CONTENT, CHARACTER, AND COLOR II:
A BETTER KIND OF REPRESENTATIONALISM
Sydney Shoemaker – Cornell University [Draft 4/14/02]
[This material is slightly revised and reorganized from two Whitehead lectures given at Harvard in April 2002. The first NEH presentation corresponds to the first Whitehead lecture and the beginning of the second. The second NEH presentation corresponds to the rest of the second lecture (this file), plus “Introspection and Phenomenal Character” (which elaborates the view here at greater length).]
From now on I will assume that it is possible in principle for there to be cases of spectrum inversion in which the invertees are equally good perceivers of the colors. What I want to show next is that while allowing this possibility is incompatible with standard representationalism, it requires acceptance of a different version of representationalism.
Consider the standard way of describing a case of spectrum inversion. Returning to Jack and Jill, we say that red things look to Jack the way green things look to Jill, blue things look to Jack the way yellow things look to Jill, and so on. Of course, we might also express this by saying that the phenomenal character of Jack’s experience of red things is like the phenomenal character of Jill’s experience of green things, and so on. Or by saying that “what it is like” for Jack to see red things is “what it is like” for Jill to see green things, and so on. But “phenomenal character” is philosophical jargon, and “what it is like” is on its way to being that. We need to be able cash these locutions in terms that we are sure we understand. And I think that the best way of doing that is in terms of how things look.
Now the sense in which red things look different to Jack and Jill cannot be that they look to have different colors in the epistemic sense. We can suppose that both perceive red things as being red, and therefore that to both red things look red in the epistemic sense. Nor can it be the comparative sense – to each, we can suppose, red things look like standard red things under standard conditions. The remaining sense of “looks” is supposed to be the phenomenal sense. Now those who employ this notion typically speak of things as looking red, blue, yellow, etc., in the phenomenal sense. But if Jack and Jill are both accurate perceivers of the colors of things, it can’t be that the difference in how things look to them is a difference in what colors things look to them, even if “looks” is used in the phenomenal sense. Yet it must be that there is a way red things look to Jack that is different from the way red things look to Jill. Something’s looking a certain way to someone would seem to be a matter of that person’s experience representing that thing as being that way. And to represent something as being a certain way would seem to be a matter of representing it as having a certain property. So Jack’s and Jill’s experiences of red things must represent them as having different properties. But we are assuming that their experiences do not represent red things as having different colors. It falls out that the properties represented by their color experiences include properties that are not colors, that their experiences differ in which of these properties they represent, and that the difference in the phenomenal character of their color experiences consists in this difference in their representational content. These other properties will be closely related to the colors. Since how something looks, phenomenally, determines, in normal circumstances, what color one perceives it to have, one could be said to perceive the colors of things by perceiving properties of this sort. For a particular subject, perceiving the same property of this sort will normally constitute perceiving the same color. But in principle people can differ with respect to what properties of this sort they perceive when perceiving a certain color. So it is with Jack and Jill in our example.
In earlier work I referred to these as “phenomenal properties.” I now prefer the term “appearance property.” When I first introduced these properties I thought that there was nothing for them to be except relational properties that things have in virtue of producing experiences of certain sorts. Later I saw that there was nothing to prevent these properties from being dispositional properties of a certain sort. And I now distinguish two sorts of appearance properties – occurrent appearance properties, which things have when they are actually appearing certain ways to perceivers, and dispositional appearance properties, which things have in virtue of being disposed to appear certain ways to perceivers of one or more sorts. More on these later.
Spectrum inversion, of the sort I champion, involves things of the same color looking different to one perceiver than to another without either of them misperceiving, and things of one color looking to one perceiver the way things of a different color look to another perceiver, again without either of them misperceiving. This would involve in the first case the experiences of the perceivers representing things of the same color as having different appearance properties, and in the second case the different subjects perceiving different colors by means of experiences that represent the same appearance property. But it is not only in the hypothetical case of spectrum inversion that how the colors of things appear, phenomenally, can come apart from what colors they are perceptually represented as having without there being any misperception. This is in fact an everyday occurrence. In many cases of color constancy, while there is no difference in how things of the same color look epistemically with respect to color, i.e., they look in that sense the same color, there is a difference in how they look phenomenally. When part of a uniformly colored object is in shadow, the shadowed part and the unshadowed part will look, epistemically, to be the same color. But the two parts will look different. And the part in shadow may look the same, phenomenally, as a darker object in another part of the room that is not in shadow, even though it does not look, epistemically, to be the same color as that object. In such cases we could say that things of the same color are perceived as having different appearance properties and things of different colors are perceived as having the same appearance property.
In some cases of simultaneous contrast, there may be a color illusion – identically colored things may look, in the epistemic sense, to have different colors because of the effects of their being juxtaposed to things of different colors. However that may be, these things will also appear different in the phenomenal sense, and will have different appearance properties. The representation of them as having different appearance properties will not be an illusion.
