p. 414).
Smith's theory is precisely intended to offer such an alternative. So, even though everything Smith says about the constancies may be consistent with Indirect Realism, this does not impugn his central claim that the Argument from Illusion provides no good grounds for the falsity of Direct Realism.
Though Smith restricts himself to this limited defensive task, there would remain a further step required for someone who wishes to defend Direct Realism. Recall that Smith considers the Argument from Illusion to present an explanatory challenge: to explain how there can be illusory perceptions involving sensations which possess sensory properties. In the absence of a cogent Direct Realist response to this challenge Smith argues we should favour the ontologically more profligate, but explanatorily more satisfactory Indirect Realist account. In the scenario envisaged by Siegel, however, the Indirect Realist co-opts those parts of Smith's Direct Realist account that are intended to render it explanatorily satisfactory. However, when we have equally satisfactory explanations, as in such a scenario, we must turn to other considerations to 'break the tie'. There are, though, at least two that are available here: we could observe that Direct Realism is to be preferred to Indirect Realism either because it is closer to our pre-theoretical thinking about perception, or that it is ontologically more economical.

The possibility of illusion
Having outlined the way that Smith's theory attempts to defend Direct Realism, I will now proceed to argue that the theory, ingenious though its strategy is for blocking some of the traditional considerations in favour of Indirect Realism, is nonetheless unsuccessful. This is because the reason given in the argument from illusion for not identifying the external, or what Smith calls the 'normal', objects of awareness with the immediate objects of awareness, construed as sensations, applies equally to any attempt to identify the normal, external object of awareness with Smith's own proposed objects of immediate awareness, i.e., those introduced as being 'characterised' by the perceptual constancies. The reason given to block the identification in the case of sensations by the argument from illusion was simply the possibility of perceptual illusion.
To see why there is a difficulty here for Smith, let us grant his proposal that in phenomenally perceptual conscious states which register objects in sensation 7 whatever is to be characterised as the immediate object of awareness is characterised by the phenomenal constancies. The question that naturally arises is whether such objects can have properties different from the normal, external objects of awareness. If this is possible then, by Leibniz's Law, they cannot be identical and Direct Realism is false. If, on the other hand, Smith's proposed immediate objects of awareness cannot have properties different from the normal objects then his version of Direct Realism will be defensible. The problem facing Smith's defence of Direct Realism is that a good case can be made in favour of the following metaphysical possibility: a situation where the objects characterised by the constancies do not have the same properties as the normal, external object of awareness. My argument is different from Siegel's in the following important respect: it does not turn on the (legitimate) observation that sense-data could be three-dimensional and characterised by the constancies. Instead the idea is that we grant Smith the hypothesis that whatever is the object we are immediately aware of in perception is characterised by the constancies and then show that this object is one which can have properties other than the normal, external object of awareness.
Before I present the argument, I need to be clear that my objection is not the same as one that Smith has already made provision for in his theory: the recognition that it is possible for the constancies to fail. As a matter of fact, such constancies can and do fail us and some scenarios can be envisaged where they do not even feature in the phenomenal character of our sense experience. It is to account for this that Smith introduced a distinction between 'perceptions' and 'phenomenally perceptual conscious states'. A sense experience counts as the latter if it is actually characterised by the constancies; it counts as the former if it is not, but if upon movement the perceptual systems of the creature would give rise to a phenomenally perceptual conscious state. In the case of 'perceptions' the characterisation of the immediate object of awareness is still given by the constancies which would feature in experience, were the subject to move. This provides an elegant solution to the difficulty posed by possible failures in the operation of the constancies. 8 However, my argument does not appeal to the possibility of failure of the operation of the constancies, but instead appeals to the possibility of a creature with a sensory field that is phenomenally three-dimensional and characterised by perceptual constancies, but which is systematically non-veridical. The kind of case I have in mind can be illustrated by thinking about a kind of experience that humans can have. I suffer from astigmatism in one eye; when I first began wearing corrective glasses for this my visual experiences of movement at the edges of my visual field, outside the area covered by my corrective glasses, was extremely disorientating. As I walked along, objects appearing to move toward me at a regular speed would appear to suddenly leap closer to me as they passed across the boundary set by the rim of my glasses. Although the effects of astigmatism were disorientating, the effects were sufficiently minor that it was no impediment to navigating my locomotion. It is not important to the example, though, that there be variability in the sensory qualities of such an experience as the objects appear to move; what is important is that any attempt on my part to reach out and catch such objects would systematically result in error. The errors in this kind of case arise from precisely the non-sensory constancy features that are home ground for Smith's theory. These kinds of abrupt shifts in apparent position were the result of putting on spectacles, but there is no reason to suppose that there could not be a creature whose experiences were like this as a result of the normal operation of the perceptual systems that underwrite the constancy features of its sensory experience. Moreover, there is no reason to suppose that there could not be such a creature for who these shifts and discontinuities were quite uniform across its sense fields.
