Abstract and Keywords
This chapter defends a surprising implication of the causal theory of perception: we see silhouetted objects by virtue of seeing their back surfaces. The primacy of surfaces is reviewed to support the principle that we see opaque objects by seeing their surfaces. Since the front surface of a silhouetted object is idle, the causal theory rules out the possibility that we see silhouettes by seeing their front surfaces. Only back surfaces are available.
O Moon, when I gaze on thy beautiful face,
Careering along through the boundaries of space,
The thought has often come into my mind
If I ever shall see thy glorious behind.
—Quatrain attributed by Edmund Grosse to his housekeeper
Chapter 1 dawned with the eclipse riddle: do I see Near or Far? In allegiance with the causal theory of perception, I answered that I see Far even though Near is closer to me. This follow-up chapter develops the most paradoxical implication of this solution: we see the back surfaces of silhouetted objects.
Silhouettes as absorption surfaces
In the case of opaque objects, the silhouette is the surface of the object that makes the object visible by virtue of the light that it blocks. Therefore, to see a silhouette of an object is to see a part of it. Seeing a relevant attached part of an object suffices for seeing the object. We count seeing the silhouette of an object as seeing the object (and seeing its shadow does not count) because we are seeing a surface of the object.
Leonardo da Vinci propounds a whittling objection to the assumption that the surface of an object is the outermost material part of an (p.45) object (Stroll 1988, 40–46). If the surface is the outermost part of the object, then how can it have any thickness? If it were one millimeter thick, then only the outermost half millimeter could be the outermost part of the object. That half millimeter could not be the outermost part because only its outer half could qualify as the outermost part of the object. That outer quarter millimeter can itself be halved. And so on. Consequently, surfaces have 0 thickness and thus cannot be a material part of the object.
Leonardo's whittling argument illicitly privileges immaterial surfaces. He manipulates the domain of discourse for the quantifier “outermost.” What counts as the parts of an object varies with purposes. Sometimes the parts include a coat of paint, the enamel, the peel, and so on; sometimes not. The domain is nondissective: parts of the parts need not be in the domain of discourse. If I say that the surface of my couch is plaid, I am not refuted by the observation that none of the small parts of the surface is plaid. Leonardo da Vinci's whittling argument can proceed only by shifts of context to smaller and smaller parts. Thus, the argument that material objects lack material surfaces is much like Peter Unger's (1975) argument that no material object is flat. Both of these arguments are refuted by the kind of contextual analysis David Lewis proposes in “Scorekeeping in a Language Game” (1979).
A surveyor is free to concentrate on the abstract surfaces of objects; he is interested in geometrical relations. A vision scientist must concentrate on concrete surfaces because only they satisfy the causal requirement for seeing. Consequently, most psychologists and philosophers agree that we see opaque objects only by seeing their surfaces. J. J. Gibson writes:
The surface is where most of the action is. The surface is where light is reflected or absorbed, not the interior of the substance. The surface is what touches the animal, not the interior. The surface is where chemical reaction mostly takes place. The surface is where vaporization or diffusion of substances into the medium occurs. And the surface is where vibrations of the substances are transmitted into the medium. (1979, 23)
(p.46) The scientist will go to whatever depth is necessary for the action. But he will go no further. For opaque objects, this means he stops early. Rigid objects can be almost entirely hollowed out without disturbing their appearance. Architects and set designers exploit this superficiality of vision.
When objects are covered, we become ambivalent about whether we see them. If we count the cover as part of the object, we say we see the object. But it is often a stretch to count clothes and tarpaulins as outermost parts. So if challenged, we will retract the claim to see the object and only claim to see the covering. This flexibility is predicted by the causal theory of perception (Firth 1966). Causal descriptions are highly sensitive to shifting purposes. The standard example is the description of what caused an automobile accident. The police officer blames the excessive speed of the driver, the mechanic blames the worn brakes, and the road engineer blames the steep grade of the road. There is no disagreement because the causal attributions focus on different aspects of the same causal chain. People tend to focus on whichever links in the causal chain they can best manipulate.
Covers are less troubling to silhouetted seeing. In frontlit circumstances, a ski mask greatly decreases the detail of a head. There is little loss for backlit viewing.
The loose hood of a Ku Klux Klan wizard prevents us from seeing his face. But if he stands in front of a blazing cross, we can see through the cloth to his silhouetted profile. Thanks to our excellent face recognition, this simple shape provides enough information to accurately judge race, gender, age, attractiveness, and to cross identify with front view photographs (Davidenko 2007).
