Chapter 3
The Invisible Content of Visual Art*
“All there is to thinking,” he said, “is seeing something noticeable, which makes you see something you weren't noticing, which makes you see something that isn't even visible.
”Norman Maclean
In his characteristically perceptive way, Arthur Danto has focused on a central issue about pictures and shown that, in order to resolve it, some distinctions have to be made. As it happens, the issue stands at the intersection of art and science, an appropriate target for a philosopher such as Danto, who has long employed his analytical skills in both domains. His argument bears on the histories of art and science; it leads us to ask why each has taken the course that it has.
With regard to art, the answer Danto rejects is: because of fundamental changes in the ways we see. There is, of course, an undeniable diversity of pictorial techniques for representing the things we see – schemata, as E. H. Gombrich would call them. Those are discovered or invented, then taught and learned. But it does not follow that there have been significant changes in our perceptual abilities.1 It is more likely, on Danto's account, that pictorial techniques vary because of the development of artistic skills and because of cultural differences in how the ends of art are conceived – pictorial attitudes, as they have been termed.2 It is therefore crucial for his argument that picture perception is unaffected at some basic level by those very attitudes, and the beliefs, desires, or theories they reflect.
Thus Danto claims, partly on the grounds of cognitive science, that (1) picture perception can be explained in terms of the same mechanisms that operate in ordinary perception, and (2) some of those mechanisms are cognitively impenetrable: vision is modular, in the sense of Jerry Fodor and David Marr.3 The first point implies that we do not have to learn to recognize objects in pictures through experience with pictures per se, that is, training in the deciphering of pictorial techniques. The second point implies that we do not have to learn to recognize objects in the undepicted world at all. Because basic recognitional abilities are not affected by learning and experience, they cannot be the product of learning and experience. They must be innate. It follows that people everywhere see – and have for thousands of years seen – the world and pictures of it in pretty much the same way.
I agree with Danto's first point about the reliance of picture perception on ordinary perceptual abilities. And I think his argument that pictorial diversity does not reflect a deep perceptual plasticity is an important insight, the sort of distinction that we very much need to make. Danto is right to hold that pictorial diversity is often due more to artistic skills and know-how than to fundamental changes in visual processes. Even where representational skills are developed in conjunction with theory and employed in its service, it does not follow that picture perception is theory laden. As Danto notes, much of the evidence that has been cited in aesthetics for the theory ladenness of picture perception does not show that at all, and the same is true of similar arguments in cognitive science.4
At the same time, however, that failure does not show that the modularity thesis is true. Indeed, I think that empirical evidence beyond mere pictorial diversity strongly suggests that the modularity thesis is false, at least in the strict sense set forth by Fodor. If so, then Danto's important claim – that we should distinguish between pictorial diversity and perceptual plasticity, and explain the former in terms of something other than the latter – must be grounded in another way. At the end of my discussion, I will suggest such a way. But contesting the modularity thesis, which is largely an empirical matter, is not really my aim. For one thing, I think that nothing in Danto's philosophy of art actually depends on the modularity thesis, congenial though it may be to his views. Yet he has said that it does, and while I will indicate why I am inclined to disagree, a reconstrual of his views is not what I wish to propose. Instead, my reason for setting the modularity issue aside is an argument that, even if the modularity thesis is true, perception can be shown to be highly plastic. This is so because modules are supposed to be informationally encapsulated: vision is protected from penetration only in the form of effects produced by theoretical knowledge, data, or beliefs. But there is another kind of perceptual plasticity, one that has, in fact, figured in the history of art. It has to do with perceptual strategies, taken not as depictive techniques, but as devices employed by the mind and brain. When Danto says that advancements in pictorial representation have been due to progress in the hand, not the eye, my argument will be that the eye – and the brain of which it is an extension – is affected by progress in the hand and by analogous changes of its own.
I begin first, however, with some remarks on the role of modularity and resemblance in Danto's philosophy of art. My goal in this section is to show that the two ideas should be separated, even though they may seem to go hand in hand. It will follow that Danto's appeals to the modularity of vision need not constitute an argument for resemblance. As we will see, modularity plays more than one role in Danto's account, but my claim will be that invoking resemblance between a painting and something else is unnecessary in any case; where the something else is another painting, an object that the painting represents, or any ordinary non-art object, whether the painting represents it or not. But having separated modularity from resemblance in this way, I will then argue that modularity does not really matter.
