Direct Realism and Perceptual Consciousness

Susanna Siegel * May 2005

Forthcoming in Philosophy and Phenomenological Research


In The Problem of Perception, A.D. Smith’s central aim is to defend the view that we can directly perceive ordinary objects, such as cups, keys and the like.[1] The book is organized around the two arguments that Smith considers to be serious threats to the possibility of direct perception: the argument from illusion, and the argument from hallucination. The argument from illusion threatens this possibility because it concludes that indirect realism is true. Indirect realism is the view that we perceive mind-independent ordinary objects, but can only do so indirectly, by perceiving mind-dependent objects: objects whose existence depends on being perceived or thought about. The argument from hallucination draws a similar conclusion: if we perceive mind-independent ordinary objects at all, then our perception of them is indirect in the same way.

In responding to these arguments, Smith develops an account of perceptual consciousness. Perceptual consciousness is a kind of experience, distinct from what Smith calls ‘mere sensory experiences’, or equivalently, ‘mere sensation’. Perceptual consciousness is experience that is properly perceptual, in which one has the phenomenology of perceiving things in the external world (including one’s body) that exist independently of one’s mind. Perceptual consciousness on its own does not suffice for actually being in perceptual contact with mind-independent reality, although it suffices for it to seem as if one is.  It follows that perceptual consciousness does not suffice for direct perception of ordinary objects, or for direct realism.  Nevertheless, Smith holds that the correct account of perceptual consciousness is a crucial element in blocking the arguments from illusion and hallucination, and therefore in supporting the possibility of direct perception.

This is an extraordinarily engaging book. Within a single, unified narrative, one encounters the views of many philosophers - Husserl, Fine, Broad, Sextus Empiricus, Loar, Schopenhauer, Meinong, Burge, Dilthey, Russell, Dennett, Sartre, O’Shaughnessy, Evans, Berkeley, Craig, Brentano and many others. All five sense-modalities are discussed: smell and taste get as much attention as color experience and pain, and there is an extensive discussion of bodily awareness. Smith touches on virtually every philosophical question surrounding perception and many concerning intentionality. One can learn so much about the history of these topics from this book that readers are likely to feel exhilarated by the book’s breadth. They may also be left hankering for more precise formulations of theses and arguments in some parts of the book, and some of these parts will be the focus of this essay.  

The plan of action is as follows. Section 1 introduces the argument from illusion, and the following three sections discuss Smith’s responses to the argument. Section 2 assesses Smith’s first response, which rests on an appeal to perceptual constancies. Section 3 assesses Smith’s second response to the argument from illusion, which rests on his account of a special kind of tactile perceptual experience.  Section 4 extends these assessments by discussing Smith’s notion of a sensible quality, which plays a crucial role in these arguments. Section 5 argues that Smith takes his target to be a significantly restricted version of indirect realism, that he has no substantive argument against views that many would consider to be versions of indirect realism, and that his own view about the structure of perceptual experience is compatible with such views. Section 6 introduces and criticizes Smith account of perceptual consciousness and its role in his defense of direct realism, and argues that this account is, in some parts, plausible independently of its dialectical role in the rest of the book. Section 7 criticizes Smith’s attempt to refute the argument from hallucination and concludes the discussion.


1. The argument from illusion

Smith does not present the argument from illusion in a form that makes explicit exactly what its premises are. But it is clear that he thinks one of its premises is closely related to a core commitment of the sense-datum theory – Smith’s paradigm of indirect realism. Sense-data, for the purposes of Smith’s discussion, are by definition entities that depend for their existence on being perceived, and in that sense are mind-dependent.[2] A bit of extrapolation is required to work out exactly what Smith thinks the core commitments of the sense-datum theory are. His first response to the argument is targeted at these core commitments.

According to Smith, the argument from illusion has the structure of a generalization from the case of illusory perception to all perception. (The names of the premises are his, though the four-step structure is a reconstruction):


Sense-datum premise:

(1) ‘Whenever something perceptually appears to have a feature when it actually does not, we are aware of something that does actually possess that feature’ (p. 25).


(2) If (1) is true, then illusory perception is indirect.


Generalization premise:

      (3) If illusory perception is indirect, then all perception is indirect.


Conclusion: All perception is indirect.


            As Smith notes after formulating the sense-datum premise (p. 50), this premise is supposed to apply only to a limited class of features. It is only when public objects look to have a feature in this limited class that there will be, according to the sense-datum theory, a sense-datum that really has that feature (or one closely associated with it – there will be more on this modification shortly). These limited features are the only ones that a subject can perceive sense-data to have.

            Which features are these? That is a matter of controversy. But traditionally sense-datum theorists appealed to the idea that there is such a thing as perceptually basic properties. Some defenses of this theory – such as Jackson’s in his 1977 – begin with the notion of a special, limited ‘phenomenal’ sense of ‘looks’. If something can look F in the phenomenal sense, Jackson suggested, then F is a perceptually basic property. In addition, traditionally sense-datum theorists thought that the class of perceptually basic properties was limited to what psychologists sometimes call ‘low-level’ properties. These include color and shape, but exclude natural kind and emotional properties (such as being a tiger or being sad). As we will discuss in section 4, Smith himself accepts both ideas: that some properties are perceptually basic, and that the class of them is quite limited.

Although the notion of perceptual basicness is exploited by some sense-datum theorists, it is not specific to the sense-datum theory, and is independent of any commitment to indirect realism. But it will be useful to have a label that combines the idea that there are perceptually basic properties with the thesis that perceptual experiences involve the having of sense-data. Let’s say that a property F is presented in the sensory core of an experience, just in case F is a property had by a sense-datum that can be directly perceived and F is perceptually basic. The sense-datum theorists thought that the properties that could be presented in the sensory core of experience were all and only ones that sense-data can be directly perceived to have, and that these properties in turn were all and only the perceptually basic ones.

 Strictly speaking, the sense-datum premise (in the construal of it that Smith ultimately targets) needs to be restricted to the class of perceptually basic properties.  So it is compatible with a sense-datum theory that sometimes objects perceptually appear to have properties that are presented outside the sensory core of experience, and so are not properties of sense-data. This possibility will figure centrally in this essay.

In response to the argument from illusion, Smith convincingly rebuts proposals for how to re-describe putative cases of illusion as non-illusory, and points out that if the generalization premise didn’t hold, then any transition from a veridical to illusory perception would – implausibly – be a transition from full contact with reality to ‘radically losing touch with it’ (p. 27). This leaves the sense-datum premise as the only disputable part of the argument. Smith thinks this premise has a great deal of initial plausibility. Its central attraction is supposed to be an account of what it is for a thing to appear the way it does: namely, a thing appears to a subject to be (say) yellow only if there is a sense-datum, immediately perceived by the subject, that really is yellow.


When something [white] appears yellow to me, it is, or could be, with me visually just as it is when I veridically see something that really is yellow. For if it were not, why should I make reference to the same colour in the two situations? (p. 37). only information about the color of the object I am seeing comes from my current perception of it, I shall, other things being equal, take the object that is in fact white to be yellow. But why do I take it to yellow, rather than any other color, unless as C. D. Broad put it something yellow is ‘before my mind’? (p. 36).


 What the Argument from illusion attempts primarily to achieve is the recognition that sensible qualities can be present to consciousness in perception despite the fact that they do not characterize the normal physical objects we are said to be perceiving. It then challenges us to make sense of this fact (p. 37).


Smith divides responses to the sense-datum premise into the ‘swift’ and the ‘laborious’ (p. 34). The swift responses say that the premise is based on a simple confusion, such as thinking that ‘it appears as if there is a bent stick in the bucket’ either means the same as or entails what’s expressed by ‘there is something bent that appears to be a stick in the bucket’. This rebuttal of the premise, Smith rightly says, ignores the central attraction of the premise for philosophers such as Russell, Broad, O’Shaughnessy, Robinson, and others.  The ‘laborious’ responses are supposed to be attempts to meet the challenge to make sense of how sensible qualities can be ‘present to consciousness’, without there being, in the case of illusion, any mind-dependent entity that has those qualities. Smith considers and rejects several such attempts: ‘naïve realism’, where this is the view that completely successful perceptual experiences consist in being in perceptual contact with a scene in the external world that one veridically perceives (and cases of illusion and hallucination get treated differently); ‘reductive intentionalism’, attributed to Armstrong, Dennett, and Pitcher, where this is the view that experiences are dispositions to form beliefs about the external world; the ‘dual component theory’, attributed to Schopenhauer, Reid and Sellars, according to which experiences consist in a ‘sensory’ and a ‘conceptual’ component (Smith calls the components ‘psychological operations’ (p. 67), either of which could in principle occur without the other); and ‘non-reductive intentionalism’, attributed to Craig, according to which ‘certain sensory states are internally, intrinsically directed to normal objects’ (p. 123). The discussion of these views is detailed (in most cases several versions of each view are considered) and engaging. Smith’s ultimate verdict is that none of them are correct, hence none meet the challenge posed by the argument from illusion.

