Content-Free Pictorial Realism

Alon Chasid

Content-Free Pictorial Realism*

[penultimate version; published in Philosophical Studies 135 (2007): 375-
405]

ABSTRACT. What is it for a picture to be more realistic, or more depictive,
than another? Without committing to any thesis as to what depiction
consists in, I show that degrees of depictiveness are grounded in a certain
relation between two basic kinds of differences between pictures:
configurational differences and content differences. A picture is thus more
depictive just in case it is seen as having fewer nondepictive features,
whereas a nondepictive feature is individuated through the susceptibility
of the picture's configuration to change without entailing any change in
the picture's content.

It is commonly argued that realism in pictures has several different
characterizations, and also that there are several distinct notions of
pictorial realism.[i] In one important sense, however, a picture is said to
be more realistic than another when it looks more depictive than the other.
In this sense, a picture that depicts more successfully, a picture that
deserves more to be called "a depiction," has a higher degree of realism.
In what follows I will explain this notion of realism: I will characterize
pictorial realism in terms of the degree of a picture's "depictiveness." In
order to do that, I explore the way degrees of depictiveness are normally
determined; I try to make sense of the idea of being more or less
depictive.
Here are a few preliminaries. First, although realism is taken here to be
a matter of depictiveness degrees, I will not commit to any account of
depiction, simply because identifying realism degrees with degrees of
depictiveness does not depend on any special theoretical premise about the
nature of depiction. As the theses I will argue for are justified by fairly
trivial facts about pictures—facts that any theory of depiction can
accommodate—the following account accords, in one way or another, with any
reasonable theory of depiction.
Second, degrees of depictiveness are not determined by what pictures
depict. The suggested notion of realism is therefore content free. This
claim will become clearer later, but it basically means that it is rather
the way pictures are configured and assigned content that determines their
degree of realism. I will not deny that, for some reason, one kind of
content may sometimes contribute more than another to a picture's
depictiveness. Whatever makes a picture more depictive is, among other
things, various artistic techniques as well as decisions as to what can be
depicted more successfully; yet, content cannot form any criterion for
degrees of depictiveness.
Third, pictorial realism is prima facie a matter of comparison between
pictures: we usually think of a picture as being more or less realistic
than another. The strategy I will use is therefore to locate the specific
way we differentiate between pictures when we judge their degrees of
realism. I will start by pointing out basic ways by which pictures are told
apart, and gradually sort out the ways that are relevant to realism
judgments. There is also a deeper reason for pointing out kinds of
pictorial differences: as realism degrees will be mapped on the relative
number of "vacant" configurational features, these features need to be
individuated through a kind of pictorial difference.
Finally, it is widely held that realism judgments in pictures are
relative to a perceiver or to a group of perceivers. Indeed, perceivers in
different times and places may have different standards regarding a
successful depiction. This element of relativity will also be explained,
basically by showing the manner degrees of depictiveness are related to a
perceiver.


1. Two Kinds of Differences
There are many kinds of differences between pictures, as there are many
kinds of differences between any two objects. Pictures can be made of
different materials, have different temperatures, bear different relations
to objects, and so on. There are, however, two kinds of differences that
are particularly applicable to pictures: differences in configuration and
differences in content.
First, pictures may simply differ in visible marks on their surfaces,
namely in their design or configuration: they may have different colors,
shapes, or textures; they may differ in degrees of brightness, in thickness
of lines, in size, and so on. In specifying a configurational difference,
say, that one picture has a green spot where another has a yellow spot, we
refer to the pictures' colors. Alternatively, we may distinguish between
what pictures depict by saying that, for example, one depicts the greenness
of some object whereas another depicts yellowness. But in this case we
differentiate between the pictures' contents; we refer to what I shall call
a "content difference": this is simply a difference between what the
pictures depict. Of course, pictures may sometimes be configurationally (or
content) identical: this means that there may sometimes be no
configurational (or content) difference between pictures.


T1: Pictures can differ either in configuration or in content.


