From Towards a Science of Consciousness 3 Section 2: Color -- Introduction CogNet Proceedings
Is it possible that a person who behaves just like you and me in normal life situations and applies color words to objects just as we do and makes the same color discriminations, see green where we see red and red where we see green? Or, to put the same question from another perspective: Is it possible that you are yourself red-green inverted with respect to all or most other people and that you thus are and have always been radically wrong about what other people see when looking at a sunset or the moving leaves of a tree?
Philosophers normally discuss the possibility of Qualia Inversion by considering thought experiments. But there is, in fact, scientific evidence for the existence of such cases. Theories about the physiological basis of color vision deficiencies together with theories about the genetics of color vision deficiencies lead to the prediction that some people are "pseudonormal." 2 Pseudonormal people "would be expected to have normal color vision except that the sensations of red and green would be reversed-something that would be difficult, if not impossible, to prove." 3 This inversion would affect the perception of any color that contains a red or green component. A greenish blue river would appear violet to a pseudonormal person. Remember, however, that this description will give you a correct idea of what pseudonormal people experience only if you are not yourself one of them. But there is a chance that you are. According to a model of the genetics of color vision deficiencies that was first presented by Piantanida in 1974 pseudonormality occurs in 14 of 10 000 males. 4 It is instructive to see how the prediction of pseudonormal vision follows from the combination of several empirically supported assumptions.
There are three types of photoreceptors on the retina that play a central role in human color vision and that are called B-cones, G- and R-cones. Although this is in many ways misleading, one should keep in mind that these labels derive from "blue" (in the case of B-cones), "green" (in the case of G-Cones) and "red" (in the case of R-Cones). These three cone types contain in normal people three chemically different photopigments. The photopigment contained in a receptor determines its characteristic sensitivity to light of different wavelengths. This assumption is central to the following discussion. It implies that a G-cone will react to a specific light stimulus exactly like a normal R-cone if it is filled with the photopigment normally contained in R-cones and vice versa. The amount of redness, greenness, yellowness and blueness in a perceived color depends on the average stimulation of the cones of the three types on the relevant area of the retina. The perceived color contains some amount of redness iff the relevant higher areas of the brain get the information that R-cone activity prevails G-cone activity and it containes a component of green iff G-cone activity prevails R-cone activity. In an analogous manner, a comparison between the activity of B-cones and the activity of R- and G-cones together determines whether the perceived color contains a component of blue or of yellow.
Let's see what happens if for some mistake of nature a given person has her R- and her G-cones filled with the same photopigment. The average stimulation of R- and G-cones will always be the same independently of the wavelengths composition of the light stimulus and so nothing will appear greenish or reddish to the person at issue. The mistake can occur in two ways: G and R-cones could both be filled with the R-cone-photopigment or could both be filled with the G-cone-photopigment. These two conditions actually correspond to the two main forms of red-green-blindness. According to Piantanida's model mentioned before, both genes, the one that causes production of the G-cone photopigment in R-cones and the one that causes production of the R-cone photopigment in G-cones, may be active in one single individual. In this case two mistakes are much better than one: The wrong filling in both cone types simply leads to an exchange of the photopigments in the two cone types. As a result, the discriminative capacities of the person at issue will not be impaired and his or her visual life will not be impoverished. But, whenever a pseudonormal person looks at a red tomato, her brain will get the information that G-cone-acticity prevails, so the red tomato will appear green to him or her. In general one can predict: What appears red to a normal person to a certain degree, will appear green to a pseudonormal person to roughly the same degree and vice versa. In this sense pseudonormal people are red-green-inverted. Contrary to this result, several philosophical theories imply that pseudonormal people, if they existed, would not be red-green-inverted. This is, as I will argue, reason to reject the philosophical theory at issue.
Any philosophical theory about mind should meet the following prima facie constraint: No hypothesis accepted or seriously considered in color vision science should be regarded according to a philosophical theory to be either incoherent or unstatable or false. A view that denies that red objects appear green to pseudonormals obviously violates this principle. The constraint just formulated is, however, only prima facie. Surely it may follow from a convincing philosophical theory that the terminology of some specific science is somehow confused and should be revised. However, the philosopher whose theory does not meet the above formulated constraint has the burden of proof on his side. In particular, he should be able to argue that the empirical theory at issue can be reformulated in a way compatible with the philosophers theory without change of its empirical content and without contradicting its most central assumptions. But, as we will see in a moment, this is not possible with respect to color vision sicence for a philosopher who denies that pseudonormal people are red-green-inverted.