In what sense, if any, does the content or meaning of a work of visual art depend on resemblance for Danto? Although it is clear that resemblance is not sufficient for artistic content, could it be necessary or figure into content in some way? Danto's language occasionally suggests that the answer is “yes.” For example, he says that pictorial competence requires us to be able to pick out horses in pictures on the basis that they look “enough like horses in the real world,” and that “pictures . . . resemble their denotations.” These ideas are connected. Pictorial competence is the ability to recognize what a picture denotes on the basis of its similarity to its denotatum.
However, it is important to be clear about exactly what work resemblance is supposed to do in Danto's philosophy of art. Two related points should be emphasized. First, Danto suggests that the ability to recognize similarities between a picture and its object is central for pictorial competence. But pictorial competence does not constitute the ability to recognize what a picture represents. To be pictorially competent is just to be able to recognize the objects that are displayed in the picture. That may reveal very little of what the picture represents, in the sense of what the picture is about. Pictorial competence, defined as the ability to recognize similarities, may be necessary for interpretation to proceed; it is, as it were, the first stage in the process. Such a stage is causally necessary if it is, as the modularity thesis tells us, a universal ingredient in the process of object recognition. And interpretation may determine content. But it does not follow that a perceived similarity has to figure into a correct interpretation, that is, that it mandates one interpretation, or even a well-defined range of possible interpretations, over others. Pictorial competence is simply perceptual competence applied to pictures.5
Second, Danto says that pictures resemble their denotata. But “denotation” in Danto's discussion just means perceptual categorization: “Pictures pick out a class,” for example, “horse” if they are pictures of horses. Thus a perceived resemblance tells us only what type of thing a picture represents, in the form of some actual or possible individual. Such perceptual categorization lies at the heart of pictorial competence; thus, it is only a causal antecedent of the ability to interpret the picture. Even if pictorial competence imposes some constraints on the correctness of an interpretation, there will be wide latitude in the meanings that can be assigned. A picture that resembles a reclining man, for instance, might convey a complex meaning, as brought out by the title, Woman in Childbirth. Appreciating the irony, if that is what it is, depends on recognizing the figure of a man; yet the route to the interpretation is so circuitous as to imply that such recognition could lead almost anywhere.6
Indeed, Danto offers an argument in his earlier work that suggests pictorial denotation and pictorial competence have little to do with pictorial meaning, beyond imposing some causal constraints on the processing of visual information. He notes that scientific evidence has shown that pigeons can distinguish different types of objects in pictures. But he argues that, lacking language, pigeons cannot possibly understand the meanings of works of art, which depend on interpretation. This suggests that “what a picture represents” for pigeons is entirely a matter of what it denotes. Thus the pigeons and we can both recognize a bird in a painting in which the bird represents the Holy Spirit; but only we can see it as the Holy Spirit. Danto speaks of this as a “level” of comprehension and says that there is “more to art than perception can account for.”7 This leads naturally to the conclusion that the kind of pictorial competence that pigeons possess is also present in humans, but simply as a stage in the understanding of art that precedes and leads up to a cognitive grasp of artistic meaning. Pictorial competence is, so to speak, the pigeon within us all, a phylogenetically primitive capacity that we require in order to interpret pictures, but one that plays no real role in the interpretation itself.8
It is true that Danto says it is part of the analysis of pictorial meaning that pictures resemble their denotations. Moreover, he seems to see his view as a descendant of atomism (or “externalism” as he calls it), according to which higher-order thoughts are composed of lower-order sense impressions, which are themselves unaffected by entering into the combination.9 And finally, he tells us that artistic perception transcends picture perception; it is of a different order, in which thought is given sensuous embodiment. We might think, therefore, that the forms of embodiment are the residues of picture perception in the experience of art, and that what they resemble makes some positive contribution to the meaning of the thought that is embodied.
Nonetheless, I do not think it follows from any of these claims that resemblance has to be an ingredient in meaning. Rather, resemblance is part of the analysis of meaning in the sense that it is a fact that must be reckoned with (if Danto is right) in a theory of pictorial meaning, as part of an argument. And what remains of atomism in Danto's account is really just the cognitive impenetrability of modules. Moreover, while it is appealing to think that the subtleties of meaning that are communicated depend on the mode of expression, it is surely not true, on his view, that the medium is the message (at least in most cases), or that what is said includes (as we might put it metaphorically) explicit remarks about what is shown. The idea I resist is that Danto must hold a “two-factored” theory of pictorial content in which a denotative factor is signaled (if not determined) by resemblance, and a fuller meaning is supplied by interpretation, with content somehow being a combination of the two. Such theories face serious problems in explaining how to fit the two factors together, problems that I think Danto's account can avoid.