While a difference in configuration is a difference between features of
marks, a difference in content takes many forms. Pictures may first differ
in depicting different objects. One picture may depict a human face,
another may depict a forest, and a third a building. Pictures can also
differ in depicting one object as having different properties: they may
depict, for example, the same face as smiling or as not, as having a
different number of freckles, or as having different colors. Any difference
in depicted properties is thereby also a content difference (we may say,
e.g., that one picture depicts blackness whereas the other depicts
brownness). Finally, whether or not differences in appearance are
differences in properties, the fact that an object is depicted as having a
different look or as appearing differently will obviously count as a
content difference: pictures that depict different light conditions, say,
bluish versus yellowish ones, differ in content. Insofar as a
characterization is of something as it is depicted, it refers to the
picture's content, hence it is relevant to a content differentiation.
Ontological questions regarding the nature of the depicted, we may assume,
do not impinge on the bare distinction between depiction and depicted, nor
on content differentiations between pictures.
T1 does not require approaching the two kinds of differences in terms of
experience, but it is useful to do so because, as it will later be shown,
there is no ground from which to speak of pictorial realism independently
of the way a picture is experienced. What matters for our purpose is not
whether there really is a configurational or a content difference, but
whether such a difference is properly experienced.
A configurational difference between pictures is normally perceived. In
order to perceive such a difference, the marked surfaces need not be seen
as pictures at all. A picture-blind perceiver, namely a perceiver who lacks
the ability to perceive a picture as depicting something, may still be able
to see, qua a competent perceiver, configurational differences, possibly
without having the thought that these are differences between pictures. She
may notice that one surface has a curved line on its upper part whereas the
other one lacks it, or that the printed shapes are not congruent,
although—she may observe—they resemble each other. A perceiver who can see
a picture as depicting is also capable of configurationally distinguishing
between pictures independently of the differences in content that such
differences determine. It may be that a perceiver is deluded to believe
that she sees the depicted scenes only, not any surfaces whatsoever. Thus
deluded, she will not see that the pictures configurationally differ.
Whereas a configurational difference is to be perceived, a content
difference, I will assume, is to be pictorially experienced. I take the
pictorial experience (henceforth: the p-experience) of x to be the seeing
of some surface as depicting x. In normally looking at a picture, we p-
experience trees, a forest, greenness of leaves, bluish light conditions,
twilight, and so on. Assuming a certain notion of normal or appropriate p-
experience, probably by applying some standard of correctness to
pictures,[ii] we may say that a content difference amounts to a difference
between the contents of the proper p-experiences: a content difference
simply is a difference in what should be properly p-experienced. We may p-
experience a twilight in one picture and p-experience a dawn in another,
thus we may come to believe that the pictures differ in content; granted
that we had the proper p-experiences, we noticed a content difference
between the pictures. I will not say more about this issue, however, as it
is irrelevant to this discussion. By assuming that we deal with proper
experiences, I will ignore possible gaps between experiences and reality
and refer to a difference in what is p-experienced as a real difference in
content and, likewise, to a difference in what is seen on the surface as a
real difference in configuration.[iii]
Notice that I remain neutral with regard to what the p-experience amounts
to. Although the nature of the p-experience has been deeply disputed, by
being minimally defined here, the suggested notion of p-experience accords
with any reasonable theory of depiction: it is certainly compatible with
both visual theories of depiction, on the one hand, as well as with Nelson
Goodman's theory, on the other. Following Richard Wollheim, visual theories
of depiction basically identify the p-experience with a unique experience
called "seeing-in": to p-experience x is, on these theories, to see x in a
surface. Without going into various explanations that were given to seeing-
in,[iv] we can say that for all these theories a content difference is a
difference in what is seen-in the pictures. Surprisingly or not, the p-
experience plays some role, mutatis mutandis, according to Goodman's theory
of depiction as well. Goodman thinks that the pictorial relation is
basically reference fixed by arbitrary plan of correlation rules; there is
no necessary visual connection between the seer and the picture's content,
just as there is no such connection between a seer and a verbal
description's content (they can both be determined by any set of rules).[v]
There is nevertheless nothing in this claim that is incompatible with the
fact that content differences between pictures are differences between
contents of some experience. Of course, Goodman's account involves various
oddities, and Goodman is well aware of this fact.[vi] So, given his view
that the pictorial relation is identical to reference, a further, similarly
odd thesis follows: p-experiencing a face is basically of the same kind as
"verbally experiencing" a face; that is, it is like the experience (of a
face) that we have in reading a description of a face. Whether or not we
should see the latter as an experience of a face is for Goodman to explain.
Undeniably, the fact that it looks more intuitive to accept the notion of p-
experience in depictions than to accept an analogous notion of "verbal
experience" in descriptions follows from the pictorial phenomena that
Goodman wishes to ignore—and on which visual theories focus—in explaining
depiction.

2. Underdeterminacy
These two "basic" kinds of differences are not straightforwardly related to
realism. Being different in realism degrees is not just a matter of
differing in configuration or in content. A difference in realism is a more
elaborated sort of difference, a sort that follows from a certain relation
between the basic kinds of differences.
In many cases, if there is a content difference between pictures there is
also a configurational one. Monroe Beardsley thought that this is also a
matter of necessity: "If two paintings depict different things, they must
differ somehow in their design."[vii] Yet this claim is false. Looking at
configurationally identical pictures, we may still p-experience different
things: looking at the famous drawing, we can p-experience a duck or
alternatively p-experience a rabbit. We look at a single configuration and
(properly) p-experience different properties. A context might affect this
ambiguity as well: if one copy of this picture is placed in a series of
duck pictures and another copy is placed in a series of rabbit pictures,
they will probably be systematically seen as a picture of a duck and as a
picture of a rabbit respectively, although they are identical in design.
The duck-rabbit picture clearly shows that pictures can have different
contents even when they do not differ in design.[viii]
However, the fact that a configuration does not completely determine
content, like in the case of exclusive contents, is hardly relevant to this
discussion, because a degree of realism is assigned to each of the contents
separately. A somewhat curved line that depicts a snake or alternatively a
mountain can be more realistic as a depiction of a mountain than as a
depiction of a snake. A degree of realism is thus to be related to one
content in the case of exclusive contents.
Whereas having a content difference without a configurational one is of a
broad, familiar kind, the opposite case is peculiar to intentional objects
and is hardly discussed with respect to pictures.[ix] Can we p-experience
no difference by seeing differently configured pictures? Beardsley, again,
thought that "if two representational designs differ as designs, no matter
how little, they must differ in what they depict."[x] Although Beardsley's
claim is true of a unique kind of picture, it is, once again, not a general
truth. A slight difference in brushstrokes may sometimes entail a content
difference, but there are many sorts of configurational differences that
entail no content differences. I will call a difference in configuration
that entails no content difference a "purely configurational difference"
(henceforth: a PCD). The following cases are examples of PCDs between
pictures:
1. Consider a set of line drawings of a face, where the only
configurational difference between them is either in the thickness of the
lines or in the lines' color. Such drawings are obviously different in
design; the difference in their configuration is clearly observed. Yet in
many cases they depict a face as having the same properties and the same
appearance; that is, they are content identical. Sometimes differences in
thickness or color might be significant enough to change the content: we
may p-experience, for example, a nose in one drawing as grosser than it is
in the other. But cases in which there is no difference in content—even
though a clear difference in line-thickness or in color is observed—are
clearly possible.
2. A black-and-white photograph can differ in content from its sepia
counterpart in some respects: sometimes it may seem as if the depicted
light conditions are different. But in many other cases PCDs clearly occur.
In such cases we p-experience the one and the same overall content in both
the black-and-white photograph and in the sepia photograph, although they
differ in configuration.
3. A blue, striped picture of a chicken configurationally differs from a
green, dotted picture of a chicken, and they both differ from a similar
chicken-drawing that is neither colored, striped, nor dotted, but blank
("blank" means here having the same configurational features of the sheet).
4. A blown-up picture is clearly different from the original in its
configuration. When a picture is blown-up by enlarging each of its pixels
or by similar ways, it usually has the same old content: we p-experience
the same things as having the same properties and appearances, although we
see that the pictures are of different sizes. An apple depicted in a one
square-inch region may thus be p-experienced as having even the same size
as of an apple depicted in a five square-inch region. This typical blowing-
up, which entails no content difference, is possible because usually a
depicted size is independent of the size of the configuration that depicts
it. Given that a configuration's size does not depict size (nor anything
else), it can change without changing the content: it therefore produces a
PCD.
Given these examples, we can patently argue that:


T2: A configurational difference between pictures does not necessarily
entail a content difference.


