V

The second question about similarity asks whether resemblance is or is not conventional. Both Goodman and Gombrich agree that it makes no sense to speak of perceptual similarities in the real world independent of human perception. This is the point of Goodman's denying "any constant or absolute relationship between a picture and its object" and any "constant and independent sources and criteria of representational practice." It is also the point of Gombrich's rejection of the "innocent eye."

But Goodman rejects as well the unique "success" of the evolution of linear perspective in two-dimensional representations; and Gombrich insists on it.52 For Goodman, there is no empirical basis for

claiming that particular modes of representation are rightly favored over others—species-wide: no "natural" resemblances are ever discernible. For Gombrich, there is a great deal of favorable evidence.

Goodman challenges the logic of the claim—both Gombrich's and J. J. Gibson's.53 Gombrich concedes: "Nelson Goodman is certainly right when he protests that the behavior of light does not tell us how we see things."54 But Gombrich moves on, more promisingly, to theorize about the way in which ecologically shifting perception (in Gibson's sense) inclines us to draw our pictures into the phenomenal world in which we live: although "the world does not look like a picture [he says] a picture can look like the world."55 Both miss what is essential to the relationship between perception and interpretation: namely, that spontaneously perceived resemblance is possible and, in effect, unavoidable; it is the consequence of a culturally entrenched (but biologically viable) form of life within a symbiotized world. Put another way: the biological constraint is primarily holistic; any set of operative predicates that accord with the conditions for the survival of the race and, more narrowly, the survival of particular societies is suitably "realist" in its mode of functioning. Beyond that, within our symbiotized world, we may conjecture about our perceptual proclivities in favor of certain notably stable resemblances. Furthermore, in the arts and history, perceived similarities tend to rely less and less on subcultural constraints. This is another way of confirming the sui generis forms of cultural emergence.

Once we grasp the argument, both the history of linear perspective and the theoretical quarrels about pictorial representation fall into place. As Gombrich firmly and correctly remarks, natural perspective and the rendering of physiognomic likeness are guided by the phenomenal appearance of the world—not by the geometry of light. But the phenomenal appearance of the world is itself elastic, variably shaped by the different perceptual and nonperceptual practices of different historical cultures.

There's the remarkable lesson, the one constantly in danger of being lost in the refined air of Gombrich's and Goodman's quarrel. The perception of nature is as much Intentionally "prepared" as the perception of art; it is only that, in the space of a painting, further complex "objects"—perceptual representations—are discernible "there" that are not "discernible in nature." Even there, the success of spontaneous recognition cannot be merely an artifact of "representational systems [or] pictorial purpose alone."56 There is no need to deny whatever biologi-

cally favored perceptual dispositions obtain. But, relative to the praxis of our predicative categories, "natural" resemblance is always initially inchoate and ultimately interpretive. Gombrich seems unwilling to admit the point. (Still, the biology of perception does make its contribution—which Goodman will not admit.)

Speaking of art and nature suggests a fixity that can never be found. Naturalism in painting is a matter of salience only. Pictorial realism tolerates considerable fuzziness regarding the "rules" of representation, which Gombrich somewhat ignores. But that tolerance, it must be said, is consensual, historicized: it neither converges toward objective resemblance nor establishes similarity by fiat. That is why the classic forms of realism and nominalism (regarding "universals") fail: there is no ulterior ground for extending predicates beyond consensual practice. Yet admitting that does not disallow "correction." Thus, in the matter of horizon line isocephaly (lining up the heads of human beings in the horizonal distance, as in the Masaccio and Masolino frescoes), familiarity with the isocephalic rule, together with fixing the vanishing point on the isocephalic line itself, enables us to tolerate (through our effective memory of a general practice) small but quite definite variations in the legibility of the compared sizes of particular figures: we are often aided more by opportunistic occlusion and the perception of intended spatial relations than by the requirements of strict scale. The "natural" is itself a culturally tolerated perceptual precipitate.

One could make an amusing puzzle of Velázquez's Las Meninas along these lines—by intruding a sense of trickery regarding isocephaly. Thus, the use of the dwarf and the midget at the right of the painting, more or less in the same isocephalic line as the Infanta, raises odd questions about the relative size of the figures represented—a matter that is complicated by the kneeling of one maid on the Infanta's right and the apparent breach of isocephaly by the tall maid on her left, as well as by the "witty" uncertainty of the size of the dog. A similar isocephalic puzzle appears in the line that links Velázquez himself, the mirror image of Philip and Maria Ana, the court officer in the doorway, and the other two adults at the right of the painting; for, now, the illusion is clearly an illusion, since the distance of the mirrored image baffles the isocephalic order if not first correctly interpreted.

