7

EPISTEMOLOGICAL CONSTRAINTS What do you know about colors, not as a student of physics or physiology, but simply in your capacity as a subject of visual experience? We think that you know, for example, that red and orange are properties; that they are different properties, though of the same kind-different determinants of the same determinable; that they are not as different from one another as they are from blue; and that they cannot simultaneously be instantiated in exactly the same place. Finally, you know that red and orange are properties that things visually appear to have, and you know when things appear to have them. All but the last two items of knowledge are necessary propositions. Red and orange-that is, the properties that things appear to have in looking red and in looking orange-not only are distinct, similar determinates of the same determinable but are essentially so. A property that wasn’t a determinate of the same determinable as red, or wasn’t distinct from red, or wasn’t similar to redsuch a property simply wouldn’t be orange. And vice versa. What’s more, mere reflection on color experience provides all the support that might ever be needed for all of the knowledge cited above. That is, you need only reflect on the experiences of seeing things as red and as orange in order to know that they are two distinct, incompatible, but rather similar determinates of a single determinable property; you need only reflect on particular experiences in order to tell which of these properties they represent; and there are no possible circumstances under which more evidence would be needed. We wish to remain neutral on the explanation for this phenomenon. The knowledge in question may be delivered in its entirety by introspection on the contents of the relevant experiences. Alternatively, it may require the recognition of relations among the contents of these experiences, so long as the relations are such as can be recognized a priori. It may even require empirical support, so long as the support required is no more than what’s provided by the experiences themselves. All we claim is that the experiences of seeing red and orange provide whatever is necessary for this rudimentary knowledge about those properties. Consider the consequences of denying that your knowledge about colors has this status. If the experiences of seeing red and orange didn’t provide all of the support required for the knowledge that they’re distinct but similar determinates of the same determinable, then your knowledge of these matters would be hostage to future empirical discoveries. You would have to consider the possibility of obtaining evidence that red and orange are in fact the same property or, conversely, that they aren’t similar at all. And given how the references of “red” and “orange” are fixed, evidence that red and orange are the same property, for example, would amount to evidence that the property that objects appear to have in looking red is the same as the property that they appear to have in looking orange. Does visual experience leave room for the hypothesis that things appear to have the same property in looking red as they do in looking orange? We think not. Nor does it leave room for the hypothesis that red and orange are less alike than red and blue, or that something seemingly seen as red on a particular occasion is being represented as having a property other than red. Your knowledge on these matters is such that nothing would count as evidence against it. Meeting the Epistemological Constraints Yet would such knowledge be possible if physicalism were true? We believe that the answer may be yes in the case of Fregean realization-physicalism, but that in the case of all other versions of physicalism-that is, Russellian theories and identity theories-the answer is no. What sets the latter theories apart from Fregean realization-physicalism is their implication that visual experiences like yours represent colors only as a matter of contingent fact. Under the terms of these theories, an experience internally indistinguishable from your experience of seeing something as red might fail to represent its object as having that color. The reason is that red is represented by your experience, according to these theories, only by virtue of facts incidental to the internal features of the experience. Which facts these are depends on the physicalist’s conception of visual representation. Under the terms of Russellianism, they are the causal or correlational facts by virtue of which some mental item, or some behavioral disposition, introduces the microphysically constituted property red into the contents of experiences. Twin-earth examples, in the style of Putnam,’s will readily demonstrate that the same mental item or the same classificatory behavior might have been correlated with objects of a different kind, sharing a different property-in which case, internally similar experiences would not have represented the property that, according to physicalism, is red. Under the terms of Fregeanism, the facts in virtue of which visual experience represents a microphysical property are the facts in virtue of which instances of that property uniquely satisfy the characterization by which things are visually represented as red. And these facts, too, are bound to be contingent if red is identical with a microphysical property, for reasons illuminated by the naive objection discussed above. Although the naive objection cannot defeat physicalism, it does force the Fregean identity-physicalist to concede that the characterization by which things are visually represented