114

Shifted Spectra The paper now shifts gears to two arguments for the possibility of spectra that are inverted or shifted. I have not said what sense of ‘possible’ is at issue. The relevant notion, I think, is metaphysical possibility, but there is no need to go into the matter because I will be arguing for a stronger thesis. In the case of shifted spectra, my conclusion will be that there is reason to think they are actual and in the case of inverted spectra I will be arguing for nomological possibility withan eye to actuality. Thus far, I have argued that we do not take our attributions of looking red to be shown false or otherwise problematic by the discovery of qualia, thus indicating that our tacit semantic policy is to key “looking red” to intentional contents rather than qualitative contents. But I have not yet argued that there really are states with qualitative contents, that is I have not argued for qualia. In this section I argue that individual variation in the phenomenology of color visionWittgenstein and Qualia / 87 at least sometimes goes beyond variation in the use of public color terminology. I argue that this fact can be used to argue for the existence of qualia and further that its appreciation by us provides further support for the claim of the last section that attribution of looking red is keyed to intentional contents rather than qualia. (See (Block, 1990, 1999) and Putnam’s example of the Ixxzians (Putnam, 1999, p. 162–166)). The first premise in the argument is the fact that there is enormous variation in the physiology of color vision in the normal population which leads to enormous variation in color phenomenology. For example, there are large differences among normal perceivers in peak sensitivity of cones in the retina. For example, there is a 51.5%/48.5% split in the population of two types of long wave cones that differ by 5–7 nm, roughly 24% of the difference between the peak sensitivities of long and middle wave cones. ((Neitz & Neitz, 1998)) This characteristic is sex-linked. The distribution just mentioned is for men. Women have smaller numbers in the two extreme categories and a much larger number in between. As a result, the match on a common test “most frequently made by female subjects occurs where no male matches” ((Neitz & Jacobs, 1986), p. 625). (See Block (1999) for further explanation.) Further, variation in peak sensitivities of cones is just one kind of color vision variation. In addition, the shape of the sensitivity curves vary. These differences are due to differences in macular pigmentation, whichvary with“bothage and degree of skin pigmentation” (Neitz & Jacobs, 1986, p. 624). Hence races that differ in skin pigmentation will differ in macular pigmentation. There is also considerable variation in amount of light absorption by pre-retinal structures. And this factor also varies with age. I emphasize gender, race and age to stifle the reaction that one group should be regarded as normal and the others as defective. Premise 1 said that that there is enormous variation in the physiology of color vision in the normal population which leads to enormous variation in color phenomenology. Premise 2 says that the variation in the phenomenology of color vision must vary even among those whose use of color terms is the same. For the borders for our application of color terminology to objects must to some extent be constrained by needs of communicating and cooperating withother people. Imagine, contrary to fact, that we were all given rigorous training in the application of accepted color terminology to the point where variation in application of color terms among us was minor. Things we almost all categorize as ‘borderline between red and purple’ will still look quite different to us, just as, in the inverted spectrum scenario, a fire hydrant that the members of the inverted pair both categorize as ‘red’ looks different them. My premise 2 is that to some extent, that is the way things are. That is, color terminology is to some extent subject to social and linguistic constraint. How do we know that? Normal people see colors as having the same similarity relations. In the Farnsworth-Munsell 100 hue test, subjects are asked88 / Ned Block to arrange 100 chips in a circle with one chip fixed as the starting chip. Normal subjects make nearly the same arrangement (Hilbert & Kalderon, 2000). One of the things social and linguistic influence does is to impose a set of categories on that similarity space. Before the introduction of oranges into England, the color we call ‘orange’ was considered a shade of ‘red’. (The OED’s first listed use of ‘orange’ as the name of a color is from 1512.) Cultural and linguistic influences no doubt affect the borders of our color categorizations. And of course culture also affects our shade categories. I am sure that those who were school age in the USAafter 1949will agree pretty much on the forty shades of Crayolas introduced then and in use (with more added and a few subtracted in 1958, 1972, 1990, 1993 and 1998) since then, such as turquoise blue—despite variation in the way those colors look to us. The upshot is that there are likely to be many normal cases of same use of public color terminology, different phenomenology—and that is what I am calling shifted spectra. If red things look slightly different to you than to me, there