Chapter 7 of Color for Science, Art, and Technology, ed. Kurt Nassau. Elsevier, 1998.

A problem that has long preoccupied philosophers is whether color qualities are to be located in the physical world, independent of the consciousness of perceivers, or whether they are mind-dependent phenomena. This is not merely a matter of the terminology that we choose to employ, as in the moldy riddle, "If a tree falls in the forest and nobody is around, does it make a sound?" That question is readily answered once the questioner is obliged to specify whether ‘sound’ is taken to mean "vibration of the air" or "auditory sensation."

What intrigues philosophers is quite different. It may be expressed by a pair of related questions: First, how can we reconcile the picture of the world presented to us by science with the view of the world that naturally suggests itself to common sense? Second, are color sensations (and other sensations and feelings) identical with brain states, or different from them?. The first of these issues was addressed by Galileo (1) in his essay of 1623, The Assayer:

Galileo’s sentiments had been voiced some 2000 years earlier by the atomist Democritus, and repudiated by Aristotle, who preferred to locate colors outside of living creatures, in the interaction of light and matter. Medieval as well as ancient thinkers had followed Aristotle in this. For them, the qualities that objects are seen to have are by and large the qualities that they do have; they held that the business of our senses is to reveal Nature’s finery rather than to clothe things with raiment of our own devising. But Galileo’s words heralded the birth of a new view of the Book of Nature: it is, he was fond of saying, written in the language of mathematics. The objects of mathematics are quantitative: number, structure, and motion. The powerful new mathematical physics seemed not to allow for the qualitative features of experience. Colors, sounds, and odors were thus swept from the physical world of matter in motion into the dustbin of the mind, where they joined thinking, purposiveness, and consciousness, which the new physics had likewise dispossessed from the world of matter. This "bifurcation of nature", as philosopher-mathematician Alfred North Whitehead (2) later dubbed it, served physics well, but left obscure the nature of mind and its relationship to the physical world.

Science and philosophy have both become far more sophisticated than they were in the 17th century, yet the basic picture of a bifurcated nature remains with us. As some of its early critics remarked, this picture offends common sense, for nothing seems so evident as that color is an inherent feature of the world outside our bodies: grass is green, the sky is blue, coal is black. How could it be that shape is an inherent feature of the physical world, whereas color is a construction of the mind? Can we even imagine a shape with no color whatever? Furthermore, as we shall shortly see, there are grave difficulties in understanding in what sense colors could be "in the head". For these reasons, several philosophers in recent years have tried to show how colors could in fact be identified with items in the physical world. If they were successful, color would be detected by the senses rather than being created by them. So let us briefly consider some of the proposed physical candidates for identification with the colors (3).

The first of these is that colors are to be identified with certain wavelengths of light. Thus, an object is said to be yellow if it predominantly reflects light of about 580 nm. Many physicists use the terms ‘red light’ or ‘green light’ in such a way as to suggest their acceptance of this point of view. This is commonly traced back to Isaac Newton, who is said to have shown that white right is really made up of "light of all colors". Newton himself was far more cautious, remarking in the Opticks that "the rays, to speak properly, are not colored. In them there is nothing else than a certain power and disposition to stir up a sensation of this or that color"(4). The problem with the proposed identification of light with wavelength is, of course, that it flies in the face of metamerism [Chapter 1, section xx]. Metamerism occurs whenever two stimuli with different spectra look the same because they stimulate the same responses in the cones of the retina. A monochromatic light of 580 nm, for example, can be precisely matched by a mixture of monchromatic lights of 540 nm and 670 nm, neither of which is seen as yellow. Similarly, indefinitely many pairs of monchromatic lights can match a reference white. If a perfectly good white can be "made up of two colors" and in indefinitely many ways, it cannot be true that white light is identical with "light of all colors."

A second proposal acknowledges that colors are not natural physical kinds, but are, instead, heterogeneous classes of spectral reflectances or emissions grouped together by how they affect the human visual system. By this account, an object is red just in case it reflects light that produces a certain ratio of cone responses. This obviously meets the challenge of metamerism, but one might seriously question whether it captures the central features of color as we experience it. We have learned from the physiologist and psychologist Ewald Hering [Chapter 1, section xx] that there is an elementary red that contains no perceptual trace of any other color, whereas a color such as orange, for example, is never elementary, but is always seen as a perceptual mixture of red and yellow. There are four elementary, or unitary hues–red, yellow, green, and blue–and innumerable perceptually mixed, or binary hues, such as orange, purple, or chartreuse (5). However, it makes no sense to say that a spectral reflectance or emittance is elementary rather than mixed. The unitary-binary hue structure, so central to the colors as we know them, has no counterpart in the domain of spectral relectances or emittances, therefore spectral reflectances and emittances cannot be identical with colors, since they lack an essential property that the colors have. Spectral reflectances and emittances are of course causally essential to our perception of color, but that does not entitle us to say that theyare colors.

