A Guided Tour of Color

Jonathan Cohen

One of the most salient facts about our experience of the world is that objects appear to have colors. This feature of our experience is both striking and pervasive. Indeed, representations of colors of objects are among the most notable deliverances of the visual modality, which is perhaps our most important source of information about the world. For this reason, among others, questions about the nature of color have crucial significance for a variety of philosophical subjects including perception, ontology, epistemology, semantics, and philosophy of mind. But the nature of color is a fascinating philosophical topic in its own right, and there has been a significant increase in the philosophical attention paid to this matter in recent years. In this essay I'll survey some of the main views about the nature of color in the contemporary literature and attempt to lay out some of the arguments that have been used to support or reject various of these accounts.

1. Some Initial Distinctions

Perhaps the most fundamental distinction among theories of color is that between realist accounts, which hold that color properties exist and are instantiated by objects in the actual world, and irrealist accounts, on which no actual objects instantiate color properties. Irrealism about color is, of course, at odds with the common sense view that colors are pervasive features of ordinary objects in the world. Still, defenders of the view, such as [Hardin, 1988] and [Maund, 1995], have argued that this is a case where common sense is in error, and stands in need of correction.

There are roughly two reasons philosophers who have realistic views about other properties (i.e., those not sympathetic to idealism or nominalism about properties in general) have been attracted to color irrealism. First, some have begun with the conviction that the only real properties are those recognized by science, and have rejected color realism on the ground that red, blue, and the like do not appear in the inventory of properties recognized by the sciences.1 Second, some irrealists have come to their position by a process of elimination, only after becoming convinced that no substantive account of color properties can satisfy certain intuitively and empirically motivated constraints about what colors must be.2

Among realist theories of color, we may distinguish sense-data theories, on which colors are properties of the mental entities that are the immediate objects of perception, and what might be called externalist theories, according to which colors are properties of ordinary external objects --- the tables, coffee cups, and hippopotami that populate the extra-mental world. Sense-data theories have a long and venerable tradition in empiricist theorizing about perception and knowledge, and have been defended in such contemporary works as [Jackson, 1977] and [Perkins, 1983]. However, many philosophers remain unconvinced about both the ultimate tenability of the ontology of sense-data and their necessity in understanding perception.3 For philosophers who share this doubt, it has seemed more promising to take seriously the naive intuition to the effect that the bearers of color are extra-mental. This is just to say that most color realists have been tempted by some form of color externalism.

Among externalist accounts of color, we may distinguish between those that take colors to be intrinsic --- henceforth, intrinsic theories --- and those that take colors to be relational properties --- henceforth, relational theories.4 Most of what follows will center on the dispute between intrinsic and relational theorists.

Intrinsic theorists, who understand colors as intrinsic properties of external objects, think that colors are objective and mind-independent; in particular, they insist that colors are not constituted in terms of relations to subjects or minds. A typical account of this sort is one that takes colors to be physical properties --- usually some kind of reflectance property of surfaces, transmittance property of transparent surfaces and volumes, emittance property of luminous sources, or some combination of these. However, other versions of intrinsic theories are possible as well; for example, one might hold that colors are intrinsic but non-physical (whatever that comes to), or intrinsic but unanalyzable (a fortiori not susceptible of analysis in terms of physical kinds). That said, the prevailing intrinsic accounts of colors take them to be physical, and this has encouraged writers to discuss these accounts under a variety of seemingly non-equivalent labels, including `color objectivism', `color physicalism', and (adverting to the distinction among primary and secondary qualities of matter adumbrated by modern philosophers such as Galileo, Boyle, and Locke) 'primary quality theory of color'.5 Intrinsic theories of color have become increasingly popular in recent years; versions of them are defended by [Armstrong, 1968], chapter 12, [Hilbert, 1987], [Byrne and Hilbert, 1997a], and [Ross, 1999].

In opposition to these views, relational theories have it that colors are constituted in terms of relations between objects and subjects, and therefore proponents of these views deny that colors are objective and mind-independent in the sense in which intrinsic theorists think they are. A standard relational account, often associated with Locke, holds that colors are dispositions to cause certain sensations in certain kinds of minds; for example, on one version of this theory, red is the disposition to look red to normal observers. This view -- henceforth color dispositionalism -- understands colors as what the modern philosophers mentioned above would regard as secondary qualities, and therefore has come to be known as a secondary quality theory of color.6 I think it is safe to say that dispositionalism is the received view about color in philosophy; versions of it have been defended not only by moderns such as Galileo, Boyle, Newton, and Locke, but also more recently by [McGinn, 1983], [Peacocke, 1984], [Wright, 1992], and [Johnston, 1992], among others.

Still, dispositionalism is not the only relational theory of color. Another is the functionalist view, defended by [Jackson, 1998c], [McLaughlin, 1999b], and [Cohen, 2000], that colors are the properties that dispose their bearers to cause particular types of sensations in certain kinds of minds. Functionalism is in many respects closely related to dispositionalism, but the two views cannot be identified: functionalists say that colors are the properties in virtue of which things have their dispositions to look colored, not the dispositions themselves, while dispositionalists identify colors with the dispositions in question.7 A further relational account is the so-called "enactive" view of [Thompson, et. al., 1992] and [Thompson, 1995], which attempts to understand colors in terms of the varying (ecologically described) functions performed by the visual systems of different species of organisms.

2. Motivations for Intrinsic Accounts

Why might one be tempted to think of colors as intrinsic? One reason is that ordinary perception (uninformed by philosophical theory or scientific experiment) seems not to mark any distinction between colors and other properties that we take to be intrinsic, such as mass. Another reason is that such a view permits us to understand why we take objects to maintain their colors when the ambient lighting, viewing distance, or other circumstances of the viewing situation are modified, and why we expect that one and the same color can be detected by multiple perceivers: if colors really are intrinsic properties of objects, then modifying the parameters of the viewing situation cannot change objects' colors, and literally the very same intrinsic colors will be available for detection by distinct perceivers (cf. [Hilbert, 1987], 2). A further reason for wanting to treat colors as intrinsic properties (discernible in [Hilbert, 1987], [Matthen, 1988], and [Shepard, 1992], for example) is that, if they are, we can understand color perception as a computational process of recovering the intrinsic properties of objects from the undifferentiated tangle of information arriving at the retina, and this way we may conceive of color perception as of a piece with the highly successful and extremely influential computational accounts of shape and motion perception broadly in the tradition of [Ullman, 1979] and [Marr, 1982]. Finally, some authors ([Smart, 1975], [Armstrong, 1987], [Tye, 1995], [Dretske, 1995]) have been attracted to intrinsic accounts of color because they want to understand color experiences relationally --- in terms of the color properties these experiences represent, and therefore (on pain of circularity) need an analysis of color properties that does not, in turn, appeal to color experiences.8

3. Motivations for Relational Accounts

What motivations support a relational view such as dispositionalism? One of the most powerful motivations (discussed by Galileo, Berkeley, and Locke, among many others) concerns the wide interpersonal and intrapersonal perceptual variation in respect of color: x can look red to me and fail to look red to you, and x can look red to me under one viewing condition and fail to look red to me under another viewing condition. Of course, this is not enough by itself to show that colors are relational; something may look square to me and fail to look square to you, but no one supposes that shape properties are constituted in terms of a relation to a perceiver. However, the thought goes, the perceptual variations in the two cases are not strictly analogous, because in the case of color (unlike shape) two incompatible perceptual variants can both be correct:
... suppose a given range of objects look systematically red to us and systematically green to Martians, and suppose our and their colour discriminations are equally fine. Then there will be no choosing between these groups of perceivers in respect of whose experience determines the colour of the objects in question ([McGinn, 1983], 9--10).
The idea, then, is that perceptual variation in respect of shape does not show that shape is relational because we have objective, observer-independent, well-motivated (geometric) criteria for deciding whether something is an instance of a certain shape (for example, it is square just in case it has four right interior angles and sides of equal length), and this justifies us saying that a perceiver's perception that x is a square is correct just in case x meets the (independently certified --- in this case, geometrically certified) conditions that are necessary and sufficient for being a square. In contrast, the suggestion goes, there is no independent standard on the strength of which we can choose among the perceptual variants in respect of color, since there is no motivated, independent standard that we could use to select from among the variants. If this is correct -- if there is no ground for ruling out any of the perceptual variants, then x could be red for me and fail to be red for you, or look red to me under one viewing condition and fail to look red to me under another viewing condition. Of course, this is just to say that red is not an intrinsic property of objects, but is some sort of relation between objects, perceivers, and viewing conditions; and this, in turn, is just to say that red is a relational property.