A real world case that is closer to the case of spectrum inversion involves the well attested differences among normal human subjects as to what combinations of wavelengths are perceived as “unique hues” – e.g., the fact that what some perceive as a unique green others will perceive as a slightly bluish green and still others will perceive as a slightly yellowish green. Here the different subjects give different descriptions, using color terms, of what they perceive, yet there seems no good basis for saying that some are perceiving correctly and others are misperceiving. Plainly their different descriptions do give expression to differences in how things look to them. Assuming that we are not prepared to say that the colors themselves are perceptual system relative, the situation seems best described by saying that normal subjects perceive the same determinate shades of color by perceiving slightly different appearance properties, and that the phenomenal distinction between unique and binary hues resides in the first instance in the appearance properties and applies to the colors themselves in virtue of their association, in certain sorts of observers, with certain appearance properties.
I have said that when we characterize appearance properties as ways things look, the use of “looks” is not its epistemic sense or its comparative sense, but is instead its “phenomenal” sense. And as I have mentioned, philosophers who employ this notion typically assume that sentences like “It looks blue” can be used to report how something looks in the phenomenal sense. And so they can, if there is an implicit relativization to a type of perceptual system. What would be wrong is to suppose that if a thing looks blue to Tom, and the way the thing looks, phenomenally, to Tom is the same as the way it looks to Dick, then it must be that it looks blue to Dick. A particular phenomenal way of looking will be related to a particular shade of color only in virtue of a certain sort of perceptual system being so constructed that in a standard sort of case things of that color look that way to a possessor of that sort of perceptual system. And in principle, the same phenomenal way of looking can be related in this way to different colors relative to different sorts of perceptual system. As noted earlier, even if we hold the perceptual system fixed the way things of a given shade of color look can vary with lighting conditions and the like. But of the various ways things of a given shade of color can look to a creature of a given sort, there will be one way, or a small range of ways, that has a privileged status. Roughly, this will be the way things of that color look to such a creature in optimal viewing conditions. Call this the canonical way of looking for that shade, relative to that sort of perceptual system. When we say that something looks sky blue to a person, using “looks” in the phenomenal sense, we typically mean that the way it looks is the canonical way for sky blue things to look to a creature of the sort he is. It is compatible with this that the circumstances are such that the person has no inclination at all to think that the thing in question is that shade of blue – thus, for example, an expanse of snow can look blue without one’s having any inclination to think that it is blue.
My name for appearance properties gives a good indication of what sort of properties I take them to be. An appearance property is a property something has just in case it appears a certain way. Saying that something appears a certain way can mean either that it actually is appearing that way to someone or that it is such as to appear that way under conditions that actually obtain. And thus we have my distinction between occurrent appearance properties and dispositional appearance properties. I take something’s actually appearing a certain way to someone to be a matter of its producing in that person, via the processes and mechanisms involved in the relevant sort of perception, a perceptual experience of a certain sort – one in which the person is “appeared to” in a certain way. So occurrent appearance properties will be relational properties that things have only when perceived. Dispositional appearance properties will of course be dispositions whose manifestations are instantiations of occurrent appearance properties.
As I have said, on this view the phenomenal character of experiences is that part of their representational content that consists in their representing appearance properties. It is what appearance properties the object is represented as having that fixes the phenomenal character of the experience, and fixes “what it is like” to have the experience. The view respects transparency. Indeed, a natural way of expressing transparency is to say that if asked to attend to the phenomenal character of a perceptual experience, all one can do is attend to how something appears to one, or how one is “appeared to,” in virtue of one’s having the experience – and to attend to that is to attend to what appearance properties are represented by the experience.
Michael Tye has objected to this view that it implies that we do not directly perceive the colors; we perceive them only indirectly, by perceiving other properties, what I now call appearance properties and what in the work Tye discusses I called phenomenal properties. I hope that my change of terminology helps to defuse this objection. To say that it is by perceiving appearance properties that we perceive the colors of things is equivalent to saying that it is by perceiving how a thing appears that we perceive what color it is. It would be bizarre to claim that things appearing certain ways to us amounts to there being a veil of perception between us and them. Someone might allow that of course things appear phenomenally certain ways to us when we perceive them, but deny that it is by perceiving things to appear in certain ways that we perceive their properties. And certainly it would be correct to deny that in perceiving a thing we normally attend to how it appears to us, phenomenally, or normally infer from this to what properties it has. But it is compatible with this that we cannot perceive something without its phenomenally appearing somehow to us, that this contributes, along with our background beliefs, to how it appears “epistemically” and how we take it to be, and that when something appears to one a certain way, phenomenally, its so appearing is part of the nonconceptual content of one’s perceptual experience. So, in the case of color constancy mentioned earlier, the different “looks” of the shadowed and illuminated parts of a surface are both represented in one’s experience, and are input, along with the information about lighting contained in the rest of the scene, to the epistemic appearance that the surface is uniform in color.
