Colour and qualia

There are two basic philosophical problems about colour. The first concerns the nature of colour itself. That is, what sort of property is it? When I say of the shirt that I am wearing that it is red, what sort of fact about the shirt am I describing? The second problem concerns the nature of colour experience. When I look at the red shirt I have a visual experience with a certain qualitative character - a ‘reddish' one. Thus colour seems in some sense to be a property of my sensory experience, as well as a property of my shirt. What sort of mental property is it? Obviously, the two problems are intimately related. In particular, there is a great deal of controversy over the following question: if we call the first sort of property ‘objective colour' and the second ‘subjective colour', which of the two, objective or subjective colour, is basic? Or do they both have an independent ontological status? Most philosophers adhere to the doctrine of physicalism, the view that all objects and events are ultimately constituted by the fundamental physical particles, properties and relations described in physical theory. The phenomena of both objective and subjective colour present problems for physicalism. With respect to objective colour, it is difficult to find any natural physical candidate with which to identify it. Our visual system responds in a similar manner to surfaces that vary along a wide range of physical parameters, even with respect to the reflection of light waves. Yet what could be more obvious than the fact that objects are coloured? In the case of subjective colour, the principal topic of this entry, there is an even deeper puzzle. It is natural to think of the reddishness of a visual experience - its qualitative character - as an intrinsic property of the experience. Intrinsic properties are distinguished from relational properties in that an object's possession of the former does not depend on its relation to other objects, whereas its possession of the latter does. If subjective colour is intrinsic, then it would seem to be a neural property of a brain state. But what sort of neural property could explain the reddishness of an experience? Furthermore, reduction of subjective colour to a neural property would rule out even the possibility that forms of life with different physiological structures, or intelligent robots, could have experiences of the same qualitative type as our experiences of red. While some philosophers endorse this consequence, many find it quite implausible. Neural properties seem best suited to explain how certain functions are carried out, and therefore it might seem better to identify subjective colour with the property of playing a certain functional role within the entire cognitive system realized by the brain. This allows the possibility that structures physically different from human brains could support colour experiences of the same type as our own. However, various puzzles undermine the plausibility of this claim. For instance, it seems possible that two people could agree in all their judgments of relative similarity and yet one sees green where the other sees red. If this ‘inverted spectrum' case is a genuine logical possibility, as many philosophers advocate, then it appears that subjective colour must not be a matter of functional role, but rather an intrinsic property of experience. Faced with the dilemmas posed by subjective colour for physicalist doctrine, some philosophers opt for eliminativism, the doctrine that subjective colour is not a genuine, or real, phenomenon after all. On this view the source of the puzzle is a conceptual confusion; a tendency to extend our judgments concerning objective colour, what appear to be intrinsic properties of the surfaces of physical objects, onto the properties of our mental states. Once we see that all that is happening ‘inside' is a perceptual judgment concerning the properties of external objects, we will understand why we cannot locate any state or property of the brain with which to identify subjective colour. The controversy over the nature of subjective colour is part of a wider debate about the subjective aspect of conscious experience more generally. How does the qualitative character of experience - what it is like to see, hear and smell - fit into a physicalist scientific framework? At present all of the options just presented have their adherents, and no general consensus exists. 1 Physicalism and objective colour The question ‘What sort of property is the redness of the shirt?' is sensible only against a background of assumptions regarding what counts as an informative answer. Otherwise, the following naïve response seems appropriate. Colour is one of the features physical objects have, and red is a type of colour. What else is there to say? What makes the question substantive, however, is the background assumption of the doctrine of physicalism. Physicalism is a doctrine accepted by a wide variety of philosophers, despite wide disagreement about its detailed characterization. There are two basic tenets: (1) that there is a privileged level of basic physical properties, its precise inventory being the business of physical science to determine; and (2) that all causal transactions in nature occur by virtue of mechanisms that are ultimately realized by the properties of this basic class. This doctrine is sometimes expressed by the slogan, ‘no change without a (basic) physical change' (see Materialism in the philosophy of mind; Supervenience of the mental). Acceptance of physicalism generates a problem about colour in the following way. The colour of my shirt - its being red - has causal influence; most notably, on the visual experiences of sentient beings. By the second principle of physicalism there must be physical mechanisms that mediate this influence. So how do these