These examples are relatively clear-cut because we can easily point out
the kind of PCD between them. Yet PCDs are often possible. As we shall see
later, almost every picture can have a counterpart that differs from it in
configuration only. This simply means that in most cases there are several
configurational ways to depict a given content, although we are actually
presented with only one of them. Notice that I do not refer here to the
somewhat trivial claim that there are several ways to depict an object—say,
as appearing differently, from different perspectives, and so on—because
these ways involve content differences. I argue, rather, that a different
array of marks can depict the same, overall content. Of course, certain
pictures may not have any other picture that differs from them in
configuration only; I will discuss this kind of picture later. But usually,
even when we find it hard to tell some possible PCD—perhaps just because we
are not presented with an appropriately different picture—there might be
such a difference; there might be a different, configurational, way to
depict the one and the same content.
An important corollary of the PCDs phenomenon is that p-experiencing
something does not imply that the configuration is perceived as having an
overall appearance of the depicted object. That a picture can appear unlike
its subject, however, is less crucial here, because this fact has to do
with the nature of depiction, whereas this account has to remain neutral
with regard to this issue. Instead of examining what bearings PCDs have on
the nature of depiction, I will try to show how they are relevant in
determining degrees of realism.
An initial thesis about realism is that pictures that differ in
configuration only are depictive (i.e., realistic) to the same degree. The
above examples may intuitively convince us that this is the case. Those
pictures share their level of depictiveness simply because they have the
same "loose" parts: changing the same sort of configurational feature in
them (a color, a line thickness, a texture, or a size) does not entail any
content difference.
That PCDs imply the same degree of nondepictiveness will become clearer
if we generalize the above examples. For whether or not PCDs are rare,
depictions must have a certain structural quality that allows for such
differences to occur. Finding this quality will help to explain why
pictures that can configurationally change "in the same way"—without
entailing a content difference—share a degree of nondepictiveness.

3. Vacant Features
How are PCDs possible? The answer has obviously to do with the way content
is assigned to configuration. Theoretically, there are two ways that allow
for PCDs to occur: first, when content is not assigned to a feature, and
hence this feature can change without altering the content; and second,
where features are assigned content, but in radically different ways so as
to produce different designs of the same content. I will refer to the
latter, somewhat hypothetical, possibility only later. Focusing on the
former possibility is necessary not only because it is the common one but
because it is the only one that is relevant to determining degrees of
depictiveness.
Consider, again, the bare, noncontroversial minimum that by seeing a
surface we p-experience some content. Now, it clearly does not follow that
every feature of the surface is seen as depicting some part of the content:
when we normally p-experience something, certain pictorial features might
be seen without being seen as depicting anything. Some such features are
almost never seen as depicting: flatness, being framed, or being hung on a
wall are usually perceived as not having any content. What I rather wish to
focus on is configurational features that are sometimes seen as having
content, sometimes not; that is, they are features by which we
configurationally differentiate between pictures.[xi] In some pictures,
certain lines, colors, textures, or any other configurational features
might still be seen as not depicting anything. Viewing, for instance, a
blue, striped picture of a chicken, we p-experience the chicken as having
no specific color or texture even though we see blueness and stripes; we p-
experience the chicken just as we p-experience it where the area between
the contours is blank. This latter fact is apparently sufficient to argue
that the p-experience in such cases is not constituted by the seeing of
these configurational features (although, one might rightly claim, seeing
the blueness and stripes affects one's aesthetic judgment of the picture).
Of course, this kind of case is not a matter of necessity: with some
instruction, knowledge, or perceptual habits we can p-experience a blue
chicken with stripes. Yet there are cases in which we do not. Sometimes
content is not assigned to some features of scattered arrays of lines or
dots, to the specific size of a mark, to degrees of brightness, to a
specific shape, and so on.
I will call the configurational features of a picture that are not seen
as having any content "nondepictive" or "vacant" features. Briefly, then:


T3: Some features of a picture's configuration may not be seen as
depicting something; in such a case they are vacant.


Notice that the nondepictiveness of a feature is relative to one
picture.[xii] Features that lack content in one picture can still be
assigned content in another: if the lines' thickness or color is vacant in
one picture, it can still be depictive in another.
Individuating a vacant feature is not simple. In some cases, the
nondepictive feature seems overt: a spot that, for some aesthetic reasons,
was intentionally put on a picture's surface in order to "suspend" our p-
experience may be characterized as vacant. The difficulty in telling what
feature is nondepictive begins where vacancy applies to a more specific
feature, to something that might hardly be couched in common terms. We may
look at a drawing of a face and notice that something in the way the face's
shape is depicted implies nondepictiveness. We may admit that it is by
virtue of having a shape that the picture depicts the face's shape, and,
moreover, that it is by virtue of having a rounded shape that the face's
shape is depicted. But we may also notice that the exact shape is
nondepictive. The same may occur with other, relatively specific kinds of
configurational features: although a color is seen as depicting, the
specific degree of brightness or saturation may not be depictive; and a
dotted texture may depict, say, the face's texture but without being
"fully" depictive of the face's texture, for instance, where the paint dots
are "too big" to depict beard stubble.
Those who hold that perceptual content can be nonconceptual will probably
use the experience of such fairly specific nondepictive features as
evidence. For, arguably, we can see that a fairly specific configurational
feature does not depict although we lack a concept for it. Now whether or
not experiencing vacant features requires having appropriate concepts,
these features still need to be individuated.
I will now suggest an individuation criterion that accords, in my
opinion, with our intuition regarding the nondepictiveness of a feature.
The criterion is to be drawn from Nelson Goodman's claim about pictorial
repleteness. Goodman argues that pictures, unlike words and diagrams, are
relatively replete: nothing can be ignored in them, or seen without being
thereby seen as depicting something.[xiii] Arguing that pictures are just
relatively replete, Goodman probably does not deny that vacant features are
possible. But being mainly concerned with the relatively replete structure
of pictures, he claims that (in pictures) "any thickening or thinning of
the line, its color, its contrast with the background . . . none of these
is ruled out, none can be ignored."[xiv] On his view, then, what at least
could make a feature nondepictive is its susceptibility to be ruled out
without altering the content. The thickness of the lines does not depict
anything, insofar as thickening or thinning does not matter: this is
precisely what makes thickness vacant in, for example, a diagram.
Pictures, as we saw, can differ in configuration only, and they are
therefore replete only to some degree. Yet Goodman's criterion still
applies: that a configurational feature has no content is simply captured
by its susceptibility to be replaced content-independently (or CI-ly).
Clearly, this susceptibility is common to the blueness in the chicken
picture, to the lines-thickness in line drawings, to size in appropriate
cases, and also to the more specific nondepictive features—features that
can hardly be couched in common terms. Their vacancy is necessarily
connected to the fact that they can be altered without affecting what is
already depicted. More precisely:


T4: A configurational feature F of a picture p1 is nondepictive just in
case there can be a picture p2 that does not differ from p1 in content, yet
differs from it in not having F; when this is the case, p1 and p2 differ in-
configuration-only with respect to F.


Because we deal with contingent pictorial (configurational) features, or
features by which we configurationally differentiate between pictures, then
not having F amounts to having a different configurational feature G that
is likewise vacant in p2. Sharing their overall content and differing with
respect to F (or G), p1 and p2 differ only in configuration.
The nondepictiveness of a feature is thus related to having PCDs. Granted
that pictures can have nondepictive features, these features are to be
individuated through their being susceptible to change CI-ly, whereas these
changes result in nothing but various configurationally different pictures
that share their content. Of course, a nondepictive feature F might be
replaced CI-ly not only by one but by several configurational features: for
example, the blueness in the chicken picture can be replaced CI-ly by
redness, greenness, or yellowness. In this way the vacancy of F gives rise
to an array of PCDs, and PCDs govern the vacancy of these features in the
various content-identical pictures.
The suggested criterion sharpens the manner by which vacancy is ascribed
to features. Recall the case of the alleged nondepictive spot on a picture.
We may think that having a spot at that location is nondepictive. Yet this
is not true, for having a spot, at least in some pictures, cannot change to
not having it without entailing a content difference; moreover, it might be
argued that all pictures have spots, so there is no way of not having a
spot and still be a picture. It therefore seems that what is nondepictive
in such cases is a more specific feature; that is, having a certain color
at that location. Having a specific color is clearly nondepictive according
to the suggested criterion, as it can be replaced CI-ly by other colors.
Notice finally that a nondepictive feature F can also be replaced by some
non-F to the effect that there is a change in content. The nondepictive
blueness in the chicken picture can obviously be replaced by a black color
and thereby to change something in what the picture depicts (the chicken
will no longer be depicted as having no color but rather as being black).
In such a case we may say that the vacancy of the feature is "filled out"
in the new, appropriately blackened, picture. Of course, that some
replacements of an F by non-Fs have bearings on content does not entail
that F is depictive, nor that it is nondepictive. The suggested criterion
requires that a nondepictive F can be replaced CI-ly by some Gs, not by
every G.
As T4 shows how depictiveness and PCDs are necessarily related, the
initial thesis about realism is clear. Pictures that differ only in
configuration correspond in their level of nondepictiveness: they all have
the same sort of nondepictive feature. In each one of them, either the
specific degree of line thickness, the color of the lines, the color
between the contours, or the size are nondepictive; such pictures can all
change CI-ly to not having their specific degree of line thickness,
specific color, and so on. The same level of nondepictiveness is thus
guaranteed through their being different only in configuration in one
certain way. For now, a PCD between pictures simply necessitates that they
are "infected" with the same sort of nondepictiveness.
What happens, then, when differences in configuration do entail content
differences? Arguably, realism degrees are to be determined similarly; yet
how? What makes it true that a line drawing of a face depicts less
realistically than, say, a color painting of a mountainous view?
What seems to be central, and should therefore be extracted, is that a
degree of depictiveness is determined by the number of ways in which—or
likewise, the number of features with respect to which—a picture can change
CI-ly. A difference in realism degrees is thus to be determined by the
number of vacant features a picture has relative to another. What makes a
picture more realistic is that it has fewer ways of having other pictures
differ from it only in configuration. A line drawing would be less
realistic than a color painting simply because it has more features that
are susceptible to be replaced CI-ly (i.e., that are nondepictive) than the
painting has. Now, when pictures differ only in configuration, they have
the same number of vacant features because they can change CI-ly in exactly
the same ways (let alone in the same number of ways). But it can likewise
be recognized that pictures with different contents have more or less the
same number of vacant features and are thereby depictive to the same
degree; or, alternatively, that one picture has more such features, and
hence is less realistic, than another.
Mapping degrees of realism on the number of vacant features is yet
problematic, as this actual number does not seem to be recognized in any
normal realism judgment. Of course, we do not normally count these
features; we merely estimate which picture has a bigger number of vacant
features and judge it as less realistic. What is problematic even in this
rough estimation is the various ways vacant features appear. One vacant
feature, such as a rounded shape, is actually composed of various spots at
different locations on the surface, each of them with a certain hue,
saturation, and brightness degrees; these features are vacant as well.
Likewise, different features—for example, color and shape—can be combined
to form a further, composed feature that is susceptible to change CI-ly,
and hence is vacant too (given this option, the entire pictorial
configuration is a vacant feature, as it can also change properly). The
number of vacant features is perhaps not infinite: recall that we
approached depiction in terms of experience, so only features that are
properly seen are considered. But this would barely help to solve the
problem, as that number might be big enough not to be taken into account in
any normal judgment of realism. No one considers every seen dot on a vacant
rounded shape and every combination of features as vacant too. We usually
consider a shape or a whole certain region, rather than every part or
combination, as one feature. So if normal realism judgments do not include
all vacant features, how is nondepictiveness estimated after all?
I argue that vacant features, or the various ways a picture can change CI-
ly, are perceptually individuated more roughly than they really are. We
indeed do not take into consideration every vacant feature: by obeying
perceptual regularities as well as habits with respect to the p-experience,
we single out certain vacant features and estimate only their relative
number. Habits, of course, change in different times and places, but so do
realism judgments. Clearly, this account would not be adequate if it did
not explain the relative nature of realism judgments.