A third proposal makes no attempt to identify colors with any particular physical property or process. Rather, it says that colors are identical with whatever causes color sensations in normal observers under standard conditions. Thus, a bird’s feathers would be blue just in case they would look blue to a normal observer under standard conditions. Although this works well enough as a rough-and-ready account of when we are entitled to attach color words to objects, it fails when made to perform more exacting tasks. Normal conditions will vary considerably with the type of object that is observed (compare the conditions for seeing the blue of a rainbow, a star, and a Munsell sample, respectively), and normal observers do not in fact have identical visual responses. As professional color-matchers know only too well, there can never be such a thing as a metameric match that will satisfy all normal observers, and it is an established fact in visual science that the spectral locus of a unique hue is a statistical construct taken over populations of normal observers, populations which show wide variance in their locations of unitary blue and green.

A final line of defense by some who wish to locate colors outside the head is to urge that colors are indeed properties of the physical world even though they correspond to nothing in the world that is described by contemporary physics. This simply shows, say they, that contemporary physics is incomplete. This is likely to seem to most readers to be a council of desperation, for although no competent person would claim that today’s physics is able to capture all of the phenomena of nature, it seems highly unlikely that the phenomena now left out include the colors. Unlike the situation in the 17th century, the physics and chemistry of color stimuli are well understood. The physical mysteries of color that are left to explore all seem to turn around what goes on once light is absorbed by the photoreceptors of eyes.

Faced with these serious objections to any attempt to locate colors in the physical world outside of nervous systems, our search to find the locus of color must next turn to the processes that transpire within living creatures, just as Galileo had supposed. Unfortunately, our task has now become even more difficult. It is obvious that colors are not, literally speaking, in brains as raisins are in cookies. The question is what relationship color perceivings have to brain processes. This is in fact the second of the two questions that were raised at the beginning of this essay. Those who hold that subjective sensations of color, or pain, or feelings of envy are reducible without remainder to neural processes and biochemical events are mind-body materialists, and those who deny this are mind-body dualists. Although psychophysicists and physiologists commonly call themselves materialists and indeed look like staunch materialists when they putter about their laboratories, catch them in the off-hours, get them to use such words as ‘subjectivity’ and ‘consciousness’, and many of them will sound like closet dualists. True, few if any of them suppose, as did some dualists of old, that the bits of mind-stuff that compose sensory qualities are fragments of an autonomous domain, with causal powers independent of brain function, but many of them are persuaded that the event of sensing a patch of red in their visual field, although it may be caused by brain processes, could not be identical with a brain processes, or at least that we could never have satisfactory grounds for deciding that it is identical with a brain process.

From the 17th century to the present day, arguments on behalf of dualism have ultimately involved appeals to a small underlying set of intuitions which commonly take the form of little pictures of how the world must be. All of them purport to show that materialist accounts of subjective experience must inevitably leave something out. Let us undertake our examination of the issue between dualism and materialism by looking at three such pictures, along with their associated arguments, and see how well they survive analysis.

We shall call the first picture Blind Mary, the second Leibniz’s Mill, and the third, Chromatic Inversion. In each case, we are to suppose, for the sake of argument, that neuroscience has attained a utopian state of perfect knowledge, so that there are no scientific issues left to be resolved and only logical and philosophical questions remain. According to the supposition, we not only know how the whole physics and chemistry of the nervous system works, but just how its functional architecture enables the organism to extract and utilize information in the environment. We can predict just how someone will respond to whatever stimulus is thrown at her, no matter how complex that stimulus might be.

