Introduction

How is it possible to see depth in a flat picture, or a mass of water such as a lake in its dry surface? The purely physical characteristics of pictures, such as flatness and dryness, are often distinct from the represented properties of depth and watery mass that one can see in them. But how can this seeing of such typical pictorial content properties be genuine seeing, if what one sees is not, strictly speaking, actually being physically presented to oneself?

I won't rehearse standard attempts to handle such problems. Instead, I'll present a brief summary of one novel approach that is argued for in my just-published book The Double Content of Art 2005a (DCA), then outline a more comprehensive view that grows out of the DCA approach.

The Double Content of Art presents a sustained series of arguments to the effect that artworks must be identified with the representational content of concrete artifacts such as physical paintings, manuscripts, concrete musical performances and so on. But since clearly we can in general perceive artworks--e.g., see paintings, or hear pieces of music--it follows that we can genuinely perceive any of the kinds of content that are associated with artworks, including their conventional subject matters. A wide range of other arguments for the legitimacy of this straightforwardly perceptual view of our experience of artworks are also provided in the book--see also Dilworth 2005c.

Another central element in the argument of DCA is, as the title suggests, that artworks involve a double or two-level content structure. Artworks must be distinguished from their subject matters, so that the normal case is one in which a concrete artifact such as a physical picture represents the visual artwork itself, which in turn represents its subject matter, with two different kinds of representation being involved. Typically each kind of representation plays a complementary role, with the artwork itself being made up of stylistic, expressive and intentional content, which in turn represents in the conventional sense the subject matter of the picture. A comprehensive theory of these two kinds of representation--aspectual versus intrinsic--and of their relationships is also provided in the book.

I The Double Content of Normal Perception

However, even if all of these claims are accepted for artworks, a more complete understanding of how perception of content is possible would require a broader and more fundamental theory, in which perception of artworks or other representations were closely related to normal perception of non-representational objects. Also, the crucial issue of how exactly perceptual reference to content items is similar to, or different from, standard cases of perception of concrete objects or properties would also need to be fully explained. In addition, since on my account perception of artworks always involves a double content structure, it would be highly desirable to find similar double content structures in normal non-representational perception, as a preliminary to giving a unified account of perception generally.

Fortunately it is possible to show that normal perception must also have a double content structure, as demonstrated in two articles, Dilworth 2005b and 2005d. One basic argument for this view is that retinal or other sensory information is a mixture of information both about the properties of objects, such as their shape or color, and also information about various relevant aspectual conditions, such as ambient lighting, or the angle at which a surface is viewed. Thus perception of the color or shape of an object is achieved only after a decoding operation in which the intermixed object-related versus aspectual information components are separated out, but both components persist in normal perception, such as when one simultaneously sees both the circular shape of a disc, and also the aspectual elliptical shape it has because of the perspective in which one is viewing it.

In terms of standard perceptual and psychological theory, this double content view is a natural extension to perception generally of theories concerning perceptual constancy phenomena, such as the fact that a surface can continue to look white even under different conditions of ambient illumination. The double content view generalizes the idea of 'looking a certain way' to include both aspectual and object-related appearances, which structured combination it argues to be ubiquitous in normal perception.

II The Close Relations Between Pictorial and Normal Perception

Enough theory is now in place to explain, or construct, close links between normal perception of non-representational worldly items, and the perception of artworks. To begin, a useful way to summarize the double content view of normal perception is that it provides an explanation of the extreme flexibility of perceptual recognition mechanisms, in that even very wide variations in sensory data with respect to a worldly property F can be explained as being caused by independent aspectual factors A, which the decoding procedure can appropriately compensate for. The bridge to artworks is provided by the fact that the concrete representational artifacts, upon which a viewer's gaze is directed when she perceives an artwork, also often have very different properties from the subjects that they represent. On the present view, our abilities to perceive artistic contents are based upon the same general-purpose perceptual mechanisms, which enable us to perceive the properties of ordinary objects even under extremely variable aspectual conditions.

For example, the ability to perceive a simple pencil drawing on white paper as having a person as its subject matter arguably is of a piece with our ability to normally perceive a faint outline on a foggy day, in which virtually everything looks white, as nevertheless being the outline of an actual person. Arguably the same, or at least closely similar, perceptual abilities are mobilized in each case. While at the same time, the very different epistemic conditions under which each perceptual episode occurs--with an actual person causing the foggy perception, versus the pencil drawing on white paper causing the perception of the subject matter of the drawing--can serve to explain why the former case is a case of perception of content, while the latter is a case of perception of an actual person.