The line of argument in favor of relationalism sketched above turns crucially on the claim that, unlike the case of shape properties, there is no perceiver-independent, well-motivated standard for choosing among the perceptual variants with respect to color properties. This claim, however, has seemed dubious to some. After all, one might suggest, there are several scientific and industrial recipes for deciding which among a range of observers is correct in her color judgments, and which among a range of viewing conditions is appropriate for making correct color judgments. Therefore, an anti-relationalist might think that she could appeal to these recipes as a way of blocking the argument for relationalism sketched above. Unfortunately, there are several reasons for thinking that these recipes won't serve the philosophical purposes to which an anti-relationalist might hope to put them.

Consider the recipes for selecting among perceivers first. First, as noted in ([Hardin, 1988], 76--82), the scientific and industrial specifications that have been articulated are typically statistical constructs over a range of distinct actual individuals, but that differ significantly from most (perhaps as many as 90% of) human visual systems;9 therefore, fixing the colors by appeal to such standards would commit one to the unpalatable conclusion that the color discriminations of most (perhaps as many as 90% of) human visual systems are non-standard, and therefore erroneous. Second, these definitions are stipulatively chosen for particular purposes (mathematical convenience, industrial standardization), and reflect the particular purposes of their users; therefore, there is no reason for hoping that such standards will reveal the objective nature of color. Third, the large number of distinct standards that are adopted for different purposes itself suggests that no one of these standards will serve to fix the colors authoritatively, but rather that each is only useful in certain delimited contexts.10

Analogous considerations prevent us from choosing a standard for viewing conditions, even though there are scientific and industrial recipes for standard conditions (cf. [American Society for Testing and Materials, 1968], [Judd and Wyszecki, 1963], [Kelly and Judd, 1976], [Wyszecki and Stiles, 1967]). To see why, consider a notion of standard conditions based on one such recipe -- the instructions for the Munsell color chips.11 These instructions state that "the samples should be placed against a dark achromatic background and `colors should be arranged under North Daylight or scientific daylight having a color temperature of from 6500 degrees to 7500 degrees Kelvin. Colors should be illuminated at 90 degrees and viewed at 45 degrees, or the exact opposite of these conditions'" ([Hardin, 1988], 68).12 While these conditions are adequate for standard uses of the Munsell chips, they cannot be taken as a general specification for The Authoritatively Correct Conditions For Color Perception. One reason is that these conditions could not be used to specify the colors of stars, neon tubes, rainbows, and other cases where the Munsell-specific conditions cannot be met.13 Moreover, this set of conditions is not amenable to the colors of directionally reflective materials, materials whose color is dependent on use, or translucent materials.14 For these reasons, the Munsell-inspired specification of standard conditions is inadequate as a fully general proposal. Of course, we could answer some or all of these complaints by clinging to the specification from the Munsell instructions if we are willing to declare that the phenomena falling outside its scope are color-illusions; however, without independent motivation, such declarations will seem disturbingly ad hoc.

If the Munsell-inspired proposal fails because of the wide range of ecologically valid situations in which we want to talk about colors, a natural suggestion would be to divide objects into different types, and then specify standard conditions for each type. Thus, on such a proposal, we could specify separately standard conditions (i.e., conditions appropriate for veridical color judgments) for opaque surface colors, volume colors, film colors, self-luminous colors, and so on. However, this procedure, too, runs into trouble quickly. Consider just surface color, for example. It is well known that, because of contrast effects, a given surface patch looks different in respect of color to a given observer as a function of the region surrounding the patch. We can eliminate such effects if we specify that the standard conditions include viewing the patch through a reduction tube; however, this choice will have untoward consequences. First, it will follow that surfaces in the vast majority of ecologically valid settings will not look to the vast majority of ecologically valid observers to have the colors they in fact have; this by itself seems a rather ad hoc sacrifice of plausibility in the service of an unmotivated preference for specificity. Second, (as a limiting case of the first problem) a specification relying on the use of reduction tubes will have the consequence that, necessarily, nothing could ever have any of the contrast colors --- colors that cannot appear in the absence of contrast --- such as brown, olive, pure white, and pure black. I suggest that a theory of color on which it is necessary that nothing is brown, olive, pure black, or pure white, is not a theory we should endorse (at least not without first demanding a compelling argument). Furthermore, contrast effects are not the only worries; how a surface looks to a given subject depends also on the choice of magnitudes for such values as size and resolution the field of view --- choices for which there are no obviously principled criteria.15 Similarly, a set of standard conditions for surfaces would need to include a choice of illumination, but each CIE standard illuminant (ostensibly the best candidates for a principled choice of standard conditions of illumination) has metameric pairs that are not metamers under other illuminants; again, it is hard to see how to make a non-arbitrary choice between the illuminants.16

It seems, then, that the standardized specifications of standard condition that are designed for use in particular laboratory or industrial purposes are not adequate to our needs. And once again, since such scientific and industrial standards are our best motivated candidates, it seems unreasonably optimistic to think that a more successful specification of standard conditions is in the offing.

To summarize, perceptual variations in color have encouraged relational theorists because it has seemed to them that there is no principled way to select among the variants --- to claim that one of the variants is correct, at the expense of all the rest. On the other hand, there is dramatic perceptual variation in respect of color: x can look red to me but fail to look red to you, and x can look red to one observer in one viewing circumstance but not in another. But if there is this range of variation, and if there is no principled ground for choosing between the variants, then it seems difficult to resist the conclusion that x is red for me and also that the very same x is not red for you, and that x is red for me in one viewing conditions but not in another. And this is just to say that red is a relational rather than an intrinsic property of x.

4 Objections Against Intrinsic Views

What arguments have been used to argue against intrinsic theories of color? Aside from the arguments just summarized in favor of relationalism, two kinds of argument against intrinsic theories of color have figured most prominently in recent literature. The first concerns the fineness of grain of the colors, and the second concerns structural properties of the colors.

4.1 Intrinsic Theories and Fineness of Grain

A standard form of objection against any proposed philosophical analysis claims that the proposal individuates the entities it hopes to explicate at an inappropriate grain (either too finely or too coarsely). Some thinkers have deployed this form of argument against intrinsic theories of color, claiming that the phenomenon of metamerism shows that intrinsic theories individuate colors too finely ("the fineness objection").

To state the fineness objection as generally as possible, let us understand a metameric pair for some intrinsic theory of color as a pair of stimuli that differ in respect of their colors as colors are understood by that intrinsic theory, but that are perceptual matches for a given observer and a given viewing condition. The members of a metameric pair, then, will be perceptual matches for a certain observer under a certain viewing condition, even though one of them has and the other one of them lacks a particular color property, as color properties are understood by that intrinsic theory. Whether or not there exist metameric pairs in this sense, of course, depends on the details of the intrinsic theory under consideration. However, it is clear that, if we understand color properties as surface spectral reflectance distributions (as on the most prevalent current intrinsic accounts of color) there are metameric pairs for any illuminant. The worry raised by such pairs, of course, is that their color is not distinguished by ordinary perception, even though it is distinguished by some intrinsic theory of color; in other words, such theories seem to individuate colors more finely than perception, and therefore (according to the objector) more finely than is appropriate.

One response to this objection on behalf of the intrinsic theorist rejects the objector's implicit assumption that ordinary perception provides the appropriate criterion for individuating colors. On this line of response, we accept that the theory individuates colors more finely than ordinary perception (finer than red and blue, for example), but simply deny that the individuative standards of ordinary perception carry any authority --- we insist that ordinary perception fails to mark real distinctions among colors. That is, on this view, red, blue, and the like are not themselves colors, but superordinate categories each of which subsumes many distinct colors that should be distinguished by an adequate theory of color.

