Philosophical Studies

, Volume 156, Issue 2, pp 267–281

Tye-dyed teleology and the inverted spectrum


    • Department of PhilosophyUniveristy of Minnesota

DOI: 10.1007/s11098-010-9580-6

Cite this article as:
Ford, J. Philos Stud (2011) 156: 267. doi:10.1007/s11098-010-9580-6


Michael Tye’s considered position on visual experience combines representationalism with externalism about color, so when considering spectrum inversion, he needs a principled reason to claim that a person with inverted color vision is seeing things incorrectly. Tye’s responses to the problem of the inverted spectrum (2000, in: Consciousness, color, and content, The MIT Press, Cambridge, MA and 2002a, in: Chalmers (ed.) Philosophy of mind: classical and contemporary readings, Oxford University Press, Oxford) rely on a teleological approach to the evolution of vision to secure the grounds upon which people with normal color vision can be justly called ‘right’ and those with inverted color vision can be called ‘wrong’. I demonstrate that since the inverted spectrum thought experiment requires that both sorts of vision be behaviorally indistinguishable, no biologically acceptable concept of teleology will allow Tye to draw the distinction he needs.


TyeRepresentationalismInverted spectrumNatural teleologyEvolutionNatural selectionQualiaColor externalism

1 Introduction

Externalist representationalism1 claims that the phenomenal content of our conscious experiences (the qualia2) are a type of representational content. That is, the reason that a conscious experience of yours has the phenomenal feature of redness is because that phenomenal vehicle (redness) has the function of conveying information about the environment to you—specifically that there is a red thing in your field of vision. Externalist representationalism is attractive for a number of reasons: it tries to do justice to our conscious experience, it aims to give a completely naturalistic account of the mind (allowing us to naturalize phenomenology), and it respects the arguments of Putnam and Burge (for semantic externalism) that so many find persuasive. I will examine Michael Tye’s theory of externalist representationalism3 (especially his handling of the challenge posed by spectrum inversion4—the idea that two subjects could be viewing the same physical object but have different phenomenal experiences as a result of that viewing). Briefly, Tye claims that since the normal person and the person with spectrum inverted vision are having different phenomenal experiences whenever they look at colored items; they represent the world as being in two different ways. The items themselves cannot have incompatible colors on their surfaces, so at least one of the subjects must be seeing the world incorrectly. Tye appeals to evolutionary history to secure the claim that some perceivers get colors right, and others misperceive. I contend that the theory of evolution (and any form of natural teleology based on evolution) cannot do the job that Tye demands of it. Tye must either abandon his attempt to cope with the inverted spectrum (which will also imperil his account of color experience more generally), or he must admit that his theory is in conflict with evolutionary biology (since he would have to significantly revise evolutionary theory before it could determine whether normal or inverted color vision is correct).

2 Tye’s externalist representationalism

Tye claims that qualia should be identified with a certain sort of representational content (content about the things that the mental state represents), and not with any introspectively available nonrepresentational qualities (intrinsic, nonrelational properties as per Block and Nagel). For example, on Tye’s approach, if you look at a ripe tomato in good light, your visual experience relates you to the tomato, and the color of the tomato itself becomes part of the content of your visual experience. That is representational content, since it serves to represent features of the world to you. The phenomenal character of your experience of the ripe tomato (its redness) is present only in virtue of the features of your experience that represent the tomato as having the color red. Because your experience serves to represent the tomato as being colored red, it carries the phenomenal character of looking red. The phenomenal content just is a type of representational content. Tye says, “…[T]he view I accept is that what it is like to have a visual experience (what is sometimes called ‘the phenomenal character of the experience’) is a matter of a certain sort of representational content that the experience has,” (Tye 2002a, p. 448). So, if two experiences have the same representational content, they must also have the same qualitative features.

The problem that qualia pose for representationalism is that it seems possible for the information about the properties of the objects that we perceive to register in consciousness via any number of phenomenal vehicles. Tye takes the qualitative features of experience seriously, accepting the intuition that different phenomenal vehicles could be used to transmit the information about the objects that our senses track. I believe that Tye deserves credit for trying to do justice to our phenomenal experience, but that his attempt ultimately fails.

