Inversions spectral and bright

Comments on Melinda Campbell
Society for Philosophy & Psychology, San Francisco, 31 May 1996


Austen Clark
Department of Philosophy U-54
University of Connecticut
Storrs, CT 06269-2054


Spectrum inversion is a thought experiment, and I would wager that there is no better diagnostic test to the disciplinary affiliation of a randomly selected member of the audience than your reaction to a thought experiment. It is a litmus test. If you find that you are paying close attention, subvocalizing objections, and that your heart-rate and metabolism go up, you have turned pink: you are a philosopher. If on the other hand the thought experiment leaves you cold, and you wonder why otherwise sensible people would worry about such things, you have turned blue and you are a psychologist.

Professor Campbell's paper is an intriguing one to read to such a mixed audience, because she is using the philosophical conundrum of spectrum inversion in a novel way--one which might appeal to psychologists. Her main goal I believe is not to fashion some knock-down answer to the question of whether undetectable inversion is possible, but rather to ask what concepts of color, color experience, and qualia are necessary even to pose the question. So the thought experiment is used, not as intended by the authors--as a tool useful for probing the world knot--but rather as a piece of data about the thoughts--specifically the concepts--of those authors. What sorts of concepts do the people posing such questions employ? We can imagine cognitive ethologists several centuries hence, tip-toeing into an alien ecosystem--stacks filled with old philosophical journals--and using the thought experiments found therein not as a guide to metaphysical insight, but rather as case studies. Even a skeptic could then admit that the thought experiments have some use.

Professor Campbell goes on to suggest that if we replace our pre-theoretic concepts of color with ones that better reflect their true nature, the worries about spectrum inversion will be stilled. I believe she presents two independent arguments for this claim. The first is based on the difficulties of inverting the brightness dimension of color space. The second is the longer and more complicated argument, based on the need to reconceive the representational character of color experience. I will say a few things about the argument from representational character, but will focus mostly on the brightness argument, since about the latter I find it possible to say something both sensible and short.

I

The argument from representational character starts with a careful distinction between "qualitative character" and "phenomenal property". The former are properties of our experience; the latter properties attributed to appearances. For example, the blue appearance of the sky is explained by showing how the sky causes in me an experience that has a particular qualitative character--one in virtue of which the sky looks blue, and not some other color.

Campbell suggests that we should think of experiences of color as one and all intentional (p 12). The experience, with its particular qualitative character, represents the phenomenal property. Only in such "appearance events" or "appearings" are colors exemplified. These appearings are "relational events". Certainly the appearance presented by a surface depends strongly on its relations to neighbors: its relative luminance and spectral composition, and its relations to "global illumination phenomena" (p 5). One cannot identify such relational properties with intrinsic properties of particular objects. Campbell urges that the same morals apply to the states that represent colors. So color properties cannot be identified with any intrinsic properties, whether subjective or objective (p 12). Qualitative character becomes a mere "representational vehicle"; qualia are "empty vessels" of information about appearances. As she says, "Apart from fulfilling its functional role, there is no phenomenal character attributable to the vehicle itself..." (p. 11). This way of conceiving colors "short-circuits spectrum inversion worries" (p 12).

The account becomes very complicated. Consider four claims that are parts of it:

1 Qualitative character is just a variety of representational content.

2 The qualitative character of an experience is exhausted by specification of what it represents. Endnote 1.

3 Otherwise nothing particular is identifiable as the qualitative character of an experience. Endnote 2.

4 Qualia as such are not directly accessible to consciousness. Endnote 3.

I agree that if you subscribe to these claims, you are not likely to be worried about spectrum inversion. I also agree that empirically oriented models of color vision are likely to endorse claims that are at least similar to these.