However, one might wonder whether this response is a bit unmotivated, as it stands. It is tempting to answer that we began the project with the hope of understanding the nature of red, blue, and the like --- properties that are detected and individuated by ordinary perception: if a theory of color tells us that red is not a color but a superordinate of many colors, then it is providing an analysis of something other than our initial target. On the other hand, the intrinsic theorist might respond, this may be one of the cases where a new understanding of the facts genuinely warrants a reconfiguration of the aims of a scientific theory: this is no more ad hoc, she might suggest, than the (presumably warranted) determination that ichthyology should exclude the study of whales, insofar as whales are not fish. Of course, what warrants these proposed theoretical reconfigurations (when they are warranted) is the possibility of providing them with independent motivation in terms of the methodology and taxonomy of the field of study. When such motivation can be supplied (as, I take it, it can be in the case of the exclusion of whales from the domain of ichthyology), there is no reason to object that the reconfiguration is an ad hoc maneuver designed only to avoid charges of inappropriate individuation. What is needed to sustain the intrinsic theorist's response, then, is such a theory-neutral motivation for individuating colors more finely than by the standards of perceptual matches.

Just such a motivation is offered in [Hilbert, 1987], chapter 5. Hilbert's motivation turns on the fact that, when members of a metameric pair are perceptual matches (for an observer) under one illuminant I1, they must be perceptual mismatches (for that observer) under some different illuminant I2.17 Hilbert proposes that, while it might sound counterintuitive to distinguish the colors of surfaces that are perceptual matches under I1, this is less of an affront to intuition than it would be to identify the colors of the same surfaces when they are not perceptual matches (viz., under I2). Of course, for the intrinsic theorist, the colors of surfaces do not vary when we modify the illuminant under which we view them, so we must choose between saying that the members of the pair share a color (because they are perceptual matches under I1) or that they fail to share a color (because they are not perceptual matches under I2). If we choose to say that the members of the pair share a color, then we must hold that their apparent difference in color in I2 is an illusion. But if the appearance of difference is an illusion, it is a case where an illusory apparent difference reveals a real intrinsic difference between the surfaces; and that, claims Hilbert, sounds more like a veridical appearance than an illusion! Consequently, Hilbert suggests, we are wiser to regard the surfaces as differing in color after all, despite the appearance to the contrary in I1.18

4.2 Intrinsic Theories and Structural Properties

A different argument against intrinsic theories of color that has been extremely prominent in the recent literature centers on structural relations that hold among the colors.19 The argument in question (henceforth, the argument from structure) alleges that colors bear structural relations to each other that the intrinsic properties identified with the colors by intrinsic theorists do not bear to each other, and therefore (by Leibniz's Law) that the contemplated identification between colors and intrinsic properties cannot be sustained.20

What are the structural relations holding among the colors that form the cornerstone of the argument from structure? The first set of relations involves the set of similarity and exclusion relations among the colors; following [Johnston, 1992], I shall refer to these as 'unity relations'. For example, one of the unity relations is that red is more similar to orange than it is to blue; another is that no shade of yellow is a shade of blue; others include that red and green, orange and blue, and yellow and purple are pairs of maximally dissimilar hues. By gathering all such relations, we can construct a Quinean similarity space of the colors, and assign to each color a place in this space. A second set of structural relations among the colors involves the idea that there are precisely six phenomenally elementary colors, which are experienced as being perceptually unmixed: red, green, blue, yellow, black, and white (of these, we are particularly interested in the four chromatic colors red, green, blue, and yellow). That is, there is a shade of red which is experienced as not at all bluish and not at all yellowish, there is a shade of green which is experienced as not at all bluish and not at all yellowish, there is a shade of blue which is experienced as not at all reddish and not at all yellowish, and there is a shade of yellow which is experienced as not at all reddish and not at all greenish. In contrast, the vast majority of colors are experienced as being mixed: for example, no shade of orange is experienced as not at all reddish and not at all yellowish, and no shade of violet is experienced as being not at all bluish and not at all reddish. Indeed, even for the colors which have unmixed shades, most shades will be experienced as mixed: thus, although there is an unmixed shade of red, most shades of red are experienced as being somewhat yellowish or somewhat bluish. Colors experienced as perceptually unmixed are called unique; the others are called binary.21

The argument from structure, then, alleges that colors have certain structural properties (they are organized by the unity relations, they admit of a distinction between unique and binary) that the intrinsic properties proposed as identical to the colors lack, and therefore that the colors cannot be identical to those intrinsic properties.

Several answers to the argument from structure have been proposed. For example, Tye proposes that the unique/binary distinction among colors is nothing but a set of associations with the colors that we learn when we discover how various pigments combine (e.g., by mixing paints in kindergarten; see [Tye, 1995], 148). Since there is nothing to block the association of such learned categories with the intrinsic properties identified with colors by intrinsic theories, the thought is that there could be a unique/binary distinction for intrinsic properties, so the alleged distinction between colors and such intrinsic properties would evaporate. Unfortunately, this proposal is implausible for empirical reasons. First, evidence that the unique/binary distinction made by (pre-kindergarten) pre-linguistic infants and non-human primates ([Hardin, 1988], 40ff, [Teller and Bornstein, 1987], [Hardin, 1997], 293--294) suggests that the distinction in question is not learned at all (a fortiori, not learned in the way that Tye suggests it is learned). Second, since kindergarten experimentation informs us that blue and yellow pigments combine to form green pigment, Tye's proposal would have the erroneous consequence that green is binary. For these reasons, it seems that Tye's proposal fails to capture the unique/binary distinction that applies to colors.

A different attempt at answering the argument from structure occurs in ([Hilbert, 1987], 117-118), where Hilbert defines a similarity metric over spectral reflectance distributions intended to preserve the facts of unity. Hilbert's proposal builds on the (plausible) assumption that the perceived similarity or difference between two colors must be understood as a similarity or difference in respect of how those two colors affect at least one of the types of photoreceptors in our eyes. There are three such types of receptors in human beings with normal color vision (S-cones, M-cones, and L-cones), and we can represent the (overlapping) regions of the electromagnetic spectrum to which these receptors are sensitive as RS, RM, and RL (respectively). Assuming, with Hilbert, that colors are spectral reflectance distributions, the total contribution made by a color (i.e., reflectance) to one of the photoreceptor types will be equal to the integral of that reflectance curve over that receptor's spectral region (RS, RM, or RL). Indeed, we can represent the total contribution of a reflectance to all the receptor types as a triple of such integrated reflectance values, each of whose components is the contribution of the reflectance to a single receptor type. Then, given two different colors, we can think of the similarity/difference between them in terms of the distance (Hilbert envisions a three dimensional Euclidean distance metric) between the triples of integrated reflectances corresponding to the two colors. Unfortunately, however, this proposal is inadequate to the empirical data; as noted in ([Thompson, 1995], chapter 3), Hilbert's metrical relation correlates with subjects' similarity judgments only in extraordinarily constrained circumstances --- those in which samples are all of a constant lightness level viewed against a homogeneous surround.22

However, while these attempts by intrinsic theorists to answer the argument from structure are unsuccessful, I think there is an alternative, and more compelling, response that can be given on behalf of the intrinsic theorist. The key to this line of response is the idea that we may regard the structural properties of the colors primarily as structural properties of color experiences, and only derivatively as structural properties of the colors themselves.23 This deserves some explanation.

The data supporting the claims about the facts of unity and the unique/binary distinction come from subjects' judgments about the colors. Presumably the subjects who make these judgments are perceiving the colored stimuli presented to them; and, however we ultimately understand this process, it seems that perceiving colored stimuli is at least correlated with undergoing token color experiences. Indeed, it is plausible that subjects' similarity judgments about colors are made on the basis of phenomenal similarities between the color experiences that accompany their color perceptions.24 If this is right, then it is reasonable to take the data concerning the structural properties that we have reviewed as showing that color experiences have a similarity structure, and admit of a distinction between unique and binary.25 Significantly, to interpret the data in this way neither requires nor precludes the claim that the colors themselves have these structural properties as well.

Moreover, notice that, while the intrinsic theorist insists on distinguishing between colors (according to her, intrinsic properties of surfaces and the like) and color experiences (mental states that are, according to her, metaphysically independent of the color properties), she will surely agree that color properties are the de facto bases for the dispositions to cause color experiences. That is, the intrinsic theorist will say that a tomato's redness is not constituted by its relation to color experiences, but she will readily admit that the tomato's color is (contingently) the material basis for its disposition to cause red experiences in perceivers.

