Authors' Response

continuing commentary on "Color Realism and Color Science"


Alex Byrne

Department of Linguistics and Philosophy, 32-D808, Massachusetts Institute of Technology, Cambridge MA 02139, USA


David R. Hilbert

Department of Philosophy and Laboratory of Integrative Neuroscience (M/C 267), 601 South Morgan Street, University of Illinois at Chicago, Chicago IL 60607, USA


Abstract: Our reply is in four parts. The first part addresses objections to our claim that there might be "unknowable" color facts. The second part discusses the use we make of opponent process theory. The third part examines the question of whether colors are causes. The fourth part takes up some issues concerning the content of visual experience.

Our target article had three aims: (a) to explain clearly the structure of the debate about color realism; (b) to introduce an interdisciplinary audience to the way philosophers have thought about the issue; (c) to argue that colors are certain sorts of physical properties ("productances").

We are very grateful to the commentators in this continuing commentary for their criticism and constructive suggestions.

R1. Ordinary intuitions and unknowable colors

As Dennett notes, the target article emphasizes that the problem of color realism is not about color words, or the folk category of color. As we conceive it, the problem concerns certain properties that are represented by the human visual system and those of a variety of other animals. However, Dennett thinks that our methodology does not fit happily with the touted "scientific" nature of the problem: we place too much weight, he claims, on "ordinary intuitions about color".

But, in fact, we do not place any weight on ordinary intuitions about color. (Dennett himself seems to think that ordinary intuitions should be accorded someweight: an account of color should be "largely consonant with everyday usage". We disagree.) Dennett's one example of our alleged reliance on ordinary intuitions is our rejection of the idea that there are different kinds of colors: surface, volume, and illuminant (see sect. 3.1.2. of the target article). Although the ordinary person might well find this idea odd -- tomatoes, glasses of Burgundy, and stoplights are all red! -- this was not our reason for rejecting it. Rather, our objection was that surfaces, volumes, and illuminants can all look the same in respect of color -- a fact which might be revealed by training an animal to press a bar in the presence of a type of light, and then noting the animal generalizes the rule to certain surfaces and volumes. This fact is best explained, we said, by supposing that there is a common property visually represented. Whether or not we are right about this, the argument certainly does not make any appeal to ordinary intuitions.

            Dennett suggests that our claim that it may be unknowable whether a chip is unique green is "counterintuitive". He might mean by this that ordinary intuition rules out such a possibility (thus implicitly accusing us of arbitrarily picking and choosing between intuitions), but we doubt it, because surely ordinary intuition has no clear opinion on the matter. In any case, whether or not Dennett himself is resting any weight on ordinary intuition here, he offers another consideration entirely. Dennett claims that the conclusion that there are no unknowable color facts follows from the "coevolutionary coordination of color vision and reflectance properties", and obviously this argument does not appeal to everyday usage or the opinions of the folk.

            Dennett does not spell out his argument in any detail; still, it is instructive to set out an argument that is naturally suggested by his remarks. (We emphasize that we are not attributing this argument to Dennett. See also Dennett 1991, pp. 375-83, which we lack space to discuss.) 1) Evolution fine-tuned the colors of certain fruits so they were readily visible to certain fruit eaters, and evolution also fine-tuned the fruit eaters' color vision (in particular, the spectral sensitivities of their cone pigments) to detect the colors of the fruits (see Regan et al. 2001). Hence: 2) in normal conditions the fruit eaters will correctly perceive the colors of fruits. Hence: 3) in normal conditions the fruit eaters will correctly perceive the colors of things generally. Hence: 4) our claim that most people misperceive unique green chips in normal conditions is incorrect.

            There are three main problems with this argument, apart from the fact that the coevolution hypothesis is not an established fact. First, and perhaps most seriously, on any remotely defensible elaboration of 1), it does not imply 2). The most 1) could imply is that the fruit eaters are by and large correct: any evolutionary fine-tuning of fruits and cone pigments would leave plenty of room for minor misperceptions of determinate shades, and minor variations between individuals (see sect. R2.6 of the first replies). Second, 2) does not imply 3). A system specialized for infallible detection of the colors of ripe fruits might well make mistakes with other colors. (In this connection, it is worth pointing out that the coevolution hypothesis just concerns the red-green opponent system, not the yellow-blue system.) Third, 3) does not imply 4). The main empirical evidence for coevolution concerns certain species of Old World monkeys, not humans. Granting for the sake of the argument that some remote trichromatic ancestor of humans never misperceived colors in normal conditions, it by no means follows that humans have retained this capacity.

