It is quite clear that Helmholtz here focuses on an explanation of the qualitative aspect of colors, and this is exactly the point where today’s view departs from the Young–Helmholtz’ theory.

Hering’s opponent colors theory

Hering vehemently opposed the Young–Helmholtz theory, and he claimed that color vision was based not on three primary sensations, but on four chromatic and two achromatic elementary, or unique, color perceptions (Urfarben) and their corresponding physiological processes. This idea had support in Leonardo da Vinci’s (1452– 1519) observation in A Treatise on Painting (1906) that six particularly simple colors are found in nature. They are the four unique, or elementary chromatic hues, yellow, red, blue and green, together with the two achromatic colors, white and black. These colors serve as six qualitative references in subjective or phenomenological color space (Hering, 1964; Hurvich, 1981). The physical stimuli associated with these percepts depend on the viewing situation, and they vary from person to person. In the spectrum, unique blue is found around 470 nm, green at about 500 nm, yellow at 575 nm, and a saturated unique red is usually a mixture of the red and blue ends of the spectrum, with a complementary wavelength of about 495 nm. Unique yellow can be determined with an extraordinary precision. Although one person’s selected wavelength for unique yellow may be found anywhere between 565 and 590 nm, the precision can be at the order of a few nanometers (Richter, 1969; Mollon and Jordan, 1997). These identifications seem to have little to do with culture and language (see peer commentaries to Saunders and van Brakel, 1997).

In Hering’s opponent theory, the two pairs of chromatic unique colors, together with the achromatic pair, black and white, were associated with three pairs of hypothetical, antagonistic physiological processes. In agreement with Hillebrand (1888, p. 70), Hering came to associate unique red with the breaking down, or wearing out (‘dissimilation’) of a particular ‘visual substance’ – and unique green with the building up, or restoration (‘Assimilation’), of the same substance. Similar antagonistic processes in two other substances gave rise to unique yellow and blue, and to white and black. The opposite nature of these paired hue qualities, as displayed in the hue circle of Figure 5.3, were thus associated with processes which mutually excluded one another. The ‘visual substances’ were not the photopigments of the receptors; they were unidentified physiological substrates at an unspecified level in the visual pathway (Trendelenburg, 1943, p. 81).

For quite some time, the threeand four colors theories were strongly opposed, even after the Austrian physicist Erwin Scho¨dinger (1887–1961) showed, in 1925, how they could be reconciled into one. Schro¨dinger’s idea was that the three-color theory could be valid at the level of the receptors in the retina, and that the receptors’ reaction to light stimulation could be transformed to four color-coding signals later in the visual system.

In this context, it is of interest to note that G.E. Mu¨ller, in his zone theory of color vision, developed a more differentiated and modern view of the physiological processes underlying unique hues. He reserved the primary antagonistic ‘neuronal processes’ for the explanation of color contrast (such as simultaneous and successive contrast, induction, adaptation, etc.). With regard to unique colors, he envisaged four ‘psychophysical excitations’ [obviously referring to Fechner’s (1966) ‘inner psychophysics’] at a central level that received their inputs from the lower-level antagonistic neural processes (Mu¨ller, 1930).

Dorothea Jameson and Leo Hurvich’s extensive studies of the psychophysics of color opponency in the 1950s were based on the above concepts of physiological substrates, and the revival of Hering’s ideas through ‘hue cancellation’ experiments (see Hurvich, 1981). In these experiments, additive mixtures of colored lights were used to determine a unique hue, and the theoretical framework implied that a neural mechanism was in an ‘equilibrium state’ whenever an opponent mechanism had been balanced. In this theory, unique yellow, for instance, was viewed as such an equilibrium state between a ‘red process’ and a ‘green process’. However, the residual sensation (yellow in this case) did not need to be relevant, since the judgements were based only on the absence of redness and greenness, independent of other color attributes of the stimulus (it could also be white or blue).