Smith does not give more than a sketch of an alternative account of how sensible qualities can be ‘present to consciousness’ without giving up on direct realism. So the two responses he gives to the sense-datum premise are neither swift nor laborious.

At this point the dialectic becomes more complicated. Smith thinks the sense-datum premise is the part of the argument to attack; but his objection is directed toward a more general thesis from which the sense-datum premise as stated above follows. Whereas the sense-datum premise links specifically illusory appearances to direct awareness, the more general thesis links all appearances to direct awareness. The following is a reconstruction of a claim Smith takes to be a core commitment of the sense-datum theory:


(SDT1)  When something appears to a subject S to have a property, S is directly          aware of something having that property.


Like Smith’s own formulation of the sense-datum premise, (SDT1) says nothing about whether we perceive external objects directly or indirectly. Since the sense-datum theory is meant to be the paradigm of a theory stating that perception is indirect, its core commitments must include more than just claim (SDT1).  The claim (SDT2) reflects this other core commitment of the sense-datum theory:


(SDT2) The only properties of which we are directly aware are properties of



Since sense-data are by definition (for Smith’s purposes) mind-dependent entities, (SDT2) is incompatible with direct realism. On my reconstruction, the target of Smith’s objection is the combination of (SDT1) and (SDT2).  Because of this, I will speak of Smith’s objections to the sense-datum theory, rather than to the sense-datum premise.


2. Smith’s first response to the argument from illusion: perceptual constancies

Smith’s first response to the sense-datum theory is that it can give no account of constancies in perception, such as size-constancy. Perceptual constancies pose a general challenge for theories of experience. The challenge is to describe what changes in experience, given what stays the same. An object moving toward you appears to stay the same size, yet it also changes the way it looks along the way, and the challenge is to describe this change. The straightforward proposal that it looks to change size is clearly wrong. A better proposal is that it looks to be coming closer.

However, Smith says that a defender of the sense-datum theory is committed to saying that in a case in which an object is seen to approach you, the thing of which you are immediately aware will get bigger: “[W]hat must a sense-datum theorist say of the typical situation in which an object is seen to approach me? He must say that the sense-datum, …that of which I am most fundamentally and immediately aware, gets bigger.” (p. 178).  Smith’s objection is then that it makes a false prediction about the phenomenology one typically has when one sees things that are coming closer, since given size constancy, things do not seem to get bigger as they approach:


The key to an answer to our Problem [i.e., the problem posed by the argument from illusion]…is the recognition that we are not…aware of perceptual sensations as objects because if we were, perceptual constancy would be wholly absent: the object of awareness would appear to change whenever there was a change of sensations, because such sensations would be our objects (p.178). 


Let’s say that something has a constant size (color, position, etc) property in a situation just in a case it maintains the same size (color, position, etc.) property in that situation. And say that a ‘constancy situation’ is one in which a public object maintains a constant size, but in which there are nonetheless changes in how it looks. Then Smith’s objection to the sense-datum theory can be reconstructed as the claim that the following four claims are inconsistent:


(i)                  In constancy situations, public objects appear to remain constant in size.


(ii)                In constancy situations, sense-data don’t remain constant in size.


(iii)               When something appears to a subject S to have a property, S is directly aware of something having that property. (SDT1)


(iv)              The only properties of which we are directly aware are properties of sense-data. (SDT2)


If (i) and (ii) are true, then (SDT1) and (SDT2) are not both correct.


            It seems to me that the sense-datum theorist has two responses to this argument.

            The first response is to deny (ii). Smith’s case for (ii) is that the sense-datum theory is committed to holding that as a thing in the external world looks to come closer, the sense-datum that one is directly aware of gets bigger.  Some versions of sense-datum theory hold that in the visual case, sense-data occupy a visual field that has only two dimensions, so that there is no such thing as a sense-datum moving closer to you (or moving farther away from you). This version of the theory is defended by Berkeley and by John Foster (2000). But there seems to be nothing about the status of sense-data as intermediaries that obligates them to exist in such a field. Perception of public, external things could be indirect, even if sense-data occupied a three-dimensional visual field and could move closer to the perceiver. In that case, the sense-data themselves could both remain constant in size and change the way they look as they get closer. This would account for constancy phenomenology while keeping perception indirect. And then premise (ii) would fail, at least for some versions of the sense-datum theory. 

Smith does not argue against the idea that there could be three-dimensional sense-data. But Foster has so argued in the course of defending a sense-datum theory. Foster argues that the status of sense-data as mind-dependent entities rules out the possibility of their having three dimensions. Foster presents this as the ‘fundamental reason’ to think that there could not be any three-dimensional sense-data.


The only way in which we think of a visual sense-datum as genuinely structured in depth is by supposing that it exists in an external space, and that its various parts stand, within that space, in certain distance relations to the subject; and this is precisely what is ruled out by the fact that the sense-datum is internal to the subject’s awareness (2000, p. 159).


Why is it ruled out? Since public objects look to exist in external, objective space (and, let’s assume, spatial properties are part of the ‘sensory core’ when they are presented in experience), three-dimensional sense-data would really have to exist in that same space.  Then the sense-datum theory would say that when someone views a public object and it looks to be a certain distance away, there really is a sense-datum located where the public object looks to be. But, Foster says, a sense-datum cannot really occupy such locations “if it is internal to [the subject’s] awareness” (ibid). External space, Foster is assuming, does not include any part of the ‘mental’ space that is supposed to be ‘internal to the subject’s awareness’.

            However, this problem is not unique to the experience of three-dimensionality.  The sense-datum theory already faces an analogous problem for direction – even if the sense-data are supposed to be two-dimensional. If there are two-dimensional sense-data, then they will have to have certain positions in whatever field they exist. More specifically, they will be in a certain direction from the subject (e.g., directly-in-front-of, to-the-left or -right-of), even if they are not at any distance from the subject. To handle this problem, it is open to a sense-datum theorist to hold that these relational spatial properties are not exactly the same as the analogous ones had by public objects in the external world. This would be to hold that if a public object and a sense-datum were both positioned directly in front of the subject, they would nonetheless not thereby be positioned in the same direction. To mark this difference, the sense-datum theorist might say that the sense-datum is related to the subject by being in-front-of* the subject, whereas the public object is simply in front of the subject. More generally, sense-datum theorists need not claim that the properties of sense-data of which one is (supposedly) aware are exactly the same as corresponding properties had by public objects. For the case of color, for example, some claim that the color-related properties had by sense-data are ‘redness as it appeared to be’ (cf. Peacocke 1983, chapter 1). There does not seem to be special reason to block this move for the spatial properties of distance while allowing them for color and direction.

            Now, some sense-datum theorists, notably Jackson 1977, denied that any special analogs for spatial properties of sense-data were needed, claiming that sense-data were indeed located in external, objective space (though they were still meant to be private and therefore unobservable by anyone other than the subject on whose mind their existence depended). Let us grant Foster’s assumption that this is not an option. Then one reply to Foster’s argument on behalf of the proponent of three-dimensional sense-data is that the relation that holds between the subject and the sense-datum is not distance in objective space, but an analog of this: distance in the visual field (where the visual field is a private mental space). This move prevents sense-data from ending up in objective space, which was supposed to be incompatible with their being mind-dependent entities.

            So proponents of two-dimensional sense-data – or those who hold that the only acceptable notion of sense-data is one on which they are two-dimensional - are already committed to positing analogs of spatial relations in the case of direction. The claim that there are three-dimensional sense-data thus seems to survive Foster’s challenge that there could be no such thing. And surviving this challenge counts against Smith’s premise (ii). One might try to pose a more general objection to the very idea that there is a mental visual field hosting sense-data (it is a fair question exactly what, and where, ‘private mental space’ is); but this objection would differ from the one Smith is making, and it would have no special force against three-dimensional sense-data.

A second, quite different, reply to Smith’s argument is also apparently available to the sense-datum theorist. This reply accepts claim (ii), that sense-data don’t remain constant in size in constancy situations, as well as accepting claim (i) and (SDT2). But it does not thereby give up indirect realism or the sense-datum theory. For this reply revises (SDT1) so that it applies only to perceptually basic properties, and it denies that the properties that appear to remain constant in constancy situations are perceptually basic. 


(SDT1*) When something appears to a subject S to have a perceptually basic property, S is directly aware of something having that property.


As I noted earlier, and as Smith himself notes (in defense of the sense-datum theory against an objection by Pitcher (pp. 47-50)), sense-datum theorists already have independent reasons to limit (SDT1) in this way. The independent reason is that there is (by their lights) a distinction between properties presented in the sensory core of experience, and other properties presented in experience but outside its sensory core.

With this distinction in hand, it is possible for the sense-datum theorist to hold that the relevant sense-data do indeed shrink as one sees a public object move away, but that the public object nonetheless appears to remain constant in size. If the defender of the sense-datum premise followed this strategy, then she would hold that although we are not directly aware of objects remaining constant in size in constancy situations, they are present in experience nonetheless, by being presented outside the experience’s sensory core. And that blocks the punch line of Smith’s objection, which is that if we were directly aware only of sense-data, then “perceptual constancy would be wholly absent” (p. 178).