But the problem is that proponents of spectrum inversion have arguments against all four claims. Those arguments seem relatively independent of the details of how color vision works. For example, we find explicit arguments in Block (1990), Shoemaker (1991), Peacocke (1989), and others that qualitative content is not exhausted by a description of intentional content. We find a recent explicit argument distinguishing "phenomenal consciousness" from "access consciousness" (Block 1995). These philosophers think they are sometimes aware of qualia. Unlike the typical subjects of the cognitive ethologist, these defend their concepts, often with ingenious arguments.

Here is one way to highlight the conceptual dispute. Consider the following description. Two individuals are behaviorally and psychologically indistinguishable--they are "functionally isomorphic" and have identical discriminations--yet some stimulus presents a different qualitative character to one of them than to the other. That's the "spectrum inversion hypothesis". Is it possible? There are two very different senses in which you might say "no":

(1) It is inconsistent with what we know about the operation of the human visual system; or

(2) It is (somehow) conceptually incoherent: it lacks sense or uses the concept of qualitative character in an incoherent way.

The first sense is the sense that attaches to perpetual motion machines or mass traveling faster than the speed of light. They violate laws of nature. It is in this sense that Campbell argues that brightness inversion and hence spectrum inversion is impossible. But many philosophers hanker after the second variety of modal claim. They cheerily admit that inversion perhaps could not take place in us, constituted as we are. But as long as there is no conceptual incoherence in supposing that we might have had visual systems in which inversion could take place, then, in the sense that interests these philosophers, spectrum inversion is still possible. Endnote 4.

II

These issues can be clarified by focusing on the intriguing puzzle of brightness inversion. This would be inversion of the color quality space along its achromatic core: the axis ranging from white, through the gray scale, to black.

Professor Campbell argues that an undetectable brightness inversion is inconsistent with how the human visual system works. Sensations of relative brightness serve two different roles. First, they register contrasts between differing surfaces in a scene. Under constant illumination, the surfaces that look brighter are ones that reflect more light. Practically anything that reflects 70% or more of the incident radiation will look "white", while anything that reflects 4% or less will look black (see Boynton 1979, p 164). Since the nervous system has no independent access to the illumination incident upon such bodies, the only way to pick up such differences is to compare the luminance of a surface--the intensity of radiation reflected from it--with that of its neighbors. A clever design would extract ratios, not differences. If the energy incident on both surfaces is roughly the same, then luminance ratios would yield relative reflectances. In fact this is more or less what the visual system does. Endnote 5. The perceived lightness of a patch is a product of the relation between its luminance and that of its neighbors.

But sensations of relative brightness serve a second role too: they are correlated with the global illumination of the scene. The range of energies with which the visual system must cope is enormous, varying by a factor of 100 million to one. That's the range from sunlight at noon on a clear day, to stumbling around in the woods late at night by starlight alone. Endnote 6.

These two principles of operation of the visual system are not entirely independent. Contrasts between lightness become more salient as luminance energy increases. We can see why. Suppose one surface reflects 15% of incident illumination, and its neighbor 10%. As the lights go up the actual numbers of photons reflected will increase enormously. We get the same ratios, but the larger difference makes the lightness contrast more salient. It also makes it possible to discriminate additional variations in lightness between those two patches. They might both look black in dim light. In bright light one might be able to insert five more gray scale "steps" between them. Endnote 7.

But now we can see why the human visual system could not suffer an undetectable brightness inversion. An inversion cannot satisfy both principles simultaneously. In any scene in which we can see something white, we can see contrasts between white, gray, and black. As the lights dim we can see fewer of such contrasts, until finally everything looks black. But to the brightness invert, everything would look white when the lights are off. So either the brightness invert can make discriminations of lightness contrast with the lights off--rather mysterious!--or the salience of those contrasts to the invert vanishes in scenes where some things look white--violating the other principle.

I agree with Campbell that the hypothesis of an undetectable brightness version is inconsistent with what we know about how the human visual system works. Endnote 8. But now I can return to the question: to what extent will this finding satisfy the various parties to the dispute?