3 Representationalism and the inverted spectrum

Consider Tom the color-invert, behaviorally indistinguishable from a person with normal color vision. If Tye is correct, then his phenomenal content must be cashed out in representational terms. Since his qualia are different from those of a person with normal color vision, he represents objects in the world as having colored surfaces different from the colors that we would ascribe to those same objects. How do we know who gets the colors right? Considering the possibility that color inverts might outnumber the color normals, Tye says, “It may seem that there is no nonarbitrary way of picking out a subpopulation of normal perceivers whose color experiences do not misrepresent. Any choice of a subpopulation may seem as good as any other. Unfortunately, if that is the case, then there is no fact of the matter about who is misrepresenting,” (Tye 2002a, p. 452). Externalist representationalism needs that fact of the matter to obtain. The question is, how to secure it?

This is not an epistemic point: Tye does not need some way to empirically test for color spectrum inversion. He needs some way to fix, in principle, who is right about color, and who is wrong. Tye turns to the evolutionary history of color vision in order to establish the claim that color-inverts are mistaken in their phenomenal color perceptions. This is the crucial stage in his argument, so I will quote the passage at length:

Suppose, for example, there is a genetic defect in certain humans that are alive today, the result of which is that wires are crossed in their visual system, thereby inducing in them color experiences opposite to those that were present (in the same conditions) in most of their ancestors. Originally only a small subpopulation of the human species had the given defect, but now it has spread so that a sizeable number of us have it. These people have an experience of red when they see green things in daylight… and so on. Their experiences are now tracking colors that are opposite on the hue circle to those tracked by their biologically normal ancestors. Since the visual systems of the members of this human subpopulation are not functioning as they were designed to do, the colors their sensory states would track, were they discharging their biological function, are not the colors they actually track. This is how misrepresentation arises. The nonconceptual visual states of these humans are not tracking the colors they were designed to track. So error enters. Likewise, for other possible subpopulations. (Tye 2002a, p. 452.)

So, for Tye, a particular phenomenal vehicle (greenness, for example) has the biological function of tracking that color when it is encountered out in the world (on green objects). Phenomenal greenness acquired that biological function via natural selection, by conferring an advantage (in survival and reproductive success) to the organisms that possessed it over those that lacked it. Taking all the phenomenal color vehicles together, Tye would say that normal color vision evolved to represent the colors of the surfaces of objects to us using the particular phenomenal vehicles that correctly correspond to the colors of objects. Tom the color-invert might be able to get along in the world just fine, but in his mental life the phenomenal vehicle that evolved to represent green is evoked when he looks at red things, and vice versa. According to Tye, Tom gets colors wrong. There is a difference in the representational content of Tom’s experience, just as the usual intuitions about the inverted spectrum demand, even though this difference does not register in Tom’s outward behavior. If this teleological approach to color perception allows Tye to draw the distinction between normal and inverted color vision, then his externalist representationalism would provide a satisfying account of phenomenal experience.

Unfortunately for Tye, natural selection can only operate on features that make a difference to survival and reproductive success. Since normal and inverted color vision provide exactly the same advantages to the creatures that have them, natural selection cannot be used to draw any distinction between the two. To secure this claim, I now turn to the key concepts of biological function and natural teleology.

4 What is natural teleology?

There is a school of thought in biology (and philosophy of biology) that permits some forms of teleological explanation for biological features, called etiologicalfunctionalism. The goal is to describe the purpose of a biological feature (a structure, organ, or behavior) in terms of the adaptive benefit that the feature delivers to the organisms that have it. In this sense, a wing is for flying, lungs are for breathing, an opposable thumb is for grasping, and so on. These features are teleological because they serve some purpose, not because they were intentionally created. Following Francisco Ayala (1977, 1998a), we can draw a distinction between the purposes that drive the intelligent creation of artifacts (a jug is for carrying water) and the purposes served by the natural adaptations of living organisms. Let us call the first sort artificial teleology, and the second natural teleology (Ayala 1977, pp. 189–191). Tye’s emphasis on evolutionary history for color vision indicates that natural teleology is the relevant concept.