One response on the part of Smith to this move would be to give some reason to think that appearing to remain constant in size in constancy situations has to be part of the sensory core of experience. I will discuss this response in section 4 (as we will see, this response bears on Smith’s second objection to the argument from illusion as well – the one concerning tactile perception).

For now, the important point is that sense-datum theorists seem to have two ways to account for constancy phenomenology, and correspondingly two ways to respond to Smith’s criticism concerning size constancy. Either they can hold that sense-data themselves have constant size features as well as variant ones, and thus reject (ii); or they can invoke aspects of experience that are not part of the sensory core to account for constancy phenomenology. Given the availability of these responses, Smith’s considerations about constancy seem not to show that there is anything about indirectness of perception per se that is incompatible with constancy phenomenology.


3. Smith’s second response to the argument from illusion: the Anstoss

Smith’s second response to the argument from illusion concerns a type of experience that he calls the Anstoss.[3] The Anstoss is a kind of tactile phenomenology in which you feel yourself to be actively resisting a force that you feel to be impinging on your body from outside of it.


[The Anstoss is the experience of] a check or impediment to our active movement: an experienced obstacle to our animal striving, as when we push or pull against things (p 154)… [in which] the subject must have a sense of pushing against the foreign body, however minimally, to sensibly encounter an alien force (p. 155).


Suppose, for example, that you are lifting up a shoe from the floor. In the places where your hand makes contact with the shoe, you will feel the shoe impinging against your hand, and you will feel yourself exert some amount of force to lift it. This much seems plain. Smith makes a further observation that it is part of the tactile phenomenology of lifting the shoe that something is impinging on you, and that you are resisting something other than you.

The Anstoss is supposed to be a type of experience in which the sense-datum theory is false. Smith thinks that cases of illusory Anstoss illustrate this. For example, suppose the shoe is resting on your hand at angle A, but you experience it as resting at angle A*.[4] Then, arguably, your experience misrepresents the angle at which the shoe impinges on you, and the angle at which you actively resist it.  Smith’s charge against the sense-datum theory is that given the phenomenology of the Anstoss, there is no reason to posit an intermediate entity to explain how an external force coming from angle A* can be present to us in experience, even though the only force there to be felt is coming from angle A. There is no motivation for positing an intermediate entity of awareness, Smith thinks, because one can account for the experience by appealing to a mistake you make about your own activity.


Suppose that you extend your arm straight out in front of you, but that, blindfolded, it feels as though you are extending it out to the side. You meet a resistance half-way. Because you seem to meet a resistance where there is in fact none, must we introduce an object at that point (or at a ‘corresponding’ point in some subjective space) to account for your experience? …no. For…the experience is adequately accounted for by your mistaking the true trajectory of your active movement… [d]eparture from the true appearance [i.e., from veridical experience of the direction that the force is coming from, if any] does not need to be accounted for by bringing in some non-normal object of awareness, since this is adequately done by reference to the character of one’s activity (p. 167).


How exactly is ‘reference to the character of one’s activity’ supposed to account for the feeling of something impinging on your hand at angle A* in the shoe example, or for the feeling of your arm resisting something to the side of your body, in Smith’s example?


What we have in the a mode of perception that emerges only at the point of an exercise of activity. Objects are revealed to and through our animal striving. Passivity is in some sense involved…, but it is a passivity that only makes sense in relation to activity, since it is but a shock or check to our activity – a check that reveals the independency of something from ourselves that is the hallmark of perceptual consciousness. Because any object is so perceived only as the reciprocal of our activity, there is no need to introduce a passively received sensation to account for illusions in this area, since how an alien force is perceived will be conditioned by the character of the striving that reveals it (p. 168).


Smith’s crucial claims seem to be these: The experience of activity is one and the same as the experience of feeling an alien body impinge on you. Since these are one and the same, if positing a sense-datum is not needed to account for your experience of active movement, then it is not needed to account for your experience of an alien body impinging on you either. And positing a sense-datum is not needed to account for your experience of active movement.

            Now, so far, we seem to have two candidate explanations for the phenomenology of something external impinging on you: the sense-datum theorist’s, which says that there is a sense-datum that you really do feel to be impinging on you from angle A*; and Smith’s, which refers to the ‘character of the subject’s activity’. The sense-datum theorist’s candidate explanation is none other than the sense-datum premise of the argument from illusion. The fact that an alternative explanation is available goes some way towards defusing this premise, since to establish that premise, more would be needed than just its proposed explanation for the phenomenology: namely, some argument that the alternative explanation is no good.

But Smith’s own conclusion aims to go farther than just defusing the premise in this way. It is supposed to be that his alternative explanation is better than the sense-datum theorist’s: indirect realism, he says, ‘fails to encompass the Anstoss’ (p. 169). What is supposed to make the alternative explanation better? Smith thinks that there are some cases of Anstoss in which the sense-datum theorist’s explanation could not apply at all:


When we press against a solid object, we…usually feel pressure sensations in our body at the point of contact; here sensation is playing its usual role in perception. Such sensations are not, however, necessary for the…Anstoss. We can feel such a check to our agency even if the relevant body-part is anaesthetized, or if we use some implement to feel the object’s renitent bulk. In both of these cases, certain sensations will be…present: in the first, there will, at least usually, be muscular sensations, and in the second, there will (normally) be pressure sensations where we are holding the implement. Such sensations, however, do not occur where we feel the obstacle to our action (159-60)…[these] sensations, though ‘in’ consciousness, are not objects for consciousness. They are not objects for consciousness either in relation to the Anstoss, or in relation to the constancies. (p. 187).


When you poke a sandbag with a stick, or when you press against it with an anesthetized hand, you can have an Anstoss experience without feeling any bodily sensations at the same location as the one where you feel the sandbag to be. How is this fact supposed to bear on the sense-datum theorist’s explanation of the Anstoss? Smith is making the following assumptions:


(A)   If tactile perception in cases of Anstoss involves direct perception of sense-

data, then these sense-data must be those associated with bodily sensations.           


(B)   If tactile perception in cases of Anstoss involves direct perception of sense-data, then these sense-data must be felt to be located where the subject feels an external object impinging on her.


The sandbag cases above are cases in which the felt location of the relevant bodily sensations differs from the place where the subject feels an external object impinging on her.  Given (A) and (B), it follows that tactile perception in these cases cannot involve direct perception of sense-data. If so, the sense-datum theory cannot explain Anstoss phenomenology in these cases. Smith’s alternative explanation in terms of ‘the character of one’s activity’ is then a better explanation. And if it is a better explanation in the sandbag cases, then it will be a better explanation in all the cases of Anstoss – assuming that these should be treated in the same way.

As Smith acknowledges (p. 169), this objection to the argument from illusion has its limits. Even if its conclusion is true, for all it says, all illusory perceptual experiences other than Anstoss experiences may still be indirect. It would be simple to reformulate the Generalization premise of the argument from illusion so that it generalized to all of these experiences, rather than to all perceptual experiences – and if that generalization were true, direct realism would still be threatened.

But does the limited objection succeed? Do Smith’s considerations about the role of active movement in cases of illusory Anstoss experiences weaken any reason to posit sense-data as objects of awareness in these cases?  His argument is only as strong as the assumptions (A) and (B).  But there are well-motivated versions of indirect realism that deny each of these assumptions.

The first version of indirect realism denies (A).  It posits a sense-datum that is felt to be located wherever a subject feels there to be an external object impinging on her body – whether or not one also feels a ‘bodily sensation’ there. This version is motivated by exactly the same sorts of consideration as those in the visual case. In the case of visual illusions, the sense-datum premise promises to explain how it could look as if something was (say) yellow even when the public object one sees isn’t yellow. Likewise, in the tactile case, it promises to explain how it could feel as if something is impinging on you (say) from angle A* even when the thing impinging on you isn’t coming from that angle. Whether one also feels pressure sensations at that very same location seems to be beside the point. If there were sense-data that felt to be located where one feels an external object to be impinging, then in cases of perception, one would perceive the impinging external object by directly perceiving the sense-datum. And this would make tactile perception indirect.[5]

The second version of indirect realism denies (B).  On this view, perception of the impinging object is not perceptually basic, so there are no sense-data that correspond directly to that object, and there are no sense-data felt to be located where the impinging object is felt.  In these cases the sensory core just involves sense-data corresponding to the bodily sensations one feels in tactile experiences (in sandbag cases, these will be either pressure sensations at the point of contact with the implement, or muscular sensations). It is by directly perceiving these bodily sense-data that one can perceive a public, impinging object, making the latter perception indirect.

            These two responses to Smith’s Anstoss argument correspond closely to the two responses to his perceptual constancy argument discussed in the previous section, suggesting that these arguments share a common structure. In both cases, there is some feature presented in experience that Smith claims cannot be presented indirectly: the feature of remaining constant in size in constancy situations, and the feature of being a force impinging on a body in the Anstoss. In both cases, the indirect realist has two options. First, she can posit sense-data that really have the properties (or analogs of the properties) that public objects are (wrongly) perceived to have in cases of illusion. Call this the complex sense-data option. Second, they can hold that the contentious feature is presented in experience but is presented outside its sensory core, and so not by being a property of any sense-data. Call this second option the two-factor option.[6]

To block the complex sense-data option, what’s needed is some reason to think that something about the status of sense-data precludes them from having the feature in question – call these AC-properties, for impinging on one’s body in the case of Anstoss, and remaining constant in size in constancy situations. Smith gives no such reason in either case.  (The discussions of Foster in section 2, and of Martin in footnote 5, consider and reject such reasons.)