When the Swedish physiologist Gunnar Svaetichin (1956) published the first recordings of spectrally dependent positive and negative potentials in the retina of fish, he believed he had proven Hering to be right. Later, when Russell De Valois (1965) had already found spectral activation and inhibition of cone-opponent cells in the lateral geniculate nucleus (LGN) of the macaque monkey, these findings were regarded as further confirmation of Hering’s opponent theory in the primate. It became common among neurophysiologists to use color terms when referring to opponent cells, as in the notations ‘red-ON cells’, ‘green-OFF cells’, ‘þR–G’ and ‘þG–R cells’ (Wiesel and Hubel, 1966; De Valois and Jones, 1961; De Valois, 1965; De Valois et al., 1966). In the debate that followed, with traditional colorimetrists still clinging to the old trichromatic theory, some psychophysicists were happy to see what they believed to be Hering-opponency confirmed at an objective, physiological level. Consequently, little hesitation was shown in relating the unique and polar color pairs directly to the opponency found in these cells. In the so-called ‘R–G’ opponent cells, the activating and inhibiting inputs were mainly from L- and M-cones, and together with the cone-opponency ‘B–(RþG)’ these directions in cone excitation space were later called ‘cardinal directions’. Despite evidence to the contrary (Valberg, 1971; Burns et al., 1984), textbooks have to this day repeated the misconception of relating unique hue perception directly to the responses of such opponent cells and the cardinal axes in color space. The apparent analogy with Hering’s hypothesis was carried even further to imply that, for instance, red results from activation of ‘R–G’ cells and green from inhibition of the same cells [blue would be caused by activation and yellow from inhibition of the ‘B–(RþG)’ cells]. This use of color names on the early opponent responses was an understandable mistake in view of Hering’s postulate, and will be discussed later.

The retinex theory

Parallel to the new neurophysiological exploration of color vision, scientists’ interest in color vision had also developed in another direction. In 1958, 100 years after von Helmholtz and Maxwell, Land (1909–1991), the inventor of the polaroid filters and sunglasses, and of ‘instant photography’ (the Polaroid camera), gave some spectacular color demonstrations at the Annual Meeting of the Optical Society of America. The classical theory said that we would need three primary colors, for instance three slide projectors, one with red light, the other with green and the third with blue, in order to produce all other colors in an additive mixture on the screen. Yet Land surprised the world by projecting a black and white image of a bowl of fruit with only two primaries but nevertheless producing all colors. He had left out the red filter in one of the projectors, so that the projector projected only white light (the mixture on the screen being between green, blue and white). However, the image on the screen nevertheless contained red, although not so vivid a red as with the red filter in place. If he put the red filter back again and removed the green filter, green was still seen in the projected image, and the same happened with blue. This demonstration was not consistent with the belief that there was a close correspondence between the spectral distribution of a color stimulus and its appearance, i.e. that a patch of red color would require radiation from the long-wavelength (red) end of the spectrum, blue from the short-wavelength region, and so on. Land demonstrated that it was not necessarily so, and this came as a great surprise to many people (Optical Society of America, 1994). He claimed to have demonstrated that the Young–Helmholtz three-receptor hypothesis was wrong (Land, 1959).

In reality, Land had given another, and particularly striking, example of the strange phenomenon that had already been observed by Leonardo da Vinci, that adjacent areas of different colors within the field of view influence and change each other’s appearance. This phenomenon was in no way new; it was a particular case of simultaneous color contrast that had also been observed by Goethe (1963) and made famous by him. Colors of surfaces are in reality, although this is not always easily observed, a complex product of all the light imaged on the retina, including those in the periphery. In 1839 the French chemist M.E. Chevreul (1969) devoted a whole book to this phenomenon, which has fascinated color researchers ever since.

Chevreul worked in the famous Gobelins tapestry factory in France, and had to deal with customers’ complaints that the colors of the final product came out differently from what had been agreed upon (for instance when a pattern or texture of the same material changed its appearance due to the background).