The flexibility of causal attributions helps us deflect counterexamples to the doctrine that we see by seeing surfaces. Consider Avrum Stroll's putative counterexample of Venus: we see Venus even though the surface of Venus is completely covered with clouds (Stroll 1988, 75–76). I say that we see Venus only if we count the clouds of Venus as part of Venus. If we count Venusian clouds as not part of Venus, then we do not see Venus.
(p.47) Stroll denies that Jupiter has any surface at all because it is gaseous. Here, I concede that “surface of Jupiter” is vague. But the vagueness of the location is compatible with Jupiter having a genuine surface. Microphysicists regard the surfaces of all objects as vague.
Stroll (1988, 175) also has us consider a tomato that is suspended on a string. We can see the surface of the tomato while it is stationary but cannot see its surface when the tomato is spinning. I agree that one cannot make out important features of the surface (bruises, discolorations, etc.). But that is compatible with still seeing the surface. If I see a distant tomato as a red dot, I see the surface of the tomato even though I cannot make out its surface features. My seeing the surface of the tomato explains why the tomato looks red. When I truly stop seeing the surface of an opaque object, I lose sight of the object.
A batter can detect a “slider” by the appearance of a red dot in the upper-right quadrant of a baseball thrown by a right-handed pitcher (Byhill et al. 2005). The red stitches of the ball are rotating at the critical fusion frequency. Baseball coaches mark balls with red dots as a training aid. But the surface of a normal baseball lacks the mark. The batter is seeing an effect of spinning a surface rather than a feature of the surface itself.
When riding an elevator, we see the elevator by seeing the interior surfaces of the elevator. The situation is more confusing when we are inside of something that is diffuse. A balloonist approaching a cloud first sees it by virtue of the light its outer portion reflects. As he enters the cloud, he sees the cloud from the inside—but not because any neatly delimited surface is available to him. Light is being scattered into his eyes by an undifferentiated collection of water particles.
As the balloonist exits the cloud, he can see the sky, but not because he sees the surface of the sky. The sky is boundless, extending indefinitely into space. The sky is no more material than a hole. (I return to the sky in chapter 5.) Happily, our sky contains something material, air, which can count as part of the sky. The air constitutes the atmosphere, which is a gradually layered “object.” In the daylight, we see the sky by virtue of the lower portions of atmosphere that are made visible by the light it refracts and scatters. You cannot make out which (p.48) is the last part of the sky you can see. You are akin to a diver who can see the ocean around him even when not looking at the surface or the bottom. At night, you can see much farther because the sun's light no longer washes out much in your field of vision.
Most of these complexities are irrelevant to silhouetted objects. The observer is viewing these light blockers from the outside and can often see the precise boundaries of the object. I assimilate purely contrastive seeing to the case of seeing objects that transmit light.
Surfaces have physical properties. In frontlit conditions, we see an object by virtue of the light transmitted by its front layer. In backlit conditions, we see the object by virtue of light blocked by its back layer. A layer blocks by a different mechanism than it reflects. Consequently, the surface of a backlit object consists of the portion of the object that is just enough to block the light. This absorptive surface must be distinguished from the reflective surface. We can represent the two surfaces of the far side of the moon like so: )). The outer parenthesis is the surface relative to light reflection. The inner parenthesis is the surface relative to light absorption. An astronaut in orbit over the far side of the moon during a solar eclipse does not see the silhouette of the moon even though he sees the reflective surface of the far side of the moon. Viewers from Earth are seeing the absorption surface.
More precisely, the viewers from Earth are looking into a concave, bowl-shaped surface. It is indistinguishable from a dark disk just as the silhouette of a skeletal cube (viewed corner on) is indistinguishable from the silhouette of a flat hexagram. However, we are more apt to interpret the silhouette of the moon three dimensionally as a convex surface, in particular, the near side of the moon. This accounts for the temptation to infer that the moon or sun has turned black. Most objects are convex rather than concave. Our strong preference for convexity is probably due to the incorporation of this empirical regularity into our visual system.
Familiar seeing by transmitted light is surprisingly interpretative. The image of a square could be produced by an infinite variety of trapezoids or tilted rectangles. But silhouettes are much more interpretive because much less information is conveyed. Perceptual psychologists (p.49) are impressed by how many of these physically possible interpretations never occur to the viewer. Usually, we promptly commit to a single interpretation. David Marr concludes that we disambiguate by hard-wired guidelines:
Somewhere, buried in the perceptual machinery that can interpret silhouettes as three-dimensional shapes, there must lie some source of additional information that constrains us to see silhouettes as we do. Probably, . . . these constraints are general rather than particular and do not require a priori knowledge of the viewed shape. (1982, 219)
Marr articulates these assumptions as if the lines of sight emanate from an illumined object. If each point on the silhouette corresponds to one point on the viewed surface, we obtain a curve that serves as a contour generator. The three-dimensional shape of the object can be inferred given Marr's further pair of assumptions that nearby points on the contour of the image correspond to nearby points on the object and that all points on the contour generator lie in a single plane.