To block the two-factor option, what’s needed is some reason to think that the AC-properties would have to be properties of sense-data, and so could not be presented outside the sensory core. Smith may hold that AC-properties would have to be what he calls sensible qualities, where these are properties that would be presented in the sensory core of experience, if any properties are. To assess this response, let us turn to Smith’s discussion of this notion.


4. Sensible qualities

            Like many sense-datum theorists, Smith thinks there are two kinds of properties that may be presented in perceptual experience: properties which are in some sense perceptually basic, and other properties that are not perceptually basic but are presented in experience nonetheless. Smith marks this distinction by talking of sensible qualities. This is Smith’s term for perceptual basicness; as we will see, he builds quite a lot into his version of this notion. According to Smith, a property F is a sensible quality iff it is not the case that one can see that an object has F only in virtue of seeing that it has another property G. For example, he holds that redness and roundness are sensible qualities, while sadness is not:


Feigned sadness is possible because even when we truly see that a person is sad, we do so only in virtue of perceiving some physical attributes. On the other hand, one does not typically see that an object is red, or spherical, in virtue of seeing anything other than its color or its shape. And if an object does look red or round in this straightforward way, and one is perceiving this object veridically, it does follow that the object really is red or round. Let us call any feature of a physical object of the latter sort a ‘sensible quality’ (p. 50).


Here Smith is tacitly invoking two notions. First, there is the notion of a directly perceivable property, where this is a property F that you perceive something to have, when there are no other properties in virtue of perceiving which you perceive F. (This would be a minimal notion of perceptual basicness). Second, there is the notion of a veridicality-relevant property, where this is a property F such that if an object looks F, and one is perceiving veridically, then the object really is F.  Smith appears to hold that the two notions are co-extensive.

Smith suggests, in effect, that our intuitions about when there is misperception can tell us which properties are veridicality-relevant.  If something looks red but isn’t, Smith claims, there will be misperception; whereas if someone looks sad but isn’t, there will be no misperception. So redness but not sadness is veridicality-relevant.  At various points, Smith suggests that veridicality-relevant properties include sweetness, being curved, shrillness (for noises), coldness, sphericality (p. 267) and being dagger-shaped (p. 264); but exclude wetness (p. 50) and being a dagger (p. 264). More radically, Smith thinks that no property can be veridicality-relevant if it is such that, to have it, an object’s surfaces must continue out of view. This will rule out properties such as being a bird, being a hand, being a whole egg, and being any other sort of opaque three-dimensional volume (p. 265). (So ‘spherical’ and ‘dagger-shaped’ have to be taken as picking out something other than three-dimensional volumes).

We can now see how this view bears on his objections to the sense-datum theory. The sense-datum theory holds that the directly perceivable properties are precisely those perceivable properties had by sense-data.  Now, if Smith is right that directly perceivable properties are just the ones to which veridicality is sensitive, then he has a potential argument for the view that the sense-datum theory has no option but to treat the AC-properties as properties of sense-data. The argument goes like this:


AC-properties are veridicality-relevant properties.

All veridicality-relevant properties are directly perceivable.

If there are sense-data, then all directly perceivable properties are properties of sense-data.

So if there are sense-data, then AC-properties are properties of sense-data.


If such an argument succeeded, the two-factor option would not be open to a sense-datum theorist: it would not be open to sense-datum theorists to hold that these properties are presented in experience outside of its sensory core. They would then be forced into trying to show that there are complex sense-data – namely, sense-data with AC-properties.

The first premise seems reasonable: there is a strong intuition that AC-properties are relevant to veridicality in cases of constancy and Anstoss.  However, the second premise is more disputable.  For example, it is open to a sense-datum theorist to hold that sadness is veridicality-relevant even though it is not directly perceivable.  This is an especially reasonable line for a two-factor theorist who holds that sadness is perceived even though it is not in the sensory core of experience.  Such a theorist may reasonably hold that all perceived properties are veridicality-relevant, whether or not they are in the sensory core.  If so, then the same applies to AC-properties, on the two-factor view of these properties: they are outside the sensory core and so are not directly perceivable, but they are veridicality-relevant all the same.

In effect, Smith’s notion of a sensible quality runs together the distinct notions of direct perceivability and veridicality-relevance, without any argument that the two notions are co-extensive.  In the absence of such an argument, the two-factor sense-data theorist can reasonably hold that the two notions come apart.  So it appears that Smith has no well-grounded objection to the two-factor treatment of perceptual constancy and of tactile perception in cases of Anstoss.

There is another line of thought in Smith’s book that might at first seem to rule out the two-factor option. Historically, the 20th century sense-datum theorists who appealed to just such a two-part structure of perceptual experience thought that the second part involved concepts. H. H. Price, for instance, distinguished between the Given on the one hand, and ‘perceptual acceptance’, on the other, where the latter resulted from applying concepts to the former.[7] Smith devotes his chapter 3 to arguing against the general view that concepts must be possessed and ‘applied’ in perceptual experience (whether what they are applied to are sense-data or not). But a two-factor view of sense-data need not rely on this view about concepts. For instance, it might instead try to draw the distinction on phenomenological grounds. Or it might (also) appeal to the idea that there is some sort of inference internal to the experience from the sensory core to the second factor, where the inference does not consist in ‘applying’ concepts to the sensory core: both factors could be ‘conceptual’, or both could be ‘non-conceptual’. These forms of the two-factor option, and all others that do not rely on the idea that concepts are always employed in perceptual experience, are left untouched by Smith’s argument.

 So in the end, for all Smith says about sensible qualities, both options in response to Smith’s objections –the complex sense-data option, and the two-factor option - seem to be available to the sense-datum theory after all. 


5. Smith’s restrictive notion of indirect realism

            There is one more move Smith might try to make to oppose the two-factor option. This is to claim that despite appearances, the two-factor option is not really a version of indirect realism after all.

At the start of the book, Smith characterizes indirect realism as follows:


When one thing is perceived in virtue of some distinct item being perceived, we can say that perception of the latter mediates perception of the former, and that this former object is not the immediate, but only the indirect, object of perception…Assuming a Realist perspective on physical objects, our question is whether such objects are only ever perceived, and could conceivably only ever be perceived, in virtue of our perceiving something distinct from them. Answering this question affirmatively is definitive of Indirect Realism (p. 6).


Call ‘broad indirect realism’ the view that one perceives public objects in virtue of perceiving mind-dependent entities. The two-factor option, on which one perceives AC-properties by perceiving other perceptually basic properties, clearly is a version of indirect realism, given this gloss.

In subsequent discussion, however, Smith takes as his target a much more restrictive notion of indirect realism: one according to which the mediating entity in virtue of which public objects are indirectly perceived is the sort that can reasonably be mistaken for those very public objects. Consider Smith’s assumption (B) from the discussion of the Anstoss:


(B)    If tactile perception in cases of Anstoss involves direct perception of sense-data, then these sense-data must be felt to be located where the subject feels an external object impinging on her.


One reason to assume (B) is the idea that the only truly indirect realist sense-datum theories of touch would be theories according to which the sense-data were felt to be located where one also feels an external object to be impinging on one’s body. That Smith has such a restrictive notion of indirect realism in mind as his real target would explain why he makes assumption (B). It would also explain why Smith never considers the two-factor option, according to which one has Anstoss phenomenology partly in virtue of having bodily sense-data – even though these bodily sense-data are not felt at the same place as the place where one seems to feel a public object impinging on one’s body.[8] 

            That Smith is implicitly assuming the restrictive notion of indirect realism as his target also makes sense of why he does not consider the two-factor option in the case of constancy. That theory granted that sense-data are two-dimensional. Assuming that two-dimensional sense-data appear to be two-dimensional, and that nothing which appears to be two-dimensional also appears to be a public object, this two-factor option would not count as a version of indirect realism in Smith’s sense.

            Given Smith’s construal of indirect realism, he presumably will construe these two-factor options either as versions of direct realism (e.g. if he holds that perception of an object is direct iff it does not involve perception of an intermediary that is mistaken for the original object), or else as involving some third alternative that is neither direct nor indirect realism.  But these construals appear to be somewhat idiosyncratic.  The two-factor options are clearly versions of broad indirect realism, and most would consider them to be versions of indirect realism simpliciter. And in any case, these views accord with Smith’s initial gloss on what indirect realism is. Moreover, Smith has no substantive argument against them.

            In fact, Smith’s own view about the nature of perceptual experience seems to have the same structure as the two-factor option, and seems compatible with a version of broad indirect realism. Consider Smith’s positive response to the challenge posed by the sense-datum premise. The challenge, he says, is to account for how “sensible qualities can be present to consciousness in perception despite the fact that they do not characterize the normal physical objects we are said to be perceiving” (p. 37). (Here Smith is using ‘do not characterize…’ to mean roughly the same as ‘are not instantiated by…’).