Thus, Land’s demonstrations were a new twist to an old phenomenon. He nevertheless proposed his own theory, the retinex theory (retina þ cortex), to explain it. When these effects were again brought to attention in 1983 after a long period of silence, Land (1983) had come to regard the retinex theory as an explanation of color

constancy. Color constancy refers to the fact that, when illumination changes, say from bluish daylight to a yellowish-red evening sunset, the color appearance of the objects reflecting these different spectral distributions does not change in the expected way. A banana is yellow in either case. The bananas and the other objects resist the color change expected from the conventional three-receptor theory (which would be in the direction of the color of the illuminant). This ability of the visual system to adapt to and to neutralize the prevailing color of the illuminant is clearly not what one would expect from the Young–Helmholtz hypothesis. However, the retinex hypothesis was not able to predict the changes that actually occur any better than other hypotheses that had failed in describing color constancy and color adaptation.

According to Land, the Young–Helmholtz hypothesis was now obsolete. However, color scientists did not show much interest. Land himself admitted in private (personal communication) that, had he not been the founder of the Polaroid Cooperation, with great influence and considerable resources, the criticism would have defeated him. This fascinating story is worth closer study by someone interested in the sociology of science.

With his publications in 1983, a new chapter in the history of color science was about to be written, inspired by Land’s stubbornness, but also by the need to develop algorithms for color constancy that could be used in artificial vision and digital photography. By this time, neurobiologists had developed methods that allowed them to record from single cells in the retina and in the brain. Many of them were fascinated by Land’s demonstrations of color constancy in complex Mondrian-like images. Well-known neurobiologists became Land’s protagonists, among them influential people such as the Nobel Laureate David Hubel in the USA, and Semir Zeki in the UK. Land cooperated with them both, and he therefore became quite influential in their thinking about color and color vision (Creutzfeldt et al., 1979; Land et al., 1983; Zeki, 1993; Crick, 1994).

Color in current neuroscience and neurophilosophy

It is tempting to see a link between the renewed interest in color vision and the extensive exploration of the separate visual pathways from the retina to the visual cortex that flourished at the end of the twentieth century. We have already mentioned the particularities of the parvocellular, koniocellular and the magnocellular pathways (see p. 121), and here we only want to emphasize once more the role played by the opponent cells in color vision. These cells combine cone inputs in an opponent manner, much like what was postulated by Schro¨dinger in 1925. The neurophysiological interest in color vision has not yet brought us much closer to an understanding of the physiological mechanisms behind color constancy, Land’s important problem.

Until now neuroscientists have been concerned with the neural correlates of color perception, i.e. how cell activities could represent the intrinsic, abstract structure of color space. For instance, if a cell increases its firing rate to a patch of red spectral light of increasing intensity, and decreased its activity to mid-spectral green lights, one might conjecture that, either:

1.the cell is, via inputs from preceding neurons, excited by electromagnetic radiation from the long-wavelength end of the spectrum, i.e. to the physical property of the light; or

2.the cell responds to the red color; i.e. there is a correlation between the perceived quality and the cell’s response, independent of the spectral composition of the stimulus.

Quantitative measurements in the retina and in the LGN point to the first alternative, except for a few special cases that have not been adequately confirmed. It has been reported, for instance, that cells in area V4 of the visual cortex respond to the color and not to the wavelengths of a stimulus (Zeki, 1980, 1983a, b). Whatever the case may be, the processes behind color qualities, say the redness of red, are still enigmatic.

Now someone is likely to ask if it is at all possible to come closer to the actual phenomenon of color qualia. The activity associated with perception can be measured at different levels in the nervous system, and we can localize specific activities to brain areas by means of positron emission tomography (PET) and functional magnetic resonance imaging (fMRI), but do we get any closer to the conscious experience of qualities by such means?