III

Unfortunately not all those parties carry the same baggage. Many arriving from the philosophical districts do not draw their concepts of color from empirical science. They are interested in elucidating our ordinary concepts of color; possibility is limned by consistency with those concepts. As long as brightness inversion is possible in the latter sense, it will not particularly matter for their purposes that a visual system showing such an inversion would have to operate on principles different from our own.

I do not think our ordinary concepts of color or color experience rule out the possibility of brightness inversion. Consider the experience of walking out of a dark theater on a bright sunny day: the brightness is so dazzling that it is hard to register surface contrasts, or to see the colors of the automobiles in the parking lot. At extremes of brightness one loses hue and surface contrast discriminations almost as much as one does in complete darkness. Endnote 9. These extremes are rarely visited, however, because light and dark adaptation proceed so quickly. We need to imagine that our brightness invert, when he walks into a dark theater, has an experience with the qualitative character of the sort that we have when walking out of it into the sun. We stumble on both transitions--losing contrast and hue discriminations--but for different reasons. Sitting in the dark, our brightness invert has experiences of such dazzling brightness that he can barely make out edges, and will stumble around just as we do in the dark. When the lights go up our invert has experiences with the qualitative character of the sort that we have when the lights dim. This helps him see better, because he is less dazzled.

In us sensations of brightness are correlated with the amount of energy absorbed by retinal receptors; in brightness inversion such correlations would need to run in the opposite direction. But there does not seem to be anything incoherent with the suggested alterations. I think it is hard to find anything in our ordinary concepts that a priori eliminates the possibility of brightness inversion, or, more broadly, of spectrum inversion.

So suppose spectrum inversion could not happen in our visual system, organized as it is, but that there is nothing in our ordinary concepts of color which would render incoherent the description of a visual system in which spectrum inversion could happen. What moral should we draw from this finding? It all depends on what relation you think holds between those "ordinary" concepts and ones deriving from empirical inquiry. If you are an "analytic functionalist", intent on giving an analysis of those ordinary concepts, this finding would be damaging if not fatal, since it shows something to be possible which your analysis implies is impossible.

But these days more and more philosophers themselves turn blue if pressed to produce an analytic truth, or to defend the conceptual coherence of something which is physically impossible. A so-called "psycho-functionalist" approach is much more in keeping with the tenets of Professor Campbell's paper. On this our ordinary concepts of colors and color experience point dimly to natural kinds whose real nature is to be revealed eventually by empirical science. Those ordinary concepts manage sometimes to secure reference to important internal states and processes, even if many or most of the platitudes giving those terms their life within ordinary language turn out upon inspection to be false.

Only if you hold something like a psycho-functionalist view can you suggest that empirical inquiry will give us a better grasp of the very same colors and color experiences that we have been talking about all along, in our dim folk way. Endnote 10. Professor Campbell speaks in several places of refining our concepts of color; of providing a "fuller" account or one closer to their true nature. Clearly these refinements derive from empirical investigation. Once we adopt them, she says, the spectrum inversion hypothesis ceases to be a "live possibility".

Now the proponent of analytic functionalism, faced with this prospect, might feel somewhat the way an English speaker feels when confronted with Orwell's Newspeak. True, in Newspeak you are no longer bothered by the problem of deciding what are the limits of legitimate authority of a government. Newspeak lacks the means to express the possibility that there could be any. So the problem won't bother you any more. Similarly, in NewColorSpeak, you will no longer be bothered by the possibility of spectrum inversion. This hardly provides a good reason to abandon English. Fortunately, in the latter conceptual dispute we can provide some good reasons. The more refined concepts of color derive from a successful empirical theory, which orders and explains a large domain of sensory experience. Ultimately that empirical success is the best reason to refine our concepts. Note well: this "refining" eliminates some conceptual possibilities, and adds new ones: it redraws the very map of what is possible. But that has been going on for a long time. We might as well base such maps on what we know.

Back to Austen Clark online papers.

Back to Uconn Philosophy home page.