as red does not represent what it is to be red. For as an identity -physicalist, he believes that to be red is to have a particular microphysical property, and yet the objection forces him to concede that things aren’t seen under microphysical characterizations. The Fregean identityphysicalist must therefore believe that things are seen as red by means of a contingent characterization-a characterization that is, in fact, uniquely satisfied by instances of the property red, but not because it represents what redness is. And twin-earth examples will once again demonstrate that such a characterization might not have been uniquely satisfied by instances of red or might have been uniquely satisfied by instances of another property. Just as a mental symbol might have tracked a different property, so the visual characterization “whatever causes this feeling” might have been satisfied by a different property; and in either case, your visual experiences wouldn’t have represented red, under the terms of the corresponding theory. Thus, Fregean identity-physicalism is like Russellian physicalism in implying that your experience of something’s looking red might have been exactly as it is, in all respects internal to you, while failing to represent anything as red. And this consequence has the corollary that there are circumstances under which you couldn’t tell, by mere reflection on the experience of something’s looking red, whether it is being represented as having the property red. The physicalist may object, at this point, that something’s being contingent doesn’t entail its being a posteriori. He will argue, more specifically, that the reference of “red” has been fixed for you by a description alluding to your visual experiences: red is, by stipulation, whatever property is attributed to objects by their looking red. That red is the property that something appears to have in looking red is therefore knowable a priori, even though it is contingent, just like the length of the standard meter-bar in Paris.19 This response misses the epistemological point. The term “red” has been stipulated as denoting the property attributed to objects by their looking red; but the phrase “looks red” has been stipulated as denoting experiences introspectively similar to some paradigm experience. The problem is that under the terms of the theories now in question, there is no introspectively recognizable kind of experience for which you can always tell by introspection whether the same property is represented in all or most experiences of that kind. These theories therefore imply, to begin with, that, for all you know by reflection on visual experience, the attempt to fix the referent of “red” as the property attributed to objects by their looking red may have failed, since there may be no property that predominantly satisfies that description. They imply furthermore that, even if there is a property represented by most instances of things’ looking red, you cannot necessarily tell by reflection when a particular experience is representing that property. This problem can best be illustrated by imaginary cases of context-switch- ing.20 Suppose that your environment were to change in such a way that your mental designator for red was correlated with, or your visual characterization of red was satisfied by, a new and different property that replaced the current property red wherever it occurred. At first the content of your visual experiences might remain the same, with the result that you saw objects as having a property that they no longer had. But gradually your visual designators or characterizations would come to denote the new property rather than the old. Tomatoes would therefore appear to have a new and different color property-appear to have it, that is, in the only sense in which a Russellian or an identity -theorist can conceive of them as appearing to have any color at all. Yet in all respects internal to you, your experiences would remain unchanged, and so you would be unable to tell by reflection that you were no longer seeing tomatoes as having the color property that you had previously seen them as having. Russellianism and identity -physicalism therefore entail that without investigation into the physical causes and correlates of your visual experiences, you cannot necessarily know whether tomatoes appear today to have a different color property from the one that they once appeared to have. You might know that whatever property they appear to have is likely to be the current holder of the title “red,” if any property is. But you may not be able to tell when things have appeared to have that property in the past; and you may not be able to tell in the future when things appear to have it. Hence there remains a significant sense in which you don’t necessarily know when things appear to be red.2′ Indeed, these theories entail that you cannot always tell without investigation whether objects appear to have any color properties at all. For just as experiences internally indistinguishable from yours might represent different properties, so too they might simply fail to represent properties. Such a failure would occur if the characterizations applied to objects in visual experience were not satisfied, or if the corresponding mental designators were not systematically correlated with visual stimulation from objects of any particular kind. Consider the Russellian theories, which say that visual experience represents objects as colored by means of symbols or behavioral dispositions that designate microphysically constituted kinds. Reflection on such an experience wouldn’t