The prediction of red-green-inversion for pseudonormal people is a direct consequence of a general and highly plausible assumption about the dependence of phenomenal experience on physiological properties that may be put as a supervenience thesis:
(ST) There can be no difference between two persons with respect to their phenomenal experience unless there is also a difference in their relevant physiological properties.
Two remarks about the intended class of "relevant" physiological properties refered to in (ST) are necessary. (1) This class only includes physiological properties of the brain and not of the retina. The reason is this: If there were a difference between two individuals with respect to their retina that would not cause a physiological difference in their brain, then it would be mysterious if the corresponding phenomenal experiences were all the same different. Mysterious cases of this kind are excluded by (ST) if the relevant class of physiological properties is restricted to properties of the brain. (2) Also, it would be quite mysterious if the mere fact that a person learns to associate specific linguistic expressions with his or her phenomenal experience would suffice to change its qualitative character. Therefore, physiological properties that underly causal connections of this irrelevant kind to the language center of the brain, are also excluded from the intended class of physiological properties either. With these explanation of relevance in mind it is clear that there is no difference in the relevant physiological properties between a normal person who is looking at a red tomato and a pseudonormal person who is looking at a green tomato and vice versa. So, who denies that pseudonormal people are red-green inverted thereby denies (ST). 5
This result can be used for an argument against functionalism. According to functionalism to have a sensation of red is to be in a state that is caused by a certain kind of stimuli, leads to specific other internal states (e.g., the belief that there is something red), and causes-together with other internal states characteristic outputs (e.g., the utterance "this is red"). This rough characterization of functionalism is precise enough to clarify why pseudonormal vision represents a problem for most functionalist theories of mind. 6 According to functionalism there is a characteristic type of inputs IG that causes sensations of green (and to be caused by inputs of this type is an essential property of sensations of green) and there is a specific type of outputs OG that is typically caused by sensations of green (and to be a partial cause of output of this type is again an essential property of sensations of green). Now, most versions of functionalism are either committed to the claim that
(A) The inputs received by a pseudonormal person when looking at a red tomato are not of the type IG
or to the claim that
(B) The outputs partially caused by the internal state of pseudonormal people when looking at red tomatoes are not of type OG
or to both of these claims. In each case, the functionalist must deny that red tomatoes appear green to pseudonormal people. But, as we have seen, this shows that his theory is in conflict with a central and highly plausible assumption of color vision science.
Two questions in the debate about the so-called explantory gap thesis should not be conflated:
Q1: Is it possible to understand why certain physical properties of physical systems lead to the occurrence of consciousness at all?
Q2: Is it possible to understand why certain physical processes in the brain of sentient beings lead to experiences of a given qualitative type?
I will only discuss the second question here. If we call the physical process that actually underlies sensations of green "G" and the physical process that actually underlies sensations of red "R," then a negative answer to the second question may be put as follows:
(T1) Explanatory Gap Thesis
Even if we knew everything there is to know about the physical 7properties of R and G, we still would not understand why R is correlated with sensations of red and why G is correlated with sensations of green.
Most proponents of the explanatory gap thesis defend their claim by an intuition captured by T28:
(T2) Qualia Inversion Hypothesis
Even if we knew everything there is to know about the physical properties of R and G, we could still coherently conceive of circumstances where experiences of red are correlated with G and experiences of green are correlated with R.
It is quite generally accepted in the ongoing debate that T1 and T2 stand or fall together. In what follows I will challenge this common assumption. In my view, there is some reason to doubt the Qualia Inversion Hypothesis and still there is good reason to believe in the Explanatory Gap Thesis.9 In several papers and talks Hardin has argued simultaneously against T1 and T2. 10 I will describe what seems to me to be the abstract common structure behind these arguments and I will call any argument of this common structure "the Hardin-Argument." There are mainly two reasons why I think it is worthwhile to look at Hardin's work from this abstract viewpoint: One thereby is led to see quite clearly why (T1) does not depend on the truth of (T2) and one is led to more general insights about what kind of explanations of conscious experience can at best be expected from the empirical sciences.