mechanisms relate to the colour? There seem to be four possible answers. First, colour is itself a basic physical property, akin to mass, charge, charm and the like. Second, colour is reducible to other physical properties, in something like the way that heat is reducible to mean kinetic energy or water is reducible to H2O (see Reductionism in the philosophy of mind). Third, there is no natural physical property with which to identify colour: colours are mind-dependent properties in the sense that they can only be characterized by reference to their effects on human visual systems. Finally, colours are not real properties of objects at all: in a sense they are an illusion. The first alternative is clearly a non-starter, since colour is not a basic physical property. There is controversy about the second. To explain the controversy, we must review some basic facts about colour vision. What we normally think of as perceived colour is a combination of three properties: hue, brightness and saturation. Hue (essentially the shade of colour itself) is determined by the wavelength composition of the light hitting the eye; brightness is determined by the relative intensity of the light; and saturation is a matter of how much hue relative to white light there is in the stimulus. Now, it might be thought that a specific colour could be identified with light of a specific wavelength, intensity and saturation. Then the colour of a physical object could be identified with a tendency to reflect just that composition of light. However, it turns out that bundles of light with quite different distributions of wavelengths can produce the same effect on the visual system and therefore are perceived as the same colour. (Such distributions are called ‘metamers'.) Also, there is the phenomenon of colour constancy. As illumination changes, say from bright outdoors to indoors, or noon to late afternoon, the composition of the light reflected from my red shirt changes. Yet it continues to look red throughout. One standard explanation of colour constancy is that the visual system takes changes of illumination into account by comparing the light reflected from many objects at once, thus using the contrast as a major determinant of perceived colour. But this makes the identification of colour with any property of the light reflected from an object quite complicated, if not hopeless. In response to this problem, David Hilbert (1987) proposes that we identify colour with the complex, but quite objective, property of ‘spectral reflectance'. An object's spectral reflectance is a function that takes specific distributions of light contained in illuminants as input and specific distributions of light reflected as output. He argues that it is precisely this property that is preserved in cases of colour constancy. What we perceive when we perceive the shirt as red throughout changes of lighting conditions is its surface spectral reflectance. There are various problems with this view, the most notable being the problem of metamers. As mentioned above, two very different light distributions, even against the same background illumination, can yield the same perceived colour. Yet, in Hilbert's view, since the two objects reflecting these two different distributions have different spectral reflectances, they must count as differently coloured. This seems quite counterintuitive, especially if the background illumination in question is broad daylight. (Hardin (1988) dismisses the spectral reflectance view largely for this reason. Hilbert (1987) responds at length to the objection, and the interested reader should consult his discussion.) In view of the difficulties with the second alternative, it might be thought that colour is best thought of as a disposition on the part of objects to cause certain experiences in us. So what makes the shirt red is its disposition, however grounded in its physical properties, to cause a normal human observer, under normal conditions, to have a reddish visual experience. (The historical roots of this position can be found in Locke's discussion of primary and secondary qualities; see Primary-secondary distinction; Locke, J. §4.) Note that on the dispositional view, colour is a real, but mind-dependent property of objects. That is, objects are really coloured, though their being so is dependent on properties of us. Hardin (1988) disputes the dispositional view on the grounds that there is no principled analysis of the phrases ‘normal conditions' and ‘normal human observer'. Thus there is no unequivocal way to attribute colours to objects, since as we change observer and conditions we change the object's colour. On his view the appropriate locus of colour attribution is experience: there are chromatic experiences, but no coloured objects. In our terms, there is only subjective colour, no objective colour. 2 Subjective, ‘qualitative' colour The puzzle concerning subjective colour is an instance - perhaps the most discussed instance - of the general puzzle concerning the qualitative, or phenomenal, character of conscious mental states: the puzzle of ‘qualia'. We will continue to restrict our attention to colour experience, but most of the positions and arguments we review can be applied to the more general issue of phenomenal experience as a whole. Moreover, when it comes to visual experience, colour provides the most compelling challenge to an account consistent with physicalism. Indeed, we now need to ask the same question about subjective that we asked about objective colour: what sort of physical property could it be? One obvious way of sorting properties is to distinguish intrinsic from relational properties. To a first approximation, intrinsic properties are those that an object possesses in virtue of conditions wholly within the object. Relational properties are those that an object possesses in virtue of its standing in some relation to other objects. So, for instance, my body's mass is an intrinsic property, whereas its weight is relational, since it depends on my body's location relative to the gravitational field generated by another object. (I weigh less on the moon than on the earth, though my mass is the same.) Let us call the reddish qualitative character of my visual experience of the shirt ‘subjective red'. Is subjective red an intrinsic property of my experience? At first blush, it certainly seems to be. But, given the constraints of physicalism, if subjective red is an intrinsic property of my experience, and if my experience enters causal transactions with other states (or events) in virtue of its qualitative character, then subjective red must be somehow reducible to, or composed of, a physical property of my nervous system. 