But Tye thinks that my view does erect a veil of perception between us and the colors of things, and indeed that it makes it compatible with the content of our experience that things have no colors at all. He says that “if colors are seen by seeing other qualities, not themselves colors, then the fact that a seen object has the qualities it is directly visually experienced as having affords no epistemic guarantee that it has the color the subject of the experience takes it to have or indeed that it has any color at all” (p. 103). Now it is no objection to my view that it has the consequence that something’s being perceived to have a certain appearance property affords no guarantee that it has the color the subject takes it to have – that is as it should be. It would be an objection to it, however, if it had the consequence that perceiving an appearance property, of the sort in question, afforded no guarantee that the object has any color at all. And there is one understanding of my view on which it does have this consequence. So it is important for me to reject that understanding of it. I have characterized the instantiation of occurrent appearance properties as consisting in the production by an object of experiences of a certain sort. If no constraint is put on the nature of the causal process by which an object produces an experience, then something could count as having the appearance property I see when I see a ripe tomato if it produced in someone the same sort of experience by a process that did not involve at all the mechanisms involved in vision. And in that case, arguably, something could have that appearance property without having any color. Clearly I must rule this sort of case out. Appearance properties must be individuated not only by the sorts of experience they produce or are disposed to produce, but also by the kind of causation involved in the production of these experiences. In the case of the appearance properties involved in our perception of color, the causation must be the sort involved in color vision – it must involve the reflection or emission of light, its impact on the retina, and all the rest of it. To veridically perceive the instantiation of such appearance properties we must see the object in which it is instantiated, and the causal process by which the object produces one’s experience must be part of that involved in seeing the object. There will then be no possibility that the things perceived to have these properties lack colors.
It may seem that this does not really avoid Tye’s objection. For couldn’t there be a kind of perception involving a causal process that did not involve light and the retina and that eventuated in experiences phenomenally like our color experiences? And if so, wouldn’t the objects of this sort of perception have appearance properties that are indistinguishable from those we perceive in perceiving colors, but which something can have without having a color? And isn’t that just as bad as the consequence Tye thinks follows from my view? Probably it is just as bad. But I would deny that there is this possibility. It is possible, perhaps, that a creature could have experiences phenomenally just like our color experiences that are standardly produced in it in a way that does not involve light and eyes. Perhaps this could happen in a brain in a vat. But if, indeed, that creature’s experiences were phenomenally just like ours, it would not be veridically perceiving appearance properties different from those we perceive – it would be having illusory experiences that represent, falsely, the instantiation of the same appearance properties ours represent. This is of course what standard representationalism would have to say about this case, and on this point standard representationalism and my version of representationalism agree.
I have characterized appearance properties in terms of what kinds of experiences their possessors cause or are disposed to cause. You may well ask how the kinds of these experiences are specified. They of course cannot be specified in terms of what colors they represent, because it is essential to the account that experiences of the same kind can represent different colors when they are experiences belonging to different sorts of perceivers, and experiences of different kinds can represent the same colors when they are experiences of different sorts of perceivers. And it may seem that the account goes viciously circular if they are specified in terms of their phenomenal character, for on my account the phenomenal character of an experience is an aspect of its representational content that consists in the representation of appearance properties.
But the circularity can be circumvented, if I may so speak, by appealing to a different way in which phenomenal character can be characterized. An important part of my account involves a point made in my first lecture, namely that relations of phenomenal similarity and difference among perceptual experiences have to be relations that play a distinctive functional role, having to do with the effects of the holding of these relations on discrimination, recognition, and belief fixation. We can think of particular phenomenal characters as the features of experiences in virtue of which these similarity and difference relations hold. This ties in with the characterization of phenomenal character as the representation of appearance properties. For phenomenal similarity is a kind of intentional similarity; it is, more specifically, similarity with respect to how one is appeared to. Likewise, phenomenal difference is difference with respect to how one is appeared to. And similarities and differences with respect to how one is appeared to are, in the first instance, similarities and differences with respect to what appearance properties one’s experiences represent. I say “in the first instance,” because of course in representing objects as appearing certain ways one’s experiences normally represent things as having certain objective properties, e.g., colors. And when the appearance properties are of the sort involved in color perception, there will be a close association between representing things as similar or different with respect to appearance properties, of the sort involved in color perception, and representing them as similar or different with respect to color. But this close association is not invariable, since, as noted earlier, it can happen without there being any perceptual illusion that things of different colors appear the same, and things of the same color appear different.