But what then of Danto's appeal to resemblance to resist the conventionalist views of Marx Wartofsky and Nelson Goodman? As I have noted, he opposes the idea that diversity in art is due to (and in turn promotes) perceptual plasticity. Such plasticity has been taken to reflect differences in knowledge, beliefs, and expectations, and, thus, to support the view that “what a picture represents” is fully determined by the propositional attitudes of its diverse perceivers. Danto himself holds that artistic meaning does, in fact, depend on interpretations that are “historically indexed”; yet he resists the conclusion that artistic meanings are, as we might put it, simply the product of interpretation and nothing else. He holds that through our
central cognitive systems . . . we interpret what we sense, relative to the system of beliefs. It is through this interface that history supervenes on perception. The history of perception is the history of central systems vesting what we see with meanings that have not entirely to do with what we see.10
This suggests that while history and interpretation are distinct from perception, they are related to it – they “supervene” on it. And while meanings depend mostly on interpretation, they seem not to depend on interpretation alone. For if meanings have not entirely to do with what we see, they must have something to do with it.
However, if we take the notion of supervenience seriously in this connection, there is no reason to think that the relation of perception to interpretation is one of part to whole, or meaning-element to meaning-elaboration. If history and interpretation interface with perception through the perceiver's beliefs and thoughts, then we can say that they supervene on perceptual states, in the sense that they work through higher-order psychological functions that are “implemented by” or “realized in” lower-order perceptual functions. The latter are the medium or functional architecture of the former in the same way that brain structures provide a medium or physical architecture for the implementation of the perceptual functions themselves. Beliefs are cognitive powers that depend (in the case of understanding pictures) on the capacity for perceptual recognition. But perceived similarities do not have to enter into interpretations any more than a registration of changes of light intensity by neurons enters into the content of representations produced by the recognition of shapes.
This suggests that it is really the modularity of basic pictorial competence that is important for Danto's argument against Wartofsky and Goodman, and not resemblance. To be sure, those ideas may be interconnected. Resemblance might be due to modularity: it is the fact that the same perceptual processes are employed in the same ways that explains perceived similarities. But there is no necessary connection here. Moreover, even if we define resemblance in this way, it is not clear that the “resemblance” is the sort that seems most important with regard to picture perception, viz., resemblance at the level of conscious experience. Although Danto has suggested that the inadequacy he is concerned with is the inadequacy of perceptual phenomenology, there is no guarantee that similarities at the level of unconscious basic visual processes will translate into any commonalities at the level of conscious awareness. There is presumably something it is like to be a pigeon, so phenomenology is something even a pigeon can have. But if resemblance enters there, it is a much more rudimentary form of experienced similarity than we would normally think should play a role in the understanding of art.
One objection that might be raised against my claim that resemblance is irrelevant to Danto's philosophy of art is that resemblance seems clearly to play an important role in his famous method of indiscernibles. According to that method, several paintings can look the same and yet have non-identical contents by virtue of their histories, including culture-dependent beliefs that the artist and perceivers have held. This thought-experiment appears to introduce a reliance on resemblance in the sense that there can be visually identical paintings, works that somehow “look the same.”
Richard Wollheim, for one, has objected to the method of indiscernibles on the grounds that it assumes two objects will continue to “look the same” even after we have learned a great deal about differences in their backgrounds.11 It is not clear, he argues, that we can generalize from the cognitive impenetrability of simple perceptual phenomena to complex works of visual art. And, he notes, even if the modularity thesis is true for the early stages of vision directed at artworks, the constraints it imposes can be superceded by the effects of knowledge on later, more central processes that are conscious and not modular. I think all of these points are correct. Nonetheless, I claim that, even if Danto's method of indiscernibles depends on the modularity of vision, it does not require resemblance: the indiscernibility among paintings need not consist in their producing the same perceptual experience.