There are serious debates among etiological functionalists about whether teleological functions get their telos from the current benefits that accrue to organisms that have the feature in question (why the feature is maintained) or if the telos derives from past benefits to the organism’s ancestors, even if the feature provides no current benefits (why the feature originated). I believe that Tye can remain neutral on this debate, standing ready to accept whatever version eventually wins out.5 Rather than test Tye’s account against each variation, let’s see if it can stand on the features that are common to all biologically informed accounts of natural teleology.

In order to ascribe a natural teleological explanation to a feature (an organ or behavior), it must have come about via natural selection, in virtue of some advantage in reproductive success conferred to the organisms that possess the feature over those that do not. The concept of differential reproductive success is a complex one, which Ayala explicates in the following manner:

Differential reproduction is a compound process, the elements of which are differential survival, differential mating success, and differential fecundity. Natural selection implies that some genes and genetic combinations are transmitted to the following generation on the average more frequently than their alternates. Such genetic units will become more common in every subsequent generation and their alternates less common. Natural selection is a statistical bias in the relative rate of reproduction of alternative genetic units. (Ayala 1998a, p. 33)

Color vision certainly provides an advantage to those people who have it, so it looks like a prime candidate for teleological explanation. However, the question is not whether we can give a teleological explanation for color vision per se, but whether we can draw a teleological distinction between normal color vision and inverted color vision. By hypothesis, both normal and inverted color vision allow their bearers to draw exactly the same distinctions among the objects in the environment. Since color inversion has no consequences when it comes to relating to the environment and to one’s fellow humans (be they color normals or inverts), natural selection cannot operate on inverted color vision. There simply is no difference between inverted and normal color vision that has any impact on survival and reproductive success. Whether inverted vision spreads through a population of color-normals or not will depend on genetic drift and other factors wholly unrelated to the survival advantages conferred by visual perception. Without natural selection actually selecting between normal and inverted color vision, we cannot offer any teleological explanations of the sort that would draw a distinction between them.6 As Ayala points out, when considering the possibility of adaptively equivalent protein variants, “The presence in a population of one amino acid sequence rather than another adaptively equivalent to the first would not then be explained teleologically,” (Ayala 1977, p. 189). What holds for adaptively equivalent proteins also holds for adaptively equivalent varieties of color vision. The ability to make visual color-discriminations is selected for, but the phenomenal character that transmits each color to consciousness is not. Evolutionary biology directly undermines Tye’s attempt to fix a privileged set of phenomenal vehicles, rather than supporting it.

Once we understand that essential point, we see that Tye’s choice to call color-inversion a ‘genetic defect’ is a mistake. If spectrum inversion were to exist, it would not be a defect, but an equally successful variation. From the standpoint of natural selection, neither would be preferred to the other. Both sorts of color vision serve the same biological function, and they both serve it equally well. So there is no way for natural selection to get any purchase here, and certainly no grounds to call one or the other a ‘defect’. Inverted color vision lets Tom navigate the world and communicate with the people around him just as well as normal color vision would. So, despite Tye’s assertion, inverted color vision is discharging its biological function for the color-inverts, just as normal color vision does for the color-normals.

5 Possible responses

In Consciousness, Color, and Content, Tye addresses a version of the inverted spectrum challenge (proffered by Shoemaker): interspecies spectrum inversion. In the course of that discussion, Tye considers the possibility of large-scale spectrum inversion among today’s human population and claims that it would not undermine representationalism. Tye deploys several counterarguments to Shoemaker, and I would like to begin consideration of Tye’s responses here—to see if Tye’s counterarguments to Shoemaker might also serve to defend against my critique.

5.1 Representationalism is an empirical proposal

As Tye relates the challenge, Shoemaker offers the possibility of an alien species, whose color discriminations (and abilities to deal with colored things) are just as good as ours, but whose color qualia are inverted with respect to our own. Shoemaker concludes that the aliens have just as much right to claiming that their color perceptions are ‘right’, and that there is no way to adjudicate the dispute over who is seeing colors accurately and who is not. Tye addresses the problem in this way:

Granted, these hypothetical creatures have as much a right to accurate color experiences as we do. Without further information, it would certainly be arbitrary to say that we are right and they are wrong at the level of sensory experiences. But so what? Only if no further story could coherently be constructed under which they end up being in the wrong while we remain in the right is there any problem here for pure representationalism. (Tye 2000, p. 108, italics in the original.)