Given a notion of phenomenal similarity and difference that can be applied both intrasubjectively and intersubjectively, it would seem that we can give a “package deal” definition that simultaneously defines both an appearance property and the corresponding phenomenal state, the corresponding state of being-appeared-to. In what follows, the phrase “the state of being appeared-blue-to” will refer to the state of being appeared blue to in the phenomenal sense, and the phrase “the property of appearing blue” will refer to the corresponding appearance property. It should remembered that it is only relative to certain sorts of perceptual systems that a given phenomenal state can be described as one of being appeared-blue-to, and that the corresponding appearance property can be described as one of appearing blue – in the present context, it is relative to our sort of perceptual system (or at any rate, mine) that the phenomenal state is one of being-appeared-blue to and the appearance property is one of appearing, phenomenally, blue. We can now say that the state of being appeared-blue-to and the appearance property of appearing blue are the unique state S and unique property P such that S represents P, something is in state S just in case the way something looks to him is the way blue things look to us under optimal conditions, and something is P just in case it is producing state S in some observer (in a way that does not involve misperception), or disposed to produce S in some sort of observer. This uses the circumstances in which we are in such a phenomenal state and perceive such an appearance property, namely those in which we are perceiving or seeming to perceive blue, to fix the reference of the terms “state of being appeared-blue-to” and “property of appearing blue” to the state and property in question; but assuming that sameness of phenomenal character is well-defined for the intersubjective case it allows creatures of other sorts to be in the same state and perceive the same appearance property when they are not perceiving or seeming to perceive blue.
But while I have said something about what intrasubjective sameness of phenomenal character comes to, I have so far said nothing about how this notion – and, more generally, the notions of phenomenal similarity and difference – apply intersubjectively. Let qualia be the features of perceptual states in virtue of which they stand in relations of phenomenal similarity and difference. As I have said, in the first instance those relations will be defined for the intrasubjective case, by the role they play in discrimination, recognition, and belief fixation. And it is this that gives qualia their functional roles. Each quale will occupy a place in an intrasubjective similarity ordering, and the places in this ordering occupied by the qualia instantiated in different perceptual experiences will determine the similarity and difference relations among the experiences. But assuming physicalism, qualia must be physically realized. Supposing they are multiply realized, each quale will be associated with a set of properties that can be said to be realizers of it in virtue of the way instantiations of them are related – that is, the realizers are such that experiences in which different ones of them are instantiated, but are otherwise the same (at the appropriate functional level of description), are qualitatively alike. It will, of course, be the functional role these physical properties play intrasubjectively that makes them realizers of qualia, and realizers of the particular qualia they realize. But if two creatures are alike with respect to what physical features of their brain states serve as realizers of the qualia involved in their experiences, then an experience of one of them is phenomenally similar to an experience of the other to the extent that the qualia realizers they instantiate are ones whose instantiations would yield phenomenally similar experiences if instantiated in the same subject. Or, more simply, the experiences of the different creatures are phenomenally similar to the extent that they share the same qualia, which they do by sharing the same qualia realizers or by having different qualia realizers that are realizers of the same qualia. While it is the intrasubjective functional role of a property that makes it a quale realizer, its being a realizer means that it realizes the same quale in whatever creature its instantiation occurs – so if the same realizer is instantiated in the experiences of two different subjects, or two different qualia realizers are instantiated in them but these are ones that realize the same quale when instantiated in the same subject, the experiences of the two subjects will be to that extent phenomenally the same.
The account just sketched, which I have presented elsewhere, is an attempt to parlay a functionalist account of intrasubjective phenomenal similarity into an account of intersubjective phenomenal similarity by appealing to the idea that the physical realizers of phenomenal states are properties that can be shared by different persons. It therefore assumes the falsity of what I have called the Frege-Schlick view – the view that phenomenal similarity is well-defined only for the intrasubjective case. Robert Stalnaker has made a forceful case in favor of the Frege-Schlick view and against my strategy for dealing with the intersubjective case. I wish that I had more to say than I do in response to his arguments, but in any case I do not have the space to say any of it here. So I will bracket that issue, and offer my view as the view one ought to accept if one thinks, as I do, that there are phenomenal similarities between experiences of different persons as well as between different experiences of the same person.
When I first developed my current view, the view that invokes appearance properties, I thought I had to hold that the qualia in virtue of which experiences are similar and different are not, in first instance, accessible to introspection. I said that they are known only “by description” – a claim that left friends of qualia shaking their heads. This was because I thought that they have to be different from the representational features of the experiences that collectively make up their phenomenal character, and because it was the latter I held we have direct access to in introspection, this being my way of respecting the transparency thesis. One reason to think that the qualia and the representational features must be distinct is the belief that identifying them would involve me in an unacceptable circularity. But I think that the circularity can be avoided by means of the sort of package-deal definition I have just sketched, and I now think that there is no reason not to identify the functionally characterized qualia with the representational features that make up phenomenal character, and no reason to deny that the qualia are introspectively accessible.