At this point a critic might remind us of how much still needs to be accomplished, as follows. So far all that has been presented here is a more psychologically sophisticated recognitional theory of artwork perception than perhaps has previously been available. Broadly recognitional theories, such as those of Schier 1986 or Lopes 1996, claim that we recognize items in pictures in much the same way as we recognize actual items of those kinds. The double content view explains in addition the facility with which we are able to do this, and the continuity of artistic with normal cases, but the above account so far has provided no further epistemic, semantic or ontological insight, no matter how effective it might be as a theory of cognitive perceptual mechanisms. It remains paradoxical how one could genuinely see e.g. a lake in a picture when there is no actual lake that one sees, without illusion or error of some kind. Also, specifically ontological issues as to the existential status of the seen content, plus semantic issues as to how genuine reference is possible in such cases, also need to be addressed (though my book DCA, Dilworth 2005a, does address many of these issues).

III Semantic and Ontological Issues Resolved

It seems to me that the critic is right. A satisfactory resolution of these and related issues requires the construction and defense--at least in outline form--of a comprehensive, broadly naturalistic theory of perception, potentially capable of giving an adequate general account of any kinds of perceptual reference and intentionality in naturalistic terms. Such a theory should, among other things, also be able to explain how there could be cases of veridical perception of content without overstepping naturalistic boundaries. I shall now briefly outline the main concepts of one such theory that has recently been developed.

The basic idea is as follows. What is needed is a functionalist theory of perception that explains perception in terms of the three standard kinds of causal role invoked in any functionalist theory of mind or cognition, namely the causal roles of sensory inputs, intervening internal cognitive or dispositional states, and behavioral or other action-related outputs. My version of such a functionalist theory could be called the reflexive theory of perception, in that according to it, an organism Z perceives a property F of object X just in case property F causes Z to acquire some F-related disposition, that is, some disposition to act toward object X in some specifically F-related way. For example, on this reflexive theory, a typical case of person Z veridically perceiving an object X to be red would produce a disposition in Z to engage in some red-related behavior toward X, such as by putting X into a bin reserved for red objects. Non-veridical perception of the redness would show itself in potential incorrect color-sorting actions with respect to object X. (For further discussion see Dilworth 2004 and 2005e).

On this reflexive account, perceptual reference is explained in terms of the closed causal loop between property F, that causes the relevant perceptual state S in person Z, and Z's F-caused and F-related dispositional state S, that potentially would result in some action by Z upon that very same property F itself. The more specific concept of perceptual intentionality would allow also for different ways in which person Z might conceive of property F, whether veridically via a red-related disposition toward F, or non-veridically via e.g. a green-related disposition toward that same property F, even though both intentional states would involve perceptual reference, as reflexively defined above, to the same worldly property F.

Now prima facie, such a naturalistic theory has no way in which to accommodate veridical perception of content items as such. In this reflexive theory, any perceptual reference is entirely explained in terms of closed causal loops between actual objects or properties, and dispositions directed toward those same actual items. But when one sees a lake in a picture, since there isn't any physical lake which causes one's perception of lake-related content when looking at the picture, at best a non-veridical or mistaken perception of a lake would apparently result, leading to an illusion or error theory of content perception.

However, fortunately a functionalist perceptual theory has resources not available in more standard perceptual theories, which can be used to circumvent such problems. On standard, purely input-based theories of perception, if a real lake didn't cause a perception of a lake, then the perception must be non-veridical. But a functionalist theory can also consider as relevant factors both intermediary, internal processing factors, plus behavioral output factors as well. With suitable adjustments in those factors, the fact that an actual lake didn't cause the perception can be accommodated as follows.

To begin, consider what is involved in normal, non-representational cases of non-veridical perception of lakes. For example, one might attempt to drink from, or jump in, something X that one perceives as a lake, only to find out from the behavioral results that one was mistaken. Thus non-veridical perception of an item X as being a real lake involves characteristic kinds of disposition-formation plus behavioral outputs.

Nevertheless, it is clear that normal perception of the lake-related content of a picture involves no such characteristic non-veridical dispositions or behavioral outputs--normally no-one is under any illusion that they are looking at a real lake when perceiving a picture of one, aside from trompe l'oeil cases, and so they are not tempted to act toward the picture as if were a real lake. But such characteristic differences in disposition-formation and behavior toward pictures, as opposed to toward objects that are perceived as being real lakes, provide enough relevant functional differences to warrant a claim that veridicality standards for perception of lake content, as opposed to perception of actual lakes, are independent of, and systematically different from, veridicality standards for perception of actual lakes. As another example, we take it for granted that a picture of a lake can be handed from one person to another, without our assuming that the lake in the picture has thereby shifted its position. But only a radically disoriented person would attempt to hand a real lake to another person, again showing that the relevant veridicality criteria for real versus represented lakes are systematically different in each case.

The remaining issues to be dealt with concern how it is possible to perceptually refer to content items in such cases, and the ontological status of the content item referred to. According to the purely naturalistic reflexive theory of perception being used, all genuine perceptual reference must involve causation by physical objects and properties F of dispositions that are themselves directed toward those very same physical items. So somehow reference to content items must be explained in terms of reference to such actual physical items, if it is to be genuine, on the present account.

