6

Further Distinctions The foregoing responses to the naive objection are cogent, as far as they go; but in our opinion, they don’t go far enough. The physicalists have described a way in which microphysically constituted colors aren’t represented in visual experience-namely, under microphysical characterizations -but they haven’t yet told us how else such colors are represented. Similarly, the realizationphysicalist has described what color properties are not-namely, microphysical properties-but he hasn’t yet told us what colors are instead. The realizationphysicalist therefore owes us an account of the higher-order properties that are identical with colors, in his view; and all of the physicalists owe us an account of how the properties with which they identify colors can be the ones represented in visual experience. Once again, different physicalists are likely to respond differently. The distinction between Fregeanism and Russellianism and the distinction between the physical identity view and the physical realization view define a four-fold partition of physicalist theories. And within each cell of the partition, further variation is possible. For example, some physicalists believe that the experience of seeing something red normally has a distinctive, introspectible quality in addition to its representational content-a visual “feel,” if you will-and that what the experience represents cannot be understood independently of how it feels. Others believe that a visual experience doesn’t have intrinsic qualities, or that such qualities are in any case incidental to its content. Different physicalists are also motivated by different intuitions about how physically constituted colors are best identified and hence about how they are likely to be picked out in visual experience. Some identify colors as those physical properties which are common to various classes of objects; they consequently treat the perception of colors as the recognition of physical similarities and differences.9 Others identify colors as those physical properties which cause particular visual effects, and consequently treat color perception as the recognition, via those effects, of their physical causes.10 These disagreements might be thought to require further subdivision of physicalist territory, into eight or even sixteen regions instead of four. But we begin to wonder, at this point, whether all of the resulting regions would be occupied by theories that were even remotely plausible. We shall therefore proceed less abstractly, by developing the latter intuitions about how to identify physically constituted colors. Each of these intuitions could in principle lead to eight different theories, as it is combined with Fregeanism or Russellianism, with identity theory or realization theory, and with credence or skepticism about qualia. As we have suggested, however, not all of the resulting permutations are viable. What’s more, the lines of thought departing from these intuitions ultimately tend to converge. We shall therefore attempt to formulate only those accounts of color experience which are both plausible and distinct. The First Intuition: Similarity Classes One way of picking out an object as red is by saying that its surface shares a property with the surfaces of ripe tomatoes, British phone booths, McIntosh apples, and so on. Perhaps, then, an object can be visually represented as colored by being represented as sharing a property with certain other objects. But do references to phone booths and tomatoes crop up in the visual representation of objects as red? Surely, people can see things as red without even having the concept of a tomato or a phone booth. Of course, this particular problem could be circumvented if each person’s visual experience were conceived as characterizing red objects in terms of paradigms familiar to that person. But the resulting conception of visual experience would still be wrong, for two reasons. First, the experience of seeing one thing as red makes no explicit allusion to other instances of the color, familiar or not. No matter how conversant one is with tomatoes, and no matter how centrally tomatoes may have figured in one’s acquisition of color concepts, seeing a red fire engine doesn’t appear to be an experience about tomatoes. Second, visual experience never represents objects as having their colors necessarily or trivially, whereas it would represent tomatoes (or some other objects) as necessarily and trivially red if it represented things as red by characterizing them as sharing a property with tomatoes (or with those other objects).’ 1 The moral of these observations is not that an object’s color isn’t visually represented as a property shared with other objects; the moral is simply that if it is so represented, the other objects aren’t specified individually. The possibility remains that the experience of an object as red represents it as sharing a property with objects in a set that includes tomatoes but which is specified without reference to them or to any other individual members. Yet how can the appropriate set of objects be specified in the content of visual experience, if not in terms of its members? To suggest that it be specified in terms of a property characteristic of those members would defeat the point of the current intuition. The point of the intuition is that a color can be represented in terms of a set of objects precisely because it’s the only property common to all members of the set. Specifying the set in terms of the property characteristic of its members would therefore require an antecedent capacity to represent the color-which would render specification of the set superfluous. A Humean Proposal Nevertheless, the intuition that an object’s color is seen as a property shared with other objects can be preserved, with the help of a proposal dating back to Hume’s Treatise.’2 Imagine that the experience of seeing an object as red has the indexical character “It’s one of that kind,” wherein the reference of “that kind” is determined by the subject’s disposition, at the prompting of the experience, to group the object together with other objects. If the latter objects do constitute a kind, by virtue of possessing some common property, then the experience will have as its content that the former object belongs to that kind and hence that it possesses the characteristic property-a property that could easily be microphysical or realized microphysically. This account of color experience is of the liberal Russellian variety, since it suggests that visual content characterizes its object as belonging to a kind, but that the kind in question must be specified by direct reference rather than by a more specific characterization. Direct reference is mediated in this case by a correlation between potential classificatory behavior of a subject, on the one hand, and a microphysically constituted kind of object, on the other. As we have seen, the proposal has no Fregean version, because it requires specification of a kind, and no such specification can be found in the introspectible content of color experience. An Information-Theoretic Proposal Here is an alternative way of preserving the first intuition. Imagine that a particular mental symbol is regularly tokened in response to visual encounters with objects of a particular kind, whose members belong to it by virtue of possessing some characteristic property. The symbol may then qualify as indicating-and thus, in a sense, as referring to-the kind with which it is correlated., 