But the representational status of qualia is complex. They can be thought of as the vehicles of representation of properties of perceived objects. As such, they can be said to represent such properties. In the case of objective properties, such as colors, it will be only contingently that a given quale will represent a particular property – e.g., that a particular quale represents red. Any view which holds that perceptual experiences represent what they do in virtue of causal correlations between types of experiences and features of the environment, or (as on Dretske’s view) in virtue of an evolutionary history that bestowed on certain types of experiences the function of indicating certain features of the environment, needs a way of typing experiences that makes it a contingent fact that experiences of a given type represent the particular features of the environment they do. It may be thought that all this requires is that the experiences belong to physical types, which physicalist representationalists who reject qualia can readily allow. But I think it is clear that the typing must be functional rather than physical. The types must be so related to each other, and to the rest of the subject’s psychology, as to determine its “quality space,” and so must play an appropriate role in determining its discriminatory and recognitional capacities, and in generating beliefs about its environment. What must be true of different tokens of the same type is that they share a certain causal role; this allows for “multiple realization,” and so for the possibility that tokens of an experiential type might be physically heterogeneous. Spelling out the conditions for type membership requires a functional account of qualitative similarity and difference, and qualia are the properties of sensory states, presumably properties that are physically realizable, in virtue of which they stand in these relationships. Thus it is that it is qualia that type the experiences, and thus it is that qualia contingently represent the environmental features they do.
But it is necessary, not contingent, that qualia represent the appearance properties they do. Appearance properties are individuated in terms of the types of experiences they produce, and these experiences are typed by their qualitative character. The qualitative character of an experience, what qualia are instantiated in it, thus fixes its phenomenal character, i.e., what appearance properties it represents.
As a first approximation, one could say that this is a view which is externalist about the determination of the objective representational content of experiences, i.e., of what objective properties they represent, but internalist about the determination of the content that bestows phenomenal character, i.e., that which represents appearance properties. But this is only a first approximation. Earlier I said that the perception of an appearance property requires that the causal process that results in the perceptual experience be of a certain sort – for the appearance properties involved in vision, it must be a process involving the impact of light on the eyes, etc. If Hillary Putnam is right that a brain in a vat which had always been a brain in a vat, and which was the microphysical duplicate of the brain of someone veridically perceiving a barn, would not be having an hallucination of a barn but would instead, or at any rate might instead, be having veridical experiences of the state of the computer to which it is attached, then the experiences of such a brain in a vat could not be phenomenally just like those of the normal embodied person with the microphysically identical brain. For, if the brain in a vat is not misperceiving, his experience cannot represent the appearance properties that the experiences of the normally embodied person represent. So my view cannot allow that the phenomenal character of visual experiences supervenes on the state of the subject’s brain, and probably it cannot allow that it supervenes on what is inside his skin. In a very modest way, this content is externalist. But this contrasts with the radical way in which perceptual content is externalist on the views of Fred Dretske and Michael Tye discussed earlier. Their views allow physically identical creatures to have phenomenally different visual experiences in virtue of standing in different relations to their environments which give their experiences different representational content. My view does not allow this. The only case in which it would allow a physical duplicate of a creature not to have experiences phenomenally like that creature’s visual experiences is that in which the physical duplicate’s relations to its environment are not such as to permit any of its states to count as experiences with visual content of any sort whatever.
Where does this account leave the explanatory gap problem? I distinguished in my first lecture an “objective explanatory gap problem” and a “subjective explanatory gap problem.” The first had to do with the seeming disparity between how we perceive the world, when we have experiences of color, and what we believe on the basis of science to be its real nature. The second had to do with the seeming disparity between what it is like to undergo sensory experiences and what we believe, if we are physicalists, to be the real nature of those experiences. We saw that a standard response to the objective explanatory gap problem has been to “kick upstairs into the mind” the qualitative character of colors and other secondary qualities, and that if we are physicalists this leaves us with the subjective explanatory gap problem. We also saw that standard representationalism can be seen as an attempt to solve the subjective explanatory gap problem by kicking the phenomenal character of perceptual experiences “downstairs,” into the external world, thus leaving us with the objective explanatory gap problem. What I have called the selection gambit can perhaps be seen as a compromise between these strategies in its attempt to acknowledge a sense in which the nature of our perceptual system determines the phenomenal character of our experience while at the same time holding that phenomenal character is objective representational content.