is no saying that either of us perceives more veridically than the other, since we are both normal perceivers, and so there is no way of capturing the difference in external terms. (A number of replies are canvassed in my (2003).) So there are qualia. Anobjection to this line of reasoning: Red things may look slightly different to you than me, but that difference may be describable in ordinary language. For example, a red thing may look yellower to you than to me. Here is an argument to support the objector. Many people report slightly different color vision in their two eyes, e.g. things look yellower through one eye than the other. (Hilary Putnam mentioned that he sees colors differently in different eyes at the Putnam at 80 conference mentioned in footnote 1.) Suppose one of Putnam’s two eyes could be copied and inserted in an eyeless person, Oscar, while the other was inserted in another eyeless person, Elmer. If Putnam sees yellower through the eye that Oscar got, we have reason to think that Oscar sees yellower than Elmer. So here is a difference that is effable. Suppose Oscar and Elmer are looking at a wall painted an orange that is perfectly balanced between yellow and red. One possibility is that Elmer says it is pure orange but Oscar says the wall is slightly on the yellowish side of orange. However, another possibility—and one that my argument of this section militates towards as a reality and not just a possibility—is that both Oscar and Elmer say and think that the wall is pure orange. One would expect that if both had been trained on Crayolas! And there is no reason to suppose that they way it looks to them deviates systematically from their thought and talk. The situation is then that they both see it as pure orange but one sees it as yellower than the other. The best way to understand how that could be is to acknowledge that there are ways of seeing orange and that those ways can differ from person to person. The objection that we are considering notes that we can compare these ways in ordinary language. And here is the reply: yes we can compare them in ordinary language, but that does not show that we can fully capture these contents inWittgenstein and Qualia / 89 ordinary language. Oscar and Elmer differ in their color vision, but they may have equal claim to see orange as orange. Points similar to the ones mentioned here were made in Block (2003) where I was arguing (as against representationism) that the phenomenal character of experience outruns its representational content. I claimed that two phenomenally slightly different experiences could nonetheless represent the same color (in different phenomenal ways), even the same minimal shade. Much of that paper was concerned with the issue of whether the phenomenally different experiences that represented the same color could nonetheless be said to have different representational contents. However, the point argued for in this paper is weaker as can be seen by noting that even if the slightly phenomenally different experiences are also different in representational content, they are still qualia if those representational contents cannot be expressed in public language. That is, even if my argument against representationism fails, the argument for qualia in the present sense might succeed. The picture that emerges is that there is a structured space of color qualia that determines the structure of real world colors themselves. Normal people have pretty much the same similarity relations in their structured spaces (as shown by the experiment with the 100 hue test mentioned earlier) but differ in correspondences between that space and colors in the world, probably because of variation in the retina and in pre-retinal structures. Relations among color qualia are determined by the shared structured qualia space, and so those relations can be expressed in terms of the colors themselves (e.g. “yellower”) but there is no objective “location” of color qualia in real world color space because of the differences among people in correspondence between the space and actual colors.20 The point about qualia being comparable but not expressible in ordinary language applies also to inverted spectra. In the Oscar/Elmer shifted spectrum case, Oscar and Elmer differ by a rotation of the color circle—slightly towards yellow in Oscar. In the usual ROYGBIV depiction of the color circle, that corresponds to a clockwise rotation. In the inverted spectrum cases to be discussed below, one twin differs from the other by a 180◦ rotation, which is the same in both directions. These points will be the basis of my discussion below in Section 12 of the “Frege-Schlick” view. To summarize: the Putnam two eye example suggests that qualia can at least sometimes be compared across people. One person’s color experience is yellower than another’s or one person’s color experience is “complementary” to another’s. Does this show that the contents are expressible in public language after all and so are not qualia? No: qualia can be compared in public language without being fully capturable in public language. Even if your experience is yellower than mine, it can nonetheless be the case that there is no color term F such that your experience of red things is describable as “looks F”.90 / Ned Block 8. Behavioral Indistinguishability My arguments for qualia on the basis