In this chapter, we explore the specialization of colour vision in an evolutionary setting. We discuss how specialization evolves, and show how a proper conception has important consequences for how we should understand colour experience. We shall find that perceptual specialization is best accommodated by the Sensory Classification Thesis.

I. The Multiple Emergence of Colour Vision

It is a remarkable fact about good colour vision (trichromatism) that (a) only a few types of animal possess it (aside from birds, among which it is relatively common), and that (b) these are phylogenetically widely separated. Among mammals, only some primates have trichromatic colour vision. But good colour vision occurs in nearly every broad group of animals: insects, reptiles, birds, fish, mammals. The wide phylogenetic separation of colour perceivers suggests that colour vision is latent in almost all phylogenetic groups, a genetic possibility that can emerge with relative ease when the circumstances are right. For if you pick two colour-seeing species from distinct phyla—pigeons and humans, for instance—you will find that there is no direct line of descent that links these two species to a common colour-seeing ancestor. In other words, good colour vision arose independently in these two organisms.

It seems, then, that trichromatism emerged de novo in each phylogenetic group in which it occurs. Some evolutionary theorists assume that this is a case of convergence, that is, a case where exactly the same function evolves independently in more than one species. This is implied also by most philosophical views, for they do not admit the possibility that colour vision could have different functions in different phyla. After all, the function of colour vision is to detect colour, is it not? How could such a function be differentiated in its multiple occurrences? In this section, we argue that these tendencies are off the mark. Through a consideration of how biological speciation occurs, we will argue that the history of colour vision consists of the emergence of different functional specializations, albeit on the same biochemical substrate.

(p. 178 ) Normally, the independent emergence of any one trait in different phylogenetic groups is regarded an improbable conjunction of events. How then can colour vision have emerged independently so many times? In order to answer this question, we must first look at what all biological colour-vision systems share. The chemicals in cone cells that are sensitive to different wavebands are based on protein molecules known as ‘opsins’. Opsins evolved about a billion years ago in bacteria, where they were and are still used for photosynthesis. They were then co-opted for differential colour response relatively early in their history (Allman 1999, ch. 1). In Halobacterium salinarium, a unicellular organism which lives in salt marshes, a light-sensitive pigment closely related to rhodopsin, which is used for colour vision in vertebrates, is used to guide the organism towards sunlight. In another unicellular organism, Chlamydomonas, another pigment closely related to vertebrate rhodopsin is used to guide the bacterium towards its energy source. Significantly, Chlamydomonas's use of colour information requires short-term memory, since it demands the comparison of light intensities over short periods of time. (Thus, it has some claim to satisfying the Functional Definition of Colour Vision stated in Chapter 6, section III.) This shows that the material basis for colour vision—i.e. the opsins—is very ancient, and can likely be traced to a common origin, or at least to a few independent events that occurred very long ago in bacteria. In time, these opsins came to be localized in structures specialized for differential wavelength sensitivity, the cone cells. Cone cells incorporate stacks of photosensitive membranes incorporating the opsins; they can be activated by the capture of just a single photon. Notably, these cone-cell membranes are descended from bacterial cilia. This progression from ancient opsins to highly sensitive cone cells marks a classic case of evolutionary optimization.

Now, possessing a single opsin is not enough for colour vision. Good colour vision requires three different visual pigments, as we saw in the previous chapter (section I.1). What is required, then, is the emergence in an organism of several opsins, each sensitive to a different waveband in the visible spectrum. Now, the opsins of the short-wavelength cone differ from those of the long by the substitution of one amino acid for another. Opsins proliferate by just such small substitutions. There is, however, a limit to the number of such substitutions, and thus a limit to how many classes of cone cells are available to natural selection. On the other hand, all visual organisms possess some opsins, and thus the proliferation of photoreceptor types within the above-mentioned limits is frequently available to natural selection. For judging from the considerable variety in cone-cell sensitivities over the human population, mutations that modify opsins occur regularly, and these mutations will be selected and persist if they bring an advantage. This explains the latency of colour vision in all phyla. On the other hand, new and different colour-sensitive receptors have never evolved. The front-end of colour vision, the receptors that are differentially sensitive to light of different wavelengths, (p. 179 ) is (a) of ancient origin, (b) more or less fixed in structure, and (c) readily available.