3 And tokenings of the symbol may consequently introduce its referent into the content of visual experiences, in such a way that objects are represented as members of the kind to which the symbol refers. Such an experience will naturally be described, on the one hand, as registering the similarity of its object to other members of the kind and, on the other, as attributing to its object the property characteristic of the kind. In a sense, then, the object will be seen as having a property by being visually associated with other objects that have it. A microphysical or microphysically realized property may thus be attributed to an object by way of the object’s visually detected similarity to other objects.14 Introducing Qualia Now suppose that the mental correlate of a color category were not some item of a subliminal mentalese but, rather, an introspectible sensation or quale. To begin with, this supposition could simply be appended to the foregoing Russellian account. The visual sensation associated with the appearance of a particular object could then be treated like a numeral in a paint-by-numbers scene, assigning the object to a kind, and hence attributing to it an associated property, without characterizing the kind or property in any way. Which kind or property a particular sensation denoted would be fixed, as before, by causation or correl- ation.15 Once the mental correlate of a color category is imagined as accessible to introspection, however, the resources for a Fregean theory become available. The content of a visual experience can then be imagined to invoke the accompanying sensation and hence to characterize its object under the description “having the property that is this sensation’s normal or predominant cause.” Such an account of how colors are represented can be adopted by proponents of both identity- and realization-physicalism. An identity -physicalist can say that red is the property referred to within the proposed characterization-the property that tends to cause the accompanying sensation. A realization-physicalist can say that red is the higher-order property expressed by the entire characterization-the property of having a property that tends to cause the sensation. On the first reading, colors may turn out to be identical with microphysical properties; on the second, they may turn out to have microphysical realizations. The Second Intuition: Causes of Visual Effects At this point our development of the first intuition, that colors can be identified in terms of similarity classes, has brought us around to the second intuition, that colors can be identified in terms of their visual effects. Indeed, we have already canvassed the only plausible theories derivable from the latter intuition-namely, theories according to which colors are visually represented by, or by reference to, visual sensations that they cause. These theories can be paraphrased as saying that colors are visually represented as the properties that normally cause objects to look colored. But such a paraphrase will make sense only if looking colored is understood to consist in giving a visual appearance that’s accompanied by particular visual sensations, rather than in being visually represented as having colors. For if colors were represented as the properties that normally cause objects to be represented as having colors, the content of color experience would be viciously circular. Now, some philosophers have denied that this circularity would be vicious. One philosopher has even claimed that it would be a virtue, in that it would account for the notorious indefinability of colors. Colors are indefinable, he says, precisely because their definitions are unavoidably circular.16 We think, however, that the proposed circular definition would imply that the content of color experience is vacuous. When one describes an object as having properties that would cause it to be visually represented as red, one is describing it in terms of the experiences that it is equipped to cause, and one is describing those experiences in terms of their content-namely, as experiences of seeing the thing as red. The content of one’s description therefore includes, as a proper part, the content of the experiences that the thing is described as equipped to cause; and the content of one’s whole description depends on that component. For this reason, the description cannot express the content of the experiences in question. If the content of seeing something as red were that the thing was equipped to cause experiences of seeing it as red, then the content of seeing something as red would include and depend upon the content of experiences of seeing it as red. The content of seeing something as red would thus include and depend upon itself; it would characterize the thing, in effect, as having a property that would cause experiences containing this very characterization; and hence it would fail to attribute any particular property to the object. Circularity in the content of color experience would render that content vacuous.17 Thus, the content of visually representing something as colored cannot be that the thing has whatever normally causes objects to be visually represented as colored. As we have seen, however, the content in question can still be that the thing has whatever normally causes objects to look colored, in the sense that it causes their visual appearances to be accompanied by a color sensation. Outline of the Argument We have now developed various proposals for ways in which visual experience might represent microphysically constituted color properties. We began with the Humean proposal that colors are directly denoted by the subject’s classificatory dispositions. We then introduced an informationtheoretic proposal, which says that colors are directly denoted by mental correlates, whether they be items of mentalese or introspectible qualia. We concluded with a Fregean variant of the latter possibility, to the effect that colors are characterized descriptively as the properties that normally cause color sensations. Despite the diversity of these proposals, we think that they are uniformly unsuccessful in showing that visual experience might represent microphysical or microphysically realized colors. Each of them fails to satisfy one of two fundamental requirements for an adequate theory of color vision. First, we shall argue, a theory of color must respect the epistemology of color experience: it must be compatible with one’s knowing what one knows about color properties on the basis of seeing them. The epistemological problem for physicalism is not that the microphysical nature of colors cannot be known on sight; it is rather that other things about colors are known on sight but could not be known in this way if physicalism were true. Second, we shall argue that a theory of color must respect the phenomenology of color experience: it must be compatible with what it’s like to see the world as colored. Mere reflection on what it’s like to see colors does not reveal whether the properties being seen are microphysical, but it does yield various constraints on any theory of what those properties are. In particular, such reflection reveals that color experience is naive, in that it purports to acquaint us directly with properties of external objects. In our opinion, no physicalist theory can meet this phenomenological constraint while meeting those imposed by the epistemology of color as well. We consider these constraints in turn, beginning with the epistemological.