Given that the intrinsic theorist will allow this relation between the colors and color experiences, and given that color experiences have the structural properties of unity and the unique/binary distinction, the intrinsic theorist can now provide analogous properties for the colors themselves, thereby answering the argument from structure. For she may construct the structural properties of the colors by deriving them from the structural properties of the color experiences the dispositions to produce which the colors are the bases. Viz., the intrinsic theorist may say that red is unique and that orange is not because the former is the basis for the disposition to produce a unique color experience (a red experience) while the latter is the basis for the disposition to produce a binary color experience (an orange experience). Similarly, she can hold that red is more similar to orange than it is to green because these intrinsic color properties are bases for the dispositions to produce red, orange, and green experiences, and because the experiences stand in the phenomenal similarity relations they do.

It seems to me that this explanatory strategy should count as a satisfactory reply to the argument from structure on behalf of the intrinsic theorist.

5. Objections Against Relational Views

So far I have discussed the motivations for intrinsic and relational views of color, and have reviewed some of the prominent objections against intrinsic views. Of course, the literature also contains several objections against relational accounts, although they are most often formulated as objections against dispositionalism (this is understandable, insofar as dispositionalism is by far the most familiar relational account of color). In what follows, I shall consider three objections against relational views. The first two, in different ways, allege that relational accounts fail to do justice to the phenomenology of color, while the third concerns the place of color in the causal order.

5.1 Relational Accounts and Phenomenology

Many have objected to relational understandings of color on the strength of a phenomenological intuition to the effect that colors are not, as these views claim that they are, relational. Rather, some have insisted, everyday phenomenology presents colors as monadic properties. Thus, whatever successes they have along other dimensions, relational accounts are accused of misrepresenting the phenomenology of color (see [McGinn, 1996] for one version of this accusation).

Obviously, one way in which a relationalist might respond to the accusation that her view runs roughshod over the phenomenology of color is to deny that phenomenology is authoritative about whether colors are relational or not. Such a denial could take either one of two forms. First, a relationalist could hold that color phenomenology is silent with respect to the relationality (or otherwise) of colors.26 Alternatively, a relationalist may admit that phenomenology speaks to the relationality of colors --- in particular, that it presents colors as monadic --- but insist that this presentation is erroneous.

Both of these responses require denying a thesis (helpfully dubbed 'Revelation' in [Johnston, 1992]) to the effect that the phenomenology of color provides an authoritative, complete, and incorrigible understanding of color. But Revelation has seemed attractive to many theorists; for example, Russell tells us that, "the particular shade of colour that I am seeing... may have many things to be said about it.... But such statements, though they make me know truths about the colour, do not make me know the colour itself better than I did before: so far as concerns knowledge of the colour itself, as opposed to knowledge of truths about it, I know the color perfectly and completely when I see it and no further knowledge of it itself is even theoretically possible" ([Russell, 1912], 47).27 Many contemporary philosophers who have thought about the issue reject Revelation, and therefore would not take that thesis as an obstacle to the defenses against the phenomenological objection just considered.28

Thus, a number of relationalists have answered the phenomenological objection simply by stating their dissatisfaction with Revelation. Unfortunately, this sort of answer will not mollify an anti-relationalist, for she will insist that the phenomenological objection survives the denial of Revelation. After all, she might suggest, the objection doesn't require (as per Revelation) that phenomenology is authoritative about the entire nature of color; all that she needs for her objection is that the phenomenology of color is correct in its presentation of colors as monadic. And now the question is what independent motivation we have for accepting or denying this particular deliverance from color phenomenology.

I believe that the relationalist can provide independent motivation for denying the particular phenomenological impression that colors are monadic by showing that there is a similar impression of monadicity connected with other properties that are less controversially relational. The thought is that if an impression of monadicity is compatible with relationality in the case of properties other than color, then we should not trust the inference from the impression of monadicity to the conclusion of non-relationality in the special case of color properties either.29

Consider, then, the property too heavy to lift. As I go around in the world I judge that certain things are too heavy to lift, and others are not. Because I constantly make these judgments against the background of my own (relatively stable over time) strength and physical build and in the context of a relatively unchanging gravitational field (the field is relatively unchanging because I will hold it fixed as much as I can for purposes of judging whether objects are too heavy to lift), my classifications seem reasonably stable from occasion to occasion, and therefore I have no reason to suspect that too heavy to lift is anything but a monadic property of objects. A moment's thought, however, should convince us that too heavy to lift is a relation between objects and subjects (and gravitational fields). Here, too heavy to lift initially seems monadic, even though it is not, because the relative stability of all but one of the relata (only the object changes, while the lifter and the gravitational field remain fixed) prevent us from seeing the relevance of the other relata.

The relationalist about color will say the same thing about color properties that was just said about too heavy to lift. To wit: As I go around in the world I judge that certain things are red, and others are not. Because I constantly make attributions of redness against the background of my own (relatively stable over time) perceptual apparatus and in the context of relatively unchanging perceptual circumstances (the perceptual circumstances are relatively unchanging because I will hold them fixed as much as I can for purposes of judging whether objects are red), my classifications seem reasonably stable from occasion to occasion, and therefore I have no reason to suspect that red is anything but a monadic property of objects. A moment's thought, however, should convince us that red is a relation between objects and subjects (and perceptual circumstances). Here, red initially seems monadic, even though it is not, because the relative stability of all but one of the relata (only the object changes, while the viewer and the perceptual circumstances remain fixed) prevent us from seeing the relevance of the other relata.

If the foregoing is plausible, then we should be unsurprised to find that phenomenology can mislead us about the relationality or non-relationality of various properties. And if this is correct, then we should not allow the phenomenological intuition that colors are monadic to prevent us from endorsing a relational understanding of color.

5.2 Relational Accounts and Seeing

A second, related, argument pressed against relational accounts in [McGinn, 1996] concerns the question of what can be a direct object of the seeing relation. McGinn presents this objection against dispositionalism (a view he had endorsed in [McGinn, 1983]), rather than against relational views in general; he worries that, while manifestations of dispositions may be direct objects of our seeing, dispositions themselves may not. But, the objector continues, since colors are visibilia par excellence --- since they are paradigm examples of properties that can be direct objects of seeing, it follows that colors cannot be dispositions:
When you look at an object you do not see (de dicto) its dispositions to act in certain ways in certain circumstances, but you do see what color it is. Here, of course, I mean direct object perception, not just seeing-that --- seeing the property itself, not merely seeing that it is instantiated. You may see that something is soluble by watching it dissolve, but you do not see its solubility --- that property itself. You can see the manifestation of the disposition, and you may also see the categorical basis of the disposition in the object's molecular structure, but your eyes do not acquaint you with the property of being disposed to dissolve.... And now the point about colors is that they enter the very content of primitive visual experience, being part of how objects appear, but dispositions of whatever kind cannot themselves enter visual content in this way ([McGinn, 1996], 540; cf. [Mackie, 1976], chapter 1).
Although McGinn's formulation of this objection is directed against dispositionalism, it seems clear that the worry is applicable to any relational view: if colors are possible direct objects of seeing, then any relational account of color will be threatened by the worry that relations (as opposed to relata) cannot be direct objects of seeing.30

In fact, McGinn had already considered and offered a response to this objection in his earlier defense of dispositionalism ([McGinn, 1983], 133-135). This response (which McGinn disavows in [McGinn, 1996], but which is taken up in [McLaughlin, 1999b]) involves the contention that 'sees' introduces a hyperintensional context (similarly for 'looks') --- one where sameness of sense and reference does not guarantee intersubstitutability salva veritate. If this is right, then one could accept the dispositionalist's proposed analytic equivalence between 'red' and 'the disposition to look red', and then explain the divergence in truth value between 'I see red' and 'I see the disposition to look red' as a consequence of the hyperintensionality of the main verb 'see' (mutatis mutandis for other relational accounts, such as functionalism).

Unfortunately, this answer strikes me as unconvincing: while it is plausible that 'looks' may introduce a hyperintensional context, it seems (at least, to my linguistic intuition) not only that the context created by 'sees' is not hyperintensional, but that it is extensional. If so, the hyperintensionality proposal cannot speak to the present objection, which concerns whether colors can be the direct object of 'sees'.

