Knowledge of Color

As colors have a metaphysics, so they have an epistemology. In addition to ordinary empirical truths about what colors objects have, there are also general truths stating a priori necessities: for example, “Orange is closer to red than to blue”, “There cannot be reddish green”, “Nothing can be white and transparent”, and “Nothing can be red and green all over”.  [1] There is also knowledge of what colors in themselves are (“knowledge of things”): we know what red is, for example. We know quite a lot about colors before we even get to questions about the empirical distribution of colors among objects. Compare shapes: we also know a lot about shapes, both what they are and general truths about them, in addition to propositions about the shape of particular objects. This knowledge is also of a priori necessities. But there is a difference between shape and color: there is no analogue of geometry for color. There is no mathematical science of color comparable to the mathematical science of shape. There are no color theorems analogous to the theorems of Euclid’s Elements. There is barely a mathematics of color at all, though colors do form an abstract structure. Nor can we conceive of colors as derived from anything analogous to lines, as geometrical figures are so derived: you can form circles and rectangles with simple lines, but you can’t form red and green from a single basic chromatic primitive. Colors just don’t have the requisite degree of structure or quantitative character to allow a color geometry. Sounds are closer to shapes in this respect, as evidenced by musical theory: scales and keys are derived from mathematical relations of pitch (pitch being like line, constructively). But colors aren’t structured like closed many-sided figures, and there is nothing like angle in color space. So our knowledge of color doesn’t include anything comparable to geometrical knowledge, despite being a priori and necessary. Colors are not mathematizable in the way shapes are.

            One might speculate that this is because colors are less objective than shapes. If colors are really experiential properties in disguise, then maybe it is the non-mathematical nature of experience that underlies the absence of a geometry of color. The same would not be true of shape, since we can separate objective shape from subjective shape: geometry is about objective shape not experienced shape; and maybe experience of shape is not susceptible to mathematical treatment either. But that kind of subjectivism doesn’t seem correct in the case of knowledge of color: such knowledge is not knowledge of experience as such—we are not thinking about experiences of color when we recognize general truths about color. Perhaps there are deep a priori connections between facts about color and facts about experiences of color, but it seems wrong to suppose that truths concerning color are simply analyzable as truths about experience—a color isn’t an experience! We have experiences of color, but the color of an object is not an experience in the perceiving subject. So the character of our knowledge of color is not immediately explicable in terms of our knowledge of color experience. Further, if we ask why experience of color is not mathematizable, the answer must surely advert to what such experiences are of, viz. colors: but this returns us to the non-mathematical character of colors themselves. There is much more hope of a mathematics of shape experience, given that such experiences are of shapes, which have a developed geometry to call their own. It appears, then, that it is colors as such that resist mathematical treatment: being red is not like being an equilateral triangle—the property itself lacks internal mathematical structure. This is just a basic ontological fact.

            Here is another difference between color and shape: there is nothing analogous to space in the color case. What I mean by this has to be carefully stated: shape can be viewed as a mode of space, but color can’t be viewed as a mode anything analogous to space. Space, like matter, has extension (at least in commonsense physics), and shapes are modes of extension; but there is nothing analogous that has some chromatic property of which the several colors can be viewed as modes. A red object (isolated from other colored objects) is not surrounded by a sea of color, in the way a square object is surrounded by a sea of space. Shaped objects exist in a medium that shares their geometrical nature, but colored objects don’t exist in a chromatic medium: space isn’t colored! The color is simply in the object not borrowed from the medium in which it exists. Space and shape are natural partners, but space is not a progenitor (or twin) of color—and neither is there anything else that plays the role of space in relation to color. There is not some milky stuff, say, that houses the colors we observe in objects. Given space, shape is not a surprise; but space doesn’t prepare us for color, which appears as a radical addition. Color just seems plonked down in space with nothing comparable to support it. There is no chromatic ether.