In the first instance we are to imagine a physiologist–call her Mary–who is in possession of this prodigious utopian understanding, but unfortunately suffers from having been blind from birth. She knows all about the neural states that sighted people are in when they see red, she knows in the fullest detail how people react to seeing red, she knows all of the things that people associate with seeing red, but does she know what it is to see red? Or, as it is sometimes put, does she know what it would be like to see red? It seems that we know something that she does not, and so her knowledge of seeing, however complete it may be from a physiological standpoint, necessarily leaves out something that is central to seeing. Expressed another way, if Mary were to gain her sight, would she come to know something that she did not and could not have previously known? (6)

This argument trades heavily on the idea that having visual experience is being in possession of a certain item of knowledge. But we need not grant this. Knowledge, we might insist, is a conceptual matter, not a sensory one, even though our knowledge is gained through sensory stimulation and must constantly be subjected to sensory test. By hypothesis, Blind Mary does not have the experiences that we have, because she cannot be in the neural states that we can be in. And as a consequence, she cannot imagine, as we can, the look of a patch of red. But conceiving, thinking, and knowing are one sort of thing, imagining quite another. As the famous philosopher-mathematician Reneé Descartes, the founder of mind-body dualism, long ago pointed out (7), we can readily conceive of a polygon of a thousand sides, easily distinguishing it from a polygon conceived as having a thousand and one sides, or from one of 999 sides. But can we imagine a thousand-sided polygon, distinguishing it in imagination from one of 999 sides? Furthermore, we can see and remember and imagine things that seem to us to be singular and indescribable. Notice that this doesn’t mean that these things are not in principle categorizable or describable, just that we at the particular moment lack the capacity to do so. So we may readily agree that Mary can’t be in a sensory state that we can be in, but we need not grant that we know something about the state that she doesn’t. Indeed, unless we possess her utopian science, we will know much less about our states than she does. Furthermore, she will likely understand more than we do about what, in the literal sense of the word, seeing red is like, that is, what other sensory states the state of red-seeing most nearly resembles.

If you were attracted by the Blind Mary argument in the first place, you will doubtless not be persuaded by this response to it. The response turned upon an unwillingness to use the word ‘knowledge’ in a broader rather than a narrower sense. You may feel that there is a deep fact of some sort to which we have so far simply refused to give expression. Perhaps this feeling can be captured more successfully by our second image, taken from the 17th century philosopher-scientist G. W. Leibniz (8):

It was Leibniz who proposed a universal computing language, and devised an early calculating machine, so it is not difficult for us to transpose his image into suitably modern electronic or electrochemical terms. Some of us may remember reacting to the overzealous claims of the Artificial Intelligence hucksters, especially in the earlier days of computers, by imagining the machines for which they had written their programs allegedly for "thinking" and "seeing", and responding, "That hunk of junk couldn’t really think or see, no matter how complicated we made it and its program. All it could ever do is imitate a thinking or seeing organism." Since then, it has become rather less clear that thinking need be a consciously accessible process, and somewhat clearer that whether a program should be regarded as a program for thinking has more to do with the detailed character of the program, its scope and capabilities and the manner of its realization, than with any questions of basic principle. Seeing has seemed a different matter though, because in at least some of its aspects, particularly the perception of color, qualitative character plays a central role; wavelength discrimination is simply not enough. If we were to walk through an enlarged digital computer, electrons whizzing past our ears, we might still ask, "How could this thing experience red?"

But let us be careful here. We might equally well suppose ourselves walking within an enlarged brain, dodging depolarizing neurons and squirts of neurotransmitters, and be equally tempted to ask, "How could this thing experience red?" Cells do not look to be any more capable of supporting sensory episodes than silicon chips do, but the fact is that networks of cells do just that. So the question is not whether they support sensory episodes, but how. And there are only three answers that we need consider. The first two are dualist answers. They agree that the processes occurring in neural networks cause sensory events, and that those events are not identical with the neural processes.

According to the first answer, which philosophers have called interactionism, mental events can affect other mental events as well as neural events. From the point of view of a visual scientist, this is an odd suggestion indeed, for it requires not only that our utopian neurophysiological pictures be incomplete insofar as it leaves out the domain of subjective phenomena, but that it be causally incomplete, so that some neural events downstream from certain perceptual episodes could not be explained by means of any previous physical and chemical happenings, but would require that some subjective episodes be invoked. But how would a neuroscientist react if in the course of tracing out a complex biochemical sequence someone told her that a particular missing step would be forever missing because its causal role is filled by a nonphysical subjective event?

This sort of consideration has made many scientists of a dualist persuasion look with favor upon our second answer, which philosophers have called epiphenomenalism. According to this view, brain events cause subjective sensory events, but these sensory events are causally idle. The subjective experiences occupy the stage of the mind, but it is the nervous system backstage that pulls the strings. To change metaphors, sensory events are the shadows on the wall, or the foam on the beer. But this position won’t do either, for we are quite persuaded that those sensory events play a real role in our mental lives. Colors, for example, delight, excite, annoy, and depress us. They affect our beliefs and incite our behavior.