On the other hand, I do not believe that McGinn's objection against relational views is decisive. For the relationalist might well insist that, if colors are relations, then they are relations that (unlike solubility) can be the direct objects of seeing. For simplicity, I'll discuss this point only in terms of dispositional versions of relationalism, but the response can be generalized. The objection we are discussing allows that manifestations of dispositions can be direct objects of seeing, but insists that dispositions themselves cannot; for example, while we can see (in the direct object sense) a dissolution, we cannot see (in the direct object sense) solubility. However, this might be thought to overlook relevant differences between dispositions, some of which can be the direct object of seeing, and some of which cannot. Namely, one might hold that a disposition can be a direct object of seeing if it is a disposition whose manifestation is a visual experience. For in that case, when the disposition manifests, what happens is constitutive of seeing the disposition. Of course, dispositionalists believe that, if red is a disposition, it is a disposition whose manifestation is a visual experience. Consequently, on the view suggested, they can explain how red is visible, even if that property is a disposition.31 In contrast, the dispositionalist will hold, there are many invisible dispositions: for example, since the manifestation of solubility is not a visual experience, but a dissolution, and since dissolution is not constitutive of seeing solubility (the two are not even coextensive --- there could be a dissolution without any perceivers being present, or even existing), the present answer will not make solubility a possible direct object of seeing. If this proposal is tenable, then dispositionalists (and, by extension, defenders of other relational accounts) are safe from the objection we have considered.

5.3 Relational Accounts and Causal Efficacy

A further argument against dispositional accounts is based on considerations of causal efficacy, and has been pressed most forcefully by Frank Jackson. Jackson points out that colors must be causally efficacious in order to be the causes of our color experiences; but he argues that dispositions cannot be causally efficacious, and concludes from this that colors cannot be dispositions.32 As before, this objection can be generalized into a worry for any relational view of color: if colors must be causally efficacious, then the identification of colors with relations will be untenable for anyone who believes that relations (as opposed to relata) are causally inert; however, for ease of discussion, I'll consider only the version of the argument aimed at dispositional theories in what follows.

The causal argument just reviewed, of course, depends on the premise that dispositions are causally inert. While there are a handful of arguments for this conclusion in the literature, Jackson has tended to rely in his writings on the topic on an explanatory exclusion argument familiar from debates about higher level causation.33 Thus, Jackson writes,

Consider... a fragile glass that shatters on being dropped because it is fragile, and not (say) because of some peculiarity in the way it is dropped. Suppose that it is a certain kind of bonding B between the glass molecules which is responsible for the lass being such that if dropped, it breaks. But then it is bonding B, together with the dropping, that causes the breaking; there is nothing left for the ... disposition itself, to do. All the causal work is done by bonding B in concert with the dropping. To admit the fragility also as a cause of the breaking would be to admit a curious, ontologically extravagant kind of overdetermination ([Jackson, 1998c], lecture 4).

[McFarland and Miller, 1998] have responded to this argument by suggesting that Jackson is unjustified in thinking that the "causal work" done by the disposition is redundant with the "causal work" done by the basis for the disposition. As they point out, it is plausible to understand the disposition's causal efficacy as a complex, derivative sort of causal efficacy, constituted, in part, by the causal efficacy of its basis. Indeed, Jackson himself has provided such a derivative notion of causal efficacy for higher-level properties in [Jackson and Pettit, 1988], so it is especially ironic that he should make this sort of argument against dispositionalism.34 If this is plausible, then there is no reason to worry that the causal efficacy of dispositions is compromised by the causal efficacy of their bases, since the causal efficacy of the basis will be compatible with (indeed, partly constitutive of) the causal efficacy of the disposition. Consequently, if this line of thinking can be sustained, the causal efficacy of color provides no reason to abandon a dispositionalist (or other relationalist) theory of colors.

6. Conclusion

Needless to say, the present essay has not exhausted discussion of the advantages and disadvantages of the various theories considered in the philosophical literature on color.35 Nevertheless, I hope that it provides a useful point of entry into some of the relevant discussions and controversies concerning color.

7. Notes

1. This appears to be at least part of the motivation behind Democritus's famous exhortation that "by convention color exists, by convention bitter, by convention sweet but in reality atoms and void." Something like this argument has been propounded more recently in ([Aune, 1967], 172).

2. This is the route to irrealism pursued in [Hardin, 1988] and [Maund, 1995]. See [Stroud, 2000] for arguments against these and other forms of color irrealism.

3. One of the more compelling recent cases against sense-data construed specifically as the bearers of color properties occurs in ([Hardin, 1988], 96-109).

4. This is not an orthodox criterion for sorting theories of color, but I think it is more useful than the more familiar alternatives: as I shall point out below, more traditional ways of classifying theories of color seem to me to run together a number of orthogonal issues.

5. Care should be used with these terms, since different authors use them differently. This is one respect in which the traditional terminology for sorting theories of color seems to me to be unsatisfactory.

6. Again, caution with this terminology is in order, since philosophers have meant a number of different things by the technical expression 'secondary quality'.

7. Nor should functionalism be thought of as an intrinsic account of colors: functionalists think colors are not particular intrinsic structures, but the second order properties of having some or other structures in virtue of which their bearers are related to observers in a certain way. Consequently, functionalists claim that colors are constituted in terms of relations to observers, and therefore should count as defenders of a relational view. For more on these themes, see [Cohen, 2000].

8. [Lewis, 1997] argues that the circularity in the offing here is not vicious, and therefore that one can hold both a relational account of color properties and a relational account of color experiences. I cannot examine this matter here for reasons of space.

9. See ([Evans, 1948], 196-197).

10. As an alternative to relying on the scientific and industrial recipes, some have proposed deferring to the perceptual variant exemplified in the largest subgroup of subjects in the population. The by-now standard argument against this proposal, which I believe originates in ([Bennett, 1968], 105-107) involves phenol, which apparently tastes bitter to about seventy percent of the population and is tasteless to about thirty percent of the population. Is phenol bitter or tasteless? It may be tempting to answer that phenol is bitter because it tastes bitter to a larger sub-population. However, as Bennett points out, the numerical majority of one of the groups seems insufficient as a motivation for relativizing to that group in fixing the properties of phenol, since numerical majority could be achieved by either group through mass micro-surgery or selective breeding (cf. [Bennett, 1971], chapter IV, [McGinn, 1983], 9-10, and [Jackson and Pargetter, 1987], 71-72).

11. The Munsell color system consists of a set of color samples (chips) used as standards in terms of which the color of test samples can be described. The Munsell system is widely-known and widely-used in scientific and industrial applications. Cf. [Munsell, 1946] and [Munsell Color Company, 1976].

12. Even this apparently quite specific formulation of standard conditions leaves out many factors relevant to the way x looks. For example, Alan Gilchrist and his students have shown me several compelling examples in which the perceived lightness of x at time t depends crucially on factors such as the relative lightness of other objects seen at t, the relative lightness of x at times earlier than t, the relative proportions of the visual field occupied by x and objects of different lightness from x at t, and so on. They also have cases showing that the perceived lightness of x depends on the perceptual groups in which x is perceptually classified: the influence of the relative lightnesses of objects in a geometrical configuration G on the perceived lightness of x depends on the extent to which x is parsed as part of G.

13. Perhaps this limitation is one of practice rather than principle in certain of these cases, but it is necessarily insuperable in at least the case of the rainbow, since rainbows are not visible from a viewing angle of 45 degrees or 90 degrees. Therefore, assuming that we want our notions of color to include what we see in rainbows, appealing to counterfactuals about how things would have looked in standard conditions won't resolve this difficulty.

14. Briefly, these conditions could assign only one of what seem to be equally plausible candidates for the color of directionally reflective materials, and could not assign any well-defined color to materials whose color is dependent on use or translucent materials.

15. One natural suggestion is that the standard condition for viewing should be defined as the point at which maximum resolution is available. However, as Hardin points out, this strategy will fail if there is no point of maximum resolution, which is what is suggested by the variation in size of the receptive fields of retinal neural units ([Hardin, 1988], 71).

16. The CIE (Commission International de l'Eclairage) is an international organization responsible for recommending standards and procedures for light and lighting. The CIE has specified several standard illuminants for laboratory and industrial use in terms of their relative spectral power distributions. The most important of these are standard illuminant A, standard illuminant B, standard illuminant C, and standard illuminant D65.

Metameric pairs are pairs of physically different stimuli (typically distinguished in terms of their having different spectral reflectance distributions or spectral power distributions) that are perceptual matches for a given observer and a given viewing condition.

17. The members of a metameric pair can be perceptual matches under all illuminants only if they agree in any properties assigned to them by an intrinsic theory of color, in which case they are called isomers; but isomers are identified in color by both intrinsic theories and ordinary perception, so they are not challenges to the individuations made by intrinsic accounts of color.