Now, given that colour vision is latent in all phyla and readily available to all classes of organism, why don't all organisms possess it? Would it not be an advantage for all organisms? There has been, as we saw a moment ago, steady improvement in photoreceptor sensitivity and structure. Why has there not been a similar steady and universal improvement in colour vision? The answer is to be found in a principle enunciated by Darwin. Darwin realized that speciation occurs when a sub-population of a species specializes. This sub-population distances itself from the ancestral population by developing the ability to exploit environmental resources in novel ways. By doing this it benefits from reduced competition. Thus, if a species is able to modify a pre-existing facility to help forge a new way of life, it is enabled to exploit resources that are not available to other species, and it frees itself (and the ancestral species) from competitive pressure. This is called the Principle of Divergence. (Gould 2002,234–50, is an informative discussion.) It suggests a reason why natural selection takes up the opportunity for trichromatism so sporadically. The reason is not, as many have found obvious, that it is an advantage for all; rather, it is a specialization which enables some organisms to occupy a niche that their predecessors did not occupy. I remarked two paragraphs ago that the optimization of photosensitive receptors was a steady progression. My present point is that the history of colour vision is not similarly a story of progress across all branches of the tree of life: colour vision emerges by divergence.

When colour vision emerges, it is because it enables a particular species to distinguish itself from the ancestral population by exploiting a resource that this ancestral population could not use. Ex hyposthesi, the ancestral species lacks colour vision, and continues on without it. Among mammals, colour vision emerged among old-world monkeys. The ancestors of this group, the new-world monkeys, lack good colour vision for the most part. These two branches of the phylogenetic tree exist simultaneously: there are still monkeys that lack colour vision. This implies that the non-colour-seeing monkeys are able to exploit their distinctive niches without using colour vision. Thus, one should not assume that colour vision is advantageous for all forms of life. We should look for its value in the difference between populations that possess it and ancestral populations that do not. If most vertebrates lack colour vision, it is not because evolution could not have provided it: it is rather that in most vertebrates this facility has utility only relative to a specialized and differentiating style of life.

Colour vision is widely assumed to have the same function wherever it occurs. The Principle of Divergence shows that this is unlikely (at least if the function is fully specified). The pressures that made colour discrimination valuable for a honey bee in competition with its ancestors are unlikely to have (p. 180 ) been exactly the same as those that made it valuable for old-world monkeys in a quite different situation. Thus, we should expect that old-world monkeys use colour vision in a different way from honey bees. This argument generalizes: since colour vision supports divergent modes of life in the widely divergent species in which it occurs, it is to be expected that colour-vision systems in divergent groups will possess different characteristics. However, as we saw earlier, the front-end of colour vision—the receptors that discriminate wavelength differences in incoming light—is more or less fixed. Consequently, the differences needed by the Principle of Divergence are achieved by specializations downstream of the receptors: (a) the addition of filters, polarization detectors, and other such gadgets, and (b) new forms of colour-processing. It is here that the causes of specialization is to be found.

These considerations suggest that each different type of trichromatic colour-vision system will focus on different colour-detectable environmental characteristics. It is not the detection of colour as such that provides each colour-seeing organism with an advantage, but the use of colour differentiation for other purposes. The idiosyncrasies noted in the previous chapter are signs of the specialized nature of the niche that various colour-seeing organisms occupy. Since each species performs a different set of tasks, each must measure the environment in its own way. Each colour classification scheme must adapt to different external realities.

II. Specialization in Primate Colour Vision

Now, what is the differentiating advantage that human, or more accurately, primate colour vision provides? J. D. Mollon and co-workers (Regan et al. 2001) have advanced the hypothesis that it co-evolved with the colours of small fruits and berries.