3 The identity theory and its problems This latter line of reasoning has led many to what has been called the ‘central state identity theory'. On this view, an experience is identifiable with a neural state, and its (intrinsic) properties are ultimately neural in character (see Mind, identity theory of). Three sorts of challenges have been mounted against this identification: (1) the conceivability argument, (2) the multiple-realizability argument, and (3) the knowledge argument. According to the first, since it seems conceivable that one could be having an experience of a reddish sort but not be in the neural state in question, the experience cannot be the same as the neural state. However, put so baldly, this argument is clearly fallacious. It is conceivable that water might not have been H2O, but that is no reason to deny that it is in fact identical to it (see Intentionality; Propositional attitude statements). However, the conceivability argument seems to many to be getting at something not so easily dismissed, and this is brought out by considering the second argument, about multi-realizability: subjective red cannot be identical to a neural state because it seems possible that there could be creatures with a different sort of physical constitution (say, Martians, or robots) that could nevertheless experience reddish visual sensations like ours. This seems more interesting than mere conceivability, since the possibility seems to survive the discovery that, among all the instances we have observed, subjective red always happens to be realized by a certain neural state. This does not seem to be the case with, for example, water and H2O: if all the water we have observed is H2O, that seems to be a good reason to think that all water is (see Kripke 1980; Essentialism). The last argument, the knowledge argument, is due to Frank Jackson (1982). If subjective red is a physical property, then if one knows all the relevant physical facts pertaining to colour vision one will know all there is to know about subjective red. Imagine Mary, a vision scientist who possesses the complete physical theory of colour vision, but who has lived in a totally achromatic environment her entire life; she has only seen the world in black and white. Upon release from this restricted environment, she encounters a ripe tomato, and exclaims, ‘So that's what red looks like!'. It seems she has learned something new, and yet, Jackson argues, if physicalism is true she should already have known all there was to know about colour experience. Hence, subjective red is not a physical property. Now, one reply to this argument is the same as to the conceivability argument. ‘Knowledge', another intentional term, is referentially opaque: the fact that Mary knows all the physical facts and yet cannot recognize subjective red from its physical description does not show that subjective red is not identical to a physical property; after all, the fact that she might know all the physical facts about H2O without recognizing that it is water does not show that water is not H2O. Mary may, in one sense, know everything there is to know when she learns all the physical facts, just not under every possible description. But that constitutes no problem for physicalism. Still, however, many philosophers feel that these replies are missing the point. What provides plausibility to the three arguments is that appeal to the neural properties of one's visual state does not seem to explain its qualitative properties. The fact that these neurons are firing in this way does not really explain the shirt's looking ‘reddish'. This is known as the ‘explanatory gap' argument (see Levine 1983, 1993; Explanation). It could be put this way: once we discover the chemical composition of water, and know various chemical laws governing the states and interactions of physical substances, we can explain the features of water that we initially used in picking it out. For instance, we can explain why it freezes and boils at the temperatures it does, why it is necessary for life, and a host of its other properties. In fact, it is not really conceivable that something could be H2O and yet fail to manifest these various properties, at least so long as we fix basic chemical laws. This seems to be what at least the conceivability and knowledge arguments are getting at: that the physical facts do not ‘upwardly necessitate' facts about qualitative experience in the way that chemical facts necessitate facts about water. No matter how detailed our knowledge of the physical mechanisms by which neurons transmit their impulses from one to the other, we are still left with the question of why all this electrochemical activity should constitute an experience with a reddish (as opposed to a greenish, say) qualitative character. It still seems quite conceivable that such electrochemical activity should occur and yet not the experience of subjective red, and this residual conceivability is a manifestation of the explanatory gap. To put