For further illustration, consider what it is like to watch an old, juddery, blackand-white film of a train approaching. The train appears generally to approach at a constant speed along a single, straight path. However, the judder of the film can result in noticeable shifts in the apparent position and speed of the train as it approaches: sometimes it appears to move slightly faster, sometimes slightly slower; sometimes it seems to jump to the left, sometimes to the right. This concerns the phenomenal character of a perceptual experience, but of course our visual system can recover a general path and speed for the train-we experience the film as showing a general progression of the train toward us. Now, this recovery can be more or less precise: the case I have in mind is one where this recovery is imprecise. The case I have in mind is of a creature with perceptual systems that underwrite its visual modality whose normal operation results in a sensuously presentational perceptual state where the constancy structured experience is rather like ours when we watch a juddery, black and white film. However, the constancy features of the experience are sufficiently imprecise that most of the creature's attempts to reach for objects on the basis of this awareness result in failure. For such a creature, when objects appear to approach it they are characterised by fairly variable deformations in apparent size, shape and discontinuous shifts in position (where such deformations, though variable, are not extreme). Such a creature could certainly get around in the world, albeit with a fair amount of luck. Now consider a case where such a creature has such an experience as it perceives the approach of an object, where the normal, external object of awareness undergoes no such deformations and shifts in position. For a creature such as I have described, the objects characterised by the perceptual constancies will then not be identifiable with the normal, external objects of awareness, by Leibniz's Law. Moreover, in this case Smith will not be able to appeal to his distinction between perception and phenomenally perceptual conscious states in the way he does with actual failures of constancy in human beings. This is because for human beings the perceptual systems underwriting our sense experience would, upon movement, result in objects whose characterisation in terms of the constancies would (according to Smith) not preclude identification. In the case of the creature I have described, this is just not the case-movement by such a creature would still result in immediate objects of awareness whose characterisation in terms of the constancies differs from the properties possessed by the normal, external object of awareness. For such creatures, then, Smith's appeal to constancies will be insufficient to meet the challenge raised by the argument from illusion. Such creatures would be creatures for who Indirect Realism would be true.
In this regard, my criticism is different from that made by Georges Dicker in his review of Smith's book. Dicker's criticism focuses on the possibility that in the ordinary course of our experiences which are characterised by constancies, those constancies might fail: 'But he must allow, as seems true anyway, that on occasion constancy fails for shape and size…. How then are we to avoid commitment to sense-data based on cases where the appeal to constancy fails?' (Dicker 2006, p. 426) Smith's response to this question is just that given above: although in such cases constancy fails, there are possible movements that would result in experiences where the constancies are restored. In the cases where the constancies fail we enjoy either mere sensation, in which case according to Smith there is no object of perceptual awareness, or we enjoy a perception, where the object of awareness is characterised by those constancies that have, in this instance, failed.