18. In [Hilbert, 1987], Hilbert makes the further claim that relational theorists (he makes the argument only against dispositionalists) are unable to describe his case adequately. [Cohen and Matthen, in preparation] argue that the situation is precisely the opposite of what Hilbert contends: the relationalist can provide a plausible description of the case, while Hilbert's own description is unsatisfying. Briefly, we hold that the unpleasant choice between the counterintuitive alternatives of saying that the surfaces are alike in color (because they match in I1) or unlike in color (because they fail to match in I2) only arises for one who thinks that colors are intrinsic (and therefore that Hilbert's argument begs the question against the relationalist). If, as per relationalism, colors are not intrinsic, but are relations to situations, then we can hold compatibly (i) that the surfaces share a color in I1 and (ii) the surfaces fail to share a color in I2. That is, we can respect both of the intuitions whose rejection strikes us as unpalatable. Even if Hilbert were correct that the denial of (ii) is more counterintuitive than the denial of (i), surely it would be preferable to endorse both (i) and (ii), as one can on a relational view. From this vantage point, Hilbert's response is inferior to that of the relationalist in that it respects only half the data supplied by intuition.

19. The material in this section draws heavily on [Cohen, 2001a], where these matters are discussed at greater length.

20. Versions of this argument occur in many places, including ([Hardin, 1988], xxi-xxii, 66-67) and ([Thompson, 1995], 128-130, 135-139). It is also the main reason for abandoning intrinsic theories given in ([Maund, 1995], see esp. 42, 141).

21. The notion of perceptual mixing is often operationalized in terms of subjects' ability to describe light samples as a combination of percentages of other color names (e.g., subjects can describe orange stimuli without using 'orange' if they are allowed to use both 'red' and 'yellow', but cannot describe yellow samples without the use of 'yellow'; cf. ([Hardin, 1988], 42) and ([Boynton, 1979], 210-211)). Similarly, Clark points out that the set of unique colors is such that (i) mixtures of its members can be found which will perceptually match any color, and (ii) each of its members cannot be perceptually matched by a combination of other colors ([Clark, 1993], 126-127).

22. Hilbert has conceded the failure of his proposal in ([Byrne and Hilbert, 1997a], 285, fn. 32).

23. I have developed such a response in [Cohen, 2001a]. Proposals along somewhat similar lines are also suggested (with varying degrees of explicitness) in ([Shoemaker, 1990], 107-108), ([Byrne and Hilbert, 1997a], 274-279), [Lewis, 1997], [Matthen, 1999], and [McLaughlin, 1999b].

24. I am deliberately leaving open the interesting question whether the phenomenal similarities between experiences must be understood as similarities between the intentional objects of those experiences. On this debate, see [Harman, 1990], [Dretske, 1995], and [Block, 1996].

25. I take this claim about the structure of color experience as the proper interpretation of the data. Of course, if the claim is true, then it deserves explanation of its own. I believe it can be explained in terms of the opponent-process organization of our visual systems (see [Hurvich, 1981], chapter 5). However, I'll put this matter aside for present purposes.

26. This form of response, which has been less popular in the literature than the other alternative (but see, for example, [Ross, 1999]), is criticized in [McGinn, 1997].

27. A similar sentiment comes out in ([Strawson, 1978], 224), and, under the rubric of `transparency,' in ([Campbell, 1993], 178ff). In addition, some such view seems to have led G. E. Moore to insist that yellow is a simple and indefinable quality ([Moore, 1903], 7, 10).

28. While I am sympathetic to the rejection of Revelation, I think we are owed some sort of explanation why this thesis has seemed so attractive -- indeed, so obvious -- to so many. Briefly, I conjecture that at least some are tempted by Revelation because (i) they confuse instances color properties with instances of color experiences, and (ii) their account of color experience accords some kind of epistemic privilege to a subject's own relation to the subjective experiences he undergoes. This (sketch of an) explanation of the appeal of Revelation still attributes an error to proponents of that thesis, but it locates that error by showing how it comes from views that have at least some prima facie plausibility.

29. A similar response is offered in [Shoemaker, 1994b], 254-255.

30. [McLaughlin, 1999b] has suggested that McGinn's worry is less problematic for functionalism than for dispositionalism (he does not, however, use it to advocate functionalism over dispositionalism, since he thinks the objection is ultimately ineffective against both views). According to McLaughlin, even if the disposition to look red cannot be the direct object of seeing (that is, even if the disposition itself is not visible), there seems no reason to say that the property that disposes its bearers to look red cannot be the direct object of seeing. Consequently, he suggests, functionalism can preserve the visibility of the colors, even if dispositionalism cannot.

Unlike McLaughlin, I do not think that functionalism is any better off in this regard than dispositionalism. First, functionalism leaves it open that the colors --- the properties that dispose their bearers to look colored --- might themselves be dispositions. Therefore, the threat of the invisibility of dispositions leaves the functionalist, too, in the undesirable position of admitting that colors might be invisible. And second, functionalism leaves it open that colors are disjunctive (in the actual world), and we might worry that the present objection concerning the invisibility of dispositions would apply to disjunctions as well: one might object that, while a disjunct can be the direct object of seeing (that it can be visible), a disjunctive property built from that disjunct cannot.

31. The claim here is not that we see the disposition by seeing the visual experience that is the manifestation of the disposition --- for one need not accept that we see our visual experiences (in the direct object sense at issue). Rather, the claim is that the visual experience itself (when it is appropriately caused by something to which we are visually attending) is the manifestation of the disposition, and that undergoing such a visual experience constitutes seeing the disposition (in the direct object sense at issue).

32. See [Jackson and Pargetter, 1987], [Jackson, 1996], and [Jackson, 1998c].

33. See, for example, the papers in the second half of [Kim, 1993], the papers in [Heil and Mele, 1995], and [Kim, 1998].

34. [Jackson, 1998] responds to McFarland and Miller, insisting that the kind of causal efficacy accruing to colors cannot be the derivative sort allowed for in [Jackson and Pettit, 1988], but [McFarland and Miller, 2000] correctly point out that this insistence is unmotivated and question-begging in the present setting.

35. It has also ignored or treated hastily a number of philosophical topics connected with color and color perception, including color constancy, the ecological function of color, spectrum inversions, historically salient views about the primary/secondary quality distinction, and the role of color in perceptual similarity spaces, to name a few.

8. References

American Society for Testing and Materials (1968). Standard Method of Specifying Color by the Munsell System. ASTM, Philadelphia. Designation D 1535-68.

Armstrong, D. (1968). A Materialist Theory of the Mind. Routledge, London.

Armstrong, D. (1987). Smart and the Secondary Qualities. In Pettit, P., Sylvan, R. and Norman, J., Metaphysics and Morality: Essays in Honour of J. J. C. Smart. Blackwell, Oxford. Reprinted in [Byrne and Hilbert, 1997], 33-46.

Aune, B. (1967). Knowledge, Mind, and Nature. Random House, New York.

Bennett, J. (1968). Substance, reality, and primary qualities. In Martin, C. B. and Armstrong, D. M., editors, Locke and Berkeley: A Collection of Critical Essays, pages 86-124. Anchor Books, Garden City, New York.

Bennett, J. (1971). Locke, Berkeley, Hume: Central Themes. Clarendon Press, Oxford.

Block, N. (1996). Mental paint and mental latex. In Villanueva, E., editor, Philosophical Issues, volume 7. Ridgeview Publishing Company, Atascadero, California.

Boynton, R. M. (1979). Human Color Vision. Holt, Rinehart and Winston, New York.

Byrne, A. and Hilbert, D. R. (1997a). Colors and reflectances. In [Byrne and Hilbert, 1997b], pages 263-288.
Byrne and Hilbert argue for an intrinsic view of color properties and an intentionalist view of color experience, and consider several objections against these views. In particular, the paper contains an interesting defense against the objection that colors have structural properties that surface reflectance profiles lack and therefore that the two cannot be identified. It also contains a novel and ingenious (albeit fiendishly complex and, in my view, ultimately unsuccessful) response to the objection that the possibility of spectrum inversions refutes their intentionalism about color experience.