necessarily reveal whether the symbols being tokened, or the behavior being prompted, were correlated with objects sharing a common property and constituting a genuine kind. For all one could tell from having the experience, the objects associated with the symbol or behavior might be utterly miscellaneous, and so these purported designators might not indicate membership in a kind or possession of a property. Hence one would be unable to tell, when things looked red, whether there was a property that they thereby appeared to have. And if one didn’t know whether things appeared to have a property in looking red, one wouldn’t know whether there was such a property as red at all. The same problem attends the Fregean theory, in all but its realizationtheoretic form. According to Fregean identity-physicalism, as we have developed it, visual experience represents red by characterizing it as the property that normally causes a particular sensation. Yet reflection on a visual representation of this form would not necessarily reveal whether there was a property that normally caused the sensation, and so it wouldn’t reveal whether the associated characterization succeeded in denoting a property. Of course, the possibility of there being no colors represented in visual experience is only the most bizarre of many possibilities that introspection could not rule out if the present theories were true. A less bizarre possibility is that visual experience might represent only two color propertiesone when things look either red, orange, or yellow, and another when they look either green, blue, or violet. The correlational or causal facts could certainly be arranged in such a way as to give these experiences one of only two contents, under the terms of Russellian or identity-theoretic physicalism. These theories therefore imply that one cannot always tell without investigation whether red and orange are different colors, the same color, or no color at all. Some Defenses and Replies Now, physicalists sometimes admit that visual experience, as they conceive it, is compatible with the possibility that there are no colors.22 We wonder, however, whether the full import of this concession is generally appreciated. The statement that there may be no colors sounds as if it should gladden the heart of an anti-realist, but it is in fact different from, and perhaps even incompatible with, the views that many anti-realists hold. What these anti-realists believe is that colors are properties that visual experience attributes to objects even though no objects instantiate them. What proponents of the present theories must concede, however, is that there may be no properties attributed to objects by their looking colored, and hence that there may be no such properties as colors, not even uninstantiated ones. They must allow that color experience not only may attribute properties to objects that don’t have them, as the anti-realists claim, but may actually fail to attribute properties to objects at all, by failing to express any properties. If this possibility were realized, color experience would lack the representational competence required to be false, strictly speaking, whereas the falsity of color experience is what anti-realism is usually about. And in our opinion, the fact that color experience can at least be false is evident on the face of it. A physicalist might respond that the designators and characterizations involved in color experience can be assumed to indicate some properties, since something or other is bound to be responsible for one’s visual sensations, as specified in the characterizations, and something or other is bound to be correlated with the designators. But the liberal criteria of visual representation that would enable one to assume that some properties or other were being represented would simultaneously undermine one’s claim to other items of knowledge about those properties.23 For if one’s experiences of things as red and as orange represented whatever properties in heaven or earth were correlated with two different designators, or responsible for two different sensations, then one would be even less able to tell by reflection whether those properties belonged to the same determinable, or required extension for their instantiation, or bore greater similarity to one another than to some third property. For all one could tell from seeing colors in the way imagined here, red might be an electrical charge, orange a degree of acidity, and blue a texture. A physicalist might respond that if the similarities and differences among colors were conceived as relative to an observer, then they would indeed be revealed by reflection on visual experience.24 Let the imperfect similarity between red and orange consist in the fact that they have distinct but similar effects on normal human observers, and any normal human observer will be able to detect their relation on sight. The problem with this suggestion is that it can account only for our knowledge of contingent similarities and differences. Red and orange, as conceived by the physicalist, are properties that happen to have distinct but similar effects on human observers, but they might have had effects that were not distinct or were even less similar. Hence the similarity relation that would be accessible by reflection on visual experience, according to physicalism, is a relation that red and orange might not have had. In reality, however, reflection on the experiences of seeing red and orange tells us that if two properties didn’t stand in precisely this relation, they wouldn’t be the properties we’re seeing.25