Dicker's objection concentrates on the case of a creature whose experience is normally characterised by constancies and when these constancies are operative there is no discrepancy between their characterisation of the immediate object of awareness and the properties characterising the 'normal', external object of awareness. The case of constancy failure Dickers seems to have in mind is a situation where such a creature has an experience, resulting from the failure of these constancies, where there is a discrepancy between the sensory properties characterised by the experience and the properties of the normal object of awareness. This situation is one to which, as I have said, Smith has a response. There is, however, a more problematic case for Smith: this is the case of a creature which has experiences that are normally characterised by constancies, but such that when these constancies are operative there are systematic and regular discrepancies between the sensory properties possessed by the experience and the properties of the normal object of awareness. Where this case is distinct from that Dicker has in mind is that these discrepancies arise precisely when the constancies for that creature are in normal operation for that creature. On A. D. Smith's constancy 521 If my argument is correct then it turns out that Smith has overlooked the importance for his defense of Direct Realism of the fact that the deliverances of the constancies are veridical, in the following sense: that in their normal operation there must be no divergence between the properties such as size, shape and location possessed by the immediate objects of awareness they characterise and those properties possessed by the external, 'normal' objects of awareness. That there must be no divergence in properties is simply a consequence of Leibniz's Law: Smith appears to have overlooked that Leibinz's Law applies also to those properties possessed by the immediate object of awareness, as characterised by the constancies. The following claim from Smith is an instance where this oversight has led him into error: 'The perceptual constancies are, indeed, far from perfect; but even if they were worse than they actually are, that would be of no significance.' (Smith 2002, p. 179) Smith is mistaken here: how good the constancies are is of paramount importance to his claim that we have no reason to suppose that the objects of awareness they characterise can be identified with the 'normal', external objects without violating Leibniz's Law.
My argument also highlights the following important feature of Smith's theory: whether or not a creature's perceptual experiences are classified as instances of direct awareness or not depends upon the veridicality (in the sense given above) of the constancy features of its phenomenology. It is clear that this is not what Smith intends: after all, as mentioned above, Smith thinks that Indirect Realism is not just contingently false, but incoherent. Nor should a philosophical theory of perception have this as a consequence, as I shall argue shortly. But, the possibility of the creature I have described shows that whether experiences possessing these constancy features are veridical or not is, according to Smith's theory, an entirely contingent matter. Consequently, according to Smith's theory, whether a creature's perceptual experience is an instance of direct awareness or not must be a contingent matter of fact.
This contingency would be avoided if we could argue that the constancy features of Smith's phenomenally perceptual conscious states somehow guarantee veridicality in the relevant respects (i.e., the location of the object, the character of its movement through space, or its felt resistance to the imposition of the subject's agency). As I have said, this is in fact what is required by Smith's defense of Direct Realism in the face of the explanatory challenge raised by the argument from illusion. One line of argument that could be pursued to provide support for this claim would be to argue that the kind of creature we have been considering does not and moreover cannot have experiences characterised by perceptual constancies, despite all my claims to the contrary. A reason for arguing this way may be found in the description I gave of there being 'variable deformations in size and shape' and 'shifts in position'. It may be argued that perceptual constancies do not and cannot result in such variability: constancies do just what they say on the tin and result in constant, unchanging sizes, shapes and locations. On this account, constancies will guarantee veridicality. The problem with this argument is that it simply begs the question. It is not obvious that it is a conceptual truth that perceptual constancies cannot result in the limited variability I have described: if it is not a conceptual truth then it looks like just a re-articulation of the claim that states exhibiting such constancy features (phenomenally perceptual conscious states) guarantee veridicality. However, this was the very claim that wanted motivating.
Given that we have no reason to suppose that the constancy features of Smith's phenomenally perceptual conscious states guarantee veridicality, we also have no reason to deny that such a creature as I have described is possible. It follows that whether a creature's perceptual experience is an instance of direct awareness or not is a contingent matter of fact. This result is critical, because the purpose of philosophical theorising about perception is not to give an account of what perception consists in for a particular organism, contingently arranged in a particular way, but to give an account in very general terms of what perception is for anything that may exhibit it. It is only by antecedently having such an account that we can legitimately class different organisms as perceiving. Direct Realism and Indirect Realism are accounts of perception at this level of generality, but if Smith's analysis is correct and my argument here is correct, then these alternatives should not be rival accounts of the nature of perception, but instead alternative ways in which perception can be manifested. Given that they are rival accounts, either my argument is incorrect or Smith's analysis is incorrect as a general account of the nature of perception.
There is a philosophical view of perception that does, in one special case, deny that there can be a unified account of perception that applies both to those cases that are caused by the external objects they represent and to those cases that are not so caused. This is disjunctivism: the view that two subjectively indiscriminable perceptual states, where one is a hallucination and the other is not, are not instances of a common kind. Smith might try to make a similar move here and say that the experiences of the creature I have described should be categorised as a case of hallucination, where they are aware of an unreal object. Such a move would be entirely ad hoc, though: in the case of visual discontinuities and abrupt jumps in position, the problem is not that there is no physical object seen, but that that physical object is seen to be where it is not. The perceptual experiences of the creature I have described are not hallucinations, they are illusions.