[T]he only way to defend direct realism requires that we conceive of sensory qualities not as characterizing sense-data – entities to which by definition we stand in the relation of being aware of – but as intrinsically characterizing psychological states, or the flow of experience as such (p. 65).


Smith answers this challenge as follows:


[A]ll we have to show, in order to block the Argument [from illusion], is that we are not directly aware of whatever it is that possesses…[sensory]  qualities, so that awareness of the latter does not cognitively mediate our awareness of normal physical objects. ...I have shown the way in which sensations, though ‘in’ consciousness, are not objects for consciousness. They are not objects for consciousness either in relation to the Anstoss, nor in relation to the perceptual constancies…[w]e can interpret the constant appearance of a single unchanging object amid varying sensations….as…the varying appearance of the normal object of perception (p. 186-7).


What exactly is the perceiver’s relation to the sensations that are ‘in’ consciousness but are not ‘objects for consciousness’? There seem to be two options. One is that the subject perceives the sensations, but does not perceive them as (or thereby mistake them for) external objects. The resulting view would be a version of broad indirect realism, akin to the two-factor option. Another option is that the subject does not perceive the sensations, but stands in some other relation to them. This option would not be a version of broad indirect realism.  Nothing Smith says about his own view seems to rule out the first option.  So his view appears to be entirely compatible with broad indirect realism.

            There is another feature of Smith’s discussion that is explained by his assuming a restrictive version of indirect realism. This is the discussion of perceptual consciousness in chapters 4 and 5, in which he distinguishes between three conditions that he argues suffice for making an experience perceptual, and which he thinks actually obtain in our perceptual experiences, so that each of our perceptual experiences meet at least one of the conditions. It would make sense to include such a discussion in such a defense, if one took as a starting point that any case of direct perception  - whether of sense-data or of public objects – would have to be a case of perceptual consciousness. An account of perceptual consciousness would at the very least show, if it were correct, what conditions one would have to say obtain in the direct perception of sense-data.  One might think, as Smith seems to, that once this account is on the table, it strengthens the case that it can’t be sense-data that we directly perceive. This case is so strengthened, only given the assumption (internal to restrictive indirect realism) that whatever we directly perceive, when we directly perceive it, we have a properly perceptual experience.

            In the next section I will argue that all three conditions are in fact compatible with broad indirect realism – and so they can’t really play any role in defending direct realism against the argument from illusion after all. 


6. Perceptual consciousness 

Smith has much of interest to say about what makes experiences perceptual when they are, and all of it seems to be independent of whether direct realism is true, and independent of the role of discussions of perceptual phenomenology in Smith’s responses to the argument from illusion. So whether or not his responses to that argument succeed in defusing its force, his account of perceptual consciousness may be plausible. I will argue that some of it is.

In perceptual experiences, one seems to be perceiving things in the external world (where this includes one’s own body). Smith distinguishes perceptual experiences from ‘mere sensations’. This is a distinction between two kinds of experiences, each of which has an accompanying phenomenology.

Some readers might doubt that experiences can meaningfully be divided into perceptual and merely sensational varieties. But as Smith construes this distinction, it clearly marks out a natural division, at least if one keeps to the level of paradigm examples. The paradigm ‘mere sensory experiences’ are certain ‘fringe’ experiences, such as having phosphenes, the experience of reddish glow when you close your eyes in the sun, or the ‘brain gray’ darkness when one closes one’s eyes in less intense light. (Smith’s apt label for these is ‘the inner light show’). For some experiences, such as pains and very simple visual experiences, it will be more delicate on which side of the distinction they belong, because they are not paradigms. To categorize these, further theorizing will be needed – both about the distinction, and about the experiences.

In any case, it is indisputable that there are some experiences of the kind Smith calls the ‘inner light show’. It is also indisputable that ‘inner light-shows’ have a phenomenal character. It is somewhat more controversial whether they ever purport to represent how things are in the world external to the subject’s mind. Smith thinks they do not, and though no argument is offered for this claim, it seems plausible. An intuition in its favor is that there is no way things could be in the external world, it seems, that would make us consider any of the ‘inner light shows’ to be correct.

In contrast to mere sensations, perceptual experiences are experiences as of perceiving events in the external world, including events in one’s body. So seeing a canary fly by, or hearing it chirp, or feeling it peck at your shoulder are examples of perceptual experiences. ‘Perceptual consciousness’ is just another label for the kind of phenomenology these experiences have and visual sensations lack. Unlike the paradigmatic mere sensations, perceptual experiences clearly purport to represent the world external to the mind of the subject. Consider the visual experience you have now as you read these words, or as you look around the area that you are in. These experiences purport to tell you what words and items are in your immediate surroundings. In Smith’s terminology, the experiences are ‘intentional’ (p. 139). As he notes, perceptual experiences can be hallucinations, and so fail to be experiences in which one succeeds in perceiving the environment. What’s distinctive of perceptual experiences is that they seem to be perceptions of the environment, even if they are not. Hallucinations can have the phenomenology of perceiving; and for all Smith’s phenomenological distinction says, there could be perceptions that nonetheless lack perceptual phenomenology. 

According to Smith, there are three conditions that can make the difference between mere sensation and perceptual consciousness: ‘being presented…with objects that are ostensibly located in three-dimensional space’, or phenomenal three-dimensionality (p. 133); a condition having to do with the possibility of changing your experience by moving your sensory apparatus (pp. 140-150 - I’ll call this a  perspectival condition); and the Anstoss, or its seeming to you in tactile experience that you are actively resisting an ‘alien force’ – that is, the force of a body that isn’t yours (p. 153). Smith also thinks that all five modalities admit of the distinction between mere sensation and perceptual consciousness (p. 150).

Curiously, the Anstoss is defined in such a way that makes it trivial that it suffices for perceptual consciousness:


What is required for the Anstoss is that the subject push against a foreign body, however minimally. Or, rather, bearing in mind that we are at present engaged in a phenomenological enquiry, that the subject have a sense of so doing (p.154).


That Smith intends to define it this way seems clear from a contrast he draws. In contrast to the Anstoss, which is a felt external chanck on one’s active movement, one can experience a check on one’s active movement that comes from an inner limit – such as separating one’s index and middle fingers of one hand as far as they will go (p. 154). Another example is lethargy – an internal check on one’s active movement. The Anstoss also contrasts with awareness of some actions. Christopher Peacocke describes a situation in which someone with a jaw made numb by a dentist is told to open his mouth, and though he lacks any feeling in his face, he can nonetheless feel when he is opening his mouth, and when he is done opening it.[9] This seems to be some sort of awareness of action that does not depend even on kinesthetic phenomenology (or, for that matter, on phenomenology associated with any of the five sense modalities). But this too is not a case of Anstoss as Smith defines it, even though it involves active movement, since there is no feeling either time of resisting something external with one’s jaw.[10]

            In contrast to the claim that the Anstoss suffices for perceptual consciousness, Smith’s claim that phenomenal three-dimensionality suffices for perceptual consciousness seems doubtful. Smith makes clear that the relevant kind of phenomenal spatiality is apparent distance from a sensory organ, such as the eyes and ears:


The more precise notion of spatiality that we require…is one that essentially involves not just the spatial relationships between the objects of awareness, but the spatiality of the relationship between any such object and ourselves, more specifically a part of our body. In vision, for example, objects are characteristically seen, when genuine perceptual consciousness is involved, as more or less distant from us – specifically, from our eyes (or eye). (p. 143).


As Smith understands the relevant kind of spatiality, it seems to differ from the putative perceived object itself having three dimensions: something could look to be located somewhere in front of the eyes and a certain distance away from them, without looking to have be a three-dimensional volume. And this seems to introduce another kind of phenomenal three-dimensionality – what we might call ‘volumetric three-dimensionality’ - besides the one Smith includes in his gloss on phenomenal three-dimensional spatiality.

  Going by Smith’s gloss, the first condition seems not to suffice for perceptual consciousness. Consider phosphenes – one of Smith’s paradigms of an inner light-show (p. 129). Normally phosphenes look to be located at some direction relative to the eyes – one’s visual experience of regular phosphenes is not neutral on whether the phosphene appears to be in front of you or behind you, for instance. Once phenomenal directionality is in place, it seems a short step to phenomenal location relative to the eyes. So phosphenes (or experiences thereof) seem to have phenomenal three-dimensionality in Smith’s sense.

Even if Smith means to include volumetric three-dimensionality in the first condition, it seems possible to imagine a phosphene that retained the non-perceptual phenomenology of our familiar phosphenes, yet did look to have depth properties. Such a phosphene might still move with the eyes, and be as hard to focus on as regular phosphenes, as well as being impossible to view from multiple angles. But for all that, it seems, it could look as if it continued out of view in the way that three-dimensional volumes do. Maybe the lighting in the inner light show suggests this. It seems possible to distinguish distance from the sense-organ, from independence from the sense-organ. The imagined phosphenes would look have a certain distance from the eye, without looking to exist independently of the eye. 