Byrne, A. and Hilbert, D. R., editors (1997b). Readings on Color, Volume 1: The Philosophy of Color. MIT Press, Cambridge, Massachusetts.
One of a pair of anthologies on color -- one on the philosophy of color, and one on the science of color. These volumes contain many of the best and most important papers for the purposes of most philosophers, and are the obvious place to begin one's reading on color. The two volumes are reviewed, together with another recent anthology on color, in [Cohen, 2001b].

Byrne, A. and Hilbert, D. R., editors (1997c). Readings on Color, Volume 2: The Science of Color. MIT Press, Cambridge, Massachusetts.
This is the companion volume to [Byrne and Hilbert, 1997b], focussing on the science of color. It contains a very useful introduction by Byrne and Hilbert, and a number of introductory essays on various areas of color science including the physics, colorimetry, psychophysics, and evolutionary psychology of color. This is a geat place for philosophers to begin their study of color science. This book is reviewed in [Cohen, 2001b].

Campbell, J. (1993). A simple view of color. In Haldane, J. and Wright, C., editors, Reality, Representation, and Projection. Oxford University Press, New York. Reprinted in [Byrne and Hilbert, 1997b], 177-190.

Clark, A. (1993). Sensory Qualities. Clarendon Press, Oxford.

Cohen, J. (2000). Color: a functionalist proposal. Under review.

This paper elaborates and defends a functionalist analysis of color properties, compares that view against others that have been discussed more widely in the literature, and argues in favor of functionalism (and against competing views).

Cohen, J. (2001a). Color and Color Space: Structural and Metrical Properties of the Colors. Ms., University of British Columbia.

Cohen, J. (2001b). Two recent anthologies on color. Philosophical Psychology 14(1): 118-122.

Cohen, J. (2001c). Critical study of Stroud's The Quest for Reality. Nous, in press.
A review of [Stroud, 2000].

Cohen, J. and Matthen, M. (in preparation). A Framework for Constructing Colour Properties, Ms., University of British Columbia.

Dretske, F. (1995). Naturalizing the Mind. MIT Press, Cambridge, Massachusetts. Originally delivered as the 1994 Jean Nicod Lectures.

Evans, R. M. (1948). An Introduction to Color. Wiley, New York.

Hardin, C. L. (1988). Color for Philosophers: Unweaving the Rainbow. Hackett, Indianapolis.
The recent explosion in philosophical interest in color can largely be traced directly to this book, which contains an excellent summary of then-current color science (this summary has held up remarkably well since its publication), and shows (by example) how this material from color science can and should be used to constrain philosophical theorizing about color. Hardin argues for an eliminativist view of color by claiming that none of the substantive theories satisfy the empirically-motivated requirements that he believes theories of color must respect.

Hardin, C. L. (1997). Reinverting the spectrum. In [Byrne and Hilbert, 1997b], pages 289-301.
This is one of the two previously unpublished essays in [Byrne and Hilbert, 1997b] (the other is [Byrne and Hilbert, 1997a]); Hardin argues against the relevance of that old philosophical chestnut, the inverted spectrum. He points out that empirical methods establish asymmetries in psychological color space, so that any putative inversion would not go undetected. The upshot is that undetectable inversions may be conceptually or metaphysically possible, but they are impossible in the actual world.

Harman, G. (1990). The intrinsic quality of experience. In Tomberlin, J., editor, Philosophical Perspectives: Action Theory and Philosophy of Mind, volume 4, pages 31-52. Ridgeview Publishing Company, Atascerdo, California.
Harman's piece is a classic and early statement of the increasingly popular intentionalist/representationist conception of phenomenal experience -- the view that the only features of our experiences of which we can be aware are features of the intentional objects of these experiences, and not intrinsic features of the experiences themselves. Harman argues for this view by showing how, if we accept it, a number of otherwise troubling issues concerning phenomenal experience will dissolve.

Heil, J. and Mele, A., editors, (1995). Mental Causation. Oxford University Press, Oxford.

Hilbert, D. R. (1987). Color and Color Perception: A Study in Anthropocentric Realism. CSLI, Stanford.

Hurvich, L. M. (1981). Color Vision. Sinauer Associates, Sunderland, Massachusetts.
This excellent textbook on color vision, written by one of the originators of contemporary opponent process theory, is unfortunately out of print at the present time. Fortunately, the book's two chapters on chromatic and achromatic opponency are reprinted in [Byrne and Hilbert, 1997c]. The book also contains useful chapters on the physics of spectral radiation, the psychophysics of color mixing, the relation of opponent process theory to retinal and cortical electrophysiology, spatial and temporal contrast effects, color adaptation, and several chapters on various sorts of color deficiencies. The presentations are uniformly high in quality, and are easily understandable by the non-specialist (e.g., the philosopher).

Jackson, F. (1977). Perception: A Representative Theory. Cambridge University Press, New York.
Jackson defends a sense-datum (representative realist) theory of perception, according to which the direct objects of perception are mental objects that represent things outside the head. He has since given up this view, but his 1977 book is the clearest and most forceful case for sense-data I have encountered.

Jackson, F. (1996). The primary quality view of color. Philosophical Perspectives, 10:199-219.
Borrowing from [Jackson and Pargetter, 1987], Jackson urges that we should adopt a primary quality theory of color, but one that incorporates several important features of secondary quality theories. Still, he argues from considerations of the causal efficacy of color to the conclusion that colors cannot be dispositions, and therefore that we must maintain a primary (rather than a secondary) quality theory of color.

Jackson, F. (1998c). From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford, New York. Originally given as the 1998 Locke Lectures.
Jackson lays out his positive case for conceptual analysis as a way of doing serious metaphysics, and then gives us a series of examples of his methods in chapters that treat various topics, including color. The chapter on color is virtually identical to [Jackson, 1996].

Jackson, F. (1998). Colour, disjunctions, programmes. Analysis, 58:86-88.
Jackson responds to the ojection of [McFarland and Miller, 1998] that a dispositional analysis of color could preserve the causal efficacy of color by appeal to his own notion of program causation (from [Jackson and Pettit, 1988]. His response is, unfortunately, little more than an insistence that the causal efficacy of color demands more than program causation -- he claims that colors could not be the causes of color experiences (which, of course, they are) unless they were causally efficacious in a non-derivative sense. I cannot find in the piece any convincing arguments in support of this claim.

Jackson, F. and Pargetter, R. (1987). An objectivist's guide to subjectivism about color. Revue Internationale de Philosophie, 160:127-141. Reprinted in [Byrne and Hilbert, 1997b], 67-79.
Jackson and Pargetter want to defend the "Australian view" that colors are objective primary qualities as against a dispositionalist conception of colors as subjective secondary qualities. However, they find that there are many attractive features of dispositionalist views that they want to incorporate into their own account; their project is to show how an objectivist about color can capture the attractive features of subjectivist views without conceding entirely to the subjectivist. The view they end up with is, in this respect, a kind of middle ground between positions that many had previously suspected exhausted the logical space of options.

Jackson, F. and Pettit, P. (1988). Functionalism and broad content. Mind, 97:381-400.

Johnston, M. (1992). How to talk of the colors. Philosophical Studies, 68:221-263. Reprinted in [Byrne and Hilbert, 1997b], 137-176.
This is a long and difficult essay in defense of color dispositionalism, but it repays close study. Johnston argues that, while no plausible theory can respect all of the core truths about color we might have hoped to preserve, a dispositionalist view is closer to meeting this goal than some have alleged. In particular, Johnston defends dispositionalism against several leading objections (several of which have to do with the metaphysics of dispositions, rather than with color per se), and then argues that dispositionalism should be preferred over a "primary quality view" because only the former secures a kind of acquaintance with the colors that we should value.

Judd, D. B. and Wyszecki, G. (1963). Color in Business, Science, and Industry, Second Edition. Wiley, New York.

Kelly, K. L. and Judd, D. B. (1976). Color: Universal Language and Dictionary of Names. National Bureau of Standards, Washington, DC. Special Publication 440.

Kim, J. (1993). Supervenience and Mind: Selected Philosophical Essays. Cambridge University Press, New York.

Kim, J. (1998). Mind in a Physical World: An Essay on the Mind-Body Problem and Mental Causation. MIT Press, Cambridge, Massachusetts. Originally given as the 1996 Townsend Lectures.

Lewis, D. (1997). Naming the colors. Australasian Journal of Philosophy, 75(3):325-342.