            Knowledge of color might be cited as a basic type of knowledge that could illuminate other types of knowledge. Puzzles raised by other types of knowledge might have precursors in our knowledge of color, in particular knowledge of what is necessary and a priori. Wittgenstein compares mathematical knowledge to color knowledge, hoping thereby to demystify it.  [2] The comparison might help resist platonic conceptions of mathematical knowledge, given that colors are relatively “concrete”. Certainly we can use color knowledge to argue for the non-uniqueness of mathematical knowledge in its synthetic a priori character: mathematical knowledge doesn’t stand apart from all other types of knowledge in having that kind of status. Even perceived colors can give rise to synthetic a priori truths—not just abstract platonic universals. But the other area that might be compared to color knowledge is knowledge in ethics, which is often compared to mathematical knowledge. It is felt that ethical knowledge needs all the help it can get in securing its epistemic credentials, and mathematical knowledge is then wheeled in as a precursor or partner in crime. The idea certainly has strong appeal for a moral realist convinced that moral knowledge is a species of a priori knowledge. But the color case might be an even better model for ethics, because it is less rigorous than mathematical knowledge and, well, less mathematical. The critic will point to the asymmetry in respect of proofs and general formal sophistication, thus pooh-poohing the comparison to mathematics; but the color case mirrors ethics more closely in these respects. Ethical knowledge resembles our knowledge of color in being synthetic and a priori, and neither is at the level of mathematical science. This is not to downgrade them: it is merely the way they are given the properties and facts concerned. Colors are just not intrinsically susceptible to mathematical treatment (save very superficially), and there is no reason they should be; analogously, moral values are not susceptible to mathematical treatment, and there is no reason they should be. So color and morality belong together epistemologically (we could also bring in sounds and even tastes and smells). In both cases we have a faculty of knowledge that delivers insight into the subject matter in question, thus delivering types of knowledge traditionally described as synthetic a priori (i.e. not analytic and not derived for experience). Anyone seeking to question the status of moral propositions in these respects must explain why they decline to take the same line for propositions about color. It turns out that synthetic a priori knowledge is quite common (and indeed commonplace) and not confined to the supposedly elevated areas of mathematics and morality. Even the humble colors, perceived by human and beast alike, can give rise to such knowledge. Being right is not so far from being red, epistemologically speaking.  [3]                          

Colin McGinn
 

  [1] Wittgenstein discusses these kinds of propositions in Remarks on Colour (1977).

  [2] See Zettel (1967): “’There is no such thing as reddish green’ is akin to the sentences we use as axioms in mathematics” (section 346).

  [3] Readers of my 1983 book The Subjective View may notice some departures in what I write here, though I think they are largely a matter of emphasis.

Colors and Powers

According to Locke, colors are nothing but powers in objects to produce ideas in our minds. He writes: “What I have said concerning colours and smells, may be understood also of tastes, and sounds, and other the like sensible qualities; which, whatever reality we by mistake, attribute to them, are in truth nothing in the objects themselves, but powers to produce various sensations in us, and depend on those primary qualities, viz. bulk, figure, texture, and motion of parts; as I have said.” (Essay Concerning Human Understanding, Book II, Chapter VIII, section 14) There is, he tells us, no “resemblance” between these sensations or ideas and the objective ground of the power, as there is in the case of primary qualities, and without minds to interact with the powers there would be no color (etc.) in the world. To say that an object is red is just to say that it has the power to produce sensations of red in us: that is, colorless objects made of colorless minute particles happen to cause in us a certain type of sensation, and that is all there is to color—it has no mind-independent existence.