Smith introduces the last proposed condition (which I’ve called the ‘perspectival’ condition, though the term may be more apt for the visual case than it is for the other sense modalities) in a very engaging discussion of olfactory experiences. Olfactory experiences, Smith says, in contrast to visual and auditory ones, have no intrinsic spatiality: there could be no difference in spatial content, he claims, between an olfactory experience of smelling a rose positioned directly beneath the nose, and an experience of standing in a room full of roses – even if neither experience involved any distortion (p. 138). This example is supposed to show that olfactory experiences do not themselves represent any location for smells, or sources of smells. So if any olfactory experiences are perceptual, this will have to be thanks to something other than phenomenal spatiality.

Smith makes a further claim that may seem to be at odds with his view that there are perceptual olfactory experiences. Nothing in olfactory phenomenology, he says, presents things as external to the nose: “you typically experience smells in (or just behind) your nose” (p. 139). If this is a feature of olfactory phenomenology, then how could it ever present anything as having an odor that is located elsewhere than in the space in or behind the nose? If no olfactory experience can do that, then none will be perceptual. Smith tries to reconcile the (typical) nose-internal phenomenology with the perceptual nature of some olfactory experiences by appealing to the apparent movement of sense-organs.


the kind of movement that is of perceptual significance is the movement of sense-organs in relation to perceived objects…More specifically, the kind of bodily movements with which we should be concerned are those by which we come to enjoy different perspectives on perceptible objects…. Whether an autonomous, identifiable organ of sense can independently be moved is not the issue. The issue is whether a part of the body serving as an organ of perception can be moved in such a way as to give rise to changing sensation, whether or not this involves movement of all, or large parts of, the body. (pp. 140-42).


Smith does not say exactly what it is to have ‘different perspectives’ on the odor, except that it is something more than (or perhaps other than) olfactory experience becoming more or less intense with changes in movement. More theorizing needs to be done, then, to specify the perspectival condition.

Even if apparent movement of sense-organs in relation to perceived objects of the sort described above suffices for perceptual consciousness, this is not the most general informative sufficient condition available, because there are static cases of perceptual consciousness – cases in which neither sense-organ nor perceived objects appear to move. If any olfactory experiences are perceptual, surely stopping to smell the roses is. If one denied this, then just by stopping to smell the roses one could interrupt an otherwise perceptual olfactory experience. So the sufficient condition Smith is after, if it is going to be both general and informative, cannot be simply that an experience is perceptual if it seems to one as if one’s nose is changing location relative to an odor in such a way that results in different perspectives on the odor. Smith acknowledges that the role of constancies in the perspectival condition will have to be something other than their actual operation: 


Consider…a cataract sufferer staring fixedly at the light streaming in from a brilliantly lit window in the dark. Although this subject is aware of but a two-dimensional array of light and shade, this ought… to be recognized as a case where the subject perceives, albeit poorly, the light of the window. The reason for this is that, if the subject were to turn his head, the light would not, or might not, appear to him to move. The sheer possibility of the operation of such a perceptual constancy serves to confer perceptual status on the original static experience; for in the realm of mere sensation, such constancy is impossible. (p. 179)


In this passage, especially the last two sentences, Smith seems committed to the view that the mere fact that the disposition holds confers perceptual consciousness, so that the proposed sufficient condition would be this:


Disposition condition: An experience is perceptual if: if one substantially moves one’s sense-organ, one’s perspective on the [odor/thing seen] will thereby change.


It seems implausible, however, that the mere obtaining of a disposition could give an experience perceptual phenomenology as opposed to the phenomenology of mere sensation. Suppose a bright light is shining directly into one of your eyes, producing an intolerable white glare. If you move out of the way, let us suppose, you will be able to see the light shining in the direction where your eye previously was, and so your perspective on this light will change: it will look like a ray extending in that direction. But when the light is shining into your eyes, there is no perceptual phenomenology, just an overwhelming white glare – in particular, no experience of something located in the space in front of you shining in. This seems to be a case in which the dispositions hold, but there is no perceptual consciousness.

             What seems to suffice for perceptual consciousness is not that the dispositions hold, but rather that they seem to the subject to hold. Suppose you unwittingly stumble into a hologram chamber where the dispositions did not hold – say it is a museum display of holographic butterflies perched on a ledge. Until the lack of constancies is made manifest, you will simply seem to be seeing unmoving butterflies. What gives your experience perceptual phenomenology, arguably, is that you in some way represent the relevant dispositions to hold. This suggests an alternative formulation of the perspectival condition:


Representation-of-disposition condition: An experience is perceptual if it seems to one that if one moves one’s sense-organ, one’s perspective on the [odor/thing seen] can thereby change.


It is not clear whether Smith accepts the representation-of-disposition condition. On the one hand, it is at odds with the passage quoted above. On the other hand, in some places he seems to endorse the points that there is such a thing as appearing as if perceptual constancies hold, and that that this is what matters for perceptual consciousness:


To perceive an object at a distance is to perceive it as being in a space through which it could move without undergoing any apparent change…The phenomenal three-dimensionality of perceptual objects that gives them ‘external reality’ is a phenomenal manifestation that such spatial perceptual constancies are potentially in play (p. 180).


Even when we perceive an object moving while at the same time we ourselves move, the former movement is perceived as being such that it would still be perceived if we were to become stationary (p. 145).


Unlike the purely dispositional version of the perspectival condition, the representation-of-disposition version is quite plausible. Consider a pair of very similar experiences that nonetheless differ in whether they are properly perceptual: for instance, ‘seeing stars’ from being hit hard on the head, and seeing points of light in the ceiling of an auditorium that are made to look like stars in the sky. The phenomenal contrast between these experiences would be the very one Smith brings into focus by discussing the distinction between mere sensation and perceptual consciousness (for the visual case). Given that so many of the properties attributed to the points of light in each will be the same, a plausible proposal about how things appear to be different in the two cases is that the lights in auditorium look to be such that you could adjust your perspective on them, whereas the ‘stars’ you see after getting hit on the head do not look that way.[11] So Smith’s claim that there is a perspectival condition on perceptual consciousness is very plausible, so long as one considers the representation-of-disposition version of it, rather than the plain dispositional version of it.

All three of Smith’s proposed conditions on perceptual consciousness seem to be compatible with broad indirect realism. We’ve discussed already how this goes for the case of Anstoss. What about the phenomenal three-dimensionality? That public, external objects should sometimes appear to be spatially separate from our sense organs (eyes, ears, nose, etc.) is compatible with the claim that we are only indirectly aware of these objects, while being immediately aware of sense-data. The  (relational) property of being at some distance from a sense-organ at the very least is a property that public objects could perceptually appear to have when we indirectly perceive them – on par, by Smith’s lights, with sadness, wetness, and slimyness. 

Likewise, the dispositional version of the perspectival condition is also compatible with indirect realism: it would simply be the indirectly perceived objects that we would be disposed to perceive from different perspectives upon moving our sense-organs relative to them. Matters are more complicated when it comes to the representation-of-disposition version. In principle, the same two options for the sense-datum theorist that were discussed in connection with constancy and the Anstoss seem available here too. In principle, it seems that sense-data could have the relational property of being such that if the perceiver moves her eyes (say), the sense-data will not thereby move, and other parts of them will come into view.[12] Alternatively, these dispositions could be properties that public objects perceptually appear to have but are not had by sense-data  - on a par, by Smith’s lights, with sadness.

To sum up, there is much that is interesting and plausible in Smith’s account of perceptual consciousness, and its interest and plausibility is independent of whether direct realism is true.


7. Smith’s response to the argument from hallucination

Smith says that the argument from hallucination poses a greater challenge to direct realism than the argument from illusion (p.187). His response to it, however, is not nearly as original nor as well-developed as his response to the argument from illusion, so I will spend less time on it here.

Smith thinks that the argument from hallucination, like the argument from illusion, is best formulated as one that generalizes from a perceptual experience that is not completely successful (in this case, hallucination rather than illusion), to all experiences with perceptual phenomenology. It has three steps:


(H1) Hallucination premise:

It is possible to have a completely hallucinatory experience, subjectively indiscernible from a case of perception, in which there is no normal, public object of which the hallucinator is aware. (p. 269, p. 266).




(H2) Hallucination conclusion:

In cases of hallucination, we are directly aware of a ‘non-normal, non-public’ object (p. 230).


(H3) Generalization premise:

If we are directly aware of a non-normal, non-public object in the case of hallucination, then we are directly aware of such objects in any cases of perception as well (p.230).


Conclusion: Any perception of public objects is indirect.


If the Generalization premise holds, then whenever we perceive things that don’t depend for their existence on being perceived or thought about, we also, and directly, perceive a non-normal, non-public object. So if the Generalization premise is true, then direct realism is false. 

Smith’s official response to the argument from hallucination is to deny (H2). “What now remains to us as the only way of blocking the argument and of defending Direct Realism”, he writes, “is to challenge [the] assumption [that]…for a hallucinating subject to be aware of something is for him to be aware of a non-normal object” (p. 230).

