8

Smart’s Analogy Now, the epistemology of color similarities and differences has received considerable attention from some physicalists who are aware that their theories appear unable to account for it. Because these physicalists subscribe to Russellian or identity-theoretic versions of physicalism, they are committed to the proposition that visual experience doesn’t characterize objects in terms that would reveal wherein their color properties consist. The problem is that if visual experience doesn’t reveal wherein colors consist, it cannot reveal wherein they are essentially alike or different. In order for visual experience to represent how being red is essentially similar or dissimilar to being orange, it would have to represent what it is to be red or to be orange-which it doesn’t do, under the terms of the theories in question. These theories therefore seem unable to explain why the similarities and differences among colors can be known on sight. Physicalists have attempted to meet this challenge by disputing its premisenamely, that visual experience would have to represent the nature of color properties in order to reveal their similarities and differences. They insist upon “the possibility of being able to report that one thing is like another, without being able to state the respect in which it is like.”26 J. J. C. Smart once offered an analogy to illustrate this possibility. He wrote: If we think cybernetically about the nervous system we can envisage it as being able to respond to certain likenesses … without being able to do more. It would be easier to build a machine which would tell us, say on a punched tape, whether or not … objects were similar, than it would be to build a machine which would report wherein the similarities consisted. David Armstrong quotes this passage in application to color similarities and concludes, “No epistemological problem, then.”27 What Armstrong seems to be suggesting is that one detects the bare fact that red and orange are similar by means of a sensory mechanism that responds to their similarity and produces an awareness of it in one’s mind. This similaritydetecting component of the visual sensorium is what corresponds, in Armstrong’s view, to the similarity-detecting machine described by Smart.28 Unfortunately, such a detector, though perfectly conceivable, would not yield the right sort of knowledge about color similarities. For if the similarities among colors were detected by sight, then one’s knowledge of them would be defeasible, by evidence of an optical illusion or malfunction. The experience of seeing things as red and as orange would reveal that these colors looked similar, and hence that they were similar if one’s eyes could be trusted; but one would have to acknowledge the possibility that their apparent similarity might be an illusion, and that they might not be similar, after all. In reality, of course, the similarity between red and orange is known beyond question and could not turn out to be an illusion. One needs to have seen red and orange in order to know that they’re similar, of course, but only because one needs to have seen them in order to know which properties they are. Once acquainted with them, one doesn’t depend on visual evidence for one’s knowledge of their similarity, since nothing would count as counter-vailing evidence. Armstrong’s Analogy Armstrong has suggested that one’s ability to perceive color similarities without perceiving their bases is analogous to the ability to perceive family resemblances: How can we be aware of the resemblance and the incompatibility of the colourshades, yet be unaware of, and have to infer, the nature of the colour-properties from which these features flow? The answer, I take it, is in principle the same … as for the cases where resemblance of particulars such as faces is observed but the respect of resemblance cannot be made out. Despite the fact that the respect in which the faces resemble one another is not identified, it can still act upon our mind, producing in us an awareness of resemblance.29 Now, if we follow Armstrong’s instructions to interpret this analogy as comparing the perceived similarity of color properties to the perceived similarity of particular faces, then it does nothing to overcome our stated objection. Although one can often see that two faces are alike, one remains aware that the appearance of likeness may be illusory, and hence that the faces may turn out not to be alike, after all, whereas the appearance of similarity between red and orange is not subject to empirical refutation. Yet Armstrong’s analogy is open to a slightly different interpretation, which might seem to suggest a case in which knowledge of bare resemblance need not be defeasible, either. Let the similarity between colors be compared not to that between particular faces but, rather, to that between the contours that the faces appear to have, which are properties rather than particulars.30 When the perception of family resemblance is thus interpreted as the perception of similarity between complex shapes, it no longer seems exposed to the risk of illusion. The faces may not have the shapes that they appear to have, of course, but the similarity between those shapes remains unmistakable, even though one may not be able to articulate the respects in which they’re alike. Why, then, can’t the similarity between perceived colors be equally unmistakable and yet equally unanalyzable? The problem with this version of Armstrong’s analogy is that one’s ignorance of the respects in which perceived shapes are alike is not analogous to the ignorance that one would have of color similarities if Russellian or identitytheoretic physicalism were true. Although one cannot say what’s common to the contours that two faces appear to have, one sees those contours under modes of presentation that represent their nature, since shapes are spatial properties and are visually characterized in spatial terms.31 Information about the aspects in which shapes are similar is therefore included in the introspectible content of their visual appearance. One may just be unable to isolate that information or extract it or put it into words. Under the terms of Russellianism or identity-physicalism, however, one’s inability to tell what colors have in common isn’t due to the difficulty of processing information contained in their visual characterization; it’s due to the absence of that information, since colors aren’t characterized in terms that represent their nature. The difference between these cases is like that between purely referential concepts, which have no sense, and concepts whose sense is difficult to explicate. If one has the concept of gold without being able to say what gold is, the reason may be that having the concept consists in nothing more than standing in the right causal relation to the appropriate objects. But if one has the concept of compassion without being able to say what compassion is, the reason is probably that one’s concept has a de dicto content that one cannot immediately explicate. Thus, reflection on one’s concepts of compassion and pity may not reveal how compassion and pity are alike, any more than reflection on concepts will reveal the relation between gold and silver-but not for the same reason. In the case of gold and silver, the reason will be that one’s concepts simply don’t reflect the basis of similarity; in the case of compassion and pity, it will be that a relation reflected in one’s concepts isn’t easy to articulate. This difference is manifested by differences in one’s authority about proposed accounts