the matter in terms of the knowledge argument, the fact that Mary would learn something new upon emerging from her achromatic world is evidence that the physical theory of colour vision does not really explain the qualitative character. Even if she knew the meaning of ordinary ‘red' she could not deduce from the rest of her knowledge that tomatoes look red (in the way that she could deduce that water freezes at 32±F). One might object that appeals to neurophysiology can indeed provide an explanation of subjective red. For one thing, if we were to find a stable correlation between subjects' reports of reddish sensations and their occupying certain neurophysiological states, then we could predict the occurrence of these sensations from knowledge of their brain states. For another, knowledge of the neurophysiological properties explains the mechanisms by which subjective red performs its cognitive functions. Colour perception involves selective sensitivity to fairly complicated properties of the light reflected from physical surfaces, and we now know a lot about how that sensitivity is implemented in neural hardware. The problem with the first consideration is that in place of a genuine explanation all we are given is a brute correlation. I may be able to predict with perfect confidence that when someone occupies this particular neurophysiological state they are having a reddish sensation, but my ability to make this prediction does not manifest an understanding of why this brain state should go with reddish, and not bluish or greenish sensations. Brute facts are just that; they are not explicable. Moreover, they do not support counterfactuals: how are we to judge whether something lacking a specific physical property we possess really would not be having a reddish experience? What could satisfy us about the modal claim that not only are reddish experiences correlated with a certain neural state in fact, butmust be so correlated? Of course sometimes we admit that certain facts are brute facts; for instance, the value of the gravitational constant, the basic laws of physics, and the like. But notice it is a methodological assumption of current scientific practice, a corollary of physicalism, that only at the basic physical level are brute facts to be found. The idea that something as complicated as the brain should just give rise to reddish experience, and there be no explanation of this fact, is inconsistent with fundamental Physicalist assumptions. If there is any hope for explaining subjective colour within a Physicalist framework, it would seem to depend on the second consideration: showing how the neurophysiological states of the brain provide the mechanisms by which colour sensations perform their cognitive functions. But to make this work it is necessary to analyse subjective colour in terms of its function, for otherwise explaining that function would not count as a full explanation of subjective colour itself. In other words, we need to abandon the treatment of subjective colour as an intrinsic property of sensation, and treat it instead as a relational, or functional, property. 4 Functionalism Functionalism is the view that mental states are definable in terms of their causal roles, their causal relations with stimuli, behaviour and other mental states (see Functionalism). Thus, arguably, something is a belief if and only if it is the result of perception or reasoning and the cause, with desires, of action. So a state would be an experience of subjective red just in case it was normally caused by viewing red things, it tended to cause judgments to the effect that something was red, and it generally related to other mental states - in particular through similarity judgments - in the way that is typical of experiences of red. Let us call the functional role in question ‘functional red'. A more specific form of the functional/relational approach is to treat qualitative states as essentially representational, and qualitative content as representational content. The idea is this. When I have a reddish experience I am in the state of representing the object I perceive as having a certain property, in this case the property of being red. Immediately, of course, the question arises how to distinguish colour experience from mere colour belief. The standard reply is to appeal to the special functional role of the visual system, for example, by attributing to it some distinctive representational system all its own. The point is that to have a reddish experience is to represent an external object as red in this special visual way. This is a relational theory since subjective red's identity is determined by its representational content, which is a matter of its relation to external objects, and it is a functional theory, since the particular functional organization of the visual system provides the basis for distinguishing experience from mere judgment. One thorny problem for the representational approach involves the alleged representational content of the sensation. Above we said that subjective red is a representation that some object is red. But what is ‘being red'? As our earlier discussion showed, it is not easy to specify what property objective redness is, so it is correspondingly difficult to specify what the representational content of subjective red is supposed to be. There are various ways a representationalist can respond. If one believes that colour is identifiable with a spectral reflectance function, then that could be the content of the sensory representation. If, however, one denies the existence of objective colour, one could say that our sensory systems represent objects as having a property that they in fact do not have. This is the so-called ‘error theory'. Finally, one might be subjectivist about the content of subjective red itself. That is, one could say that a reddish