The 'not creatures like us' objection
The most likely observation in response to my counterexample to Smith's theory is that the creature I have been describing is not like us, so there is no reason to suppose that any conclusions drawn in its case can be extended to our own. This objection gets traction when we observe that we have no reason to believe that the specific features of the imagined creature that prevent the identification of immediate object with normal, or external, object of awareness obtain in our case. This leaves it open to Smith to accept, though I doubt he would wish to, that Direct Realism is true of the creatures I have been describing but that still there is no good reason to suppose that it is true of us. 9 I call this the 'not creatures like us' objection.
There is a response to this objection that is available through a modification of one of Smith's own arguments. This is an argument Smith offers to motivate acceptance of that part of the argument from illusion he calls the 'generalising step'. This step requires us to apply the account of perception we give of illusory cases to non-illusory cases. Smith argues that this step is mandatory in view of the unacceptability of holding a Direct Realist account of non-illusory cases and an Indirect Realist account of illusory cases. He argues that such a position radically misconceives perception insofar as it construes us as only rarely directly aware of the normal, external objects of awareness and that non-normal objects of awareness pop into and out of our consciousness all the time. Given how radically such a view departs from our understanding of perception, we should reject it and apply the same analysis to both illusory and non-illusory cases.
To deny the generalising step is to suppose that as you walk out of the shop while looking at your purchase, you only become directly aware of that physical item as you emerge into daylight (assuming this is when an object shows its true colours). Only then does that physical object suddenly leap into your perception in propria persona (Smith 2002, p. 27).
For my purposes, the crucial point that needs to be made concerning this argument is that it would retain all its force for a creature that contingently never actually suffered a case of perceptual illusion, but which could. So, this argument would apply with equal force in the case of a human being who was kept from birth in an environment with constant and uniform lighting conditions and so never enjoyed the relevant changes in experience that arise in changes of lighting conditions.
How then can we make use of this argument against Smith's theory? The first step is to observe that the introduction of a possible creature with non-veridical phenomenally perceptual conscious states was really a vivid way of describing a state that it is a metaphysical possibility for us to be in. Given suitable gradual modifications of a human's visual system, one could gradually come to enjoy the kinds of non-veridical experiences characteristic of the creatures described above. The non-veridical constancies that normally characterise that creature's perceptual experience would become normal for me. As someone gradually modifies my visual system, I will be in a situation similar to the person Smith describes above: i.e. enjoying a series of smoothly connected sensory experiences as he emerges into the daylight. If that is correct, as it seems to be, then Smith's argument for the generalising step applies here: the same analysis of my post-modification states should apply to my pre-modification states. As argued above, we cannot apply the Direct Realist account to the non-veridical cases; so we must apply the Indirect Realist account to the veridical cases. The same general account of perception must be given for both: the Indirect Realist account. We can conclude, then, that within the framework of Smith's analysis of perception and illusion Direct Realism should be rejected. Consequently, Smith's defence of Direct Realism on the basis of the phenomenological constancies, despite its importance as a discussion of relevant phenomenological features of perceptual states, is not successful.

Conclusion
The foregoing argument should not be confused with one that establishes the falsity of Direct Realism or the truth of Indirect Realism. It is less ambitious than that: if correct it establishes that even if we accept Smith's analysis of perception then we still have good reason to reject Direct Realism. Given the importance of Smith's theory, this point is of some significance by itself, but there is a further, wider ranging consequence of the failure of Smith's defence. The failure calls into question the emphasis that Smith places on providing an analysis of perception according to which sensations do not feature as objects of awareness in any way for responding adequately to the challenge posed by the possibility of illusion. Smith is quite right that it is incumbent upon a theory of perception to provide such an analysis, if it purports to be a Direct Realist theory and also acknowledges sensations as the bearers of sensory qualities. However, the more serious challenge to the Direct Realist, highlighted by the preceding argument, is to show that whatever is introduced as the object of direct awareness can have no metaphysical possibility of instantiating properties other than those instantiated by the external object of awareness. This is a much harder challenge to meet and, as I have shown above, Smith's analysis of perception cannot provide such a demonstration. Moreover, adverting to the difficulty of meeting this challenge is at least one more way in which the Indirect Realist can shore up support for his own position.