(i) Most primates eat fruit, and many eat it in large quantities …

(ii) At the same time, the plants whose fruits the primates eat are competing for seed dispersal. Effective dispersal of seeds is critical to reproductive success in plants …

(iii) The set of traits shared by fruits dispersed by a particular class of consumer can be interpreted as adaptation to that dispersal agent, and is known as a dispersal syndrome. Characteristics of the primate seed-dispersal syndrome are a yellow, orange or red colour, which makes the fruits conspicuous (at least, to a trichromatic consumer) … (234).

It seems that certain plants evolved fruit that are nutritious to primates, and visible to them, while primates evolved the kind of vision that makes it easy to find these fruit from a distance against a leafy background, and under leaf colour. This relationship is a specialization for both parties: it turns out, for example, that fruits less conspicuous to primates are specialized for (p. 181 ) consumption by nocturnal animals, while the conspicuous fruit are particularly well suited for dispersion through the digestive tracts of primates with colour vision. This whole system of mutual dependence starts with colour. The fruit-colour evolves in such a way as to be conspicuous to primates with emerging trichromatic colour vision; primate colour vision in turn evolved and took hold because it offered its possessors the opportunity to focus on these particular kinds of fruit. (A competing hypothesis holds that primate colour vision evolved to aid in foraging for young leaves, which are often red in colour—see Dominy and Lucas (2001) and Surridge et al. (2003). No attempt is made here to adjudicate this dispute; we stick with the fruit hypothesis for no other reason than convenience.)

Now, what exactly is the specialization that colour-seeing primates possess? In order to answer this question, we must compare these primates with their dichromatic cousins. It will be recalled (Chapter 6, section II) that the ancestral mammalian colour-vision system simply divides the visible spectrum into long-wavelength and short-wavelength halves, measuring which is dominant. This corresponds to the ‘warm’ and ‘cool’ colour sensations. It turns out that this ancestral system is ill-adapted to searching for fruits among foliage, ‘because it has poor spatial resolution, and the fruits eaten by primates, when seen at a distance, are usually too small to be resolved by this subsystem’ (Regan et al., 2001, 241). The ideal colour-vision system for the task would make the fruit ‘pop out’ against the background; that is, it would minimize the search time for such fruit, and make it relatively invariant relative to the number of green leafy distractors. This enables a monkey surveying the arboreal scene from afar quickly to detect where its chosen fruit are, rather than having to inspect each tree close up.

Now, it is likely that fruit colour evolved just so as to be conspicuous in this way to colour-vision systems with the peculiar characteristics of primate colour vision. For what is needed is that the chromatic difference between these fruit and the leaves be maximized, while the difference among the leaves themselves should be minimized. As Regan et al. say, ‘We therefore consider the optimal photopigments to be those that maximize signal-to-noise ratio for detecting fruit targets against the visual noise of the leaf distractors, rather than the number of fruits that lie one or more JNDs (just-noticeable-difference steps) from the leaves’ (2000, 240). In other words, it is not sufficient that these kinds of fruit should simply be distinct from the leaves, for instance that they should cluster in one half of a straight line in colour space while the fruit cluster in the other half. Instead of this kind of difference in degree, there needs to be some sort of sharp qualitative difference which enables them to pop out. It turns out that the peak sensitivities of primate colour receptive cones are well spaced with regard to differentiating red and orange from green.

There are, then, two contributions that primate colour-vision systems make to carving out an evolutionary niche that includes the consumption (p. 182 ) of a particular set of fruit. The added colour information provided by this system is (a) fine-grained enough that it can detect relatively small fruit from a distance, and (b) it makes the fruit pop out against a background of leaves. Notice that the niche-creating advantages provided by colour vision do not include the discrimination of colour close up in good light. The special advantage that primate colour systems possess is not that they can process relevant information about these fruit from close up.

Keeping these specialized functions in mind, a comment made by Gerald Jacobs, an eminent student of comparative colour vision, becomes highly relevant. According to Jacobs (1981):

In testing large numbers of subjects from various mammalian species, some of which have ‘good’ color vision, I have frequently observed that given a luminance difference as cue, the animal frequently uses that as a basis for discrimination even if a color difference known to be discriminable is also available. (169)

In other words, if information about a particular surface-feature is available in luminance differences, the organism will appeal to the latter. Thus, ‘given a luminance difference as cue, [an] animal frequently uses that as a basis for discrimination even if a color difference known to be discriminable is also available’ (ibid., 159). Thus, in ‘a multi-hued world in which objects appear to merge and contrast by virtue of their differences in color’, people often mistake luminance-contrast for colour contrast. We may think that we are using colour discrimination whenever we open our eyes: in fact, we often use black and white information in its place.