Thus, Marr underwrites our tendency to view the silhouette of the moon as convex with this rational reconstruction. If Marr is correct, our visual system violates the causal theory of perception. The system assumes, in effect, that we are seeing a part of the object that is causally idle, namely, the front surface of a silhouetted object. This hard-wired heresy might contribute to the counterintuitiveness of my thesis that we see the backs of silhouetted objects.
Regardless of whether I am laboring against a neurological illusion, my thesis that we see the (inner) far side of the moon during an eclipse is paradoxical. The general reason is that we tend to model vision on cases where we see objects by means of the light they transmit. This makes it seem like we are seeing through the 3,600 kilometers of solid rock between the far side and us.
To lessen the appearance of a miracle, some people suggest that I am only seeing the edge of the moon during a solar eclipse. However, the edge can only block light if it is a physical object rather than an abstract mathematical limit. To be visible from Earth, that edge would (p.50) have to be many meters wide. If I can see the backside of this giant ring of matter, then I would still be seeing a surface that is behind many kilometers of solid rock! Reducing the quantitative scale of a miracle secures no advantage: “Don't be so incredulous; Jesus only restored sight to one eye of the blind man.” Half a miracle is still a miracle.
Suppose a florescent balloon meanders in front of a solar eclipse. First, the glowing balloon blocks our view of the bottom edge of the moon, and then the middle of the moon. But if the balloon blocks our view of the middle of the moon, then we must have previously seen the middle of the moon. Therefore, we were not just seeing the edge of the moon during the solar eclipse.
A snake has crawled on the skylight above your bed (fig. 2.1). You watch the snake coil more tightly until most of its silhouette resembles a ball. In the process, you saw the spiral gap narrow until the snake's body was contiguous with itself. That required seeing the center of the snake coil.
Here is a summary of the basic argument for the thesis that I see the entire backside of silhouetted objects: When I see the silhouette of an object, I see the object. I see an object only if I see part of it. That part must cause my perception in an appropriate way. In the case of naked, opaque, silhouetted objects, the only part that can play
Implications for transparency
Ludwig Wittgenstein asked why nothing can be transparent white. Nelson Goodman challenged the riddle by claiming that “the glass in a white light bulb is as transparent as that in a red one” (1978, 504). Jonathan Westphal defended Wittgenstein's presupposition: just as two jockeys can be equally tall without either being tall, two surfaces can be equally transparent without either being transparent. The Oxford English Dictionary defines translucent as “allowing the passage of light yet diffusing it so as not to render bodies lying beyond it clearly visible.”
A white bulb can have the same degree of translucency as a red one, but for it to be as transparent as the red what lies behind it must be as clearly visible as it is through the red. (The bodies lying beyond or behind a transparent glass must be visible as normal, so that the fact that bodies flush against a white glass are somewhat visible is not enough to make the glass transparent. What is actually seen in such a case is a shadow on the glass. The OED definition requires objects lying beyond the transparent medium, not merely behind it, must be completely visible.) (1991, 16)
The distinction between shadows and silhouettes undermines Westphal's defense. True, one cannot see an object by virtue of seeing its shadow. But one can see an object by virtue of seeing its silhouette. Some black objects look the same when backlit as when frontlit. Consider a black phonograph record that hangs behind a pearl-white glass. Because the light emitted from behind the disk is strong, you can see the disk silhouetted (fig. 2.2). You can accurately see how big the disk is, see the hole in the middle, and see that it is swinging back and forth like a pendulum. You are a marksman. You shoot a bullet right through the hole!
White is transparent with respect to silhouetted objects. That is one normal way in which we see objects. Wittgenstein presupposed that all objects are seen in frontlit circumstances—by light bouncing off them. Since white diffuses light, a white screen does occlude objects relative to that mode of seeing. But there are two sides to occlusion.
Confusion about occlusion
Objects that are seen at least partially in virtue of the light they transmit can be blocked from view by interrupting their transmission of light. But a perfectly dark object does not owe its visibility to the light it transmits. Hence, these nontransmitters are exceptions to the principle that one can always conceal an object by interposing an opaque body. We do not need to see through the “blocking” object.