The Hardin-Argument starts from the following two basic assumptions:
(P1) The color qualities we are acquainted with have a necessary phenomenal structure.
(P2) Any conceivable empirical correlation between hue sensations and brain processes is structure preserving.
The following observation by Hardin clarifies the idea behind (P1): If a creature sees neither reddishness nor yellowishness in a perceived color, then this constitutes, as Hardin says, its failing to see orange. 11 The same point may be put in another way: Being composed of red and yellow is not just an actual but rather a necessary property of orange. Note that P1 can be used in an argument against (T2) only if "necessity" is interpreted in a specific way. In the sense of necessity required for the purposes at issue, the following implication holds: If a given property is a necessary property of a color, then we cannot coherently conceive of circumstances where this color does not have the property at issue. Who wishes to argue against the thesis that qualia inversion cases are coherently conceivable must refer to properties of colors that these colors have not just under the actually given circumstances but under all coherently conceivable circumstances as well. This is why the Hardin argument needs the philosophical premiss that certain structural properties of colors are necessary and cannot be based on the corresponding empirical claims about their actual structure alone.
The idea behind (P2) can be seen by means of an example: There should be some functional property P such that the amount of redness in the color perceived by a person increases with the degree to which the correlated brain process exemplifies the property P. A similar statement should be true for every phenomenal property of colors. An exact formulation of the idea behind (P2) requires using the mathematical notion of an isomorphic mapping. But the idea will be clear enough without going into these technical details. 12 One might object to the premiss (P2) that we can in some sense coherently think of psychophysical correlations that do not preserve structure. After all, it is even coherently conceivable in some sense of conceivability that a being has color sensations without having a brain at all. However, I doubt that the sense of conceivability involved here is helpful for somenone who wishes to defend the Explanatory Gap Thesis (T1) on the basis of the Qualia Inversion Hypothesis (T2). I will therefore conceed (P2).
It is quite clear how an argument against (T2) has to proceed, once (P1) and (P2) are accepted. One must show, using concrete knowledge about both, the phenomenal and partly necessary structure of our color sensations on the one hand and knowledge about the functional structure of the underlying physiological processes on the other, that there is only one way to correlate color sensations and brain processes to each other in a structure preserving way, namely the one actually chosen by nature. Hardin has presented material that makes it appear very plausible that this part of the argument can be stated successfully. 13 So, if we decide to accept the two fundamental premisses stated before, we have to admit that the Qualia Inversion Hypothesis (T2) is false. However, this does not help to close the explanatory gap. To see this it is helpful to state the result we have reached in a slightly different way. I will use the following abbreviations: "Sp" stands for the necessary phenomenal structure of our color experiences and "Sf" for the actual functional structure of the physiological processes that underly our color experiences. The material presented by Hardin supports the following claim:
(R1) If a creature has visual sensations with the phenomenal structure Sp and a visual system with the functional structure Sf, then there is just one structure-preserving way to correlate hue sensations with neural processes.
If we concede (a) that psychophysical correlations that do not preserve structure are unconceivable (see premiss P2) and (b) that we cannot coherently conceive of a creature that has sensations of red, green, yellow and blue without having color experiences that realize the phenomenal structure Sp (compare premiss P1), then (R1) leads to the stronger result (R2): (R2) If we already know of a creature that it has sensations of red, green, yellow and blue, and that its visual system realizes the functional structure Sf, then the actual psychophysical correlations between hue sensations and brain processes are the only correlations we can still coherently conceive of.
(R2) clearly contradicts the Qualia Inversion Hypothesis (T2), but, it does not undermine (T1). To close the explanatory gap, we need a quite different result. We wish to see why the functional structure of our brain processes necessarily leads to sensations of red, green, and so on. So what we need is an argument for something like the following result:
(R*) We cannot coherently conceive of a creature whose visual system realizes Sf and yet does not have sensations of red, green, yellow and blue.