I think that despite the fact that my account is a version of representationalism, it must count as a view that reduces the objective explanatory gap problem to the subjective explanatory gap problem. The question of how things appearing a certain way can be physically realized reduces to the problem of how the psychological state of being appeared-to in a certain way can be physically realized. So the question is whether my account satisfactorily solves the subjective explanatory gap problem -- whether its account of the phenomenal character of experiences makes intelligible how having experiences with the phenomenal character our perceptual experiences do have can be realized in the physical nature of our brains.
It certainly would do so if we could accept a view that I discussed in my first lecture. This is what I have called the relational view, the view that the phenomenal character of experiences is fixed by their place in an experience space, i.e., by their place in a network of similarity and difference relations holding amongst experiences. We saw in my first lecture that this view, while incompatible with the view that there can be symmetrical color experience spaces, and so with the view that there can be undetectable spectrum inversion, is compatible with the view that there can be spectrum inversion involving asymmetrical experiences spaces like our own. That view is also compatible with the version of representationalism I have presented today, which holds that the phenomenal character of experiences consists in an aspect of their representational content, viz their representation of appearance properties, where having an appearance property is a matter of producing or being disposed to produce an experience having a certain phenomenal character. For the similarity relations that this view takes to individuate experiences are relations of intentional similarity – similarity with respect to what the experiences represent. At the same time, they are functionally defined similarity relations, which hold in virtue of causal roles played by the experiences themselves. This double aspect of the similarity relationships makes appearance properties the properties that the experiences in the first instance represent. For appearance properties are properties of external objects, making them suitable targets for representation, but at the same time are such that the similarity relations between them necessarily conform to the functionally determined similarity relations between the experiences that represent them. Taking the phenomenal character of experiences to consist in their place in a similarity ordering provides a simpler way than the “package deal” account I offered earlier of avoiding the apparent circularity involved in individuating appearance properties in terms of the phenomenal character of experiences and defining phenomenal character in terms of the representation of appearance properties. Assume with the relational view that being an experience with a certain phenomenal character is a matter of being an experience that occupies a certain place in an asymmetrical similarity ordering amongst experiences, where that ordering is defined functionally. We can then say that having a certain appearance property is a matter of producing or being disposed to produce an experience with a certain phenomenal character. Intersubjective comparisons of the phenomenal character of experiences are straightforward on this view; if an experience space of mine is isometric with one of yours, then an experience of mine is phenomenally identical with one of yours if they occupy corresponding places in our experiences spaces.
This is a tempting view. It is compatible with much that I have said. And I do want to say that occupying a position in a similarity ordering is part of what constitutes the phenomenal character of an experience. But as I indicated in my first lecture, I cannot believe that it is all of it. Phenomenal similarity relations hold in virtue of the phenomenal natures of their terms. If you fix the phenomenal natures, you thereby fix the similarity relations that hold amongst them. But not conversely – fixing the similarity relations cannot be counted on to fix the phenomenal natures. If it is true that there is only one set of phenomenal natures that stand in these similarity relations, that can only be contingently true – the phenomenal natures being what they are entails, but cannot be constituted by, the holding of the set of similarity relations. And when we are perplexed as to how the physical world can appear the way it does, what we are perplexed about is not, at least in the first instance, how it can appear in such a way that the various parts of it are perceived as standing in one or another similarity relation to one another. What we find perplexing is that perceiving the holding of these various similarity relations should involve these phenomenal natures, either in the world or in our experience of it. I focus on the yellow of the lemon, or the salty taste in my mouth, or the pain in my thumb after I have hit it with a hammer, and ask how can this be a feature realized in a purely physical state of affairs.
If the phenomenal character of an experience is not fixed by its position in an experience similarity space, there can be different creatures with identically structured experience similarity spaces whose experiences differ in phenomenal character. This amounts to there being experiences that are in all relevant respects functionally alike but differ in phenomenal character. I reject as incredible, as not even beginning to address the issue, the view that the difference between these could consist in something about their evolutionary history – their having evolved to indicate, or track, different environmental features. Assuming physicalism, the only thing that could account for the difference in phenomenal character is how these experiences are physically realized – as it might be, their being realized in a carbon based physiology rather than a silicon based physiology. And here is where the residual puzzle lies. How can the phenomenal character of our experiences be fixed in part by how they are physically realized?
It is a necessary condition of the relational view’s having a physicalistically satisfactory response to the subjective explanatory gap problem that it be possible for an experience space having a certain structure to be physically realized. Let’s suppose that such realization – realization of an experience space that gives sensory states a certain similarity ordering -- is possible. But let us also suppose that, as I have argued, the relational view is false. Then any way of physically realizing an experience space having a certain structure must, perforce, be a way of realizing states that stand in the relations of similarity and difference that constitute that structure because they have phenomenal characters that are not relational. So if this necessary condition of the relational view’s having a physicalistically satisfactory response to the subjective explanatory gap problem is satisfied, but the relational view is nevertheless false, then it must be possible for phenomenal character that is not relational to be physically realized.