This point solves a puzzle about cast shadows. The first version was devised by Robert Fogelin around 1967 or 1968. It became a topic (p.53) of lunchtime conversations with his colleagues at Yale University: Charles Daniels, Robert Stalnaker, Richmond Thomason, Samuel Todes, and Bas van Fraassen. Fogelin reports that he had been dissatisfied with a colleague's analogy between “Only I can feel my pain” and “Only I can cast my shadow.” Fogelin devised a complicated enigma involving infinitely many shadow casters. He is not sure whether the simplification to two shadow casters was his doing, the doing of another individual, or the product of the Yale collective consciousness. In any case, the Yale puzzle and its solution were presented in detail by Todes and Daniels (1975) in a format that I discuss in chapter 4 (see fig. 4.9). But I present a more natural version that Bas van Fraassen mentions in Laws and Symmetry (1989). Although “shadow” is often used as a count noun, it should be read below as a mass noun (as in “How much shadow is there in the photograph?”):
1. If X casts any shadow, then some light is falling directly on X.
2. X cannot cast shadow through an opaque object.
3. All shadow is shadow of something. (1989, 217)
Imagine a barn casting shadow on a sunny day. A bird flies between the barn and the shadow cast on the ground. The shadow directly beneath the bird cannot be cast by the bird (by virtue of I). Nor can it be cast by the barn (by virtue of II). But no third thing can cast the shadow. Hence, III is violated.
Principle II is true but insinuates that the only way that a shadow can appear on the far side of an intervening object is by penetration. It discourages us from considering the alternative that the shadow appears on the far side by default. Nothing aside from the original blockage of light is needed to place shadow there.
Opaque bodies do not block shadows, nor do they let shadows pass through. It is also a mistake to picture a transparent pane of glass as letting a shadow pass through. The glass lets only light pass through. But this does not mean that glass blocks darkness. Talk of blocking or transmitting shadows is a symptom of the fallacy of reification.
(p.54) A linguistic philosopher might be tempted to trace the reification to the fact that “shadow” is a mass noun even though it has no mass. However, I suspect a stronger influence is the representational economy of treating shade as if it were colored light. This fiction lets us effortlessly extend the projective geometry that is used to depict illumined objects to the depiction of shadows. Artists sometimes draw with white chalk on dark paper. They represent shadows with undrawn regions. This omissive method of representing shadows echoes scientific reality. However, we normally use light paper. To depict a dark area, we must shade in with a pencil or some dark substance. This practice abets the reification of shadows.
Darkness is an absence of light, not a substantive force that can be stopped with shields. Once the light has been stopped, nothing further needs to be done to ensure darkness. This asymmetry of manipulation is a general difference between properties and their privations. A homeowner who insulates his house can keep the cold out only by keeping the heat in. Cold is the privation of heat and hence is not a substance that can be directly manipulated.
Dark objects must effect contrasts
Lord Brain (1965) regards perception of perfectly dark objects as a counterexample to the causal theory of perception. But a dark object can have appropriate causal connections. Sometimes it is enough that light be transmitted by the object's immediate surroundings.
Consider the trick of hiding a small suitcase in front of a backlit large suitcase. The main reason for saying that the overshadowed smaller suitcase is invisible is that the smaller suitcase does not causally contribute to what is seen. The overshadowed suitcase also violates the requirement that S nonepistemically sees O only if O “is visually differentiated from its immediate environment by S” (Dretske 1969, 20). Consider a white moth on a white tree trunk. The white moth causes part of what I see although not in a way that enables me to discern its outline. I am looking at the moth but I am not seeing it.
(p.55) Now consider a perfectly dark moth on a perfectly dark tree trunk. This case differs from the cases involving backlit occlusions. The dark moth does not have a silhouette. The scene is frontlit. Indeed, I am shining a strong light right on the moth. The moth is absorbing virtually all the photons that land on it and hence is playing a causal role in what I am seeing. For instance, if the dark moth is atop a white spot, the moth prevents me from seeing that white spot. Thus, the dark moth satisfies the causal theorist's necessary condition for seeing. The moth is not an idler like Near.
When the dark moth is on a white tree trunk, the moth is seen by the contrast it creates via its light absorption. Absorption is a well-understood physical mechanism. The first step in this understanding is to challenge the question, “Why do black objects absorb energy so well?” Objects are black because they are good absorbers, not the other way around. Good absorbers are also good emitters of energy. That is why radiators are commonly painted black. Whether the surface absorbs or emits depends on the background level of energy. If it is higher, the black object absorbs energy. If the background is lower, then the black object emits energy.