experience is a report from the visual system concerning its own state - a way of saying, ‘I'm experiencing redly now' (see Mental states, adverbial theory of). Note that it is no objection to the first proposal that the alleged content has a complex theoretical structure and therefore could not plausibly be attributed to our sensory system. There are many representations we employ which in their internal structure are relatively simple and yet they refer to properties or objects which have a complex structure, and about which experts possess sophisticated theories. Take the well-worn example of water. According to most philosophers of language and mind, when I think about water I am thinking about a substance with the structure H2O. But of course I do not have to know this, and it certainly is not plausible to claim that my mental representation of water contains representations of hydrogen and oxygen. Similarly, subjective red could function to detect, or register, a certain complicated property of an object's surface without itself possessing complex structure. 5 Objections to functionalism The basic objection to a functionalist account of subjective colour, or of any qualitative experience, is that it just seems intuitively plausible that functional red and subjective red could come apart. The mismatch goes in both directions. That is, according to the famous absent and inverted qualia hypotheses, it seems quite possible that a creature could satisfy the conditions for functional red even though not experiencing subjective red, or, for that matter, having any qualitative experience at all. On the other hand, it also seems quite possible that a creature could experience subjective red even though most of the causal relations it normally maintains were absent (this is less emphasized in the literature). An example of the first sort of problem is the famous ‘inverted spectrum' thought experiment. As we noted above, colour experience is a function of three basic features: hue, brightness and saturation. Colour quality space can be modelled then as a three-dimensional solid; in fact, a cone. The vertical dimension represents brightness, the horizontal represents saturation, and the circular dimension represents hue. Given this model, there seems to be a mapping of points in the space onto their complements around an axis that bisects the cone through the middle. Reds would be mapped onto greens, blues onto yellows, and so on. The resulting cone would be isomorphic to the original in the sense that all the distance/similarity relations among points would be maintained. Given this characterization of an inversion, take two creatures, one of whom has a colour quality space described by the original cone and the other of whom has one described by the inverted cone. If we are just looking at the relational properties - those involving causal relations to external stimuli, internal mental states, and behaviour - it seems that the two creatures would be functionally isomorphic. Yet, when looking at a red fire engine one would be having a reddish experience and the other would be having a greenish one. In other words, while they would both satisfy the criteria for functional red, one would be having an experience of subjective red and the other of subjective green. So, subjective red cannot be identical to functional red. The ‘absent qualia' hypothesis represents an even more extreme possibility. Ned Block (1980) asks us to imagine the entire nation of China (or any similarly large group of individuals) organized, say, by telephone, so as to realize the functional organization of a human brain. Let us also imagine that this vast network is connected to a robot which has a video camera for an ‘eye'. When the robot is stimulated by light from a fire engine the network will go into a state of functional red, by hypothesis. Yet, it seems quite bizarre to claim that the robot, the network, or the system comprising both, is having a qualitative experience. Thus we have a case of functional red without subjective red (or any qualitative character at all). An example of subjective red without the normal causal connections could occur if someone's normal functioning were disturbed, so that various relations between their colour experience and memory, belief and the like no longer held. It seems possible that this could happen while the qualitative character of their experience remained unchanged. One concrete case of this is colour blindness. The fact that someone cannot distinguish red from green obviously affects the structure of their colour space, yet it is not obvious that this makes their experiences of blue any different from mine. 6 Replies There have been three basic responses to these anti-functionalist arguments. The first is to grant their cogency and claim that for experiences such as subjective red, as opposed to cognitive states such as belief, functionalism is wrong and the traditional identity theory is right. But then one must confront the objections to that theory discussed above: the multiple-realizability argument and the explanatory gap. The second sort of response is to attempt to undermine the intuitive resistance to a relational account represented by the absent and inverted qualia hypotheses. Numerous such attempts have been made, but we will focus on two related strategies in particular: what might be called the ‘asymmetries' strategy and the ‘subtle role' strategy. In the end they are both problematic, but they represent serious attempts to meet the challenge and their problems are instructive. The basic idea behind the asymmetries strategy is to argue that an appropriately chosen and sufficiently rich relational description can uniquely identify a type of qualitative character, and thereby get around the sorts of counterexamples just discussed. Consider the inverted spectrum