Jacob's comments make sense of an observation of Robert Boyle reported by John Mollon (1991). Boyle had occasion to observe a young woman who had lost her colour vision after a high fever. He reported that she was nonetheless able to perform most tasks that depend on colour discrimination reasonably well. However, she lamented that she was unable to pick small flowers in meadow, ‘tho’,' as Boyle writes, ‘she kneel'd in the place where they grew.’ Along similar lines, Mollon (2001) quotes a report on colour blindness delivered to the Royal Society of London by J. Huddart in 1777. Speaking of a shoemaker named Harris, Huddart says:

He observed also that, when young, other children could discern cherries on a tree by some pretended difference of colour, though he'd only distinguish them from the leaves by their difference of size and shape. He observed also, that by means of this difference of colour they could see the cherries at a greater distance than he could, though he could see other objects at as great a distance as they; that is, where the sight was not assisted by the colour. Large objects he could see as well as other persons; and even the smaller ones if they were not enveloped in other things, as in the case of cherries among the leaves. (Mollon 2001, 10)

Diagnosing this particular pattern of failure, Mollon says: ‘We need color vision when the target is embedded in a background that is varying randomly (p. 183 ) in lightness and in form’, a ‘dappled and brindled’ background, as he says elsewhere. Mollon is suggesting that we do not need to use colour vision to discriminate large targets against invariant backgrounds of contrasting colour. True, but we can go further. With respect to such large targets, colour-blind people are not just able to discern them against the background, but can often make accurate estimates of what colour they are—and this applies, as we saw in the previous chapter, section IV, not only to dichromats, but also rod monochromats. This is why Ishihara plates are used in the diagnosis of colour-vision deficits: these plates consist of a field of equiluminous dots, with a letter or numeral picked out in one colour against a background of dots in another colour.

To summarize again: colour-blind people are somewhat hampered, but not completely disabled, in colour-naming and colour-discrimination tasks where the contrasting colour blocks are large, especially where the light is good. They are completely disabled only when it comes to performing colour-discrimination tasks in fine spatial detail. It is the latter sort of task that seems to have been involved in differentiating colour-seeing primates from their dichromatic ancestors.

III. Questioning the Colourist Intuition

The facts recounted in the previous section threaten what we may call the Colourist Intuition, a persistent fallacy in philosophical approaches to colour. Imagine a scene that contains a variety of richly coloured things: the interior of a grand cathedral, say, with intensely coloured lights, stained glass, gilt, jewels, polished wood, coloured stone, diffraction fringes from the glass, and so on. Now take a high resolution, natural contrast, black-and-white photograph of the coloured scene. What is missing? Intuitively, the answer seems obvious: colour. Colour seems to be something discrete: something taken from a palette and added to a monochromatic scene, much in the manner of a colourist jazzing up an old black-and-white movie for uncouth contemporary audiences. Intuition tells us that when we change from black and white to colour, we substitute certain experiences for others—reds, greens, etc. in places where previously there was only black and white and grey. In traditional sense-datum theory, as captured in what David Lewis (1999/1966) calls ‘colour mosaic’ theory, there is actually nothing shared between a colour photograph and one in black and white (unless there happens to be a monochrome object depicted in the former). Each and every pixel in the one has been replaced by a different pixel in the other.