Mackie, J. L. (1976). Problems from Locke. Oxford University Press, Oxford.

Marr, D. (1982). Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. W. H. Freeman, San Francisco.

Matthen, M. (1999). The disunity of color. The Philosophical Review, 108(1): 47-84

Matthen, M. (1988). Biological functions and perceptual content. The Journal of Philosophy, 85:5-27.

Maund, B. (1995). Colours: Their Nature and Representation. Cambridge University Press, New York.
Maund argues that our color concepts are "virtual" -- that they place a number of constraints on would-be satisfiers that are, as it happens, not jointly satisfied by anything. These constraints, which Maund thinks are analytically required of anything that deserves to count as a color, are not incompatible, but are simply not all satisfied by anything in the actual world. The upshot is that our color concepts might have been satisfied (by colors), but in fact are not satisfied; this is just to say that there are no colors in the actual world (although, to be sure, things look colored in the actual world).

McFarland, D. and Miller, A. (1998). Jackson on colour as a primary quality. Analysis, 58.
McFarland and Miller find a striking irony in Frank Jackson's views on color and causation. On the one hand, [Jackson, 1996] argues against dispositional views of color on the ground that colors must be causally efficacious and that dispositions are causally inert. On the other hand, [Jackson and Pettit, 1988] have argued that higher level properties can have a causal efficiacy that is derivative on the causal efficacy of their lower level realizers: on this view, the higher level property can figure in causal explanations by "programming" for the presence of some or other low level realizer. The question McFarland and Miller pose to Jackson, then, is why a dispositionalist about color might not secure the causal efficacy of colors (qua dispositions) in terms of a higher level, derivative relation such as program causation.

McFarland, D. and Miller, A., (2000). Disjunctions, programming, and the Australian view of colour. Analysis, 60(2): 209-212.
[Jackson, 1998] responded to [McFarland and Miller, 1998] by insisting that the causal relation between colors and color experiences could not be a derivative relation such as program causation, but had to be understood as a non-derivative kind of causation. Unsurprisingly, McFarland and Miller find this response question-begging, and remain unconvinced.

McGinn, C. (1983). The Subjective View: Secondary Qualities and Indexical Thoughts. Oxford University Press, Oxford.
McGinn never argues explicitly for a dispositional treatment of color in this book. Instead he assumes such an account and then goes on to note a number of interesting consequences of the view. Among them are several ways in which colors count as subjective properties, and ways in which such subjective properties share important features with indexical thoughts.

McGinn, C. (1996). Another look at color. The Journal of Philosophy, 93(11):537-553.
Here McGinn argues against the dispositional account of color he had taken largely for granted in [McGinn, 1983]. He launches a number of interrelated objections against dispositionalism, but many of them concern the worry that dispositionalism is inadequate to the phenomenology of color as we experience it.

McGinn, C. (1997). The appearance of color. Ms., Rutgers University.
The repudiation of the dispositionalism of [McGinn, 1983] that occurs in [McGinn, 1996] turns on discrepancies between dispositionalism and the phenomenology of color. One possible response to this objection, of course, is to hold that, since color phenomenology is uncommitted about the nature of color ontology, color phenomenology is by itself incapable of telling us anything (true and) informative about colors. [McGinn, 1997] takes on this response, urging that color phenomenology must be intimately tied up with color properties.

McLaughlin, B. (1999b). The place of color in nature. In Heyer and Mausfeld, editors, From Light to Object. Oxford University Press, New York.

Moore, G. E. (1903). Principia Ethica. The University Press, Cambridge.

Munsell, A. H. (1946). A Color Notation. Munsell Color, Baltimore.

Munsell Color Company (1976) Munsell Book of Color Munsell Color, Baltimore.

Peacocke, C. (1984). Colour concepts and colour experiences. Synthese, 58(3):365-381. Reprinted in [Rosenthal, 1991], 408-416.
Peacocke's question is one of conceptual piority. Must our concept of red be explained in terms of our concept of red experience? Or should the dependence go in the opposite direction? Or is neither concept prior to the other? Peacocke plumps for the first view -- the view that our concept of red experience is prior to our concept of red. In particular, he is attracted by a dispositionalist analysis according to which the concept of red is unpacked as the concept of some sort of disposition to affect perceivers.

Perkins, M. (1983). Sensing the World. Hackett Publishing Company, Indianapolis.

Rosenthal, D. (1991). The Nature of Mind. Oxford University Press, New York.

Ross, P. W. (1999). The appearance and nature of color. The Southern Journal of Philosophy, 37: 227--252.
Ross hold a view he calls 'color physicalism', according to which colors are intrinsic and "physical properties of physical objects". Many authors have resisted this sort of view because of considerations involving, in one way or another, the way color appears to us and others like us (or unlike us). Ross responds to all of these arguments by holding that the appearance of color is an inadequate guide to the nature of color, and suggests that the arguments against the view he favors depend on untenable assumptions about the revelatory character of color appearance with respect to color.

Russell, B. (1912). The Problems of Philosophy. Oxford University Press, London.

Shepard, R. N. (1992). "The Perceptual Organization of Colors: An Adaptation to Regularities of the Terrestrial World?", in Barkow, J., Cosmides, L., and Tooby, J., The Adapted Mind: Evolutionary Psychology and the Generation of Culture. Oxford University Press, Oxford, pp. 495-532. Reprinted in [Byrne and Hilbert 1997c], 311-356.

Shoemaker, S. (1990). Qualities and qualia: What's in the Mind? Philosophy and Phenomenological Research, 50, Supplement: 109-131. Reprinted in [Shoemaker, 1996a], 97-120.

Shoemaker, S. (1994b). Self knowledge and 'inner sense'. Philosophy and Phenomenological Research, 54: 249-314. Reprinted in [Shoemaker, 1996a], 201-268.

Shoemaker, S. (1996a). The First-Person Perspective and Other Essays. Cambridge University Press, Cambridge.

Smart, J. J. C., (1975). On some criticisms of a physicalist theory of colors. In Cheng, C., Philosophical Aspects of the Mind-Body Problem. University Press of Hawaii, Honolulu. Reprinted in [Byrne and Hilbert, 1997b], 1-10.

Strawson, G. (1978). `Red' and red. Synthese, 78:193-232.

Stroud, B. (2000). The Quest for Reality: Subjectivism and the Metaphysics of Colour. Oxford University Press, New York.
In this book, Stroud argues against relationalist and irrealist views of color. Especially interesting is his transcendental argument against color irrealism: he argues not that color irrealism is false, but that a rational inquirer could never convince herself of its truth by rational means. [Cohen, 2001c] is a critical study of this book.

Teller D. Y. and Bornstein, M. H. (1987). Infant color vision and color perception. In Salapatek, P. and Cohen, L., editors, Handbook of Infant Perception. Academic Press, New York.

Thompson, E., Palacios, A., and Varela, F. (1992). "Ways of Coloring: Comparative Color Vision as a Case Study for Cognitive Science". Behavioral and Brain Sciences, 15: 1-74.

Thompson, E. (1995). Colour Vision: A Study in Cognitive Science and the Philosophy of Perception. Routledge, New York.
Thompson's "enactive" view of color vision understands colors in terms of the ecologically described biological functions performed by the visual systems of different organisms. Although Thompson claims that his view is a distinct alternative to both traditional "objectivist" and "subjectivist" views, it is arguable that the enactive understanding of color amounts to a species of dispositionalism.

Tye, M. (1995). Ten Problems of Consciousness: A Representational Theory of the Phenomenal Mind. MIT Press, Cambridge, Massachusetts.
This book is an extremely well-written defense of an intentional view of phenomenal consciousness -- it proposes to analyze conscious sensations exclusively in terms of the representational content. Since Tye also believes that representational content is susceptible of a naturalistic treatment, this would clear the way for a naturalistic conception of consciousness. As the title suggests, it presents the view through the consideration of ten problems concerning phenomenal consciousness; Tye's burden is to show that his account solves these problems more successfully than competing accounts.

Ullman, S. (1979). The Interpretation of Visual Motion. MIT Press, Cambridge, Massachusetts.

Wright, C. (1992). Truth and Objectivity. Harvard University Press, Cambridge, Massachusetts.

Wyszecki, G. and Stiles, W. S. (1967). Color Science. Wiley, New York.

9. Links

The following color-related items may be of interest to readers of the present essay.

If you know of other relevant sources on the web, please let me know!