(R2) starts from the assumption that we already know that the creature at issue has our basic color sensations (red, green, yellow and blue). Therefore it obviously cannot establish (R*). But, maybe the philosophical opponent is really a bit closer to his philosophical aim than might appear after what has been said so far. It is quite plausible that an argument very much like the Hardin-Argument can show that functional structure determines phenomenal structure in the following sense:
(R3) It is not coherently conceivable that a sentient being has a visual system that realizes the functional structure f and yet does not have color experiences that realize the phenomenal structure Sp.
With an argument for (R3) we would have established the explainability of phenomenal structure on the basis of functional structure, but still this is not enough. An explanation of why our color experiences have a specific phenomenal structure is not yet an explanation of why we have sensations of the concrete qualities red, green, yellow and blue. So we are still left with an explanatory gap between phenomenal structure and concrete phenomenal qualities. One might try to close this remaining gap by proposing the following further premise:
(P3) A creature who has visual experiences that realize the phenomenal structure Sp necessarily has experiences of red, green, blue, and yellow.
Maybe (P3) is acceptable if "necessity" is read as natural or nomological necessity. I doubt this, but we do not need to decide this issue. (P3) does not help to close the explanatory gap if necessity is read in this sense. Maybe it is a law of nature that sensations of red, green, yellow and blue necessarily occur whenever there is the phenomenal structure Sp. But, if we had to accept this regularity as a brute fact about how the world really is, we still would be left with the original puzzle: Why do we have sensations of red (this concrete quality)? Who wishes to attack the explanatory gap thesis in this way must claim that (P3) is still true when necessity if read as conceptual necessity. On this reading however, (P3) is quite obviously false. There is, as Hardin has pointed out convincingly, I think, a conceptual link that leads from concrete phenomenal qualities to phenomenal structure, but there is no conceptual link that turns an explanation of phenomenal structure into an explanation of the occurrence of concrete phenomenal qualities. This observation has its parallel in the epistemological fact that knowledge about phenomenal structure does not imply knowledge about concrete phenomenal qualities. An example may illustrate this epistemological fact. Let us assume that we find somewhere in the universe a sentient being who distinguishes surfaces of objects according to the wavelengths composition of the light reflected by these objects. In some way we find out that the alien creature has not just two channels that inform the brain about the result of comparing average acticities of different receptor types, but three such channels. If we knew that the creature has visual sensations at all, we would then have reason to assume that it is acquainted not with four but with six basic colors and that the colors it perceives have one more hue dimension: Our colors are phenomenally composed of at most two hue components, the colors experienced by these creatures are composed of up to three components (the painters of this alien society would have incredibly more possibilities of artistic expression). On the basis of the functional structure of the visual system of these creatures we might gain rich knowledge about the phenomenal structure of their experiences, but this does not help us to find out which are the six basic colors they experience. 14
We are thus led to the following diagnosis: The explanatory gap, as far as the explanation of specific phenomenal sensations are concerned, basically rests on the partial epistemological independence between phenomenal structure on the one hand and concrete phenomenal qualities on the other. So maybe the proponent of the explanatory gap thesis, instead of insisting on the Qualia Inversion Hypothesis should rather concentrate on this specific independence claim and its philosophical consequences.
What kinds of explanation of phenomenal experience may we at best expect from the empirical sciences? (1) We can of course expect what one might call single case explanations for the occurrence of a phenomenal experience of a specific kind in a given situation. For example, why do you see a rose as red? Answer: The rose produces a physiological state that is, in humans generally, correlated with sensations of red. A single case explanation of this kind presupposes psychophysical laws. (2) We can expect explanations of phenomenal structure. Why is it impossible to see a surface as greenish and reddish in the same place? Answer: the underlying physiological processes have causal properties that make it impossible for them to occur simultanously. Explanations of phenomenal structure of this kind again presuppose psychophysical laws. (3) We may expect even explanations of psychophysical laws in the following sense: Any alternative psychophysical correlation would not be structure preserving. As we have seen, explaining psychophysical laws in this manner, presupposes the fact that we have sensations of the specific qualities red, green, yellow and blue, and does not explain this fact. (4) Under the assumption that by law of nature the phenomenal structure Sp of visual experiences can only be found in individuals whose basic hue qualities are those known to us (red, green, yellow, and blue), we even may get an explanation of why we have sensations of this specific phenomenal character. This explanation, however, presupposes and does not explain why a specific phenomenal structure is nomologically associated with these concrete phenomenal qualities.-The possibility to explain phenomenal structure in the senses just sketched may be more of an explanation than some philosophers might have expected. But still it does not close the explanatory gap and we are left with our original puzzle about the occurrence of concrete phenomenal qualities.