This discourse brings us to a debate that still lingers, about the role of colors and other qualia in understanding neural activity and conscious experience. On one side of the argument, we find Francis Crick, the discoverer of the structure of DNA and Nobel Laureate, together with the ‘neurophilosophers’ Patricia Churchland (1994) and Paul M. Churchland (Churchland, 1986; Churchland and Sejnowski, 1992). For them the mind was interesting and mysterious, just like matter, and their approach to the problem of consciousness was materialistic. With the rapid advances in neuroscience, the philosophical reflection about neural processes and mental activity could be related to a variety of experimental data. Crick and his colleges at the Salk institute in California mapped the functioning of the visual brain, and the working hypothesis was mechanistic. It can be described as follows (Crick, 1994, p. 3):

‘your joys and your sorrows, your memories and your ambitions, your sense of personal identity and free will, are in fact no more than the behavior of a vast assembly of nerve cells and the associated molecules in the brain’.

Crick, and many with him, would have been satisfied if they were able to develop a neuron theory that answered the following questions: ‘What are the co-variations (correlates) between the different color experiences and neural activity?’ and ‘Which neural representations of color perception exist?’

When you ponder over the link between your conscious experience of colors, e.g. the redness of an apple, and physical stimulation giving rise to it, maybe you will find that it is more appropriate to inquire into the relationship between your color experience and the activation of nerve cells in the visual system. Is there a chain of causal relationships between light, a particular physical property of the apple’s surface, the intervening evoked neural activity in the retina and the visual pathways, and perception of red? Are the same nerve cells activated when, in my dream, I ‘see’ a beautiful bush of red roses in the evening sunset? Or should we, because of the ‘explanatory gap’, i.e. the problems of dealing with color qualities (qualia) and neural processes or physical stimulus properties in the same scientific language – talk not about cause and effect, but rather about correlation, co-variation and structural correspondence?

A surface with a certain spectral reflectance can take on virtually any color, less dependent on the illumination and the reflected spectral distribution than on the surround conditions and the adaptation of the observer’s eye. These effects were well known in the last century, and they were again brought to our attention by Land’s (1959, 1983) vivid demonstrations. The same object surface may look quite different in a normal visual environment than when illuminated in isolation in an otherwise dark laboratory. Generally, a particular color can neither be correlated with nor ‘caused’ by the spectral composition of the isolated patch, or even by the responses of the three cone types excited by it. Only in well-controlled laboratory experiments it is possible to establish a correspondence such as that postulated by von Helmholtz (1911, p. 119; see also p. 278).

At a first glance, the observation that color perception is a conscious, subjective and qualitative experience is less problematic than the attempts to relate it to some physical property of objects, or to neural activity. For those who regard consciousness as a subject for scientific investigation, now is the time to attack the problems of qualia by experimental methods. In Francis Crick’s own words (Crick, 1994, p. 9):

the problem of qualia – for example, how to explain the redness of red. . . is a very thorny issue. . . The problem springs from the fact that the redness of red that I perceive so vividly cannot be precisely communicated to another human being. . . This does not mean that, in the fullness of time, it may not be possible to explain to you the neural correlate of your seeing red. In other words, we may be able to say that you perceive red if and only if certain neurons (and/or molecules) in your head behave in a certain way.

If we substitute ‘you perceive red’ in the last sentence by ‘you perceive a quality which you have learned to call red’, we see that such an explanation of color qualia is indirect. It relies on hypothetical neural behavior and on representations – on linking hypotheses – in a symbolic language. Crick’s proposal is not concerned with qualia in itself (‘the redness of red’), but with the idea of a neural state leading to the same subjective percept in different situations.

A large part of primate cerebral cortex is devoted to processing information received from the retina, and much progress has been made in elucidating the organization and function of the visual cortex (Van Essen et al., 1991). However, the

various aspects of subjective qualities are less well understood (Gazzaniga et al., 1998). Color perception is well suited to bring forth the pertinent aspects of qualities and conscious experience (see for instance Chalmers, 1995b; Crick and Koch, 1995). Today, the study of color perception is closely tied to experimental psychophysics and neuroscience, and even philosophical reflections about this issue are submitted to rather strong experimental constraints (Hardin, 1988; Thompson, 1995; see also the discussions of Saunders and van Brakel, 1997; Palmer, 1999). A neuroscientific program searches for collections of neurons in the retina, the LGN, and higher brain areas whose activities, in one way or another, are correlated with the dimensions of color perception. One records the activity of these neurons when color stimuli change along the psychophysical dimensions luminance, dominant wavelength and purity, in the hope of finding links to the related perceptive properties lightness/brightness, hue and color strength (chroma). Even if it should be demonstrated that a simple neural– perceptual parallelism or isomorphism is too naive a hypothesis, such attempts give us useful overviews of the neural representations of different color stimuli. This project would be completed when unequivocal correlates between the perceptual properties and accompanying neural activities were found.