4. A Note on Pictorial Systems
Before showing how we perceptually individuate vacant features, it is
necessary to consider a more radical sort of difference between pictures;
that is, what is sometimes called a difference between pictorial "systems."
Such a difference seems to jeopardize T4's adequacy, for theoretically,
pictures can differ only in configuration—not only when their configuration
lacks content, but also when they are assigned content in radically
different ways. With hardly any constraint on the nature of depiction, it
might be that just as words can share their meaning but differ in their
configuration—that is, when they belong to different languages—so, too, can
pictures share their content but differ in design by falling under
different "mappings," where no vacancy is involved.
Unsurprisingly, Nelson Goodman allows for such PCDs to occur. A symbol
system for Goodman consists in a symbol scheme correlated with a field of
reference, where plans of correlations may limitlessly differ between
systems.[xv] That the notion of pictorial system becomes noteworthy for
Goodman is because he does not have any restriction as to the visual in
pictures. With no such restriction, there are many ways to depict
something; there can be as many pictorial systems as there can be natural
languages. If this is true, then for every picture we may get a content-
identical picture with a different surface: we can stipulate a different
plan of correlation to the effect that a different surface depicts the same
content. Such different mappings make PCDs possible whether or not the
pictures have vacant features.
Goodman's concept of pictorial systems is a hypothetical one, and it is
doubtful whether there can in fact be pictures that differ only in
configuration just through having their content assigned to them by
different correlation rules. Goodman does mention a case to this effect in
his discussion on realism: a realistic picture versus a picture in which
the perspective is reversed and each color is replaced by its
complementary.[xvi] He argues that this case proves that realism is not a
matter of information, as both pictures, when appropriately interpreted,
yield exactly the same information, although the first is realistic and the
"complementary" one is not. Waiving the issue of realism for the meantime,
if the complementary picture can really be seen as having the same content
as the original picture, we get PCDs independently of having nondepictive
features. Yet can such pictures really depict the same content? Is it
possible to p-experience the same content in, for example, a photograph and
in its negative? I would say that we can p-experience the same objects in
both, but hardly the same colors and overall appearance. Pictures that
differ even slightly in mapping will hardly be seen as sharing the same
content (a negative is clearly infected with nondepictive features that the
photograph does not have).
For generality's sake, however, I will regard having PCDs through
different plans of correlation (thereby independently of vacancy) as
possible. Also, I admit that without an elaborated account of the p-
experience, I have no argument against the feasibility of such differences
according to visual, non-Goodmanian, theories of depiction as well: on a
reasonable account of seeing-in it might perhaps be true that perceptual
habits change to the effect that the same content is seen-in visually
different configurations where no vacancy exists (there is certainly no a
priori restriction preventing us from p-experiencing the same content in a
photograph and its negative). But insofar as these differences are
possible, T4 should be taken as not referring to them. T4 refers thus only
to PCDs that do not result from such differences in content assignment: it
refers only to differences in systems that do not differ in correlation
rules (or, according to visual theories of depiction, to differences
relative to a given perceptual apparatus). The price for employing
hypothetical notions such as unrestricted pictorial systems (or, perhaps,
abnormal perceptual skills) is therefore spared, as T4 deals only with PCDs
within one Goodmanian system.

5. Pictorial Realism
My main idea is that what makes a picture more realistic is that it has
fewer vacant features. A picture is less depictive simply when it has more
ways of having other pictures different from it CI-ly. A line drawing that,
in order to depict, uses only contours is less depictive than a drawing
that unlike the former assigns content also to shades between contours; a
color picture is usually more realistic than a line drawing because unlike
the drawing its various hues are depictive; and so on. This notion of
realism is thus not a matter of the content's nature, at least not
primarily; it is only a matter of the extent to which a picture's
configuration depicts or is assigned content, where the degree of such an
assignment is governed by the various susceptibilities to change CI-ly.
It is simple to start with T5:


T5: A picture is realistic to the highest degree just in case any change
in its configuration entails a change in its content.


A supremely realistic picture necessarily differs from any differently
configured picture in content; there cannot be a picture from which it
differs only in configuration. Indeed, what Beardsley had in mind when he
claimed that pictures that differ in design must differ in what they depict
is the notion of a supremely realistic picture. His mistake is therefore a
wise one, for a supremely realistic picture is nothing but an "ideal
depiction" or a "completely depictive" picture: it is a picture that does
not fall under any PCD, and therefore thoroughly lacks nondepictive
features. Granted that T4 is true, all of the configurational features of
such a picture are assigned content. Beardsley's claim can thus be
interpreted as referring to a fully depictive picture, or perhaps to the
depictive part of every picture. For whereas it is false that any
configurational difference entails a content difference, it is true that
insofar as pictures are depictive or lack vacant features, such an
entailment holds. T6 thus follows from T4 and T5:


T6: A picture is realistic to the highest degree just in case it has no
nondepictive features.