What is involved in denying (H2)? (H2) makes a claim about the metaphysical structure of hallucinations. It says that they have a structure in which one is aware of a ‘non-normal’ object. To deny (H2), then, is to deny that hallucinations have this structure. One way to do this is to provide an alternative account of the structure of hallucinations.

Smith’s strategy is to argue that there is such an alternative. Smith thinks (H2) is false because perceptual experiences have intentional objects. In the case of veridical experience, the intentional object is none other than the public object that is perceived. In the case of hallucination and illusion, in contrast, the intentional objects do not exist (p. 234ff). Smith’s account of intentional objects is thus supposed to explain what it is for certain hallucinations (namely, those that are phenomenally indistinguishable from perceptions) to have the phenomenology they do, without appealing to any awareness of ‘non-normal’ objects.

According to Smith, intentional objects are things that exist in the veridical case but do not exist in the non-veridical case. Smith says he uses ‘object’ in ‘intentional object’ in a special way (here he says he follows ‘Germanic and Scholastic traditions’):


[T]he term ‘object’ specifically connotes being an object for a subject…any entity that is not in any way cognized is not to be termed an ‘object’ (p. 236).


So an intentional object, for Smith, is something that is represented by the subject. That is definitive of intentional objects, and it is what the existent and non-existent ones are supposed to have in common. The only way to deny that there are intentional objects in Smith’s sense is to deny that hallucinating subjects are aware of anything at all. Smith finds this line extreme and implausible, attributes it to Evans and McDowell, and devotes an entire chapter to its rebuttal.       

Intentional objects, for Smith, are thus defined in terms of their psychological role: to say that a mental state has an intentional object is just to say that the subject represents something. Going with this, intentional objects are supposed to be ontologically innocuous - contrary to accusations famously leveled against Meinong: “an appeal to intentional objects,” Smith writes, “should not be seen as offending anyone’s ontological scruples” (p. 243). The supposed status of intentional objects as ontologically innocuous is the core of Smith’s case for them.  (There is also a detailed discussion of some standard challenges to the coherence of the claim that there are non-existent objects, including a long footnote - note 52 -helpfully summarizing the main defenses of Meinongian views against one of Russell’s objections. I’ll focus on the core of the defense here.)

The claim that intentional objects is ontologically innocuous is straightforwardly correct for existent intentional objects, since these are none other than the public objects one sees in veridical perception.

What about non-existent intentional objects – what is their ontological status? A non-existent intentional object is supposed to be an apparatus for describing perceptual phenomenology. “To speak of an experience’s intentional object is simply to talk about the ‘descriptive nature’ of that experience, to advert to its specific intentional character” (p. 243). “[W]hen we say of a subject that he is…hallucinating a dagger, we are simply characterizing the subject psychologically. No other entity is involved” (ibid). “the intentional directedness of even hallucinatory states of perceptual consciousness is simply a phenomenological fact—one that needs to be recognized in any phenomenally adequate account of the mind. This fact alone sustains the intentionalist position. The theory is not an ontological, but a phenomenological one” (p. 245). Another way Smith has of describing his account of the intentionality of perceptual experience is that it is supposed to be ‘ontologically reductive’ without being ‘psychologically reductive’ (p. 244).  The case for non-existent intentional objects is that they are ‘the only phenomenally adequate way of talking about perceptual experience’ (p. 244).

In these passages Smith proposes a deflationary treatment of the claim that in hallucinations, one is aware of an intentional object. Such treatment might help avoid commitment to ‘non-normal’ entities. But to the extent that it avoids that, it cannot provide a positive alternative account of the metaphysical structure of hallucinations. This can be seen by considering two ways to precisify the idea that non-existent intentional objects are ontologically innocuous. The passages quoted above suggest two different views about exactly what it is to be aware of a non-existent intentional object. Each of them has a different upshot with respect to the argument from hallucination.

This first view, suggested by Smith’s remark that “To speak of an experience’s intentional object is simply to talk about the ‘descriptive nature’ of that experience, to advert to its specific intentional character” from p. 243, is that the claim that in hallucinations one is aware of a non-existent intentional object is simply neutral on the metaphysical structure of the hallucination itself.[13] That is, the claim entails nothing at all about the structure of that mental state of hallucination has, or what sort of state it is. Call this the ‘ontologically neutral’ version of the view that in hallucinations, one is aware of a non-existent intentional object.

According to Smith’s official statement of his strategy for responding to the argument from hallucination, the response is supposed to provide a view on which (H1) is true but (H2) is false.  However, the ontologically neutral version of Smith’s view about hallucinations does not entail that (H2) is false. (H2) is a claim about the metaphysical structure of hallucinations: it says that in having a hallucination, one is aware of a ‘non-normal’ object. If the claim that hallucinators are aware of non-existent intentional objects does not entail anything about the metaphysical structure of the hallucination, then a fortiori it does not entail that the experience lacks the metaphysical structure that (H2) says it has. The ontologically neutral version of the intentional-object view is thus compatible with (H2). Given this view, it will be an open question what hallucinations are like metaphysically.  The sense-datum theorist holds that they have the structure suggested by (H2), and the ontologically neutral version of the intentional-object view does not provide any alternative. So given the ontologically neutral version of Smith’s intentional object view (suggested by the quotes from p. 243 and p. 245 above), Smith’s official response to the argument from hallucination does not succeed in refuting it.

 Other things Smith says, however, suggest that the claim that in hallucinations, one is aware of a non-existent intentional object is not completely neutral on the metaphysical structure of hallucination after all. On this version of the claim, it entails a negative ontological claim about the structure of hallucinations, to the effect that they do not involve any ‘non-normal’ object. Call this version of the claim the ‘negative ontological’ version of Smith’s claim about hallucinations. It is suggested by the remark from p. 243 that  “[W]hen we say of a subject that he is…hallucinating a dagger, we are simply characterizing the subject psychologically. No other entity is involved”. If no other entity is involved, then a fortiori no ‘non-normal’ entity is involved, and so (H2) is false. Unlike the ontologically neutral version of Smith’s claim about hallucinations, then, this version really is incompatible with (H2), and so fits better with Smith’s official response to the argument from hallucination.

Now, the indirect realist will deny the negative ontological claim, holding that hallucination involves awareness of a non-normal object after all. So what reason, if any, has Smith given for accepting the negative ontological claim? The reasons are hard to find. At one point Smith says that the claim that in hallucinations one is aware of a non-existent intentional object is ‘the only phenomenally adequate way of talking about perceptual experience’ (p. 244). If we interpret him as invoking the negative ontological claim here, then we can read him as saying that the claim is supported by perceptual phenomenology.  But presumably, the indirect realist will disagree, holding that the only phenomenally adequate ‘way of talking about’ perceptual experiences is to say that in having them we are aware of non-normal objects. To assess which of these positions is correct, one would have to assess just which ontological claims, if any, are supported by perceptual phenomenology.

Whichever of these two ways one interprets the talk of intentional objects, the key question appears to come down to the inference from phenomenology to metaphysics: does perceptual phenomenology support (H2), its denial, or is it compatible with either?  The sense-datum theorist thinks it supports (H2).  Smith thinks that at least it is compatible with the denial of (H2).  It may be that he is correct about this, but it is not clear that he has given us good reason to think so. And even if there is good reason to think so, the apparatus of intentional objects doesn’t seem to be able to play any role, given its merely phenomenological status (“the theory is not ontological an ontological, but a phenomenological one”). If the appeal to intentional objects was supposed to bring into focus some aspect of perceptual phenomenology that would be disrespected by (H2), then the apparatus of intentional objects would not be dispensable after all; but Smith has given no reason to think that they are playing this role. For all he says, talk of intentional objects has the status that A.J. Ayer thought talk of sense-data had: it means nothing different from what is meant by ordinary ascriptions using ‘looks’ (cf . Grice (1989), where he uses ‘sense-data’ in the same deflationary way).[14] So long as talk about intentional objects is just a way of talking about perceptual phenomenology, one might think that it leaves the issues about the inference from phenomenology to metaphysics just as they were.

In the end, one might in principle be able to resist the indirect realist’s inference from phenomenology to metaphysics simply by giving reasons to think that the facts about the nature of perceptual phenomenology are equally well respected by a variety of proposals about the metaphysical structure of experience. Such a response to the arguments might go as follows. Compare two proposals about the structure of experience. According to the act-object proposal, illustrated by (H2) and the sense-datum premise of the argument from illusion, perceptual experiences are fundamentally structured by an act of awareness, on the one hand, and an object of awareness, on the other. On this view, the actual presence of something yellow is supposed to explain why we seem to be in the presence of something yellow. According to the propositional attitude proposal, in contrast such experiences are fundamentally structured like propositional attitudes, where the attitude – call it ‘experiential entertaining’, and say that this differs from belief  – by definition is one that one can be taken toward a proposition that (say) something is yellow, only when one seems to be seeing something yellow. The propositional attitude of experientially entertaining is by definition irreducibly phenomenal, and an experience does not need to have any other structure (such as an act-object structure) in order to be an experiential entertaining of a proposition.