Transmission of light is not a sufficient condition for being seen. The stars shine as brightly during the day as at night. Transmitters of light rely just as much on contrast as dark objects. Human beings have exquisite contrast sensitivity that strains the capabilities of electronic display devices such as television and computer graphics displays. People can detect a change of modulation of less than 1% across a border (Shapley and Lam 1993, xii). This extreme sensitivity to differences in light intensity explains how the “darkness” of an object can itself be a pure contrast effect. Sunspots look dark solely because they are relatively cool spots on the sun. They would shine brightly if they could be removed from the even hotter photosphere of the sun. The universality of this reliance on contrast explains why any object can be camouflaged by muting its contrast with its surroundings. Seen objects normally make a holistic contribution to the scene. They are visible by virtue of the differences that they make with their surroundings. The differences can be achieved by going up or down the scale of light intensity—including all the way down to zero.
(p.56) The contrast between the silhouette and its background is muted by incompletely opaque intermediates such as window curtains. Since our eyelids allow some light to penetrate to the retina, it is possible to see some objects with our eyes closed. Close your eyes and position a bright light bulb so close to them that you can feel the bulb's heat. Now wave a pen just in front of your eyelids. (Contrast sensitivity is enhanced by movement.) Seeing the pen's diffuse silhouette may make you receptive to the hypothesis that purely contrast seeing is biologically prior (phylogenetically and perhaps ontogenetically) to seeing objects by their transmitted light.
All evolutionary accounts start with a proto-eye that is merely sensitive to the presence of light and then develops into a detector of passing shadows. The first seen objects should have been silhouetted figures. Only later would we expect objects to be seen by virtue of the light they transmit. Instead of being a marginal form of vision, purely contrastive seeing is the primal form of seeing. Stereotypical seeing is an elaboration of this more basic ability.
Black holes amass the immediate surroundings that make themselves discernible. The textbook scenario features a black hole that develops from a large star in a binary star system. The larger star has been pulling matter from the smaller star. After the large star accumulates a critical amount of matter, it collapses into a black hole. Since the black hole has the same mass as it had when a large star, it continues to pull matter from its partner star. Since this new matter conserves angular momentum, it forms an accretion disk that revolves around the black hole. The matter from the inner part of the disk accelerates up to nearly the speed of light. The friction creates tremendous heat that in turn emits X-rays. Hence, the black hole makes itself visible by piecing together an environment for itself.
Alvin Goldman (1977, 282) rightly denies that there is a “transfer of energy or force” from the black hole to the observer. But he wrongly denies that there is a physical mechanism by which the black hole makes itself seen. The black hole causes other things to transfer energy to observers. During World War II bombing raids, high-flying bombers could see the silhouettes of low-flying bombers below. The (p.57) low bombers were bottom-lit from enemy searchlights. The low-flying bombers were not directly transferring energy into the eyes of highflying observers. But the low-flying bombers were still causing themselves to be visible by rousing the enemy below.
Dark objects are a varied lot. Nevertheless, neither the overshadowed suitcase nor the overshadowed heavenly body, Near, satisfy the causal requirement. True, I would see Near if Far vanished. But I see in virtue of actual causation, not hypothetical causation. Consider Harry Frankfurt's (1969) counterexample to the principle that I am responsible only if I could have done otherwise. An evil scientist has rigged up a device that will make me do an evil deed if and only if I fail to do it by my own accord. I do the evil deed, so the device does nothing. I am responsible for the evil deed even though I could not have done otherwise.
Compare Frankfurt's case with David Lewis's (1986, 285) censor. The censor is a device that will make me have the visual experience of a scene if and only if I do not have the visual experience in another way. I am having a visual experience of the scene by opening my eyes and receiving reflected light waves in the normal way. Am I seeing the scene?
Lewis says I see if and only if “the scene before my eyes causes matching visual experience as part of a suitable pattern of counterfactual dependence” (1986, 285). The censor illustrates a lack of suitable dependence:
The case is one of causal preemption. The scene before my eyes is the actual cause of my visual experience: the censor is an alternative potential cause of the same effect. The actual cause preempts the potential cause, stopping the alternative causal chain that would have otherwise gone to completion. (1986, 285–86)
Substitute “Near” for “the censor.” Although the censor is idle, its “idleness is an essential factor in the causal process by which matching visual experience is produced. . . . We cannot uniformly ignore or hold fixed those causal factors which are absences of intervention” (1986, 286). Lewis concludes that I do not see the scene.