Tye then considers two such ‘further stories’. In the first, the alien population suffers from a genetic defect (just as he asserts with Tom—so this can’t serve as a response to me, it merely repeats the position I’ve just critiqued). In the second ‘further story’, the alien population finds itself in an environment that is very different from the environment in which the alien species developed color vision. This scenario is surely conceivable, but so is one where the aliens’ current environment is no different from their ancestral one. It may seem as though Tye were begging the question here, but he is actually trying to shift the terms of the debate. What does the conceivability or imaginability of an inverted-spectrum-with-the-ancestral-environment scenario really show, and what sort of challenge does it present?

That scenario, Tye claims, poses no threat to representationalism. “…[I]t seems to me that the representationalist can happily allow the conceptual possibility of interspecies and rife, intraspecies inverted spectra even where there are no biological or genetic defects. For it is conceptually possible that sensory representation has nothing to do with natural teleology,” (Tye 2000, p. 109). It seems that Tye takes the challenge that the inverted spectrum poses to his externalist representationalism to be the same challenge that the inverted spectrum posed to standard functionalism. Since representationalism is not an a priori or conceptual truth, it does not have the same vulnerabilities as functionalism. Tye claims that representationalism is an empirical proposal, and could only be bothered by an empirical showing of real (non-defective, etc.) spectrum inversions. On its face, this seems reasonable enough. We wouldn’t want empirical research programs held hostage to mere conceptual possibilities.

Unfortunately for Tye, I am not using the inverted spectrum in the way that he envisages. Tye’s ‘further story’ must tell us who gets the colors right when some variety of color vision appears, and he claims that natural selection will do that job. But as I have shown, any variety of color vision will provide dramatic survival advantages and be strongly selected. If the benefits provided by the two variations are identical (as hypothesized), there is no way that natural selection can ever favor one over the other. The inverted spectrum does not threaten externalist representationalism by mistaking it for functionalism, the inverted spectrum shows that externalist representationalism cannot use natural selection to privilege one set of phenomenal vehicles over any others that confer the same behavioral advantages. So, however well Tye’s ‘representationalism is empirical’ move might work against Shoemaker, it just won’t serve as a response to my challenge.

5.2 Temporal priority

Perhaps the ‘further story’ might just assert that whichever sort of vision develops first gets the colors right. That would be more than a little arbitrary, but even if that bullet were bitten, it wouldn’t solve the problem. Imagine our ancestors, just at the point before the transition from seeing in black and white to colored vision.7 They are poised to develop the mutation that produces color sensitive pigments in some of their retinal cells (we’ll start with red, but the same story could be told for other colors, or all of them). Imagine that there are two groups of these creatures, separated by some natural boundary that keeps them from meeting and mixing, but that the environment on each side of the boundary is the same (same predators, same food sources, same sorts of nesting material, etc.). At about the same time, two individual creatures (call them Alpha and Beta) each develop the mutation that provides red-sensitive pigment in some of their retinal cells, one in each group (Alpha is a member of Group A; Beta is a member of Group B). Creature Alpha develops the red-sensitive pigment, and it produces the same neural activity that causes and constitutes our experience of red. Creature Beta develops the same red-sensitive pigment, but her neural pathways are a bit different, so downstream those signals produce the same neural activity that causes and constitutes our experience of green. Both creatures behave the same way, and utilize their new type of vision to get around in the world in the same ways. Beta has inverted color vision, compared with Alpha. Let us further suppose that both creatures have a big advantage over their peers, leave behind lots of offspring, and the respective mutations quickly become established in Groups A and B. Let us also presume that there are no side benefits to having one genetic complement over the other. Now Group A and Group B are inverted with respect to their visual experiences of color. For convenience, I’ll call this scenario Generation Zero Spectrum Inversion.8 Since there is no temporal priority, what follows? Should Tye say that there is no spectrum inversion here, and that both Groups have the same phenomenal experiences, despite the differences in their vision-processing neurology? That would be to give up on the claim that externalist representationalism can accommodate the inverted spectrum.