To see this, consider a variation on Danto's thought-experiment in which the objects in indiscernible pictures are presented only to the left visual field of a commissurotomy patient. Such a person will be unaware of the nude figure in the appropriate region; he will not experience the figure, at least if reportability is a condition on conscious experience, and having a certain sort of phenomenology is a condition on having a particular kind of experience. Yet, the viewer clearly registers the relevant information, as shown by his non-verbal behavior, for example, his knowing smile or blushing face. The pictures thus represent something to him, even though he is unaware of it. Presented with several such pictures – pictures that are identical in form – it seems reasonable to say that they will be as indiscernible to this viewer as to a person with an integrated brain. He “sees” them as the same, in the sense that he registers the same information via his eyes. Yet, by Danto's hypothesis, the pictures could have different histories. Our intuitions (and his analysis) tell us that, in that case, the pictures will have different contents. While these patients and we, too, may all respond to the pictures in at least some of the same ways (if the modularity thesis is true), that does not determine pictorial content. It follows that Danto's thought-experiment does not depend on perceived resemblance in any interesting sense.
If resemblance is not necessary for the indiscernibility of paintings, then what has Danto's argument actually shown? How can he demonstrate that resemblance is inadequate, unless he has provided evidence that, in cases where there is resemblance, pictorial contents differ? I do not deny, of course, that there is a perceived resemblance among the paintings in Danto's envisioned gallery. My point is that it is possible to generalize from that fact, in a way that brings out a larger issue. If Danto's argument is correct, the result is not just that perceived similarities are not enough to explain pictorial content. It is also more generally that pictorial representation cannot be explained in terms of perceptual psychology. That point is important, because it allows him to appeal to cultural, social, and historical factors, on the one hand; and, on the other hand, it supports his rejection of conventionalism. Conventionalist accounts are wrong, from Danto's perspective, in part because picture perception is, at some basic level, impervious to the effects of acquired knowledge and variations in experience. But resemblance theories are wrong too, because perceptual psychology does not explain how pictures have content. If there is a psychological basis for the transmission of cultural values, goals, or beliefs, a set of mechanisms through which cultural forces operate at the individual level, it is to be found among higher-order, cognitive processes. These are constrained, in some way, by lower-order perceptual modules. But knowing how those modules operate will never be enough to explain our understanding of pictures.12
This leaves us with the question of whether, or in what sense, modularity matters for Danto's argument, once we set resemblance aside. Of course, if the modularity thesis is false, and the thesis is as central to his argument as it appears to be, then his argument must fail. However, I am going to argue that modularity does not matter much, in two respects: first, conventionalism can be resisted without it; and second, an important form of perceptual plasticity remains untouched by the modularity of vision. I conclude that Danto's argument against conventionalism does not fail, and yet it overlooks an important aspect of our perception and experience of visual art.
I begin by considering Danto's application of the animal argument to the history of art. With regard to the claim that picture recognition in pigeons proves that picture recognition is not theory laden, I would note that this is so, only if theory requires language – a point that some philosophers of science would dispute. Paul Churchland has argued, for example, that aesthetic and scientific theories can be identified, instead, with relations among activation patterns in a neural network.13 Whether that is right or not, it points to the possibility that animal behavior might be best explained by positing some type of abstract mental or neural representations that allow the animals to categorize objects in pictures – animal pictogens, as one might call them. If so, then those representations might be interdefined items in a holistic framework, which is modifiable by experience. In that sense, picture recognition could be “theory” laden, even for pigeons. Claims about “theory ladenness” actually raise two issues: (1) whether picture perception depends on interpretation, language, and concepts; and (2) whether it depends on mental contents that are individuated holistically, rather than atomistically – content that is thereby cognitively penetrable. These claims can be separated; arguments that pertain to (1) do not necessarily pertain to (2).
Further, Danto has argued that the animal argument fits nicely with Gombrich's use of a Popperian model of art history. According to that model, art history proceeds by “making and matching” schemata: matching them to visual experience and discarding (as falsified) those that do not fit. This assumes that seeing is, in some ways, constant through changes in showing, as a fixed observational standard against which to measure schemata. One reply is, of course, that the Popperian model should be rejected for both art and science in favor of something closer to Thomas Kuhn. As someone working in the latter tradition might say, the history of both art and science is driven, not by matching theory to theory-neutral observations, but by the construction of new, internally coherent theories that are epistemically and technologically richer (and more closely linked to problems of current interest) than those that went before.