What is left to explain is how degrees of realism are mapped on the
relative number of nondepictive features. As above, many vacant features
are not counted at all in any normal realism judgment. One usually thinks
of a shape as vacant—although usually not of every seen part of it, or of
every combination of it with other features, as vacant (although these
features are vacant). Also, being blank (having the color of the sheet)
might sometimes be perceived as several vacant features; for example, when
the blankness appears on different locations and thereby depicts different
properties. Vacant features, we must conclude, are perceptually
individuated in a fairly coarse way relative to the way they occur,
especially when judging realism degrees is at stake. We thus get a further
explanation: that perceptual individuation varies among different
perceivers explains the relative nature of realism judgments.
There are two competing factors involved in perceptually individuating
vacant features. These two factors follow from the two kinds of experiences
in which the notion of realism is anchored: seeing (the configuration) and
p-experiencing (the content). The first factor is our perceptual habits as
they apply to configurations. Some Gestalt grouping principles exemplify
such habits, as they suggest that we perceive configurations as organized
into "groups" or "wholes." Because estimating the degree to which a picture
depicts requires first of all the perception of the pictures'
configuration—it should not come as a surprise that the way we perceive
configurations takes a part in this estimation. If one's perceptual habits
prevent her from grouping a certain dotted texture as just one
configurational feature, she may also fail to perceive it as one vacant
feature. And clearly, if someone perceives a circle as different small
parts combined together while perceiving regions between contours as just
one configurational feature, she may conclude that a picture with a vacant
circle is less realistic than a picture in which the color between the
contours is vacant, simply because she perceives the former as having many
more vacant features than the latter.
The second factor involves habits in experiencing the picture's content.
As T4 argues, a vacant feature is individuated through the way it can be
replaced by another feature CI-ly. But sometimes different parts of a
feature involve fairly different arrays of content-independent changes.
This fact, I claim, causes us to individuate it as more than one vacant
feature. Consider a drawing of a mountainous view where the region between
the contours is blank. The color of the sheet in this case is regarded as
one vacant feature: we normally perceive it as one "whole" configurational
feature, and it can clearly be replaced CI-ly by a different color. Yet
being familiar with the nature of a mountainous view, we realize that there
are necessarily different colors that can "fill out" this feature's
vacancy: the nondepictive color in the parts that depict a bush is to be
replaced CI-ly (or, likewise, to be filled out) by different colors. In
such a case, although one configurational feature is noticed, it
contributes to the picture's vacancy in different ways, according to our
habits with regard to what we p-experience, and hence it is counted as more
than one vacant feature.
It turns out that perceiving the number of nondepictive features, at
least in case of realism, is guided not just by habits in perceiving the
picture, but also by habits in experiencing the content. When there is no
uniformity in the way a picture can change CI-ly with respect to a feature
F, then F may be regarded as a few vacant features. To sum up, then:
although a picture can have a great number of vacant features, we
perceptually individuate them roughly. As the idea of realism is grounded
in the above two kinds of pictorial differences, there are likewise two
factors that affect perceptual individuation. One has to do with the way we
perceive a configured surface, and the other with the variety of ways a
feature seems to change CI-ly.
The two sorts of habits are far from creating a stable criterion, as they
vary with any difference either in perceptual grouping principles or in
regularities in p-experiencing the content. Moreover, these habits may
compete and thereby create ambiguity with regard to the number of vacant
features a picture has. Whereas having a certain texture is easily
perceived as one vacant feature, it might raise a problem when judging
which of two pictures with this vacant texture is more realistic, as
estimating the relative number of ways this texture can be replaced CI-ly
in the two pictures would be far from precise.
Realism judgments, however, apply in the same "sketchy" ways, just like
the instability of this criterion suggests. Comparing between degrees of
realism is ambiguous also when perceptually individuated vacant features
take radically different forms. It may be difficult to determine whether a
black-and-white picture of the Pacific Ocean is more infected with vacant
features, and hence is less depictive than a detailed drawing of a human
face. We certainly do not enumerate vacant features to form a fine scale,
but this is equally the case with degrees of realism. Only when a
difference in quantity of vacant features is fairly recognizable—for
example, when one picture's perceived vacant features are included within
the other's—is a difference in degree of realism achieved.


T7: A picture is more realistic than another just in case it has fewer
perceptually distinguished vacant features.