In fact, a number of Velázquez's paintings seem to query the significance of creatures of anomalous size in realistic representations. They also query the bearing of the imputed size of representations of

representations (mirror images, other paintings) lined up with "actual" creatures (the painter and the Infanta) represented within one and the same visual image. Velázquez's own training, through Pacheco, makes the imputation entirely reasonable.57 These games favor our construing Veláizquez's painting as a representation of representing by painting (Foucault's point) rather than (or merely) as a representation of a represented scene, a would-be painting (as has been recently suggested by John Searle).58 It also confirms, therefore, the usefulness of the distinction between aboutness and representation, and it warns us of the deep informality, the contextedness, of interpretation itself. (I favor an analogy here between Foucault's treatment of Las Meninas and Barthes's treatment of Sarrasine— in terms, that is, of an adequate theory of interpretation.)

But we must not lose sight of the point that the biological grounding of natural perspective does not entail the unique validity of any particular perspectival canon. If, as Gombrich originally supposed (following Gibson), linear perspective and visual resemblance depend primarily or essentially on the geometry of light, then a uniquely correct canon might have been possible. But if, as Gombrich concedes to Goodman—and, in conceding, supersedes Goodman's conventional-ism—these phenomena depend on the regularities of phenomenal perception, then there is room enough for incommensurable modes of perspectival rendering that may be judged to be equally "correct" in naturalistic terms. Thus: van Eyck's perspective precedes Brunelles-chi's; but if, as Panofsky observes, the Arnolfini has a number of distinct vanishing points, it hardly follows that the Arnolfini is not perspectivally correct. One might even argue that ecological perception, even in paintings, may sometimes be more favorable to alternative vanishing points than to a fixed point.59 This goes against Gombrich's use of Gibson's thesis.

The puzzle about perspectival realism begins, of course, with Brunelleschi's famous mirror experiment.60 The Brunelleschi experiment depends on the (historically original) surprise that follows when—standing in an appropriately fixed place inside the middle door of the Santa Maria del Fiore, one peeps through a small hole located in the back of Brunelleschi's own painting (now lost) of the Santo Giovanni (the Florentine Baptistery) at the represented spot corresponding to one's actual place (one's eye)—one first sees what purports to be a per-spectivally accurate picture of the Baptistery reflected in a mirror and then, when the mirror is removed, one sees the Baptistery itself, which

seems not to have been altered or to be different from the painting in any essential particular. Brunelleschi's experiment, even admitting the limitations of fixed-point perception, certainly confirms the "naturalness" of the perspective he introduced; and certainly, the fixed point of the peephole should not be confused with the representationally fixed vanishing point of Masaccio's and Masolino's frescoes—influenced by Brunelleschi's instruction though they are.

Their work clearly shows the import of Brunelleschi's discovery for a moving or ecological perspective represented two-dimensionally at a pictorially fixed point. It goes without saying that pictorial representation rarely tries to capture the difference between focal and peripheral vision at a fixed point.61 Nevertheless, regarding Brunelleschi-like possibilities involving a moderately restricted ecological perspective in an ambient world, one can find a devilishly clever confirmation in René Magritte's Les Promenades d'Euclide. Magritte's device (which is an improvement on Brunelleschi's) helps to clarify the sense of natural similarity, without yielding to the extremes of conventionalism or transparency; and, perhaps by allusion, it helps to confirm the sense in which Velázquez's painting might be a witty commentary of some sort on Brunelleschi's original experiment rather than a merely realistic representation. How could we rule out such an option, particularly if we incline (however mildly) in the direction of the radical possibilities endorsed by Gadamer's hermeneutics and Foucault's genealogies? Fou-cault's interpretation of the Velázquez begins to look like the mate of Barthes's interpretation of Sarrasine, which, of course, is a welcome economy. For it means that we have now brought the theory of sensory perception sufficiently into line with the theory of predication; the puzzles of historicizing interpretation can now be seen to infect painting and literature (and history) in the same fundamental way, and the sciences and arts can be seen to function, predicatively, in exactly the same way.

In this same sense, it is the insistence on the causal sources of the photograph that threatens its potential inclusion in the artworld. For that insistence privileges a certain regulative model of "natural" perception and thereby excludes the Intentional and artifactual mode of functioning of the photograph's pictorial powers. To speak of art in these terms would then yield nothing but a fictive (or heuristic) internal accusative answering to a certain purely extrinsic interest in some (properly) "real"—some physical—natural, "commonplace" things.62