But one could also argue that what is needed for making and matching is only a relative stability in basic concepts that make up the theory, a stubbornness borne out of past success in negotiating with one's environment. Those concepts support experiences that are relatively constant, and it is against those experiences that new schemata are tested. Indeed, Churchland has argued that this phenomenon is evident in the very persistence of illusion that Fodor points to as evidence for modularity.14 If one looks at a picture of an actual scene in which the Müller-Lyer illusion is at work, and not just at isolated pairs of lines, one sees that it could plausibly be said to be the product of a valuable unconscious inference based not on a “hardwired” linear perspective, but on depth cues that depend on real-world knowledge. If so, then only a significant amount of time and effort would tell if the illusion could be suspended.
This line of argument can be taken a few steps further. Danto mentions the case of Magritte's painting, usually called Ceci n'est pas une pipe, as evidence for the fact that objects in pictures are recognized by using the same mechanisms that are used in ordinary perception, an ability, he argues, that is not learned. I want to suggest that the impact of the painting could rather be due to something analogous to the Stroop effect. In Stroop-type studies, subjects find it hard to say the color of the ink in which a word is printed, when the word names a color that is different from that of the ink; for example, when “blue” is printed in red. One explanation is that reading is a highly overlearned, automatic process that interferes with color naming (which requires attention), precisely because it is over learned and automatic. Reading constitutes a virtual module, the operation of which is hard, but not impossible, to penetrate cognitively.
Suppose, then, that object recognition is knowledge dependent, and also automatic (because we have overlearned how to apply the relevant knowledge to objects). In that case, Magritte's picture is not only thought provoking but difficult to process. The words call attention to the fact that we ordinarily see the depicted object as we do the object itself, while implicitly instructing us to ignore that tendency and notice the nonpipe-like qualities of the picture. Object recognition automatically kicks in (this automaticity having been acquired and not inborn), despite the instruction to disengage from it. Thus a feeling of tension or effort is created that plays a role in the provocation of thought.
That is one sense in which modularity is irrelevant: it is unnecessary to explain the stability of experience or to provide a foundation for knowledge. That is because even if vision is, in principle, highly vulnerable to the vicissitudes of change, in practice it actually changes very little. This is due to our psychological make-up and the way the world is. Thus, conventionalism is shown to be false in one important respect, even without modularity: what pictures are taken to represent are not simply arbitrary assignments; they depend on general features of human psychology. Although this does not show that conventionalism is wrong to ascribe a high degree of perceptual plasticity to us in principle, it does take much of the wind out of its sails.
There is also a second sense in which modularity does not matter much for gauging how variable basic perception might be. It is related to the habits of vision, as they might be called, of which Magritte's painting makes us aware. And it points to a larger respect in which the feelings of tension and effort produced by his painting, or the sense of harmony and ease produced by others, play an important role in perceptual phenomenology. But the point in this case is not that the stability of experience can be grounded on perceptual habits rather than the native abilities that are vested in modules, it is rather that there can be such habits, even if there are also such native abilities, and they can vary widely across individuals and cultures. Moreover, their impact extends beyond the qualitative character of picture perception. They also bear on the accuracy, precision, and speed with which we recognize objects in pictures, as well as the level and type of detail we notice. To be sure, these are not matters of life and death when we stroll through a museum, as they might be in the forest, where we thumb through a mushroom guide. But they do tell us something important about how people can see a picture in different ways, and, thus, about aesthetic behavior and education.
To put the point briefly: A number of recent accounts in vision science claim that perceptual tasks like object recognition are not always performed, as David Marr suggests, by constructing a detailed representation through a standard sequence of stages.15 Such things as edge detection and shape representation might play a role in some cases, but they are not necessary in others. The evidence suggests that the visual system includes a number of subsystems, and these may be combined in various ways to perform the task at hand.16 Thus, even if a subsystem is informationally encapsulated, with regard to its own internal operation, it can still be harnessed together with other subsystems to produce a distinct perceptual strategy. In that case, a particular task can be accomplished in different ways, depending on the subsystems that are brought into play.