One common feature of the four kinds of explanation just mentioned should be kept in mind: They all presuppose and do not explain the fact that the creature at issue is a sentient being, that is has a subjective perspective, that there is consciousness at all. This aspect of the explanatory gap debate certainly concerns the more fundamental and still more puzzling mystery about consciousness.
1. When invited to the Tucson III conference, I was asked to talk about pseudonormal vision and about Hardin's arguments against the explanatory gap thesis. This is why the chapter consists of two relatively independent parts.
2. See T. P. Piantanida, "A replacement model of X-linked recessive colour vision defects." Annals of Human Genetics 37 (1974): 393-404 and Robert M. Boynton "Human Color Vision," New York, Holt Rinehart and Winston, 1979, pp. 351-358.
3. Boynton in "Human Color Vision" op. cit., pp. 356.
4. I don't know the estimation for females.
5. A more detailed discussion would, of course, have to include (a) a precise account of (ST) and the involved concept of relevance, (b) more evidence for the claim that (ST) is actually accepted in color vision science and (c) independent philosophical reasons for the acceptance of (ST).
6. Some versions of functionalism have a reply to the argument that follows but have other problems with the case of pseudonormal vision (see my paper [1996], in particular section 4).
7. Note that "physical" is used here in the very broad sense common in the relevant debate: Chemical, physiological, and functional properties are included.
8. Compare, e.g., Joseph Levine [1983].
9. The case of pseudonormal vision is, of course, irrelevant to the present question. The qualia inversion hypothesis states the conceivability of cases where two individuals differ in their phenomenal experience while there is no relevant physical difference between the two. But there are relevant physical differences between normals and pseudonormals when looking at red objects.
10. See Hardin [1987], [1996], and his contribution to this volume.
11. Compare Hardin [1996], p. 101 and his contribution to this volume.
12. For a more detailed examination of (P2) compare my paper [1998].
13. Compare his contribution to this volume and his *
14. The observation that knowledge about physical properties (including functional properties) does not guarantee knowledge about qualitative states has been discussed in detail (see, e.g., Jackson's famous paper [1982] and the discussion that followed). The point here is slightly different: Even knowledge about phenomenal structure (not to confuse with functional structure!) does not include knowledge about concrete qualia.
Boynton, R. M. 1979. Human Color Vision. New York, Holt Rinehart & Winston: 351-358.
Jackson, F. 1982. Epiphenomenal Qualia. In Philosophical Quarterly 32:127-136.
Hardin, Clyde L. 1996. Reinverting the Spectrum, in Carrier and Machamer: Mindscapes: Philosophy, Science, and the Mind. Pittsburgh, Konstanz: University of Pittsburgh Press, Universitätsverlag Konstanz, 99-112.
Hardin, C. L. 1987. Qualia and Materialism: Closing the Explanatory Gap. In Philosophy and Phenomenological Research 48:281-298.
Levine, J. 1983. Materialism and Qualia: The Explanatory Gap Pacific Philosophical Quarterly 64:354-361.
Nida-Rümelin, M. 1996. Pseudonormal Vision. An Actual Case of Qualia Inversion? In Philosophical Studies 82:145-157.
Nida-Rümelin, M. 1998. Vertauschte Sinnesqualitäten und die Frage der Erklärbarkeit von Bewußtsein. In Frank Esken and Dieter Heckmann (eds.), Bewußtsein und Repräsentation. Paderborn, Ferdinand Scnöningh: 299-324.
Piantanida, T. P., 1974. A replacement model of X-linked recessive color vision defects. In Annals of Human Genetics 37: 393-404.