            Ross offers a "pragmatic account of veridical color" that , he claims, has the "advantage" of avoiding "unknowable color facts". As explained in our first replies, we think this is no advantage at all. But in any case, we do not think Ross's account is one of the right sort. Ross holds that "an object's veridical color is relative to favored conditions of perceptual access" -- these favored conditions being a matter of our interests and purposes (Ross 2000). This form of relativism is inconsistent, however, with our commitment to intentionalism (see sect. R4 below). To apply his theory to visual experience Ross needs to assume that the phenomenal character of color experiences is independent of their content. Without such an assumption his theory, if it can be applied at all, results either in the conclusion that the apparent colors of things vary with the choice of pragmatic standards, or that we never see the true color of anything. Ross's pragmatic physicalism may be adequate as an account of some parts of color language, but it is inadequate as an account of what we see.

R2. Opponent-process theory

In his interesting commentary Broackes raises a number of issues concerning opponent-process theory and its relation to our thesis that colors are represented in visual experience as proportions of hue magnitudes (sect. 3.2 of the target article). He points out -- as did the previous commentators Kuehni, Jakab & McLaughlin, and Pautz -- that the very simple physiological model we use for illustration is empirically inadequate. As Broackes recognizes, however, we do not endorse this model; our discussion of it was merely intended to show how the hue magnitudes might be given a rough-and-ready characterization in physical terms.

Broackes thinks that our theory presupposes some version of opponent-process theory. In particular, he thinks that we "claim a precise parallelism between the structure of phenomenal color space and that of the physiological processes underlying it." There are two questions here. 1) Does our theory require the kind of straightforward mapping that Broackes describes? 2) Is there reason to doubt the existence of such mappings? Broackes rightly holds that the answer to the second question is "yes." There seems to be no evidence at present for the existence of distinct populations of cortical neurons with the characteristics required to support the various psychophysically identified color processes. In particular, there are no identified populations of cortical neurons that have the tuning required by a simple implementation of the psychophysically determined chromatic response curves (Lennie 1999).

            So everything hangs on the first question, and fortunately for us this has a negative answer. Broackes may have been misled by the details of the last paragraph of sect. 3.2 of the target article. We chose to give an approximate specification of the magnitudes in terms of physiology partly to make vivid the compatibility between our theory and the physiology, not because we require the connection to be made at this level. Our claim that colors are represented as proportions of hue magnitudes is simply a claim about the content of color experience, not about physiology.

R3. Color and causation

It has seemed obvious to many philosophers that color perception involves some sort of causal transaction between the colors and our visual systems. We agree -- otherwise it is difficult to explain how visual systems could recover color information. (This widely held causal claim, according to us, is simply a plausible empirical assumption, not a conceptual truth about perception.) Botterellargues that this assumption is inconsistent with our view that the perceivable colors are to be identified with reflectance-types rather than determinate reflectances. According to Botterell, it is the determinate reflectances that are causally efficacious, not the perceivable colors themselves.

            We have two points to make about Botterell's argument. First, we often do in practice give causal explanations that appeal to determinables rather than determinates, even though the determinates are individually sufficient for the effect. Why did the bridge collapse? Because a truck weighing more than 10 tons was driven across it. This explanation appeals to a determinable, weighing more than 10 tons, rather than to its determinates, like weighing 10.13 tons. According to Botterell this "explanation" must be understood as gesturing towards the genuine explanation that appeals to the precise weight of the truck. But -- and this is our second point -- Botterell's insistence that only the determinate property is causally efficacious is not well-motivated. What matters is that the weight of the truck was too large, not which determinate weight the truck possessed (Yablo 1992). Any weight greater than 10 tons would have produced the same effect. Similarly, in the case of color, it is the type which is causally relevant to the visual effect rather than the determinate reflectance (Smart 1975; Jackson and Pargetter 1987; Jackson 1996).

R4. The content of visual experience

In the target article we used the philosophical apparatus of propositions to elaborate on the fact that color experiences can be accurate or inaccurate, depending on whether objects have the colors they visually appear to have. For example, it might visually appear to someone that there is an orange sphere on top of a blue cube. The subject's experience represents that there is an orange sphere on top of a blue cube, and this representation can be assessed for truth or falsity. In other words, the experience has propositional content -- namely, the proposition that there is an orange sphere on top of a blue cube.

            Skokowski challenges our claim that color experiences have propositional content, on the ground that non-linguistic animals have color experiences but not experiences with propositional content. But we see no reason to accept Skokowski's unargued assumption that possession of a language is a necessary condition for undergoing experiences with propositional content, or for being in mental states with propositional content. (As we read Tye and Dretske -- whom Skokowski cites as fellow-travellers -- they do not accept this assumption either (Dretske 1995, pp. 25-27; Tye 2000).)