(p.58) Lewis's verdict has been unpopular. His alter ego, “Bruce Le Catt,” has tinkered with the counterfactual dependency theory to illuminate the possibility of securing a better match with intuition (Le Catt 1982). But these changes do not address the main problem: underweighting the importance of physical mechanisms. As Brian McLaughlin (1996) stresses, I see because the censor is an external fail-safe device that does not prevent me from exercising my capacity to see.
Let me turn the topic temporarily from nonepistemic seeing to perceptual knowledge. The eclipse riddle does not fit into the tradition of skeptical counterpossibilities. It is not like cases involving dreams, brains in vats, or fake barns (Goldman 1976). I have access to all the relevant facts. When I observe Near and Far approach and then intersect, I know they are behaving just as astronomers predicted many years ago. My situation differs from that of a passerby who had only heard of Near and just happens to look up at the moment of intersection. He might be surprised to learn that the dark shape might be caused by Far. Not me. I have been preparing the observation for years and know all the empirically relevant facts.
The only threat to my perceptual knowledge of Far is conceptual unclarity. If I fail to believe that I am seeing Far, then I do not know that I am seeing Far. Astronomers saw dark nebulae long before they had perceptual knowledge that dark nebulae are clouds of interstellar dust. They had difficulty relinquishing the wonderful possibility that the dark regions in the sky are tunnels through which we look out beyond the stars of the Milky Way into intergalactic space.
We see more than the outlines of silhouetted objects
We are tempted to say we see Near because we seem to see its outline (and so appear to satisfy Dretske's condition of visual differentiation). The influence of the outline can be gauged by considering a scenario that increases the relative size of Far. The huge shadow would make Near invisible. Those tracking the movement of Near would say that (p.59) they had lost sight of Near as it is submerged into the massive umbra of Far.
When Near and Far have their original dimensions, the trackers of Near would be able to see that Near is aligned with Far even though they cannot see Near. To see an object, the object must be causally responsible for the visual information. This requirement explains why I don't see Mary-Kate Olsen when looking at her identical twin Ashley Olsen. (As child actresses, they played Michelle Tanner in the television series Full House.) A photograph of Ashley is not a photograph of Mary-Kate even if their resemblance is exact. In the case of the double eclipse, the astronomers are seeing an object that has the same look as Near. This resemblance ensures that what they see gives them all the visual information about Near that would have been available had Near caused the image itself. Hence, the astronomers tracking Near are reliably guided by the resemblance just as a reporter tracking Mary-Kate Olsen is reliably guided by a photograph of her identical twin, Ashley Olsen.
Visual detection does not suffice for seeing. Astronomers first visually detected planets outside our solar system by observing perturbations of stars through ground telescopes. Only later, with the help of space-based telescopes, will they see a planet outside our solar system.
Is there any advantage in saying we see the fusion of near and far?
The Far answer is unaffected by the third possibility that I am seeing a composite object whose parts are Near and Far. This mereological sum, Near + Far, is arbitrary but no more arbitrary than scattered objects such as constellations. We do speak of seeing the moon even when we do not see the far side of the moon. Seeing a relevant, properly attached part of an object counts as seeing the object simpliciter (McLaughlin 1984, 580–85). “See” behaves just like most transitive verbs in this respect; to scratch a relevant attached part of an object is to scratch the object itself. Near is the front of Near + Far, and Far is (p.60) its back. If I see the front of Near + Far, then I see Near. But then the simple answer, that I see Near, would also be correct. The appeal to mereological sums would also be otiose if the simple “I see Far” answer is correct. For if I see the back of Near + Far, I see Far. The composite answer is correct only if one of the simple answers is correct.
A polarizing light filter casts no shadow. But if a second filter of opposite polarity is placed beneath, then the pair jointly cast a shadow. Someone viewing an eclipse involving two such filters might qualify as seeing their mereological sum. Both the near filter and the far filter would each contribute to what is seen. In contrast, the heavenly body Near is completely idle. The same phenomenon arises for some objects that reflect light. Glass mirrors also polarize light. Hence, glass mirrors darken when viewed through a polarizing filter even though everything else in the scene looks perfectly ordinary. The dark mirrors are visible by virtue of their contrast with their surroundings. But unlike Near, the mirrors causally contribute to the scene.
Why we see the backs of silhouetted objects
Picture Near and Far coming closer together until they fuse. I still see Far. This reinforces the point that we see the backs of backlit dark objects. When the moon eclipses the sun, only the far side of the moon blocks the light. The front half is in the shade. Thus, we see the far side of the moon during a solar eclipse.