            But two questions may be raised about this doctrine, appealing though it is. The first is that a given object has several such powers: it may produce sensations of red in one set of perceivers and sensations of green in another set (and so on for the other colors). It doesn’t have a unique idea-producing power but many such powers (it could even produce ideas like those produced by the other senses). Yet there is only one set of primary qualities underlying this multiplicity of powers. This prevents us from identifying the color with the primary quality basis, on pain of identifying the colors with each other (same basis for each power and hence sameness of color). It isn’t the power in the object that is fixing the color of the object but its relation to the minds of perceivers. Second, how could a colorless object have the power to produce sensations of color? Nothing about the object itself explains the power it has to excite color sensations, let alone specific color sensations. It is, we might say, chromatically impotent. It would be a type of miracle if primary qualities had the power to produce sensations of quite different qualities. The only sense in which objects have such powers is that minds have corresponding powers: the mind has the power, when interacting with colorless objects, to generate sensations of color—the power comes from it not the external object as such. The power of which Locke speaks is really a relational power not an intrinsic power. Intrinsically the object is powerless to produce color sensations; whatever power it possesses is conferred on it by perceiving minds. Locke should have said that colors consist in the power of our mind to impose colors on the world—powers in the mind not powers in objects. Does a square object have the power to produce sensations of a round object? Well, it can produce such a sensation if the perceptual system misfires, but it has no intrinsic power to do any such thing—as it has to produce a sensation of a square object. It just happens to cause (partially) a sensation of roundness; the real work is done by the perceiving mind—it has the power to respond with roundness perceptions to a square object. Similarly, colorless objects can elicit sensations of color, but only because minds are so set up that they can generate sensations of color in the presence of things that have no color. The external primary qualities play a minimal causal role, and considered in themselves have no power to produce sensations at all. Indeed, it is conceivable that there be no such objects and yet the mind has the power to generate the full panoply of colors from within its own resources (color sensations in a vat). In fact, we are already in a situation close to this in that we have colored mental imagery that is elicited by no external object—no primary qualities are triggering this kind of “perception”. The external object in the perceptual case merely triggers a pre-existing power of the mind; it is not the locus of the power to bring color sensations into the world.  [1]

            If we say that water has the power to dissolve salt, we mean that water has objective properties that explain how the power is exercised; but if we say objects have the power to cause color sensations, we can’t provide any such explanation. This is because objects are powerless in this regard; the mind is the origin of the power in question. Imagine a world in which there are simple objects having just two properties and yet these objects are perceived by minds as being rich and complex, endowed with (say) a thousand properties. It would be bizarre to suggest that the objects have the power to produce the full range of sensations available in this world—that power properly resides in the minds that exist in it. It would be quite wrong to say that the properties perceived are nothing but powers in the objects to produce sensations, with their impoverished two-property profiles. The objects have no such intrinsic power, though they have the weak relational power of being able to interact with minds that do have the power to perceive the full range of a thousand properties. In a way Locke undersells his own subjectivist position, which is that the mind is the origin of secondary qualities not the external world. At the least he should have distinguished between the weak sense of power and the strong sense, maintaining only that objects have only a weak power to cause sensations. Primary qualities, by contrast, have a strong power to produce sensations because of that “resemblance” he mentions, but secondary qualities are only weakly connected to the objective nature of external objects. They are connected in roughly the sense in which sugar has the power to taste bitter if the taster’s sense organs are deranged in some way, or in which red objects have the power to appear yellow if there is something wrong with your eyes. The relation expressed by “x can trigger y” is much weaker than that expressed by “x has the power to y”.

            This matters because it affects the strength of subjectivity of Locke’s basic doctrine. If we identify colors with powers-in-objects, then we accord them a degree of objectivity not intended by the basic metaphysics; but if instead we identify colors with powers-in-the-mind, then we fully endorse the fundamental thesis that colors are purely subjective, i.e. originate in the mind and are then imposed or projected onto external things. Here is Locke in full flight: “The particular bulk, number, figure, and motion of the parts of fire, or snow, are really in them, whether anyone’s senses perceive them or no: and therefore they may be called real qualities, because they really exist in those bodies. But light, heat, whiteness, or coldness, are no more really in them, than sickness or pain is in manna. Take away the sensation of them; let not the eyes see light, or colors, nor the ears hear sounds; let the palate not taste, nor the nose smell, and all colours, tastes, odours, and sounds, as they are such particular ideas, vanish and cease, and are reduced to their causes, i.e. bulk, figure, and motion of parts.” (ibid, section 17). That use of “reduced” is perhaps ill advised, suggesting as it does that secondary qualities, and ideas of them, are reducible to primary qualities of external things, which would make them as real as primary qualities in general. But it is clear that Locke intends to maintain that colors (etc.) are mere subjective projections not at all inseparable from matter—they arise from powers of the mind not from powers of matter. It is only in a very weak sense that we can say that objects have the power to produce ideas of secondary qualities, as weak as saying that manna has the power to produce sickness and pain, or a spear has the power to be thrown. The truth is (according to Locke) that colorless particles interact with our colorless sense organs in such a way as to activate the latent power of the mind to generate colors in all their glory. The power to produce color sensations is a mental power not a power of material objects considered in their own right. Perhaps Locke was subliminally influenced by his opposition to innate ideas: for if colors originate in the mind, how can ideas of them be derived from perception of external objects? At any rate, one who sympathizes with Locke’s metaphysics of color has reason to dislike his official object-centered formulation of the doctrine. The powers that give rise to colors are in the mind not in the external world.  [2]