Perhaps Tye might claim that I have been unfair to him, and that all he needs is that one of the phenomenal vehicles (not the whole set), at some time in the past, provided some advantage over its inverted variant. That would meet the requirement of having “some possible story”, wouldn’t it? But look at what sort of story it is—if Tye took this line of response, he would be claiming that his externalist representationalism can handle spectrum inversion as long as (and only if) the spectra were not truly inverted (with behavioral equivalence) at some time in the past. Suppose a functionalist claimed that he could account for spectrum inversion as long as (and only if) the two forms of vision were not behaviorally equivalent from 9–10 a.m., GMT, January 12, 2008? That wouldn’t garner much praise, and neither should this line of response. In fact, this line of response is actually an admission that the theory in question cannot account for the inverted spectrum as long as it is true to the original thought experiment (and preserves behavioral equivalence).

5.3 Change the theory of evolution

Tye could propose that natural selection has a hitherto unnoticed component—that a feature will be favored if it provides an advantage to survival, reproductive success, and (in the case of perceptual systems) matches between external colors and phenomenal vehicles. This solution will allow Tye to draw the teleological distinctions that he needs (normal color vision will be favored over inverted color vision), but the cost is quite high. Tye would have to claim that our whole understanding of natural selection must be rewritten to include particulars that make absolutely no difference to survival or reproductive success. Given the extremely fruitful and successful track record of the current conception of natural selection, this response doesn’t look very promising. This response is a paradigm case of the tail wagging the dog.9

5.4 History trumps biology

Tye might assert that over the generations, the color experiences of any creatures with inverted color vision will shift to match the correct colors out in the world, without any biological changes required (let us call that the Phenomenological Convergence). Externalist representationalists are already poised to abandon the intuition that local neural activity determines phenomenological content, and Tye does so in ‘Phenomenal Externalism, Lolita, and the Planet Xenon,’ (Tye 2011, forthcoming, pre-publication version available on Michael Tye’s home page). There, he claims that a pod on the planet Xenon, which (due to a fortuitous lightening strike) temporarily becomes the microphysical duplicate of a real human brain: “Does XP1 [the lucky pod], for the period of time during the storm in which it is microphysically identical to a particular human brain, undergo experiences, all of which are phenomenally identical to the experiences of the relevant human on Earth? I say no,” (Tye 2011, forthcoming, p. 4, italics in the original). So Tye might well adapt that strategy to address my challenge.10

In order for this response to work, we would need answers to the following questions:
  1. 1.

    What is the mechanism responsible for the phenomenological change?

  2. 2.

    If the Phenomenological Convergence is gradual, what are the intervening stages supposed to be like?

  3. 3.

    If the Phenomenological Convergence is not gradual, there must be some special number of generations before history overpowers biology. What possible facts would determine that tipping point?

  4. 4.

    What would happen if we were to transplant the visual system of one creature in Group A to a member of Group B, post-Phenomenological-Convergence?


To the first question, natural selection is not an available answer. If Tye were to appeal to natural selection here (leaving aside worries about circularity), he would either be claiming that there is an evolutionary advantage in having normal color vision over inverted color vision (that’s 5.3, above), or he would have to claim that evolution can produce phenotypic changes (the phenomenology) without any genetic or biological changes. In particular, whatever sort of vision emerges in Generation Zero (using any phenomenal vehicles you might care to imagine—red, green, the sound of bagpipes bleating, the scent of limburger, etc.,11 to provide awareness of red things in the world) will eventually converge on the correct phenomenal character of looking red without any biological or genetic alterations at all. Whatever Tye might wish to call this mechanism, it requires revisions to our understanding of biology that are at least as great as those needed for 5.3, and it incurs the same penalty. Overturning such a successful branch of science in order to preserve a theory about representational content smacks of special pleading, and abandons any veneer of credibility that externalist representationalism might have gained by initially appealing to natural selection.

To the second question, what would a member of Group B (midway between Generation Zero and the Phenomenological Convergence) see when she looked at a ripe tomato in good light? Would it look yellow? Blue? Grey? Periwinkle? Chartreuse? Some pixel-mixed combination of red and green? This trades one spectrum inversion for a much larger number of behaviorally identical variations (possibly hundreds or thousands, depending on how many generations this is supposed to take)—hardly progress towards a solution.