True, little of the far side can be differentiated by purely contrastive seeing. Only the outer rim offers positive detail. The middle supplies only the negative information that there are no huge tunnels running straight through the moon. The silhouette of a net has a much higher proportion of positively detailed surface.
We distinguish between observing an effect of an object and seeing the object. The star Algol, in the constellation Perseus, has a companion star that is too dim to be seen. Since the plane on which these two stars revolve is oriented nearly edge on to our line of sight, each star is eclipsed by the other during every revolution. Although the (p.61) effect of Algol's companion can be discerned by how it periodically dims Algol's light, astronomers describe Algol's companion as invisible. Astronomers cannot even discern its outline. The object must look some way to the perceiver (Dretske 1969, 20).
The analogy with Algol's companion raises the challenge that during a solar eclipse we see an effect of the moon but not the moon. The distinction between seeing and seeing effects can also be drawn in another way that appears to yield the commonsense answer that we see the moon but not the far side of moon. “Seeing the search beam of a lighthouse may count as seeing the lighthouse and not as seeing the bulb inside the searchlight” (McLaughlin 1984, 584). The suggestion is that the moon stands to its far side as the lighthouse stands to its bulb.
However, the far side is not as obscured as the bulb. Aristotle argued that Earth must be round because it always casts a round shadow on the moon during lunar eclipses. We have doubts about whether we see Earth this way because its shadow is displaced; Earth's shadow is on the surface of the moon but Earth is not. Similar reservations affect seeing objects by mirror reflection. But during a solar eclipse, the silhouette of the moon is not displaced. The information we get from the moon's silhouette is normally “discounted” by our ample background knowledge of the moon. However, the empirical content of the silhouetted moon is salient when the observers have unusual background beliefs. Suppose we are worried that a giant meteor struck the moon during the daytime when it is invisible. Instead of waiting until night to find out, we fortunately have the opportunity to check the moon's condition by observing an afternoon eclipse: “What a relief! There the moon is, right on schedule, just where it belongs, entirely intact. The meteor must have missed!”
Seeing a surface does not require seeing much detail. Just before a rocket passes out of view, we see the rocket as a glint or speck in the sky. What chiefly matters is our ability to track its position. We do not even require movement. The arctic anthropologist Gontran de Poncins recalls the excitement of spotting a human being in the distance: (p.62)
Suddenly I saw a black dot against the white background.
“Inuuk!” (A man!) I called out.
“Na-oo?” Shongli had not seen it.
I pointed, we moved towards it, and there was no doubt about it: that black dot was a man. In this land of one million square miles inhabited by six thousand men, the sight of a human being is overwhelming; it creates an emotion that is like a seizure, difficult to describe. (1941/1988, 232)
Since the man is ice fishing and oblivious to de Poncins, he remains completely motionless. Despite the poverty of shape information, the man is seen.
The distinction between seeing an object and seeing the effect of an object is sensitive to a standard of detail. As a boy, I fed lightning bugs to my pet toad. To my delight, the bugs would continue to light up after being consumed. I could see the toad's belly light up periodically. In the dim light of a porch lamp, I was seeing only an effect of the consumed lightning bug. I let my toad hop into the darkness because I could track him by seeing the flashes of the lightning bug. The hitch was that other lightning bugs would alight near the toad. I could not see which was the lightning bug inside the toad and which was the lightning bug outside the toad. From a distance, I was seeing both lightning bugs. Thus, what formerly counted as seeing only an effect of the lightning bug (seeing its flashes through the translucent body of a toad) now qualified as seeing the lightning bug.
The detail from the far side during an eclipse can be greater than that offered by the moon's reflected light during a hazy evening. In an 1836 eclipse, Francis Baily noticed some beadlike features along one side of the moon. “Baily's beads” are the peaks of mountains that lie along the profile of the moon. If the moon revolved with respect to Baily and if the eclipse lasted long enough to complete one revolution, Baily could have surveyed all of the moon's peaks. However, the moon always presents the same side to us. Why? Because the near side bulges about half a kilometer and so is pulled more strongly by Earth's gravity. If a meteor had knocked off this bulgy excess, the moon might (p.63) have been sent spinning fast enough to make silhouette astronomy informative.
Granted, even Baily's hypothetical survey hardly compares to the rich visual information relayed by the spacecraft Luna III in 1959. But even the proud Soviet astronomers granted that, under certain conditions, we very dimly see a portion of the “dark” side of the moon by earthlight (a phenomenon popularly known as “the old moon in the new moon's arms”).