6. Related Characteristics and Accounts
As Kendall Walton maintains, "realism is a monster with many heads
desperately in need of disentangling."[xvii] Nelson Goodman similarly
argues that "no one account of [realism] will do."[xviii] These claims
suggest that the problem with analyzing pictorial realism is twofold.
First, there might be several different notions of pictorial realism; that
is probably what Goodman means when he says that more than one account is
required.[xix] Second, even when one notion of realism is at stake, realism
can be characterized differently, as it has various contributing factors or
"heads." As Walton describes it, explicating one veridical characteristic
amounts then only to "disentangling one head of the monster."
In view of that, the suggested account has several merits. First, because
it defines realism without committing to any thesis with regard to the
nature of depiction, it is compatible with any theory that does commit to
such thesis and explains realism accordingly. As the nature of depiction
was characterized in various different ways, many familiar characteristics
of realism fit in with what has been argued here.
Some characteristics of depiction are mentioned by visual theories that,
unlike Goodman's theory, do not take them as epiphenomenal or theoretically
worthless. Consider the tenet that some kind of resemblance holds between a
depiction and its content. If the p-experience consists in, or is rather
structured to entail, an experience of resemblance between picture and
content, then the more we see various features as depicting, the higher the
degree of pictorial resemblance. Seeing fewer features as nondepictive—that
is, as not resembling some part of the subject—we experience a "more
depictive" picture. We thereby arrive at the familiar idea that degrees of
realism accord with the extent to which this (alleged) kind of resemblance
holds between picture and content.
Another famous characteristic of realism may follow from the view that
depiction involves some kind of illusion. Pictures, especially highly
realistic ones (such as the legendary Zeuxis's and Parrhasius's ones), are
said to possess a kind of illusive force. Although we are usually not
deceived by realistic pictures, supremely realistic ones are likely to
mislead us by making us take them for their subject. This common claim
follows from the present account: for given that the p-experience involves
some kind of illusion, the greater the number of configurational features
that are depictive, the greater the illusive force of a picture. A
maximally realistic picture, one that is completely depictive, is thus the
perfect one for the alleged pictorial illusion. The ability of pictures to
mislead simply reaches its highest degree when any change in configuration
entails a change in content.
Some visual theories argue that pictures, such as visual aid devices
(e.g., mirrors, telescopes, and binoculars), bear some kind of transparency
(this claim probably follows from identifying the p-experience with
ordinary perception).[xx] Once again, whether this is a literal truth or
just a layman's metaphor, it clearly explains the belief that the more a
picture is realistic, the more it reflects its subject. For the more the
picture's depictive features, the more the picture's subject is "reflected"
in it.
That these theses about depiction explain common characteristics does not
mean that the phenomenon of pictorial realism justifies these theses and
contradicts other accounts of depiction (such as Goodman's). At the end of
the day, realism is identified with the absence of nondepictive features,
and as such it can be interpreted on every theory and on Goodman's theory
in particular. Lacking vacancy thus functions as the only constraint on
Goodman's and on other theories regarding the way they explain pictorial
realism. This shows that visual theories of depiction do not gain support
from the phenomenon of realism, for the p-experience hardly inherits
anything from ordinary visual experience insofar as realism judgments are
concerned. Visual theories of depiction should therefore rely on phenomena
other than pictorial realism in order to justify a robust kinship between
the p-experience and perception.
A second merit of this account is that it is intimately related to other
notions of realism and possibly has contributed to their formation. Being
independent of the nature of a picture's content, realism as depictiveness
is to be contrasted with content-based realism, yet content-based realism
strangely seems to relate to realism as depictiveness.
Consider the idea that for a picture to be more realistic is for it to
provide a greater amount of information. Whether this criterion can be
applied with no further premise is doubtful, because an amount of
information depends on many parameters. It is, however, possible to measure
information relative to a given criterion, and in the case of realism, the
number of vacant features may be used in just such a way: depicted
information contributes to realism insofar as it is correlated with the
absence of vacant features; for when a vacant feature in one picture is
"filled out" in another, some information is thereby added to content; but
through having a vacant feature filled out, the revised picture simply
becomes more realistic than the other. A further piece of information
correlates in this way with a higher degree of realism.
Consider also the idea that pictures are more realistic when they depict
something real or familiar rather than unreal or fantastic. This content-
based notion is related as well to our notion. For depicting a familiar
object rather than an unfamiliar one tends to leave us with fewer vacant
features, at least in some cases. A blue picture of a chicken has a vacant
feature—its blueness—simply because (given that, in "our system," this
blueness should depict blueness) chickens are not blue. For some reason,
familiarity with a picture's content is often correlated with the picture's
degree of depictiveness. Impressionists, for instance, were blamed for
depicting unsuccessfully because "trees aren't really purple" and "the sky
isn't really the color of butter."[xxi] Impressionistic paintings were thus
considered, for certain perceivers, as less depictive—due to the idea these
perceivers had with respect to the paintings' content.
Recall also how vacant features are perceptually individuated: if the
ways to "fill out" what looks as one vacant feature diverse, this feature
is regarded as several ones. When the depicted object is unreal, the number
of ways a feature might be filled out tends to be fairly high. With no idea
as to what is depicted or what appearance the depicted object has, we
easily come to see features at various locations as nondepictive. Clearly,
the intimate relation between the correctness, actuality, or
familiarity—and the nature of the depictive features—is to be accounted by
a theory of depiction, but that such a relation exists can hardly be
denied. In general, then, our tendency to p-experience things "better" when
they are depicted as we know them may induce artists who see realism as
their goal to depict the actual, the real, or any object as familiar,
thereby leaving the observer with fewer vacant features or with less
ambiguity with regard to their existence.
Kinds of contents, however, do not systematically contribute to reducing
the number of vacant features: a picture of an unfamiliar object can be
highly realistic, and familiar objects can be depicted in a nonrealistic
manner. Content-based realism differs from realism as depictiveness, but it
might be that by being central to pictures, realism as depictiveness
contributed to the formation of content-based notions.[xxii]
"Realism" has other uses; some are content-based and some are not. Of
course, not every use is related to realism as depictiveness, but when it
is not, it seems to be a fairly rare use. Goodman's notion of realism is a
good example to this effect. For Goodman, degrees of realism are determined
by our familiarity with the correlation rules (this familiarity should not
be confused with familiarity with content). Goodman contends that realism
is primarily a property of a pictorial system, not of a particular picture;
and the fact that a system's pictures are realistic is relative to a given
culture or person at a given time, as it depends on how standard the system
is.[xxiii]
Thus, unlike the suggested notion, Goodman's notion of realism applies to
what Dominic Lopes rightly calls "inter-systemic realism."[xxiv] On this
view, if pictures differ in degree of realism, they necessarily belong to
different systems. Intrasystemic realism is thereby ignored by Goodman. His
criterion for comparing degrees of realism does not refer to pictures
within one plan of correlation.
That entrenchment really determines a higher degree of realism is
doubtful.[xxv] And, as above, Goodman's idea of unrestricted systems is
also problematic, at least for the fact that this idea is fairly
hypothetical and hardly fits the way pictures appear to be. Goodman
therefore explains an abstract, dubious thesis by another dubious thesis,
and thereby leaves the suggested idea intact, for if there is no
restriction on pictorial correlation rules, Goodman must face the variety
of pictorial systems that his account allows so as to assign degrees of
realism to pictures in different such systems. He thus "narrows" this
variety by the entrenchment constraint: insofar as being more or less
realistic applies to pictures qua their belonging to different systems, the
more the system is entrenched, the more the picture is realistic. Yet
Goodman says nothing either on intrasystemic realism or on intersystemic
realism insofar as systems are equally entrenched. We may argue then that
pictures in such systems are realistic in a different sense, that is, when
they have fewer nondepictive features.
It is now clear why the suggested notion is more commonly used than
Goodman's notion. Recall the case of a realistic picture versus another
with the complementary colors and reversed perspective. Goodman argues that
both pictures, when appropriately interpreted, yield exactly the same
information, yet the first is realistic, whereas the second is not. Let us
refer to the similar case of a photograph versus its negative. Goodman's
explanation will be as follows: a negative is less realistic because in
order to have the same content as the photograph, a fairly irregular set of
rules should apply to the negative. Insofar as it is seen as a depiction,
the negative has to be interpreted in an uncommon way, hence its low degree
of realism.
However, this is clearly not what we mean when we say on a negative that
it is not realistic, for normally we do not interpret it by a different set
of rules (so that it will have the same content as the photograph) and then
judge its degree according to how well these rules are familiar. We
interpret the negative within the same rules that apply to the photograph.
We perceive a difference in realism because our pictorial-experiential
habits equally apply to looking at the photograph and to looking at its
negative, so that they do not yield the same information. We then conclude
that the negative is less realistic simply because we perceive it as having
more nondepictive features. Thus Goodman's notion of realism is perhaps
applicable, but just like his notion of pictorial systems, it is not the
notion we use.