To the third question, why should we think that there is some force of history, building up over the generations until it overcomes biology’s role in shaping our phenomenal experience? The implausibility of this move is probably part of why many theorists are attracted to a mixture of wide and narrow content, with color qualia on the narrow side. But just to try to pin down the intuition here, imagine that while the rest of Group B reproduces normally there is a single individual, Barbara, who gets cloned for each new generation. So, through the history of Group B, there is always a Barbara clone, genetically identical to the first. Let us also imagine that the original Barbara lived before the Phenomenological Convergence. What should Tye say about the Barbara clones after the Phenomenological Convergence? Are they also phenomenologically inverted with respect to their pre-PC clones? If so, we need an explanation for why that particular time is when the weight of history overpowers Barbara’s genetic inheritance. If not, why should we grant that the population at large Converges at all?

For the transplant question, imagine the following. After the Phenomenological Convergence, take Alan and Barry as normal members of Groups A and B, respectively. Their visual systems are neurologically different (different enough to have been inverted with respect to each other in a bygone age), but by the Convergence hypothesis, they have exactly identical phenomenological visual experiences when they look at identical colored objects under identical lighting conditions. If we were to transplant Alan’s visual system into Barry, it seems that Tye would have to say that Barry would notice no difference between his pre-operative and post-operative color experiences.

Now, let us suppose that another member of Group B, Bugsy, is unlucky enough to get the random mutation that gives him inverted color vision with respect to the rest of Group B (that is, his visual system is wired up like the members of Group A). He is color-inverted with respect to Group B, and genetically and neurologically identical with all the members of Group A. Tye must claim that he sees colors wrong with respect to both Groups A and B, even though his visual system is identical with that of Group A. Further, Tye must claim that Bugsy is color-inverted with respect to Group A, even though he has the same brain states that are activated in Group A members when they look at the very same colored objects. In fact, every test that science could run would say that Bugsy is a member of Group A. Yet if Bugsy’s visual system were transplanted into Alan, Tye would have to claim that Alan, after his operation, would report that all his color experiences became reversed!

Even worse, suppose we take two Barbara clones, one from each side of the Phenomenological Convergence, and transplant the visual system from Elder Barbara into Younger Barbara. Tye would be forced to claim that after her operation, Younger Barbara would claim that all her color experiences were inverted. I can only imagine the lawsuits to follow. This attempted solution doesn’t provide a naturalistic explanation for qualitative experience—it makes qualitative experience utterly mysterious.

As far as I can determine, there is no good solution that Tye could adopt in response to the problem that I’ve presented. Each of the proposed solutions either fail to address my objection, requires us to fundamentally rewrite evolutionary theory, or renders us unable to give any natural explanation for qualitative experience. Those price tags are so heavy that I would be remiss if I were to try to choose among them on Tye’s behalf. Rather, Tye should simply admit that externalist representationalism cannot provide any account of spectrum inversion. We will now explore the consequences of such an admission.

6 Conclusion

If I am right, Tye’s externalist representationalism must give up its claim to be able to handle spectrum inversion. So perhaps that is what externalist representationalists should do. Tye’s earlier account of color perception (the PANIC theory without any teleological addenda), in his Ten Problems of Consciousness (Tye 1995), provided just such a theory. If we run the Generation Zero scenario through Tye’s 1995 position, he would claim that both Alpha and Beta had the same phenomenal experiences from the outset, despite their neurological differences. So, next, I’ll show that Tye would have to make that claim, and close by considering the consequences that would follow.

In Tye’s 1995 position, the representation relation (for perception, which is what concerns us here) was supposed to be causal covariation under optimal conditions (Tye 1995, p. 101). Optimal conditions depend on the specification of biological functions, “…[I]n the case of biological creatures, it is natural to suppose that the pertinent optimal conditions are the ones in which the sensory mechanisms are discharging their biological functions,” (Tye 1995, p. 153). The attendant accuracy conditions are these: a representation is accurate if it causally covaries with the thing represented so that it discharges the biological function determined by the environment where the creature evolved (which, presumably, are the optimal conditions). If we apply this to Generation Zero, we see that from the very outset, normal and inverted color vision both have phenomenal characters that causally covary in structurally isomorphic ways to the very same features in the world. If both variants respond to the same features in the world equally well, then they track the same feature, and so represent the same feature. And if they both represent the same feature, and phenomenal character just is representational content, then normal and inverted color vision must have the same phenomenal character. What’s so bad about that?