Here I disagree with Joseph Levine, whose term “explanatory gap” I have appropriated, about the nature of this gap. He thinks that in the sense in which, given all of the physical and chemical facts about H2O, it is not “epistemically possible” that H2O should fail to be water, it is epistemically possible that creatures physically just like us should lack qualia, or phenomenal states – although he insists that this epistemological possibility does not imply metaphysical possibility. On my view, in whatever sense the total physical story about H2O rules out the epistemological possibility of H2O not being water, the total physical story of the brain rules out the epistemological possibility of creatures physically just like us who lack phenomenal states. It rules out the possibility of what some call zombies. What is not epistemologically guaranteed by the total physical story about the brain plus what we know about the functional roles of mental states is not that the creature has phenomenal states – that is guaranteed -- but that it has these phenomenal states. The explanatory gap is not between the physical facts about a creature and its having states with phenomenal character; it is between the physical facts and its having states with certain particular phenomenal characters, specified as we would ordinarily specify them.
If we had an account of how an experience space of the sort we have can be physically realized, and knew that the realizer states are instantiated in us, we would be in a position to know that a particular phenomenal state had a certain physical realizer. But merely knowing what the physical realizers are, and knowing that collectively they realize an experience space having a certain structure, would not be sufficient for knowing the phenomenal character of the experiences they realize. On the relational view it would be sufficient for this, since on that view there is nothing more to an experience’s phenomenal character than its position in an experience space. So if we came across aliens in which an experience space with the same structure was realized differently, it would be true on the relational view, but it would not be true on the view I am defending, that in knowing that a certain alien state is a realizer of an experience with a certain position in that space, we would know its phenomenal character.
How disturbing should it be that knowing what physical states are realizers of phenomenal states, and knowing what functional roles are played by the realizers, is by itself insufficient for knowing what phenomenal states are realized by particular realizers? It seems to me that it should not be disturbing. It will be part of the functional role of the phenomenal states that they intimate themselves to their possessors in a special way, i.e., introspectively, and that their possessors will have concepts of them that are applied in introspective judgments. These will be concepts that can only be possessed by creatures to whose repertoires of phenomenal states the phenomenal states in question, those of which they are concepts, belong. It will follow from the functional role that there is such a concept, but it is in the nature of the case that the functional role cannot itself specify what the concept is. For one thing, we are supposing that the functional role could be occupied by different phenomenal states in creatures that are differently constituted physically, and thus that the concept associated with it could be different in different creatures. I am supposing that together with the physical nature of the realizer, the functional role does determine the phenomenal character, and so determines what introspective concept possessors of states with that phenomenal character have. But that doesn’t mean that in knowing the functional role and the physical nature of the realizer one thereby has the concept. For of course one could know all this even if, given one’s own physical nature, the realizers, and so the particular phenomenal states they realize, could not be instantiated in oneself, and so even if it were impossible in principle for one to have the requisite sorts of concepts of the phenomenal states.
What we perhaps hanker after, as a solution to the subjective explanatory gap problem, is an account that would take us from the nature of the physical realizers and their functional role to the phenomenal character of the experiences they realize. We can’t have that for the simple reason that knowledge of such physical and functional facts could not possibly give us the introspective concepts that knowledge of the phenomenal character would require. But the unavailability of such an account is not only compatible with, but is arguably implied by, an account according to which the physical and functional nature of the physical states makes them realizers of, and so states that metaphysically necessitate, phenomenal states of specific sorts. If an account of the latter sort is acceptable, that is perhaps all that we can reasonably ask for as a solution to the subjective explanatory gap problem.
Of course, getting what one can reasonably ask for doesn’t always amount to getting what one would like to have, and so it is in the present case. Candor requires me to confess that when I attend to the phenomenal character of my states of being appeared-to, and remind myself that on my view these are physically realized, a sense of mystery is still there.
REFERENCES (FOR LECTURES I AND II)
Block, N. 1999. “Sexism, Racism, Ageism, and the Nature of Consciousness.” In The Philosophy of Sydney Shoemaker, Philosophical Topics, 26. 71-88.
Boghossian, P and Velleman, D. 1989. “Color as a Secondary Quality.” Mind, 98. 81-103.
Boghossian, P and Velleman, D. 1991. “Physicalist Theories of Color.” The Philosophical Review, 100. 67-106.
Byrne, A. and Hilbert, D. 1997. “Color and Reflectances.” In Byrne and Hilbert, eds., Readings on Color, Vol. I, The Philosophy of Color. Cambridge, MA: Bradford/MIT.
Campbell, J. 1993. “A Simple View of Colour.” In Reality, Representation and Projection, ed. J. Haldane and C. Wright, Oxford: Oxford University Press.
Dennett, D. 1993. “The message is: There is no medium.” Philosophy and Phenomenological Research, 53. 919-931.