When I see the moon in normal conditions at night, I see the moon by virtue of seeing its front. Sticklers deny that I see the moon; I see only the front surface. The same sticklers would say that, on my account, we do not see the moon when it eclipses the sun; we see only the back surface. I admire the insight and attention to detail that motivates the stickler. But I am not a stickler. I say we see the moon during normal conditions at night and that we see the moon during an eclipse. But I need only insist that the stickler's standards be applied uniformly to both cases.
Suppose Near and Far have the same size. Since Far is farther away, it then only prevents light from striking the middle of Near. The stickler will say that I see a “doughnut” composed of an outer ring of Near and an inner core of Far. However, it is doubtful that I see Far in this circumstance. I cannot differentiate Far from its immediate environment. Do I see a little star if it is stationed in front of the sun? If the two are equally luminescent, I cannot differentiate the little star from its immediate environment. Suppose a wall has beige wallpaper. If someone neatly pastes a patch of leftover beige wallpaper on the wall, do I see it? Following Dretske (1969, 23), I answer no.
If I am standing with my nose against a concrete wall, then I see the wall or at least a portion of it. Yet I am not visually differentiating the wall from its immediate environment. Dretske treats this anomaly as a limiting case (1969, 26). He reasons that since there is no immediate environment, the requirement of visual differentiation does not apply.
Visual phenomenology is an important factor in first person reports of seeing—especially when the task is to break camouflage. Bird (p.64) watchers have a vivid “Aha!” experience when they make out a willow ptarmigan in the tall grass. Random-dot stereograms involve camouflage that can be broken only by stereoscopically fusing two individually meaningless images. This conjuring effect is nearly magical. Random-dot stereograms are spectacular illustrations for philosophers who think there are phenomenological constraints on what it is to see an object (Siegel 2006).
One of my astronomer friends, John Thorstensen, discovered that he could see Venus in the daytime. He had assumed that Venus was invisible during the daytime because its light is swamped by the sun's light. But in the course of aligning some equipment, he saw a bright spot that he realized must be Venus. The only visual aid needed is a straight stick to mark the exact line of sight.
The slight difference in brightness is enough to make Venus visible in the daytime. The observation conditions for seeing Near are more generous. I can clearly discern the outline of Near. True, I do not see its center because it is blocked by Far. But the block is not the sort that makes a difference to the visual match with the scene; if Far were absent, the scene would look the same. Since I am a nonstickler, I say that in that circumstance, I see Near because I see a relevant, attached part of it.
Parallel, duplicate blockers are unexpectedly common in our carpentered world of rectangular volumes. Consider a silhouetted box. Although its front side has the same dimensions as its backside, the front side is farther from the light source. Hence, the shadow generated by the backside fails to include the outermost edge of the front. Consequently, the perimeter of the silhouette is actually caused by the front side of the box. The stickler concludes we see a rectangular doughnut. But since the outline of the backside is invisible, the backside fails to satisfy the visual differentiation condition. I conclude we see the front of the silhouetted box. (More precisely, the relevant absorption layer starts from the inside of the front side.) Boxy shapes are rare in nature. Nevertheless, they constitute an im- portant exception to the rule that we see the backs of silhouetted objects.
I now summarize my main theses. Instead of adjusting principles to fit the intuition that I see Near, we ought to disown this piece of common sense as the effect of an overgeneralization. The causal theory of perception unexpectedly but correctly rules out the possibility that we see Near. Its support for Far forces us to recognize that, typically, we see the backs of dark, backlit objects, not their fronts. Thus, “occlude” must be relativized. This forces a corresponding relativization of “field of vision.” Thus, the causal theory of perception stirs an ancient species of seeing to the surface. Purely contrastive seeing is a living fossil that has everyday seeing as a descendent. Some minor adjustments are needed to accommodate this recovered form of seeing. Our conception of the physical mechanisms that support vision must be broadened to encompass the various kinds of visible dark objects. To include backlit objects, we must allow partial light blockage to count. Objects are not seen simply by virtue of a contrast with their environment. They must cause the contrast. That is the difference between Near and Far.
Intersecting eclipses can form from a fortuitous convergence of heavenly bodies. But they also form in a cyclical fashion. As the moon orbits Earth, Earth eclipses the sun. The scene would be reminiscent of the famous photograph “Earthrise” taken from the vantage point of the moon. Now take a giant step back from the moon by imagining how the scene looks to an astronaut in a high orbit around the moon. Earth is then aligned with the moon, and the astronaut has both of them in his line of sight. Near is the moon. Far is the earth.