A few things. First, by Tye’s own admission, the representationalists can only admit the metaphysical possibility of spectrum inversion if they can also tell which creatures perceive accurately and which are making mistakes (Tye 2000, p. 113). Without the appeal to natural teleology, there is no such distinction. All the causally-covarying phenomenal qualities are equally accurate, and so there is no room for error. Accordingly, the representationalists can no longer be so sanguine about the possibility of spectrum inversion. They must deny it outright.

Second, if we go back to causal covariation, then any state that causally covaries with the external colors will produce the very same phenomenal experiences. Suppose we took a normal person (Norm), as an infant, before he has actually seen anything, and dye his corneas so that only shades of red light reached his retinas. Then we do the same to another infant (Orne), but we also cross the wires in his visual system to produce spectrum inversion—he should see everything as green, yes? Not according to the causal-covariation theory. We could have tinkered with his visual system to our hearts’ content without producing any phenomenal difference!

Third, at least in Generation Zero, the causal covariation account would make synesthesia impossible. A synesthetic phenomenal vehicle that covaried causally with the appearance of red things would have to have the phenomenal experience of redness, and nothing more. Synesthesia poses a stronger challenge because it is not just a conceptual possibility, it really happens. However, it won’t produce behavioral identity—so I only introduce it at this juncture. These three considerations are only sketches of problems, and serve to show that natural teleology really was doing some heavy lifting for Tye—the loss of natural teleology is the loss of a very useful tool.

Tye’s externalist representationalism relies on natural teleology to secure facts about who sees colors correctly and who is mistaken. I have shown that natural teleology cannot be used to draw the distinction between adaptively equivalent varieties of color vision. Color vision in general is a feature that is selected for, but when Tye tries to capture the phenomenal experience of color using natural selection, he is stymied.12 From the perspective of natural selection, any phenomenal variations that do not come with a price to be paid in the coin of reproductive success are all as one.13 So we cannot appeal to natural selection when giving an account of the obvious experiential differences among these various conditions.


Representationalism is also sometimes called ‘intentionalism’, but I prefer ‘representationalism’ for purely personal reasons. Minimal representationalism requires that the representational content of an experience determines its qualitative character, and that any two experiences that share the same representational content will also have the same qualitative features. Externalism (or ‘objectivism’, or ‘physicalism’) about colors locates the colors themselves out in the world, on surfaces or films. If that is the correct account of the nature of colors, then no surface can have contradictory color properties. No patch of a surface can be entirely red and entirely green at the same time, from a particular observer’s viewpoint. For contrast, internalism about colors locates the nature of colors in the minds of perceivers—we ascribe colors to a world that is essentially colorless.


Qualia are sometimes called the ‘what it is like’ feature of our experience. For example, the way that a sip of red wine tastes would be the quale of that tasting experience. Likewise for the way that the sky looks on a clear blue day, the way that a trumpet blowing a C-sharp sounds, and so on.


As presented in Consciousness, color, and content and ‘Visual Qualia and Visual Content Revisited’. Tye is an excellent exemplar of externalist representationalism.


The idea that different people might see colors differently goes back to Locke, at least. In its modern form (Shoemaker 1975), spectrum inversion is a thought experiment with the following features. We can imagine a person whose color experiences are opposite to our own (where we see red, he sees green; likewise for yellow and blue, etc.). We can also imagine that this person has learned their native language in the normal way, and behaves just like a person with normal color vision (some tinkering at the edges might be required to make sure the relationships with dark and light colors, warm and cool colors, and other emotional associations with colors were also swapped—thus preserving the inability to detect people with inverted color vision via any behavioral test).

It seems possible that some person, call him ‘Tom’ (Tye does, and I will follow for convenience, with apologies to any Toms in the reading audience), could have his color experiences inverted with respect to yours. Tom’s phenomenal experience of green, for example, is caused always and only by the same physical objects that cause you to have red visual experiences, his experience plays the same role in his psychological economy that red experience plays in yours (it causes him to believe that ripe tomatoes are red, etc.), and his experience causes the same behavioral outputs (measured by colorblindness tests, color sorting and matching tests, and uttering verbal reports like, “That tomato is red, its probably ripe.”).