  [1] Much the same can be said about the dispositional formulation of color subjectivism: objects have many color dispositions and no object has such a disposition intrinsically. Rather, minds have dispositions to see the world as manifesting various colors—that is where the disposition originates. Objects are merely disposed to trigger such mental dispositions in the sense that they can trigger them in certain conditions. The ontological work is done by the mind not the world.   

  [2] None of this is to deny that matter might be the origin of the mind’s power to produce colors; it is just that the material objects of perception are not the locus of the power to bring colors into being. Put broadly, Locke (and his followers) are putting too much emphasis on the powers of objects and neglecting the vital contribution made by the mind.

Color and Perception

For color, to be is to be perceived: not so for shape. That was a central tenet of the modern philosophers. Color depends on perceivers for its existence, but shape does not. Thus it makes no sense to suppose that colors exist in a possible world and yet no one perceives them—not humans and animals, not God, and not the colored objects themselves. For what colors would they be? There is nothing to determine what color an object is except how it is perceived (in normal conditions by normal perceivers). The same is not true of shape: shape is determined by how the object behaves in relation to other objects, i.e. by its causal powers. For shape, to be is to be causal. There is something other than perception to fix the shapes of objects. But there is nothing other than perception to fix the colors of objects–hence the vacuity of the idea of a colored world devoid of any perceivers whatsoever. We can certainly subtract a sub-class of color perceivers from a world and leave color intact, but we can’t take allperceivers away and expect to be left with color in all its glory. Suppose colored objects all had eyes and could see their own color: we could remove human and animal perceivers, and even any divine perceivers there might happen to be, leaving only the sighted objects themselves; but we had better not remove the eyes of these remaining perceivers if we want to leave color intact. Nothing like this is true of shape, however, since the causal profile of objects can survive their not being perceived (it is “objective”). This is why people call shape a “physical” property and color a “mental” property: it is a matter of perceiver-dependence, or the lack of it. Most modern philosophers thought the necessary perceivers were human (or possibly animal); Berkeley thought they were human and divine; someone else might suppose that the objects themselves are the perceiving subjects of their sensory qualities. The common thread is that colors need perceivers, because their esse is percipi. Shapes, on the other hand, don’t need perceivers, because their esse is not percipi, but rather causal or mechanical or physical. This is the crux of the distinction between primary and secondary qualities, as it has been traditionally understood. Secondary qualities are the ones that are inextricably linked to mental acts of perceiving in such a way that perception is both necessary and sufficient for their existence (by normal perceivers in normal conditions, etc.). Without suitable perceivers, then, color has no ground of being. It is the same for other secondary qualities, perhaps more obviously so: no world has tastes and smells in it that lacks all perceiving subjects—human, animal, divine, or objectual. For what tastes and smells might these be? Tastes and smells don’t affect inter-object causal powers, so they have no ground beyond perception: to have a certain taste is to taste a certain way to perceivers—not to propel objects in a certain direction that interact with the tasty object.