argument. As presented above, it seems to make the assumption that there exists a natural axis of symmetry dividing the colour cone. However, there is empirical evidence to the effect that this is not so. First, there is the question of the location of the primaries. It seems evident that some hues are experienced as combinations of others, while some seem not to be. Contrast a pure red or green with orange or purple. One might have thought that pure examples of each of these primaries would occupy points in the three-dimensional colour solid equidistant from the origin, representing equal amounts of brightness and saturation, and equidistant from complex hues. But not so. Yellow, for instance, occupies a point of higher perceived brightness than do the other primaries. Also, some primaries occupy a larger region than do others; that is, there are more steps to be taken before we describe what we see as a combined hue. Furthermore, there is the anomaly of brown, which results from darkening yellow, yet is not thought of as merely dark yellow but another hue altogether. The point is that if you were to invert along the red-green and blue-yellow axis, it does not look as if all the relational judgments would be the same. Normals would consider dark and light blues to have the same hue, but inverts would not. Normals and inverts would differ in when they considered hues to be combinations, as opposed to primaries. Functional identity between normals and inverts could not be maintained. Hence, they would not be functionally identical to us, and spectrum inversion would not constitute a counterexample to functionalism. (See Harrison 1973; Clark 1993; Hardin 1988 for arguments along these lines.) The second, ‘subtle role' strategy is aimed more at the absent qualia hypothesis. The idea here is to attack the notion that a creature could be functionally identical to us and yet lack qualia by demonstrating that qualitative character is itself essential to normal functioning. Van Gulick (1993) notes that there is some evidence, for example, from blindsight cases - in which a patient claims not to see anything within a certain region of their visual field and yet can correctly ‘guess', upon prompting, whether or not something is there - that consciousness is necessary for carrying out certain functions (see Consciousness §7; Unconscious mental states). For instance, blindsight patients tend not to initiate action with respect to the objects that they can passively detect in their blind field. If, as such cases suggest, consciousness is essential to the performance of certain functional roles, then it is not possible for a non-conscious state to play the same functional role. Hence, absent qualia are not possible. We said above that there are problems with both strategies. With respect to the asymmetries strategy, there are two basic objections. First, even if the geometric space that characterizes our colour experience is in fact asymmetrical in the way described above, this seems to be a contingent feature of it. Why should this matter? So long as it is not demonstrated that it is essential to qualitative experience that it be immune to inversion, we have no reason to believe that all forms of qualia will have this feature. But if any form of qualitative experience is susceptible to inversion, functionalism is in trouble. Moreover, as suggested earlier by the case of red-green colour blindness, there could easily be cases of individuals whose colour experience is somewhat, but not completely, different from our own. Suppose that theirs is symmetrical, and therefore subject to inversion. Would it not make sense to say that when such a person looked at a red fire engine they were having an experience that was more like our subjective red than, say, our subjective green? But, if inversion were possible for them, we could not say this solely on the basis of their functional organization. The second problem with the asymmetrical approach is that it does nothing to address the absent qualia hypothesis. Suppose that any inversion of primaries would be detectable. Still, given cases like Block's China-head, it seems possible that a creature could realize a structure onto which the complete set of similarity relations definitive of the colour solid could be mapped, and yet there be nothing it is like to be that creature. Addressing the inverted qualia argument alone does not save functionalism. The ‘subtle role' strategy is, of course, specifically addressed to absent qualia. But here too there is a problem. Suppose we have to grant that there may be jobs, or roles, that only qualitative states can fill. That still does not mean that filling that role is what it is to be a qualitative experience. In fact, that very way of putting it - that being qualitative is essential, or necessary to playing the role - seems to imply just the reverse; that being qualitative is one thing, playing the role quite another. Suppose it turned out that being red was essential to some plant's playing the ecological role it played; nothing that was not red could do it. We would not say that being red is to play that role, but rather that being red is what makes the plant in question especially suited to play that role. It seems that the same goes for subjective red - the qualitative experience of seeing red - in the scenario Van Gulick envisaged. Nor is it clear how the ‘subtle role' strategy addresses Block's counterexample concerning the nation of China. Either one would have to admit that the entire nation experienced subjective red, or one would have to argue that the lack of qualitative character in this case posed an insuperable obstacle to realizing the relevant functional description. Both positions are difficult to defend. However, some (see Lycan 1987) have argued that there are other constraints - for instance, extremely