Endnotes

1. "we can arrive at a fully representational view of color experience..." p 13. Perhaps a partial retraction of this is found on pp 12-13: "Although we may attribute some non-representational character to qualia as representational vehicles, it would be incorrect to think of this as color, since qualia do not appear in perceptual experience: qualia are the mode through which color experience comes to represent appearance properties."

2. "But this is a misguided way to conceive of this perceptual event: there really is nothing in this situation that may be identified with a quale or a qualitative character apart from the particular neurophysiological state that embodies a visual response to pure long wave light." (pp 9-10)

3. "We should think of qualia as representational vehicles, but as such, they are not directly accessible to consciousness ...they are, so to speak, transparent onto that which it is their function to make salient..." p 10

4. Here's an interesting sample, from Shoemaker 1975:

Taken one way, the claim that spectrum inversion is possible implies a claim that may, for all I know, be empirically false, namely that there is a way of mapping determinate shades of color onto determinate shades of color which is such that (1) every determinate shade (including 'muddy' and unsaturated colors as well as the pure spectral colors) is mapped onto some determinate shade, (2) at least some of the shades are mapped onto shades other than themselves, (3) the mapping preserves, for any normally sighted person, all of the 'distance' and 'betweenness' relationships between the colors (so that if shades a, b, and c are mapped onto shades d, e, and f, respectively, then a normally sighted person will make the same judgments of comparative similarity about a in relation to b and c as about d in relation to e and f), and (4) the mapping preserves all of our intuitions, except those that are empirically conditioned by knowledge of the mixing properties of pigments and the like, about which shades are 'pure' colors and which have other colors 'in' them... But even if our color experience is not in fact such that a mapping of this sort is possible, it seems to me conceivable that it might have been--and that is what matters for our present philosophical purposes. For example, I think we know well enough what it would be like to see the world nonchromatically, i.e., in black, white, and the various shades of grey--for we frequently do see it in this way in photographs, moving pictures, and television. And there is an obvious mapping of the nonchromatic shades onto each other which satisfies the conditions for inversion. In the discussion that follows I shall assume, for convenience, that such a mapping is possible for the full range of colors--but I do not think that anything essential turns on whether this assumption is correct. (Shoemaker 1975, pp 301-2).
(Emphasis added.) In "The inverted spectrum" he says
Even if our color experience is not invertible, it seems obviously possible that there should be creatures, otherwise very much like ourselves, whose color experience does have a structure that allows such a mapping--creatures whose color experience is invertible. And the mere possibility of such creatures is sufficient to raise the philosophical problems the possibility of spectrum inversion has been seen as posing. ... If spectrum inversion is so much as a logical possibility--whether or not it is a possibility for us, as we are currently constituted--then it is clear that no behaviouristic account of qualia will do. (Shoemaker 1982, p. 367)

5. It is only "more or less" because it departs from the ratio principle at extremes and in stimulus configurations that are other than simple. See Kaufman 1974, chapter 5.

6. See Schiffman 1982, p. 179. Interestingly, white paper in moonlight has about 100 times the luminance of white paper in starlight alone. Moonlight to sunlight is a ratio of about one to one million. See also Boynton 1979, p 165.

7. Differences between different dark grays all vanish; they all look equally black. Similarly the differences between the whitest white and a very very light gray also vanish. As the lights go down the whitest white looks darker and darker. So differences between the patch that was the whitest and one that was a very light gray disappear. If the gray scale is printed on white paper, you lose the distinction between the lightest end of the gray scale and the white of the paper. The paper looks darker; it comes to match what was the lightest end of the gray scale. Eventually of course everything looks black. In the very dimmest lights things that are non-black look very dark gray; they are not very different from the darkest things one sees. There might be a whitest patch under those conditions of illumination, but it won't look very white. Or at least the differences between lightnesses that were obvious in daylight vanish: what were differing points on the lightness scale all match one another.

What this means is that the distinctions among lightness available under high illumination cannot be mapped into the distinctions available under low illumination. So the arrangement of qualities of color patches under dim lights is incommensurable with the arrangement under bright lights.