The rationale for such perceptual strategies is that they permit the visual system to utilize its resources efficiently and to take representational shortcuts if necessary. For instance, properties normally appropriate for one sort of task can be employed in the performance of another, allowing the visual system to sidestep one or more computational problems. There is evidence, in that regard, that sometimes an object's shape is recognized on the basis of patterns of shading across its surface.17 Different types of shape will have distinctive lines of curvature; thus, light will be distributed over them in recognizable ways. If so, object recognition need not depend on explicit representations of the contours of objects or their parts, as we should otherwise expect. Contour and part arrangement correspond to the shape established by shading; but by leaving that information undelineated, the visual system avoids having to solve the problem of where exactly to segment parts and how precisely to represent their shapes.
The point, then, is that as a pictorial technique, using shading to represent shape has a psychological counterpart, viz., retrieving shape from shading as a strategy that the visual system might unconsciously employ. To the extent that this strategy colors visual experience and affects the performance of picture perception tasks, it contributes something to the content of visual art. But it is something that is, strictly speaking, invisible. The effect of strategy on perception has more to do with control over the flow of information – the pathways and processes into which information is channeled – than with inference and computation, that is, the generation of a conclusion by any one of those processes. In that sense, it does not figure into visual information processing proper, and its effects are – to adopt an old-fashioned term in aesthetics – felt rather than seen. But this is invisibility of a different sort than the one Danto has in mind.18
Moreover, the effects of perceptual strategy are compatible with modularity. As I have suggested, a visual function might draw only on a highly restricted domain of knowledge; but when and where and in combination with what other functions it is employed are still matters that can be modified with time and experience. The defenders of modularity may reply that these, then, are not genuine cognitive effects. Information control, they will suggest, is opposed to cognitive processes. One might dispute the restrictive construal of cognition as inference. But my response is to concede the point and argue that the use of control mechanisms in the service of perceptual strategies is extremely important for both cognitive science and aesthetics nonetheless. It is for this reason that I think Danto's distinction between perceptual plasticity, in terms of theory ladenness, and pictorial techniques is so important, while holding at the same time that there is a different form of plasticity critical for our understanding of the perception of art.
Arthur Danto has again helped us to see something we might have missed. Appropriately, he has brought what might have otherwise been invisible into view, viz., the fact that the diverse ways of picture making do not have to be due to the effect of knowledge on basic perception. I think that he has advanced our understanding by redirecting our attention, just as good works of art often do. But my claim has been that what he has made us notice has nothing essentially to do with resemblance between a picture and its object, or with the modularity of vision, onto which artistic meaning is somehow layered. I agree that artistic meaning is invisible, in the sense that it depends on interpretation. My argument has been that artistic content is also invisible in a second sense: it derives, in this case, from the diverse strategies that perceivers use that allow them to economize on mental representation, while gaining new perspectives: strategies that coincide with aesthetically important variations in effort, energy, resource use, and individual ability. The “eye” – if we include the rest of the brain's visual system, of which eyes are the external manifestation – is a pragmatist. Lacking perfection, it is driven to make the most of what it's got. In this quest, it can vary widely in what it sees; but that is only fitting for a creative and clever mind.
Notes
1. Although, for Gombrich, schemata also have psychological counterparts, i.e., mental representations of some sort, and those are highly knowledge dependent. For a discussion of this point, see Mark Rollins, “Pictorial Representation: When Cognitive Science Meets Aesthetics,” Philosophical Psychology, 12, 1999, 387–413.
2. Mark Rollins, “Pictorial Attitudes,” in Oxford Encyclopedia of Aesthetics, ed. Michael Kelly (Oxford, 1998), pp. 154–7.
3. Jerry Fodor, The Modularity of Mind (Cambridge, MA, 1989); David Marr, Vision (San Francisco, 1982).
4. Mark Rollins, “Deep Plasticity: A General Encoding Approach,” Philosophy of Science, 61, 1994, 39–59.
5. At times, Danto speaks as if pictorial competence were relative to a system of pictorial notation and language dependent. To be pictorially competent in the use of a system, he says, is to be able to both distinguish one type of thing from another in pictures “and call them by their right names” (see Danto, “Seeing and Showing,” Journal of Aesthetics and Art Criticism, 59(1), 2001, 1–9). But as he notes, this skill is precisely not a matter of learning to “decode” a notational scheme: it “transcends the differences between notational systems.” Moreover, as we shall see, some non-linguistic animals possess this skill. Perhaps there are layers of pictorial competence. But, if so, the basic level does not depend on language or interpretation.