fine-grained functional descriptions couched in teleological terms - that would rule out the nation of China as a legitimate realization (see Functional explanation). In discussing the problems for functional accounts of colour qualia, we have not distinguished among various forms of functionalism. In particular, we have not addressed the representational version of functionalism, according to which colour qualia are functionally specific forms of representations of colour properties. It is sufficient to note here that the absent and inverted qualia hypotheses cause problems for this version of functionalism as much as any other. For suppose that an inverted spectrum is possible. Then two states could differ in their qualitative character and yet carry the same information concerning the reflectance properties of distal stimuli. Or, if absent qualia are possible, then a state could carry information concerning the reflectance properties of a distal stimulus and yet there be nothing at all it is like to occupy the state. If either case is possible, one cannot identify qualitative character with the representation of such properties (see Semantics, informational). 7 Eliminativism We have seen that functional/relational analyses of colour qualia face serious difficulties. We have also seen that treating colour qualia as intrinsic properties also encounters a serious obstacle, in the form of the explanatory gap. Of course there are possible replies to these difficulties, and many contemporary philosophers hold one or the other of these views. Yet, the persistence of objections to the various accounts of qualia has led some philosophers to embrace eliminativism about subjective colour (whether or not they, like Hardin, are eliminativists about the objective sort). That is, they see the philosophical problem of qualia as our attempting to identify a real phenomenon in the world which corresponds to a conception to which no real phenomenon does, nor even could, correspond. Belief in qualia is something like belief in ghosts. Just as we do not bother to find some real phenomenon to serve as the referent of our idea of a ghost, so too we should abandon the idea that a real phenomenon could serve as the referent of our idea of subjective red (see Eliminativism). Eliminativism is quite naturally joined to the representational position just described. No one literally denies that we experience sensations of colour. That the fire engine looks red to me is a genuine fact. However, the eliminativist teases apart what is genuine about this fact from what is illusory in the following manner. That I see, and therefore judge, the fire engine to be a certain way, is undeniable. But it is wrong to infer from my perceptual judgment concerning the surface of the fire engine to the existence of a colour-like property of my experience itself - a colour quale. It is the existence of this inner property - around which inversion and absence puzzles take hold - that the eliminativist denies. A major source for eliminativist sentiment is Wittgenstein's famous attack on the coherence of any notion of inner experience that did not manifest itself in outwardly observable behaviour (see Criteria; Private language argument). While Wittgenstein had many targets in mind with his ‘private language argument' - his argument against the possibility of a language that only one person could possibly understand - certainly the idea that I could introspect and discover qualia as properties of my inner experiences was one of them. A number of philosophers have extended this Wittgensteinian critique, most notably Daniel Dennett (1991). As Dennett characterizes them, believers in qualia are tied to a picture of the mind as a theatre (the ‘Cartesian theatre'), in which mental entities are on display before the mind's eye. He has coined the term ‘figment' to capture the unjustified inference from the reality of colours as properties of physical objects to the reality of colour qualia as properties of internal states. Suggestive as this diagnosis of the problem of colour qualia is, it alone does not really constitute an argument in favour of eliminativism. However, various further arguments have been presented in the recent literature. First, the believer in qualia is accused of allegiance to a dualist metaphysics. Second, qualia are disparaged as posits of common-sense psychology, a theory that, like many common-sense theories of the world, must give way before its more rigorous and detailed scientific competitors. Third, thought experiments like the inverted and absent qualia hypotheses are attacked as internally incoherent. Finally, belief in qualia is alleged to lead to unsavoury sceptical consequences. For instance, if someone could be functionally identical to me without having qualia, how do I know I really have them? (Of course some of these same arguments could serve to support a reduction of qualia to functional states, rather than their elimination. The line between eliminativism and reductionism, either of the functional or neurophysiological variety, is not always easy to draw.) Advocates of colour qualia have responded to these arguments. Those who do not straightforwardly embrace dualism deny that their position entails it. Qualia, on their view, are not posits of an outmoded theory, but among the primary data that any adequate psychological theory must explain. The discussion above demonstrates how hard it is really to undermine the inverted qualia hypothesis. As for the charge of scepticism, many dismiss it as just that. Sceptical worries attach to most phenomena, they argue. But whether or not these replies succeed, eliminativist challenges of the sort just described will find adherents so long as a truly explanatory connection between qualitative character and neurophysiological or functional organization is still lacking.