Dretske, F., 1995. Naturalizing the Mind. Cambridge, MA: Bradford/MIT.
Hardin, L. 1988/1993. Color for Philosophers: Unweaving the Rainbow. Indianapolis/Cambridge: Hackett.
Hardin, L. 1997. “Reinverting the Spectrum.” In Byrne and Hilbert, 1997.
Harman, G. 1990. “The Intrinsic Quality of Experience.” Philosophical Perspectives, 4. 31-52.
Hilbert, D. 1987. Color and Color Perception: A study in Anthropocentric Realism. Stanford University: Center for the Study of Language and Information.
Hilbert, D. and Kalderon, M. 2000. “Color and the Inverted Spectrum.” In Steven Davis, ed., Color Perception: philosophical, psychological, artistic and computational perspectives. Oxford: Oxford University Press.
Levine, J. 1983. “Materialism and Qualia: The Explanatory Gap.” Pacific Philosophical Quarterly, 64. 354-361.
Levine, J., 1993. “On Leaving Out What It’s Like.” In M. Davies and G. Humphreys, eds., Consciousness. Oxford: Blackwell. 121-136.
Levine, J. 2001. Purple Haze. Oxford: Oxford University Press.
Loar, B. 1997. “Phenomenal States.” In Block, N., Flanagan, O., and Guzeldere, G, eds., The Nature of Consciousness. Cambridge: MIT Press.
Lycan, W.G. 1996. Consciousness and Experience. Cambridge: MIT Press.
Moore, G.E. 1922. “The Refutation of Idealism.” In Philosophical Studies. London: Routledge & Kegan Paul.
Sellars, W. 1963. “Philosophy and the Scientific Image of Man,” in Science, Perception and Reality. London : Routledge & Kegan Paul. 1-40.
Shoemaker, S. 1994a. “Intrasubjective/Intersubjective.” In Shoemaker, The First-Person Perspective and Other Essays. Cambridge: Cambridge University Press. 141-154.
Shoemaker, S. 1994b. “Phenomenal Character.” Nous 28. 21-38.
Shoemaker, S. 1994c. “Self-Knowledge and ‘Inner Sense.’” Philosophy and Phenomenological Research, LIV. 249-314.
Shoemaker, S. 2001. “Realization and Mental Causation.” In Gillet, C. and Loewer, B., eds., Physicalism and Its Discontents. Cambridge: Cambridge University Press. 74-98.
Stalnaker, R. 1999. “Comparing Qualia Across Persons.” In The Philosophy of Sydney Shoemaker, Philosophical Topics, 26. 385-404.
Stroud, B. 2000. The Quest for Reality. Oxford: Oxford University Press.
Van Gulick, R. 1993. “Understanding the Phenomenal Mind: Are We All Just Armadillos?” In Davies, M., and Humphreys, G.W., eds., Consciousness. 137-154.
 See my 1994b and 1994c.
 See Hardin 1988/1993 and Block 1999.
 One could also say here, invoking the view discussed in my first lecture, that where there are such differences the perceptual systems of the different perceivers “select” different relations among properties to be the relations of hue similarity, and perhaps in some cases select different properties to be the colors. This is compatible with what is said here in the main text. What relations are selected to be the relations of color similarity will be constrained by the structure of the subject’s color experience space, and the structure of this space also contributes to determining what appearance properties the person perceives in having the experiences. But notice that I say “contributes to determining,” not “determines.” On my view creatures whose color experience spaces are identically structured, and which have selected the same properties and relations to be the colors and relations of color similarity, could differ in the phenomenal character of their experiences and so in what appearance properties they perceive by means of them.
 I should mention an objection (which I owe to Zoltan Szabo) to the view that appearance properties, whether occurrent or dispositional, are among the properties represented by perceptual experiences. It seems a plausible principle that experience represents an object as having the property P just in case the object appears to be P. But let P be an appearance property. Then according to this principle, experience represents an object as having P just in case the object appears P. But if the object appears P, it has the appearance property of appearing P, and on my view the experience represents it as having this property. But the principle tells us that the object must appear to have this second-order appearance property; and now it must have a third-order appearance property which the experience represents it as having. And so on, ad infinitum.
One response to this would be to say that the general principle invoked in this argument, that an experience represents an object as having P just in case the object appears to be P, holds only for cases in which P is an “objective” property, one that things have independently of how they appear. If this is right, the regress doesn’t get started, since appearance properties are not objective properties. Another response would be to allow that the principle holds generally, but to hold that appearing a certain way and appearing to appear that way are identical, and thus that the regress is only apparent.
 Tye, 2000.
 See my 1994a.
 See Stalnaker, 1999.
 See Levine 1993 and Levine 2001.
 See Loar, 1997.