So we want to have our subjective sensory experiences make a difference, but we do not want them to usurp the hegemony of physical and chemical events in the brain. How can we have it both ways? By abandoning both of our first two dualist alternatives and adopting our third, materialist alternative, that the subjective sensory events are nothing but the neural happenings themselves. But here Leibniz’s Mill returns to haunt us. How can a bunch of depolarizing neurons be identical with the event of my experiencing red rather than merely occasioning it? Who, on seeing a neural network in action could ever guess that it is generating an experience of red, rather than an experience of green, or a toothache, rather than being nothing more than a bunch of neurons acting up? We shall elaborate this objection more fully in a moment, but for now let us content ourselves with observing that if it has force against materialism, it has force against dualism as well, since nobody who simply saw the neural network could have guessed that it was causing nonphysical subjective experiencings either. So the most that it could show is that the mind-body relationship is ultimately and intractably mysterious. But does it actually show this much?

Let us look at the matter more closely by considering Chromatic Inversion, our third picture of the relationship between sensory states and neural states. Although the possibility of chromatic inversion was raised in the 17th century by John Locke (9), the version that we shall consider here is by a contemporary philosopher (10), who invites us to imagine that we have before us the utopian accounts of red-seeing and green-seeing:

Such an appeal to imagination is characteristic of many philosophical arguments. The reason for the appeal is that philosophers use imaginability as a tool for separating out the necessary features of a situation from the the merely contingent ones. For instance, even though it may be false that there is life on Mars, it is nevertheless possible that there should have been life on Mars, whereas it is not only in fact false but necessarily false that on Mars, 2+2=5. One way of seeing the difference in status between the propositions that life exists on Mars and that 2+2=5 on Mars, is to notice that one can readily imagine the state of affairs in which the first would be true but that a state of affairs in which the second would be true is quite literally unimaginable. In the case at hand, what is being argued is that since one can imagine that someone sees red when G obtains rather than R, seeing red is not a necessary feature of R’s occurring. But if it were true that seeing red is nothing but R’s occurring, that seeing red is identical with being in brain state R, then to see red is necessarily to be in state R, and to conceive the one is to conceive the other. The contrary would not be imaginable since it would be impossible, and impossible states of affairs are not imaginable.

If the argument that appeals to Chromatic Inversion is correct, there can never be empirical grounds that can justify a claim that subjective sensory processes are identical with neural processes. All that an empirical investigation could establish is that the one is correlated with the other.

But if this argument proves anything it proves too much. Could one not say that heat of a monatomic gas, thermodynamically conceived, is not identical with, but only correlated with the kinetic energy of its constituent molecules? After all, we can imagine that the course of the history of science was different, and that heat turned out to be a very subtle mechanical fluid, the caloric fluid, just as the inventor of thermodynamics, Sadi Carnot, had initially supposed.

Many of us would be strongly inclined to reject such an argument, and urge that heat really is nothing but random kinetic energy. For once we come to understand the properties of matter, and the laws of thermodynamic phenomena, it becomes clear to us that there are many reasons why the two preclude the possibility of caloric mechanisms of heat. Such mechanisms never really were possible, any more than it was ever really possible that trees should speak or witches fly. The absence of adequate concepts of these matters gave rise to an appearance, or, if you will, an illusion of their possibility. Once the details are spelled out, accounts of caloric fluid, speaking trees and flying witches collapse from their own incoherence. These things can be imagined, but only as long as they are imagined schematically.

In the case at hand, what we have been given is just a schematic account, and it is far from evident that, were the R and G stories of neural functioning to be spelled out, that we could just as easily imagine red-seeing associated with G as with R. Needless to say, we don’t have in our possession anything that approaches proper utopian R and G stories of chromatic processing. Visual scientists are in no position now to make assertions about identities in this domain that are comprarable to what physicists can say about thermodynamics and statistical mechanics. However, visual science is far enough along now for us to say useful things about the neural basis of color qualities.