            Does this imply that secondary qualities are mental (subjective, psychological)? Not strictly, since they may be non-mental qualities that are fixed by mental facts: they might be a special type of primitive property that depends for its instantiation upon mental facts about perceivers. But (a) that would be a pedantic distinction in the absence of some principled reason for withholding the term “mental” from them; and (b) even granting the point we still have the conclusion that colors require mental acts of perceiving, which is all the metaphysician of color is really interested in (it will be sufficient for object mentalism certainly). True, we don’t normally call colors “mental” or “psychological” for boring conversational reasons (the implicatures are uninviting), but from a theoretical point of view there is nothing amiss with extending the use of the term in this way—since colors do depend on mental acts strictly so called. The important point is that colors (etc.) contrast with shapes (etc.) in their relations to perceivers, with the former presupposing them and the latter quite indifferent to them. We can simplify by saying that secondary qualities are mental and primary qualities are not mental. Then we can derive appropriate conclusions from this distinction, notably that colors need perceiving subjects and shapes don’t. Of course, if we were to claim that colors are reducible to wavelengths, then the distinction would collapse and colors would not depend on perceivers  [1]; but I am assuming here that this is wrong and that the choice is between a sui generisclass of non-mental properties and inclusion in the class of mental properties. And the point I am making is that for colors and other secondary qualities, to be is to be perceived by some perceiver or other: take away all perceivers and you take away color. Or better: if there had never been any perceivers, there would be no colors (tastes, smells, etc.)—just as, if there had never been any organisms, there would be no poisons or fearsome objects or healthy life-styles. And which colors an object has depends on how it is perceived: there is no experience-transcendent criterion of color distribution. The same is not true of primary qualities, since they play a role in the causal workings of the world as described by physics.

            We can imagine someone jibbing at the idea that pains are aptly described as mental on the grounds that pains are located in parts of the body and hence are “physical” (“mental” pain is more like inner anguish). The same might be said of perceptual sensations, which are located in or near the sense organs. This fastidious thinker reserves the word “mental” for thoughts, which seem to occur in the inner sanctum of the self. One can appreciate the motivation for this policy of verbal purity, but it doesn’t affect the point at issue: even if we accept it, we can still assert that pains and sense perceptions require a perceiver in order to exist, along with a mental act of apprehension. So they are “mental” by association. They are not like states of the body that require no apprehending subject in order to exist. Even if pains and sense perceptions are not rightly described as mental, they still presuppose a perceiving subject, and that is the central point at issue. Likewise for colors, tastes, etc.: even if they are declared sui generis—neither mental nor physical—we still have the conclusion that they are dependent on a psychological subject. Supervenient facts don’t have to be of the same type as the facts on which they supervene, but they still presuppose those facts—they can’t exist without them. From the point of view of object mentalism, then, it is all one whether we call colors mental or assign them to a separate category; all that matters is that they are perception-dependent. The question at issue is the identity of the relevant perceivers—whether human, animal, divine, or the objects themselves.

            The point on which Berkeley was particularly insistent is that mental qualities cannot inhere in a purely material substance, as many had supposed. If color is mental (or dependent on the mental), then it needs a distinctively psychological subject: it can’t just be the same material substance that primary qualities happily inhere in. So when an object is both cubical and red it needs two supporting substances, one for its primary qualities and one for its secondary qualities (a self in short). If we retain the ideas of primary qualities and material substance (contra Berkeley), then we need to add a suitable substance for the secondary qualities, construed as mental, viz. a perceiving subject. The question then is what or who this subject is—human and animal minds, the divine mind, some sort of immanent cosmic mind, or the mind inherent in the object that has the quality in question. But there should not be much dispute about the conception of secondary qualities that powers this debate, i.e. their essential connection to perception.  [2]

Colin McGinn     

  [1] See footnote 66 of Saul Kripke’s Naming and Necessity (1980) for a view like this.

  [2] For someone who holds that the world could contain all the secondary qualities it now contains in the complete absence of anyone to perceive them, the obvious question is what might confer these qualities on the world. Surely these qualities reflect the survival needs of organisms and are not simply found to exist in the world ab initio. Could it be the case that all organisms misperceive the real colors and tastes of things, these being determined quite independently of how the world seems? Of course, the physical basis of such qualities could exist in the absence of perceivers, but could colors-as-we-perceive-them so exist? Is the redness of red something that exists independently of the appearance of red to perceivers? What if one group of perceivers sees tomatoes as red and another sees them as green—is one right and the other wrong (or both are wrong)?