The Colourist Intuition encourages a particular form of Universalism about colour. We could introduce it by means of an analogy. Though there are many different kinds of coloured things in it—lights, transparent panes of glass, prisms, reflective surfaces—a coloured painting or (p. 184 ) photograph reduces all the diverse colour-making properties into surface reflectances: the surface red of a painting that depicts a translucent red pane of glass is recognizably the same colour as the pane. Similarly, a television image renders all of the kinds of colours in terms of luminance colours; a colour transparency or slide photograph renders them all in terms of colour transmittances. Thus, it seems possible for just one kind of thing—wavelength-differentiating illumination in the case of the television image, reflectance in the case of the colour photograph, transmittance in the case of the transparency—to capture all of the kinds of colours present in the heterogeneous scene.

The Colourist holds that in something like the same way, colour experience renders all of the different kinds of colour—reflectance, transmission, illumination—into a single medium and takes the measure of them all. Generalizing this across species, he argues that there must be a common measure of all the properties available to any colour perceiver. A bird discriminates heliocentric directions in colour, but what does this mean? Surely, the Colourist insists, that it has experiences similar to those that we have when we view the cathedral, and that it is instinctively able to use these to identify directions. We may see a variety of things in colour, and when we take other species into account, the variety expands even further. But all of these things are seen in colour: the way of seeing them all is the same. There is a palette of pre-existing experiences characteristic of colour vision, however diverse might be the properties that are delivered to us in colour. This is the Colourist Intuition. (Anybody who holds a dispositional theory of colour—the theory that colours are dispositions in external objects to evoke colour-qualia in perceivers who view them—is committed to this intuition. So too are the recent ‘relativists’, who also define colour in terms of colour experience: Jackson and Pargetter (1987), Jonathan Cohen (2000, 2003b), and Brian McLaughlin (2003a and b).

The Colourist Intuition sits uncomfortably with the kinds of data that we considered in Chapter 6 and the observations recorded in the previous section. Why is it reasonable to think that a pigeon has exactly the same set of experiences that we do, given (a) that its colour vision is tetrachromatic, and (b) that its colour awareness is holistically tied to experiences of other properties? Why should we cling to the belief that our own experiences of colour are separate in the way that the Colourist proposes when it is possible to make a lot of the same colour discriminations using information only about luminance? Is it really tenable to think that colour is something of which we find out solely through the operation of the colour-vision system? Despite these perplexities, the Intuition is extremely persistent and difficult to overcome. It is not enough to question or reject the Colourist perspective; we need to replace it with a new approach.

Consider four species: humans and a closely related dichromat species, H′, honey bees and a closely related dichromat insect species, B′. Humans and honey bees have advanced colour vision; H′ and B′ do not. There are two ways of making groups out of these species.

(p. 185 ) Functional groups

1. A. The species with good colour vision: humans and honey bees.

2. B. The dichromats: H′ and B′.

Phylogenetic groups

1. C. The primates: humans and H′.

2. D. The insects: honey bees and B′.

It might seem obvious that since we are investigating colour-vision, it is the functional groups that are important. Clearly, they are important to the comparative study of colour-vision. The species in group A can make complex discriminations of colour, which, in the case of actual biological organisms, means three or more visual pigments and opponent processing. The species in group B lack these endowments, or possess them only in an etiolated form. Nevertheless, it is a mistake to conclude right off, as the Colourist does, that there must be a characteristic kind of datum that is shared by the organisms in the functional group A, or that there is some one thing about which they possess information, in contrast with the organisms in group B. Because humans and honey bees have no common ancestor with colour vision, there is no reason to think that the two colour-vision mechanisms developed as a response to comparable environmental challenges. Humans have single-lens eyes, their colour processing extends beyond the retina into cortical areas of the brain, and they use colour concepts in language-like representations of objects. On all counts, bees are different. On the other hand, bees are tetrachromats and use polarization as an integral part of their visual representations, including colour, and use colour vision directionally. Humans lack these refinements. Why then should we suppose that human colour discrimination is phenomenally or in terms of information comparable to that of bees?

This account of specialization in visual systems leads us into a different way of thinking about the effects of adding of colour dimensions to an ancestral fewer-dimensional system. The Colourist Intuition is influenced by the idea that sensation corresponds to the retinal image. In this Cartesian paradigm, it is natural to think that colour is discrete—it is retrofitted to the ancestral system, replacing much of what in that system was represented in black and white (or warm and cool). Generalizing to other modalities: there is an ambient store of discrete features awaiting capture by the sense modalities. Evolution picks items from this store, and adds them one-by-one to an animal's repertoire as needed.