Though Tye seems inclined towards the origin-centered view of the source of a feature’s telos, as seen in the reference to ‘ancestors’ in the passage quoted above.


What should we say if it turns out that one sort of color vision provides an advantage that is wholly unrelated to its effectiveness as a perceptual mechanism? Suppose, for example, that the genetic arrangement that produces color vision also reduces the incidence of retinoblastoma (a type of cancer of the retina, often with a genetic cause) by about 20%, when compared with those people who have inverted color vision. Would we then say that the purpose of color vision is to confer resistance to retinoblastoma, and not to allow us to navigate the world visually? That certainly sounds odd, but one might be able to make a case for it. Either way, that sort of teleological explanation would be of no help to Tye—color vision would not be selected on the basis of representational accuracy.


I tend to imagine them as tree-going primates, but I have no real reason to believe that is when color vision developed. One may imagine them as Jurassic quasi-rodents without altering the thought experiment in any significant respects.


The Generation Zero Spectrum Inversion scenario might also raise problems for teleofunctionalist approaches like Millikan’s—issues that are interesting but beyond the scope of the current paper.


In personal correspondence, Tye suggests that this response is at least a live option for him. “Color vision evolved because of what it is capable of telling its possessors about the colors of things. In our distant ancestors, Mother Nature installed internal states whose job it is to track the colors.”


Tye also considers this response to be viable, in personal communication. “Spectrum inversion simply becomes normal color vision after some number of generations. If the trait for inverted color vision becomes encoded in the genes of Group B, then distant descendents of the initial population may represent ripe tomatoes as red, unlike the initial members.” Further, Tye makes a similar claim in Ten Problems of Consciousness: “So it is possible that, with variations in biological function across environments, phenomenally relevant representational differences arise without any internal physical difference,” (Tye 1995, p. 153, my emphasis added).


Almost any phenomenological vehicle will do the job, because the job in Generation Zero is just to convey the information that there is a red thing in the visual field. The qualia of colors, sounds, tastes, smells, etc., will all do that well enough to pass the trait along by conferring an advantage. Some phenomenological vehicles would not confer any advantage, and some would be decidedly detrimental (like the feeling of blinding-agony-so-fierce-that-I-must-now-curl-up-into-a-ball-and-do-nothing-but-whimper).


This result will be congenial to a whole host of non-externalist positions about colors (internalism, relationalism, projectionism, subjectivism, etc., with specific variations of each having devoted advocates), but adjudicating among them is a task for another occasion. Ruling Tye’s externalist representationalism out doesn’t automatically rule any particular one of his competitors in.


If this is true of radical differences like spectrum inversion, it will also hold true for the smaller, real-world differences between people in assigning unique hues to particular wavelengths of light, of the sort discussed in the recent “True blue” thread of articles in Analysis. Therein, one of Tye’s responses to the peril that true-blue poses to his position mirrors his response to the peril of the inverted spectrum, so at minimum, Tye is forced to the second of his true-blue responses (Tye 2006b, p. 343). I plan to explore the impact that this has on Tye’s remaining response, in the near future.



I would especially like to thank Michael Tye for his comments and discussion on an earlier draft of this paper. I am very grateful to the following people for their helpful comments and perceptive questions: Jeremy Anderson, Francisco Ayala, Derek Brown, David Cole, Donald D. Hoffman, Tristram McPherson, Michelle Montague, Mark Newman, Wade Savage, David W. Smith, P. Kyle Stanford, Sean Walsh, Martin Young and an anonymous reviewer at Philosophical Studies. I am also indebted to the audiences at U. C. Irvine, University of Minnesota, Duluth, the National Postgraduate Analytic Philosophy Conference (at Magdalene College, Cambridge), the Third Annual Hawaii International Conference on Arts and Humanities, the Southern California Philosophy Conference, the Minnesota Philosophical Society Conference, and the Pacific Division of the American Philosophical Association; who heard various versions of the paper and whose questions and comments I benefited from. I am also grateful to the College of Liberal Arts, University of Minnesota, Duluth, for support.

Copyright information

© Springer Science+Business Media B.V. 2010