For instance, suppose that instead of talking about red and green and R stories and G stories, we had talked about red and orange on the one hand, and R stories and O stories on the other Could a proper neural O story–one given in terms of central rather than peripheral processes–have been just as fit to be an account of red seeing as of orange seeing? Most of us would be inclined to deny it. The reason is of course that we demand of a proper O account that it have a structure appropriate to a binary process, whereas a proper R account must have a structure appropriate to a unitary process. Of course unitary and binary processes come abstractly labeled as such, but one must give an account of the hues all at once, and within such an account the relative simplicity and complexity of the processes ought to be salient. There is of course no guarantee that our best neural explanations of orange seeing and red seeing will exhibit this sort of structural differentiation, but should they not, we would be disinclined to think that we had accounted for the phenomenally unique character of the one and the binary character of the other. If the future efforts of visual scientists should have one result, we would have evidence for a claim that the phenomenal event is nothing but the neural process, whereas if they should have another, we would not be able to support such a claim.

The challenge of Chromatic Inversion has not been fully met. Nevertheless, using the example of how the unitary and binary colors might be modeled by appropriately structured neural processes, we can at least gain a glimpse of how future science might meet such a challenge, and in meeting it, might provide solid reasons for accepting a materialist view of how color qualities are realized by brain states. On the other hand, it might not; nature might indeed be bifurcated, or else we might be constituionally incapable of understanding why it is not. In any event, these are now open questions, and are apt to remain so for quite a long time.

1. Galileo, G.; The Assayer. In Drake (trans. and ed.), S: Discoveries and Opinions of Galileo (New York, Doubleday and Company, 1623; 1957); 231-280.

2. Whitehead, A.N.; Science and the Modern World (London, Macmillan, 1925).

3. Hardin, C.L.; Color for Philosophers: Unweaving the Rainbow (expanded edition) (Indianapolis, Indiana and Cambridge, Massachusetts, Hackett Publishing Company, 1988), Ch. 2.

4. Newton, I; Opticks (New York, Dover Publications, 1704; 1952); prop. 2, th.2, def.

5. Hering, E.; Outlines of a Theory of the Light Sense, trans. by L.M. Hurvich and D. Jameson (Cambridge, Massachusetts, Harvard University Press, 1920; 1964).

6. Jackson, F.; ‘Epiphenomenal qualia.’ Philosophical Quarterly, 32 (1982), 127-136.

7. Descartes, R.; Meditations on First Philosophy. In Haldane and Ross (trans. and ed.), The Philosophical Works of Descartes, v.1 (Cambridge, Cambridge University Press, 1641; 1931).

8. Leibniz, G.W.; Monadology. In Parkinson (ed.), G.H., Leibniz: Philosophical Writings (London, J.M. Dent and Sons, 1973), sec. 17.

9. Locke, J.; Essay Concerning Human Understanding, ed. by P.H. Nidditch (Oxford, Oxford University Press, 1689); 1975) Bk II, ch. 32, sec. xv.

10. Levine, J.; ‘Materialism and qualia: the explanatory gap.’ Pacific Philosophical Quarterly, 64 (1983), 354-361.

Dennett, D.C.; Consciousness Explained (Boston, Little Brown, 1991).

Hardin, C.L.; Color for Philosophers: Unweaving the Rainbow (expanded edition) (Indianapolis, Indiana and Cambridge, Massachusetts, Hackett Publishing Company, 1988).

Hilbert, D.; Color and Color Perception: A Study in Anthropocentric Realism (Stanford, CA, Center for the Study of Language and Information, 1987).

Hurvich, L.M.; Color Vision (Sunderland, Mass., Sinauer Associates, 1981).

Thompson, E.; Colour Vision: A Study in Cognitive Science and the Philosophy of Perception (London and New York: Routledge, 1995).

Westphal, J.; Colour: A Philosophical Introduction (Oxford, Basil Blackwell, 1991).

nature, bifurcation of

metamerism

unitary (unique) hues

binary hues

spectral reflectances

spectral emittances

materialism, mind-body

dualism, mind-body

artificial intelligence

interactionism, mind-body

epiphenomenalism, mind-body

chromatic inversion

thermodynamics

Democritus

Aristotle

Galileo

Whitehead, A. N.

Newton, I.

Hering, E.

Descartes, R.

Carnot, S.

Leibniz, G. W.

Locke, J.

Clyde L. Hardin is Professor Emeritus of Philosophy at Syracuse University, Syracuse, NY, USA. In addition to 37 philosophical articles, 21 of which pertain to color, he is the author of Color for Philosophers (Hackett Publishing Company, 1988) and coeditor of Color Categories in Thought and Language (Cambridge University Press, 1997).