The Sensory Classification Thesis of Part I offers a completely different perspective. Sensation corresponds to the answers to certain questions posed by sensory-feature detectors: it is a record of specific conditions, of whether or not they obtain. Taking this approach, we see that the questions asked by (p. 186 ) a specialized sensory system will generally add to, modify, or complement those asked by its ancestor. From this perspective, it is much more comprehensible why the phylogenetic groups listed above would share more in the way of sensation than the functional groups, even though the members of the former differ with regard to colour. The phenomenology of added sensory specialization is strongly influenced by what was there before. Evolution is not like a Colourist, painting colours into a scene already delineated. It is much more like a technician tweaking the performance characteristics of a system—changing the sensors, adding filters, rerouting data, fiddling with ‘gain’—to create new discriminations within an older format. This is why colour experience is very likely closely tied to phylogeny. And it is why the phylogenetic groups given above are at least as important in the characterization of colour vision as the functional groups.

Colour is a disunity, then. There is neither a single phenomenology of colour vision nor a set of shared concepts that defines colour wherever it may occur. There is a commonality in the informational material from which colour concepts are constructed; this is inherited from the opsins that constitute the basis for any colour-vision system. Consequently, there is a functional commonality in the mechanisms that are needed to gather this information, but, as the Disunity of Colour Thesis stated at the start of Chapter 6 implies, no one mind-independent property that all colour perceivers track or detect, no one ecological problem they all try to solve. Considered across biological taxa in all of its occurrences, colour is a heterogeneous collection of perceptual concepts generated from wavelength-sensitive data for a variety of specialized purposes by cognitive systems with different neurocomputational structures and evolutionary histories.

Philosophers are unaccustomed to such heterogeneous collections. They are more used to thinking about mental concepts as based on the duplication of information-processing function in different neural structures. They think that the function of vision pre-exists evolution, and assume that evolution worked, much like a benevolent Creator, to realize pre-existent this function in diverse animals. The truth is that such duplication of function is exaggerated. Function emerged from an evolutionary history; it is not an omega-point at which evolution aims. Hardly ever does the same apparatus emerge twice: history rarely repeats itself; apparent convergences are unlikely really to be so (cf. Matthen 2001). There are many commonalities among organisms, of course, but shared function is largely due to common descent and shared apparatus. Biological kinds are thus to be defined by evolutionary history, not just by function or descriptive essence (cf. Griffiths 1997, ch. 8; Matthen 1998). The same applies to biological information-processing kinds. Colour vision originates more than once—but colour vision performs different functions in its various incarnations. To the extent that colour vision does involve the same kind of functioning across the biological realm, there is common descent.

(p. 187 ) The Functional Definitions of colour vision and of colour accommodate this heterogeneity by defining colour vision by functions shared across all biological colour-vision systems. To recapitulate from Chapter 6, section III, they are:

Functional Definition of Colour

A colour classification is one that is generated from the processing of differences of wavelength reaching the eye, and available to normal colour observers only by such processing.

Functional Definition of Colour Vision

Colour vision is the visual discrimination capacity that relies on wavelength-discriminating sensors to ground differential learned (or conditioned) responses to light differing in wavelength only.

These definitions do not refer to the products of colour processing, whether these be colour classes or colour experiences—these being a heterogeneous collection. Rather, they rely on the ‘front end’, the receptors, which colour-vision systems share.

Colour vision comes in degrees—dichromacy and better, and varieties—differences of similarity structures, of features detected, of behavioural significance. These variations are better accommodated by a set of parameters within which variation can be measured and systematized than by some unitary framework that derives from the human case. The definitions given above mark the ground level of such a parametric approach, the weakest sense in which we can say that something has colour vision, or that a particular sense-feature is a colour. More demanding definitions, including anthropocentric definitions, mark special kinds of colour-vision systems. Such special definitions are useful in coming to grips with particular systems of colour detection. What they fail to do is provide a general account of colour or colour vision.