6. It is important to recognize that Danto's claim that pictures resemble their denotata does not entail that pictures actually denote the objects by virtue of resembling them. According to certain “recognition” theories of depiction, picture recognition draws only on ordinary perceptual abilities, just as Danto has said. (See, e.g., Flint Schier, Deeper into Pictures, Cambridge, 1986; Dominic Lopes, Understanding Pictures, Oxford, 1993.) But recognition is explained in terms of mental representations, the contents of which are determined by their typical causal covariance with items in the world. This opens the door to a theory of pictorial denotation that is based on a causal theory of perception: what a picture denotes, in Danto's sense of denotation, depends on the typical causal covariants of the mental representations it activates. This is very close to Danto's idea of pictorial competence. However, according to recognition theories, even if the mental representations that ground pictorial denotation happen to resemble the picture's denotata, it is their typical causal covariance, not the resemblance, that determines what is denoted. Resemblance may play a diagnostic role for the perceiver, allowing her or him to recognize what is denoted. But the picture need not denote what it does, just in virtue of that fact. My claim, then, is not only that pictures do not denote by virtue of resemblance, but also that resemblance does not have to play this diagnostic role, or any other part, in Danto's theory or argument. I leave aside here the question of how denotation is related to reference and reference to meaning. If reference is distinct from denotation (because it requires a relation to a real individual), that is, of course, consistent with my claim that denotation does not figure into meaning, whatever the relation between reference and meaning is supposed to be.
7. Arthur Danto, “Description and the Phenomenology of Painting,” in Visual Theory: Painting and Interpretation, ed. Norman Bryson, Michael Holly, and Keith Moxey (New York, 1991), p. 212.
8. Of course, this basic capacity must be scaled up in humans. Perceptual categorization in pigeons will either be a conditioned discriminatory response or involve rudimentary mental representations that do not depend on conceptualization, as they do in humans. Pigeons can separate the (pictures of) sheep from the (pictures of) goats without knowing much about what sheep or goats are, or, thus, what pictures of them represent.
9. Danto, “Description and the Phenomenology of Painting,” p. 201.
10. See “Seeing and Showing,” above.
11. Richard Wollheim, “Danto's Gallery of Indiscernibles,” this volume, chapter 2.
12. It might be argued that while resemblance plays no role in determining pictorial content, it is important for our understanding of what a picture is, i.e., what distinguishes pictures from words. But the modularity thesis makes resemblance unnecessary, even here. It says that there is an anatomically distinct visual system in the brain, isolated from the language module. Part of what it is to be a picture is to be an external stimulus that activates the object-recognition process in that system, in the absence of the object in question. That is only a necessary condition, of course, not a sufficient one.
13. Paul Churchland, A Neurocomputational Perspective (Cambridge, MA, 1989), pp. 255–79.
14. Ibid.
15. V. S. Ramachandran, “Interactions between Motion, Depth, Color and Form: The Utilitarian Theory of Perception,” in Vision: Coding and Efficiency, ed. Colin Blakemore (Cambridge, 1990), pp. 346–60; Patricia Churchland, V. S. Ramachandran, and Terrance Sejnowski, “A Critique of Pure Vision,” in Large Scale Neuronal Theories of the Brain, ed. Christof Koch and J. David (Cambridge, MA, 1994), pp. 23–60.
16. Stephen Kosslyn, Image and Brain (Cambridge, MA, 1994).
17. Patricia Churchland and Terrance Sejnowski, “Neural Representation and Neural Computation,” in Neural Connections, Mental Computation, ed. Lynn Nadel, Lynn Cooper, Peter Culicover, and Robert Harnish (Cambridge, MA, 1989), pp. 15–48.
18. One could interpret Danto's anecdote of the use of shading by Lang Shining in a way that fits this model well: Artists at different times and places hit upon techniques, perhaps unwittingly, that exploit different perceptual strategies – different combinations of visual subsystems – to represent objects. If the techniques are exported to places where they are unfamiliar, the objects can still be recognized, but only with a sense of effort, because the underlying perceptual strategies are not the norm. I doubt this would explain entirely the concubine's distress, on seeing her shaded figure; I think Danto is right to say that the tension is more cultural than visual. Yet our accounts are not mutually exclusive. And culture may not be mediated entirely by representational skills and background beliefs. Perceptual strategies can embody cultural commitments as well.