(If you look at a piece of paper on your desk in moonlight, and then at gray scale in daylight, it is difficult to place the lightness of the paper under moonlight at some point in the gray scale seen under daylight. That sort of comparison is similar to comparing trichromatic vision with dichromatic. There are no naturally occurring situations where you make a match in lightness across such globally different illumination conditions--though you can match a paper in the shade with one in sunlight.)

Of course eventually the color solid becomes just two dimensional (you lose yellow-blue), and finally just one dimensional: the only distinctions among patches are distinctions along a very short gray scale. (If it is really dark enough, all the patches look exactly the same, and our three dimensional ordering dwindles to a point!) Each quality space is constructed for a class of stimuli. We will actually construct different color solids--with different lightness scales, and differing possible degrees of saturation--for Munsell patches viewed under different conditions of illumination. If it is dark enough they can all be ordered one dimensionally. All the patches at a given lightness will be indiscriminable from one another. So maybe the comparison of lightnesses across illumination conditions is something like a comparison across two quality spaces of differing structure.

8. Perhaps the Bezold-Brücke phenomenon could also be cited as grounds for this conclusion. The apparent hues of most stimuli change if the intensity of the stimulus changes. The only ones that don't are the "invariant hues", and they turn out, lo and behold, to be identical with the unitary hues. The curves through quality space caused merely by altering the brightness of a stimulus do not seem to be invertible. Or at least in a brightness invert the hue shift with changing luminance would have to follow a different law than it does in us. See Hurvich 1981, pp. 72 ff.

9. Even if we are judging surface colors under a condition of constant illumination, the hue circles constrict to a point at the white and the black ends of the color solid. Find the whitest white you can under those conditions of illumination. To find any other patch that matches that white one, it will have to be white as well. Even adding the palest hue will somewhat drop the lightness of the patch; the achromatic patch that matches even the palest yellow or pink is not the whitest white that can be had under those illumination conditions. It will be very slightly darker. A very saturated color will be considerably darker than white. Similarly in order to see the difference between a dark blue and a dark purple, both color patches would have to be slightly lighter than the darkest black one can get under those conditions of illumination. Hue differences only show up among lightnesses that are less than white or more than black.

10. This follows if you admit that some of the platitudes or propositions of folk psychology are false. The analytic functionalist conjoins and "Ramsifies" those propositions to construct the definitions of psychological terms. If any of those propositions or platitudes turn out to be false, then nothing satisfies the Ramsified theory, and there is no such thing as folk color, folk color experience, etc.

References

Block, Ned. (1990) Inverted Earth. In James E. Tomberlin (ed), Philosophical Perspectives, 4. Action Theory and Philosophy of Mind, 1990. Atascadero, California: Ridgeview Press.

Block, Ned. (1995). On a confusion about a function of consciousness. Behavioral and Brain Sciences, 18 (2), 227-88.

Boynton, Robert M. (1979) Human Color Vision. New York: Holt Rinehart and Winston.

Hurvich, Leo M. (1981). Color Vision. Sunderland, Mass.: Sinauer Associates, Inc.

Kaufman, Lloyd. (1974). Sight and Mind. New York: Oxford University Press.

Peacocke, Christopher. (1989). Perceptual content. In J. Almog, J. Perry, and H. Wettstein (eds.), Themes from Kaplan. Oxford: Oxford University Press, 297-329.

Schiffman, Harvey Richard. (1982) Sensation and Perception: An Integrated Approach. 2nd edition. New York: John Wiley & Sons.

Shoemaker, Sydney (1975) "Functionalism and Qualia," Philosophical Studies 27.

Shoemaker, Sydney (1982), "The Inverted Spectrum," Journal of Philosophy, LXXIX, 7.

Shoemaker, Sydney (1991). Qualia and consciousness. Mind 100(4), 507-24.


Return to Austen Clark's online papers .

Return to the Philosophy Department home page.