These two proposals both posit a relation – direct awareness in one case, experientially entertaining in the other - such that standing in that relation to an appropriate relatum (a yellow sense-datum, or a proposition of some sort that is true only if there is something yellow in your view) suffices for it to seem to the subject as if she is seeing something yellow. By definition, both proposals accommodate the facts about how things appear when one seems to see something yellow, since the relevant sort of phenomenology is just written in to the definition of what it takes to stand in each kind of relation. The substantive question is whether there is any such propositional attitude, or relata of the awareness relation, in the first place. The pure sense-datum theorist needs reasons to think there could be no such primitive propositional attitude as the one envisioned (or at least, that there is none); likewise, the champion of the primitive propositional attitude needs reasons to think that there are no sense-data.

There are multiple possible considerations one might marshal in support for either of these views. But suppose we consider just phenomenal aspects of perceptual experience, in abstraction from all the other considerations one might bring to bear on a choice between these theories. And suppose we consider just the metaphysical structure of each theory – i.e., the structure of attitude and content, considered in abstraction from the nature of the attitude, and the structure of act and object, considered in abstraction from the nature of the relata. The phenomenal aspects of perceptual experience alone do not seem to derail either proposed metaphysical structure. At least, neither structure is obviously necessitated  - and neither one is obviously ruled out - by perceptual phenomenology alone. If so, then so far as their metaphysical structures are concerned, both proposals are equally adequate to perceptual phenomenology, and considerations besides phenomenological ones would be needed if one had to choose between them.

If this is right, then phenomenological considerations do not provide positive support for indirect realism. But they do not provide positive support for direct realism either. If so, then ultimately the status of direct realism will have to be settled by drawing on something besides reflection on the nature of perceptual phenomenology.[15]


Works cited.

D. Armstrong, 1961. Perception and the Physical World. New York: Humanities Press.


R. Chisholm, 1942. “The Problem of the Speckled Hen”. Mind 51, 368-73.


J. G. Fichte, 1982. The Science of Knowledge, tr. Peter Heath and John Lachs. Cambridge: Cambridge University Press.


J. Foster, 2000. A Theory of Perception.  Oxford: Oxford University Press.


H. P. Grice, 1989. Studies in the Ways of Words. Cambridge: Harvard University Press.

F. Jackson, 1977. Perception. Cambridge: Cambridge University Press.


M. Martin, 1992. “Sight and Touch” in The Contents of Experience, ed. Tim Crane. Cambridge: Cambridge University Press.


G. E. Moore, 1918. “Some Judgments about Perception”. Proceedings of the Aristotelian Society.


B. O’Shaughnessy 1980. The Will: volume 1. Cambridge: Cambridge University Press.


C. Peacocke, 1983. Sense and Content. New York: Oxford University Press.


--------------  unpublished. “Mental Action and Self-Awareness”


H. H. Price, 1932. Perception. Westport, CT: Greenwood Press.


S. Siegel, forthcoming. “Mind-independence in Visual Perception”


A. D. Smith, 2002. The Problem of Perception. Cambridge, MA: Harvard University Press.

[1] All page references are to Smith’s book unless otherwise noted.

[2] Some 20th century sense-datum theorists (e.g., for a brief time, Moore in his 1918) held that sense-data were parts of surfaces of public objects, rather than being mind-dependent entities.  This view that we perceive public objects by having sense-data in this sense is not Smith’s target.

[3] Smith adopts the label ‘Anstoss’ from Fichte (1982, p. 191), who uses it to describe a felt check on intellectual acts, rather than a felt check on active bodily movement.

[4] Smith thinks another sort of illusion is possible in haptic-tactile experience: illusions for degree of force with which something resists your movement.   Call the force with which the shoe presses against your hand D, and say that you experience yourself to be tactually resisting a force of one degree D + 1. (The units here are artificial; what’s important is the idea that one could experience something as heavier than it actually is.) One might doubt that one could experience something as being heavier than it actually is. Heaviness is a relational property; if one lacks strength, say because one is weak with fatigue, then the shoe really will be heavier than it would be relative to a someone stronger.

[5] Martin (1992) objects to the idea that there could be “sense-data that one would feel to resist one’s body” on the grounds that it would “posit…objects which fall outside any [tactile] sense field (since they are felt as that which is outside the body) and which are therefore not directly experienced” (p. 213). But, he says, nothing in tactile phenomenology suggests that there is a spatial field in which things can be touched without moving (in contrast with visual phenomenology, where there looks to be a limited area in which things can be seen without moving the eyes). So there is no motivation for positing a private, mental tactile field to account for the tactile phenomenology.

It is not clear, however, why there is no such thing as a tactile sense-field in the phenomenology of tactile experience that consists both of where one feels one’s body to be, and the space in which one could feel things if one moved one’s limbs without changing location. One can have a sense (illusory or not) of something being in one’s ‘personal space’, for instance. (Not every part of one’s ‘personal space’ is felt, but an analogous point holds for the visual case, if it includes opaque three-dimensional volumes. Not every part of those is seen. Martin discusses this aspect of the visual field as he construes it in section III). Then the fact that these entities would be felt to be outside the body would not suffice for being outside of the tactile sense-field. (For further discussion of the notion of a tactile sense-field, see O’Shaughnessy, vol 1).

[6] What I’m calling the ‘two factor option’ differs from what Smith calls ‘two-component theories’ in his chapter 3. The latter divide experience into a component that essentially involves concepts (or perhaps their ‘application’), and a component that does not. No commitments at all about concepts are made by the two-factor option.

[7] Price 1932, chapter 6.

[8]  Smith says, of the bodily sensations involved in sandbag cases, “It is not these sensations that according to Hume and Prichard, we mistake for physical objects” (p. 160). This seems to be his reason for saying that in the Anstoss “it is our activity, rather than our senses, that reveals something foreign to us” (ibid). Smith grants that Anstoss always involves some bodily sensations or other. So in denying that the senses reveal something foreign to us, he seems only to be counting only those sensations that one feels at the same place as the place where one feels the foreign object to be.

[9] “Mental Concepts and Self-Awareness”, unpublished.

[10] Given how the Anstoss is defined, there is no room for controversy about whether it suffices for perceptual consciousness. There is a question, though, about the exact kind of externality that Anstoss experiences represent things as having. It certainly seems to be part of some kinds of tactile phenomenology that one is encountering something other than one’s body – and, going with that, something whose movements do not depend on the existence of one’s body (and so one’s skin, the apparatus of tactile perception). But this by itself could be compatible with a lack of at least one kind of perceptual consciousness – a kind that presents things as existing independently of one’s mind, at least one aspect of it. Perhaps an experience could present something as independent from one’s body, while taking no stand on whether its movements and even its existence is independent from one’s will. It seems easy to imagine an item – say, a shoe - that was not part of a subject’s body (and did not feel any way when one touched it), but whose behavior and existence was nonetheless subject to one’s will. Perhaps witches of some sort could produce and interact with such items, yet have exactly the same kind of tactile phenomenology that we have when picking them up! Smith sometimes seems to gloss over the distinction between existing independently of one’s body and existing independently of one’s mind by speaking of the relevant sort of independence as independence ‘from ourselves’ (e.g. p.  168).

[11] A more elaborate pair of cases of this sort is discussed, and something close to the representation-of-disposition condition is defended for visual experiences, in “Mind-independence in Visual Perception”.

[12] That sense-data, if there are any, have such properties is implicit in one defense of the sense-datum theory against Chisholm’s speckled hen objection.  According to the objection, when a hen looks to have an indeterminate number of speckles, the sense-datum theory must posit a sense-datum that really has an indeterminate number of speckles; but (barring vagaries of what counts as a speckle), there is no such thing as an indeterminate number of speckles – so the sense-datum theory must be false. (See Chisholm 1942, also Armstrong 1961). The defense against the objection is that there are features of sense-data that are available upon further inspection (such as how many speckles they have), even if these features are not immediately discernible. This response assumes that sense-data are such that one can at least vary how much attention one pays to their features, without thereby changing those features. (See Foster 2000). The further stipulation that such variations in attention could not occur without moving one’s eyes relative to the sense-datum seems ad hoc. So if this response to the speckled hen objection is coherent, then the representation-of-disposition version of the perspectival condition is compatible with indirect realism.

[13] The first view is also suggested by Smith’s remark that “the intentional directedness of even hallucinatory states of perceptual consciousness is simply a phenomenological fact” and that his theory “is not an ontological, but a phenomenological one”.

[14] “The Causal Theory of Perception,” reprinted in Grice, H.P., 1989. Studies in the Ways of Words. Cambridge: Harvard University Press.

[15] For helpful discussion of general issues related to section 7, I am grateful to Maja Spener and Scott Sturgeon. Thanks to members of a reading group on The Problem of Perception book held at Harvard in the Spring of 2003, and to audiences at UCL and the Australian National University, where earlier versions of this paper were presented. For further criticisms and discussion thanks to Jake Beck, Philippe Chuard, Benj Hellie, Bernard Nickel, John O’Dea, Kranti Saran, A. D. Smith, Daniel Stoljar, and especially David Chalmers and Mike Martin.