However the neural activity correlated with my experience of a color consists of a stream of short electrical impulses within a network of neurons, i.e. a biophysical process. These correlates do not describe the nature of my subjective experience of, for instance, the redness of an apple and how it is brought about. My experience of the red quality depends on my living brain, but it cannot be explained or deduced in a causal manner from a biophysical process in and between its neurons. How brain activity gives rise to conscious experience remains an enigma.

Such questions, which were also investigated by the Austrian philosopher Ludwig Wittgenstein (1979) in his book Remarks on Color, has occupied the Australian David J. Chalmers (1995a, b; Tucson, 1996). Chalmers has provoked neuroscientists and neurophilosophers by statements such as:

‘the one who is only interested in the functioning of the brain, he deals with the easy and resolvable problem. However, behind this lies the real problem, namely the one about conscious experience’.

Chalmers’ point is that conscious experience is different from and something more than nerve activity; it is an enigma, and it will probably continue to be so for a long time to come.

Between the two extreme positions of Crick and Chalmers it is possible to claim that, when a color ‘takes place’, this is something more than a physiological reaction to a physical stimulus resulting in a percept. The color embraces both positions, it is a mediator between processes in outer nature and the subjective inner nature. In addition, it is a phenomenon in its own right governed by natural laws (like in color constancy, additive color mixtures, etc.).

It is not fair to reduce the genuine experience of a subjective reality to something else, to entirely substitute it for another experience, as in the following: ‘sound is just a train of compressive waves traveling through the air’; ‘pitch is identical with

frequency’; ‘light is just electromagnetic waves’; ‘color of an object is identical with a triplet of reflectance efficiencies’; ‘warmth and coolness of a body is just the energy of motion of molecules that make it up’. You may have heard these or similar statements from your high school teacher or read them in a text book. Phrases like these are forgivable when used in everyday speech, but in science, and especially in introductory texts in neuroscience and neurophilosophy, they can only cause confusion and serve as obstacles to understanding. Not only do they ignore the essential transformations that take place in sensory organs and in the neural system, but in equating physical concepts with subjective impressions they also deny the existence of qualia. In interdisciplinary research, one must carefully choose one’s language, be it the terminology of physics for electromagnetic radiation, that of neurophysiology for describing neural activity in the brain, or a psychophysical or psychological account of our perceptions. In every field it is important to do justice to the relevant discipline – and to the phenomenon itself.

Colors continue to fascinate and to challenge our imagination. The many different color theories and models that have been developed to this day for explaining how colors ‘take place’ call for caution and skepticism towards such theories. They demonstrate the need to ascribe to subjectively experienced qualities a reality of their own, independent of culture and the level of knowledge which at any given time sets the background for our theoretical descriptions. In this way, the history of color theories illustrates the role that we have given ourselves, the roles we have played – and continue to play – in attempts to understand the physical world.

What needs an explanation?

How can we explain the following facts about color vision?

1.Metameric stimuli with different spectral distributions look alike (they have the same color).

2.Monochromatic lights of different wavelengths have different colors, and this qualitative difference cannot be eliminated by adjustments of intensity.

3.Three colors are sufficient to match all other colors in an additive color mixture.

4.Some people cannot distinguish red from green, and others confuse yellow and blue.

Below we shall see how these experiences can be explained by receptor physiology. In addition there are some other color phenomena that cannot entirely be explained at the receptor level, such as (a) the Bezold–Bru¨cke phenomenon and (b) the Abney effect.