of the shifted and inverted spectrum do not presuppose that the shifted or inverted perceivers are behaviorally indistinguishable from one another. For example, the 51.5%/48.5% split in the population that I mentioned could, for all I know, determine the extent to which warm colors advance or are exciting compared to cool colors. If half of mankind is inverted with respect to the other half, the inversion shows the existence of qualia even if the inverted spectrum is behaviorally detectable—unless one half is thereby seen to be abnormal or unless the behavioral difference undermines the claim of inversion. To see this, note that the little dialectic I rehearsed earlier about who has authority over the word ‘red’ would apply to the two normal groups as well as to two normal individuals. Neither group could pretend to be “the” group for which red things look red. If it is alleged that for members of one group, red things look green, the other group can complain that the situation is relevantly symmetrical, whether or not there are differences in which colors advance and which recede. Virtually all arguments that inverted spectra are impossible that I have read or heard appeals to one or another sort of asymmetry in color space.21 (I am talking about arguments that spectrum inversion is impossible not that it makes no sense or is otherwise problematic as alleged in the arguments by Robert Stalnaker that will be considered later.) For example, Bernard Harrison (Harrison, 1973) argues that there are more colors between red and green going one way around the color circle (via blue) than going the other way (via yellow). A far more significant asymmetry is that the most saturated colors differ in their lightness levels from color to color. If you decrease the lightness of a sample of yellow you get a different color (brown), but the same is not true for green or blue. The most saturated reds are darker than for other primary colors. Further, desaturated red is pink but desaturated green is green, so hue is not symmetrical withregard to saturation.22 These are not just differences in categorization, but also differences in similarity. A light and dark blue look much more similar to one another than do light yellow and dark yellow, i.e. brown (Hilbert & Kalderon, 2000). Just how significant such asymmetries are is hotly debated. Stephen Palmer (1999) discusses a far larger number of ways in which the color space can be mapped onto itself than had appeared in the prior literature, some of which may avoid such problems although Justin Broackes (2007) argues otherwise. In addition, as Shoemaker notes, it is not clear why minor asymmetries ought to be significant given that we could imagine a slight variant of human color vision in which those asymmetries are absent but color experience is much the same as ours. But this argument is unsound according to Daniel Dennett (1991). Dennett notes that red is advancing, warm and exciting, whereas green is receding, cool and calming, claiming that such asymmetries in function are what make the colors look different from one another. So, according to Dennett, Shoemaker’s imagined race for whom these “minor” asymmetriesWittgenstein and Qualia / 91 are ironed out, could not have color experience much like ours. Others (Block, 1994) argue that these differences in reaction may not be intrinsic to color experience, their widespread presence in different cultures being explained by environmental regularities that are not necessary at least to the basics of color experience.23 But this whole set of controversies can be sidestepped—at least if the existence of qualia is the issue. For as I have noted, the argument for qualia that I am talking about depends not whether the inversion or shift is behaviorally indetectable but whether the groups that are inverted or shifted with respect to others are normal. If it turns out—and as I argued, the evidence points in that direction—that a person’s spectrum is very often shifted with respect to his neighbor, I don’t think anyone would conclude that this shows that defective color vision is widespread. In the rest of this paper, I will discuss an argument that an inverted spectrum is possible. I will focus on an intrasubjective inverted spectrum because of the advantage I mentioned earlier. I will be considering an inversion in which the inverted pair are roughly functionally equivalent but not necessarily exactly functionally equivalent. For example, perhaps one member of the pair experiences red things as exciting and advancing whereas the other experiences them as calming and receding. (Earlier versions are in (Block, 1990, 1994).) I say that an argument that an inverted spectrum is possible is an argument for qualia. “But doesn’t a possible inverted spectrum show at most that qualia are possible rather than actual?” The argument from the possibility of inverted spectra to qualia is this. Suppose an inverted spectrum is possible but nonactual. If we manage to create an inverted spectrum (a procedure that may work is described below), it would be wrong to suppose that we have created qualia where they did not already exist. The fundamental nature of experience does not change because we create a situation that exhibits a fundamental feature of it! So the possibility of an inverted spectrum is sufficient for the existence of qualia.