            Skokowski's positive view is that the color properties, not propositions, will serve as the contents of color experiences: "when I experience the yellow of the lemon, I experience the physcial property yellow (reflectance)". But what is the content of the experience described two paragraphs back? According to us, it is the proposition that there is an orange sphere on top of a blue cube; according to Skokowski (apparently) the content is a list of properties and relations: {the property orange, the property blue, the property of being a sphere,É}. If that is right, then Skokowski cannot distinguish the experience of an orange sphere on top of blue cube from the experience of a blue sphere on top of an orange cube: both experiences have the same Skokowski-content. And since the experiences evidently do differ representationally, Skokowski's alternative account of content is inadequate.

            Stoljar agrees with us that experiences have propositional content, and uses this fact as part of his ingenious argument that we have an unstable position. We endorse physicalism, intentionalism (or representationalism: see sect. 3.5 of the target article), and deny, as Stoljar puts it, "that color experience is systematically illusory". Stoljar argues these first two claims imply that color experience is systematically illusory.

            Intentionalism is the thesis that the phenomenology of color experience cannot vary independently of its content: if two experiences are alike in their represental content, they are alike in their phenomenology. Thus, according to the intentionalist, if you and I are "spectrally inverted" with respect to each other, the representational contents of our experiences must differ; according to the anti-intentionalist (notably Block 1990) this is not so. Although intentionalism was to a large extent left on the sidelines in the target article, it is in our view an important component of our overall theory, not least because many erroneous accounts of color are motivated by anti-intentionalism (see our discussion of Ross in section R1 above).

            The crucial step of Stoljar's argument is from intentionalism to the claim that color experience purports to inform us that colors are sui generis properties (a fortiori not reflectances or productances). Granted this step, the rest of the argument goes through: if intentionalism is true, then color experience represents that physicalism is false; hence, if intentionalism and physicalism are true, then "color experience is systematically illusory".

            How troubling is this conclusion? Perhaps not very, a fact that is slightly obscured by Stoljar's use of the expression "error theory" to characterize the conclusion. The error theory is usually taken to be the view that objects do not have the colors that they appear to have (see sect. 2.1 of the target article). But on this way of speaking, the error theory is not the conclusion of the argument: the "error" is that physicalism is false, not that lemons are yellow. We could live with the view that experience misleads us as to the true nature of color.

            In any event, we question the crucial step: we don't see how intentionalism supports the view that color experience represents that the colors are sui generis, and so that physicalism is false. In order for experience to represent this, it would have to represent the color properties themselves as having properties (in particular the property of being unanalysable or sui generis), and we don't think experience has this sort of content: lemons look yellow, but yellowness does not look any way at all (Byrne 2003).

            There is a distinction between not representing F and representing not-F, which Armstrong illustrates by the "headless woman illusion" (Armstrong 1968; Armstrong 1987). (The "headless woman" merely does not look to have a head, as opposed to looking to have no head.) The important question Stoljar raises is whether experience tells us that physicalism is false, as opposed to not telling us that it's true (Hilbert 1987, p. 37). Color experience, we think, is in this respect like the headless woman illusion.


Armstrong, D. M. (1968). The headless woman illusion and the defence of materialism. Analysis 29: 48-49.

Armstrong, D. M. (1987). Smart and the secondary qualities. In: Metaphysics and morality: Essays in honour of J. J. C. Smart, Eds. P. Pettit, R. Sylvan and J. Norman: 1-15.

Block, N. (1990). Inverted earth. In: Philosophical perspectives, Ed. J. Tomberlin. Vol. 4: 53-79. Atascadero, CA, Ridgeview.

Dennett, D. C. (1991). Consciousness explained. Boston, Little Brown and Co.

Dretske, F. I. (1995). Naturalizing the mind. Cambridge, MA, MIT Press.

Hilbert, D. R. (1987). Color and color perception: A study in anthropocentric realism. Stanford, CSLI.

Jackson, F. (1996). The primary quality view of color. In: Philosophical perspectives, Ed. J. Tomberlin. Vol. 10. Cambridge, MA, Blackwell.

Jackson, F. and R. Pargetter (1987). An objectivist's guide to subjectivism about colour. Revue Internationale de Philosophie41: 127-41.

Lennie, P. (1999). Color coding in the cortex. In: Color vision: From genes to perception, Eds. K. R. Gegenfurtner and L. T. Sharpe: 235-47. Cambridge, Cambridge University Press.

Regan, B. C., C. Julliot, et al. (2001). Fruits, foliage and the evolution of primate colour vision. Philosophical Transactions: Biological Sciences 356(1407): 229-83.

Smart, J. J. C. (1975). On some criticisms of a physicalist theory of colors. In: Philosophical aspects of the mind-body problem, Ed. C.-y. Cheng: 54-63.

Tye, M. (2000). Consciousness, color, and content. Cambridge, MA, MIT Press.

Yablo, S. (1992). Mental causation. Philosophical Review 101: 245-80.