At this point, the previous distinction between perceptual reference and perceptual intentionality becomes salient again. In the case of normal perception of a physical property F of a non-representational object, both veridical and non-veridical perception of F involve reference to the same property F, but the intentionality can be different in veridical versus non-veridical cases. Hence there is some sense in which each perceptual content or 'intentional object'--such as that involved in seeing the redness of an object as red in veridical cases, or as green in non-veridical cases--is different in each case, even though both contents involve reference to the very same physical property. Or in broadly Fregean terms, the functional differences between veridical versus non-veridical perceptions of a given property F serve to define different 'senses' or modes of reference for each perceptual content, even though each of them has the same actual reference.

Now this point can be extended to cover the more radically different content cases, in which the basic criteria for veridical versus non-veridical perception have systematically shifted from those appropriate for real cases. Presumably pictures of lakes are such that some group F of their physical properties cause normal perceivers to see lake-related content, which seeing consists in the acquiring of dispositions directed toward those very same physical properties F of the picture. However, as just discussed, the perceptual content or 'intentional object' could be different, depending on whether the viewer veridically perceives the lake content or instead perceives it non-veridically as a picture of a bathtub, even though the reference would be the same in either case, namely those physical properties F of the picture that cause the recognition of the lake content.

A more straightforward functional characterization of the situation, free of traditional intentionality or Fregean concepts, could be provided as follows. On the present naturalistic reflexive theory, the concept of reference involves a fundamental kind of referential spread or bluntness, in that reference is explained in terms of overt behavioral outputs directed toward an object or property. But any physical realization of a relevant functional perceptual state inevitably involves a wide range of interactions of a person's actions with physical objects and worldly states, so that it is never the case that a perceiver interacts with just one worldly object or property. Thus conceptualizations of objects and properties are inevitably more precise than references to them, on the present account. But an implication of this fact is that it is semantically harmless to refer to objects or properties that strictly speaking don't exist as such--as ontologically independent content objects or properties--as long as any such references arise only as part of legitimate functional perceptual states in which concrete physical objects or properties F are causing the relevant states, and which states are themselves referentially related to those same concrete objects or properties F. Or otherwise put, perceptual reference to content items is legitimate as a kind of reference that is dependent on reference to the relevant physical items that normally cause such content-related perceptual states to occur.

As might be expected from this account, the most appropriate ontological view of content items on the reflexive view would be an irrealist one: strictly speaking there are no ontologically freestanding or independent content items, but perception of them, as explained above, can play a legitimate functional role in perceptual activities that exploit the double-content-based recognitional flexibilities of perceptual systems. Or as a related way of explaining content items, appropriate groups of physical properties of pictures give rise to emergent dispositional properties to cause appropriate content-related perceptual states in culturally trained perceivers, and the seeing of content items involves appropriate functional interactions, of the kinds previously described, by perceivers with those emergent physical properties.

To conclude, here is a series of near-platitudes showing why some theory roughly similar to the present one is likely to be correct. First, one certainly can genuinely see a lake when looking at a picture of a lake, because one can immediately recognize lakes in pictures just as easily as real lakes. Second, a theory that doesn't closely relate representational seeing to normal seeing of worldly items has no hope of explaining the basic recognitional psychological data--even pigeons can recognize objects in pictures, and hence see content items. Third, our extremely flexible recognition mechanisms require postulation of two different kinds or levels of content in order to explain their powers. Fourth, no illusion could be involved in the seeing of content by normal viewers, because none of the normal behavioral signs of confusion or misidentification are present in such cases. And fifth, presumably all of the above points must be consistent with some rigorous naturalistic theory of perceptual intentionality, which explains content perception in terms of references to actual objects and properties, because virtually no-one these days seriously doubts that some such theory must both be true, and available to us if we look hard enough for it. The present reflexive view may not be correct in detail, but arguably something similar to its general outlines is theoretically unavoidable.

REFERENCES

Dilworth, John (2004) "Naturalized Perception Without Information", Journal of Mind and Behavior 25 no. 4, pp. 349-368.

----- (2005a) The Double Content of Art, New York: Prometheus Books.

----- (2005b) "The Twofold Orientational Structure of Perception", Philosophical Psychology, 18 no. 2, April 2005, pp. 187–203.

----- (2005c) "A Double Content Theory of Artistic Representation", Journal of Aesthetics and Art Criticism, 63 no. 3 (Summer 2005), pp. 249-260.

----- (2005d) "The Double Content of Perception", Synthese, in press.

----- (2005e) "The Reflexive Theory of Perception", Behavior and Philosophy, in press.

Lopes, Dominic M. (1996) Understanding Pictures, Oxford: Oxford University Press.

Schier, Flint (1986) Deeper Into Pictures, Cambridge: Cambridge University Press.