of the relevant objects or similarities. If someone proposes an account of what gold is, or how it is like silver, one cannot confirm his account simply by consulting one’s concepts. But if someone proposes an account of what compassion is, or how it is like pity, reflection on one’s concepts may indeed suffice to reveal whether he’s right, even if it wouldn’t have enabled one to formulate the account on one’s own. To judge by this test, the visual representation of shape is like a concept that’s difficult to explicate, since one can indeed confirm an account of the resemblance between two faces by reflecting on how they look. There is thus good reason to believe that one’s knowledge of family resemblance depends on visual information of a sort that is not contained in the appearance of colors, as understood by Russellian or identity-theoretic physicalism. One does see the respects in which two faces are alike, although one may be unable to isolate or describe them, whereas the versions of physicalism under discussion imply that the respects of similarity between colors are utterly invisible. Hence one’s ability to be certain about family resemblances is no indication that one could be equally certain about color resemblances if these versions of physicalism were true. Explaining the Epistemological Intuitions Away We believe that the foregoing epistemological objections rule out any theory that portrays visual experience as representing colors contingently-that is, without characterizations that denote them necessarily. They thereby rule out Russellian versions of physicalism and Fregean identityphysicalism as well. Although such theories cannot respect ordinary intuitions about the epistemology of color, some of them can attempt to explain those intuitions away. In particular, any physicalist who acknowledges the existence of distinctive color sensations, or qualia, can argue that we have mistaken introspective knowledge about those sensations for knowledge about the color properties that they help to represent. What the ordinary observer knows by reflection, this physicalist may claim, is not that there are distinct but similar properties that red-looking and orange-looking objects appear to have but, rather, that there are distinct but similar sensations that accompany these appearances. According to this response, we have displaced-indeed, projected-these items of knowledge from their true objects, which are color qualia, onto color properties. But can the physicalist extend this explanation to our most fundamental knowledge claim, that color experience can be known on reflection to represent properties? He can try. For he can say that we have mistaken the introspectible presence of color qualia in visual experience for an introspectible representation of color properties. Because reflection on visual experience does reveal that things look colored in the sense that their visual appearances are accompanied by color sensations, the physicalist may argue, we have mistaken it as revealing that they look colored in the sense of being represented as having color properties. But why would we commit this mistake in the case of color, when we have no tendency to commit it in the cases of other, equally vivid sensations? One isn’t tempted to think that sensations of pain, for example, attribute any properties to the objects that cause them. Reflection on the experience of being pricked by a pin doesn’t yield the conviction that the pin is being represented as having a pain-property. Why, then, should reflection on an experience accompanied by a color sensation yield the conviction that its object is being represented as colored? Here again the physicalist may think that he has an explanation. For as Wittgenstein pointed out, sensations of pain, unlike sensations of red, are not regularly received from particular objects or surfaces; if they were, “we should speak of pain-patches just as at present we speak of red patches.”32 Perhaps, then, we believe that visual experience attributes color properties to objects because we’ve observed the regularity with which color sensations are associated with the perception of particular surfaces. According to this explanation, the knowledge that we have claimed to possess on the basis of mere reflection is in fact derived from observed patterns and correlations within visual experience.33 Unfortunately, the patterns and correlations cited here would provide no grounds whatsoever for believing that visual experience attributes color properties to objects in the ways required by Russellian or identity-theoretic physicalism. From the fact that particular objects are individually associated in visual experience with a particular sensation, no conclusion can be drawn about whether the sensation has any normal or predominant cause, and hence about whether there is an external property that it can help to represent. Various objects regularly occasion sensations of red, but those objects are so various that they may not have any surface properties in common, for all one can tell from visual experience. Hence their observed association with one and the same quale provides no grounds for thinking that the quale has any informational potential. What’s more, the association of color sensations with particular objects is no more regular or reliable than that of pain with particular kinds of events. After all, pain serves its monitory function only because young children can learn that it regularly accompanies bumps, scrapes, punctures, encounters with extreme heat or cold, and so forth. Having obvious external correlates is essential to the evolutionary purpose of pain. If what led us to view a sensation as the representation of something external were its observed correlation with various external stimuli, we would have no more occasion to take this view of color than of pain. Thus, the point of Wittgenstein’s remark about pain patches cannot be that pain appears to have no representational content because it has no apparent external correlates. What, then, is the point? Surely, it’s that sensations like pain (and color) involve qualities that we can easily think of as located in the external world, but that this thought is blocked, in the case of pain, by there being no particular places where it seems to be located. The external correlates of pain aren’t places, and so pain isn’t subject to the sort of displacement that the mind practices on other sensations. Thus, what the association of sensations with particular surfaces produces, and what Wittgenstein was suggesting that it would produce even in the case of pain, is a tendency to perceive the sensations as located on those surfaces-an inducement, in short, to the projective error. But the result of this error is precisely that the qualia themselves, rather than microphysical properties, are attributed to objects in visual experience. Thus, if the patterns cited by the physicalist have their most likely result, they result in the falsity of physicalism as an account of the properties that visual experience represents. The physicalist explanation of our basic epistemological intuition is therefore unstable. The physicalist wishes to claim that visual experience does not project sensations onto external objects, as their perceived properties, but that reflection on visual experience does project our knowledge about sensations onto objects, as knowledge about their perceived properties. What is cited as accounting for the latter projection doesn’t really account for it, however, and would in fact account for the former projection instead.