7. Conclusion
A picture is completely depictive, or supremely realistic, when all of its
configurational features are depictive. Likewise, the greater the number of
depictive features of a picture, the higher its degree of depictiveness.
Yet how does such a criterion apply?
I argued that a degree of depictiveness is grounded in two kinds of
differences between pictures. A nondepictive feature is individuated by its
susceptibility to be ruled out without entailing a content difference, and
degrees of depictiveness are measured by the number of nondepictive
features. As normal realism judgments do not include the actual number of
vacant features, I pointed out the ways vacant features are perceptually
individuated.
The suggested notion of realism is content free. It is not determined by
preferences with regard to what a picture depicts. We nevertheless saw that
such preferences (among others) may contribute to the disposal of
nondepictive features. Realism as depictiveness is related, after all, to
content-based realism, as well as to other notions of realism.
Keeping with the argument's neutrality, I did not try to explain how
content affects the number of vacant features. Being independent of any
theoretical premise as to what depiction consists in, this account is
general enough not to explain this or any other puzzling characteristic of
pictures. Clearly, this gaping hole still needs to be filled out, but
rather by a complete theory of depiction.


The University of Michigan, Ann Arbor




References
Beardsley M. (1958): Aesthetics: Problems in the Philosophy of Criticism,
New York: Harcourt, Brace & World
Budd M. (1993): "How Pictures Look," in Knowles D. & Skorupski J. (eds.),
Virtue and Taste, Oxford: Basil Blackwell, pp. 154–175
Chasid A. (2004): "Why the Pictorial Relation is not Reference," British
Journal of Aesthetics 44: 226–247
Goodman N. (1976): Languages of Art, Indianapolis: Hackett
Goodman N. (1984): Of Mind and Other Matters, Cambridge MA: Harvard UP
Hopkins R. (1998): Picture, Image and Experience, Cambridge: Cambridge UP
Hyman J. (1997): "Words and Pictures", in Preston J. (ed.), Thought and
Language, Cambridge: Cambridge UP, pp.51–76
Hyman J. (2004): "Realism and Relativism in the Theory of Art,"
Proceedings of the Aristotelian Society 105: 25–53
Hyman, J. (2006): The Objective Eye: Color, Form and Reality in the
Theory of Art, Chicago: U of Chicago Press
Lopes D. (1995): "Pictorial Realism," Journal of Aesthetics and Art
Criticism 53: 277–285
Peacocke C. (1987): "Depiction," Philosophical Review 96: 383–410
Sartwell C. (1994): "What Pictorial Realism Is," British Journal of
Aesthetics 34: 2–12
Sèrullaz M. (1978): Phaidon Encyclopedia of Impressionism, Oxford:
Phaidon
Walton K. L. (1984): "Transparent Pictures: On the Nature of Photographic
Realism," Critical Inquiry 11: 246–277
Walton K. L. (1990): Mimesis as Make-Believe, Cambridge MA: Harvard UP
Walton K. L. (1992): "Seeing-In and Seeing Fictionally," in Hopkins J.
and Savile A. (eds.), Psychoanalysis, Mind and Art, Oxford: Blackwell, pp.
281–291
Wittgenstein L. (1958): Philosophical Investigations, Oxford: Blackwell
Wittgenstein L. (1977): Remarks on Color, Berekley & Los Angeles: U. of
California Press
Wollheim R. (1998): "On Pictorial Representation," Journal of Aesthetics
and Art Criticism 56: 217–226
Zemach E. (1999): "Look, This is Zeus!" in Krausz M. & Shusterman R.
(eds.), Interpretation, Relativism and the Metaphysics of Culture, New
York: Humanity Books, pp. 311–333
-----------------------

Notes

* I am, above all, deeply grateful to Ken Walton, whose kind advice and
comments were extremely valuable and mind-enriching. I also thank Tirtsa
Harif, John Hyman, Bob Stecker and Eddy Zemach for their helpful comments.
This paper was written during a period of postdoctoral research at the
Department of Philosophy, University of Michigan, Ann Arbor. I wish to
thank the faculty members, students and staff of the department. I am also
grateful to the Hebrew University of Jerusalem that provided me with a
postdoctoral grant and gave me the opportunity to spend this productive
period in Ann Arbor.


[i] See Walton (1990), p. 328 ff.; Goodman (1984), pp. 126–130; Hyman
(2004); Sartwell (1994); Lopes (1995).
[ii] See, e.g., Hopkins (1998) pp. 71-77.
[iii] The argument here is indifferent with regard to the notion of proper
p-experience. I could alternatively define pictorial realism relative to
one experience, whether or not it is a proper one. This strategy is
employed by Goodman (1976), Ch. 3.
[iv] Wollheim (1998) discusses some changes in his view. Others explain
seeing-in in various different ways. See, e.g., Peacocke (1987) and ibid.
note 24; Hopkins (1998); Budd (1993); Walton (1992).
[v] Goodman's well-known theory is presented mainly in Goodman (1976).
[vi] See ibid. p. xii.
[vii] Beardsley (1958), p. 295
[viii] The case of content difference with no configurational one is known
in philosophy mainly through Wittgenstein's discussion in Wittgenstein
(1958), part II, section xi. Wittgenstein uses pictorial examples to show
that we may see that nothing has been changed in an object yet see it
differently. Although depictions provide suitable cases for that
experiential shift, there may be other sorts of cases of such a shift, even
when the seen object is neither a picture nor an intentional object
whatsoever. The phenomenon of noticing an aspect simply goes beyond
pictures. Of course, a theory of depiction should take into consideration
that pictures, among other things, exemplify this phenomenon. Such a theory
should explain how the p-experience reflects the noticing of different
aspects of a surface and how it can thereby include exclusive contents.
[ix] John Hyman mentions such differences: see Hyman (1997), p. 73; cf.
Hyman (2006), Ch. 8. I will later use some of Hyman's examples.
[x] Beardsley (1958), p. 295
[xi] I thereby refer here only to configurational features that a picture
might not have and still be a picture: having a color is perhaps a
configurational feature, but because a picture cannot have no color—i.e.,
we do not configurationally differentiate between pictures by means of this
determinable—then having a color does not count here.
[xii] This notion of vacancy is completely different from Goodman's notion
of vacant inscription (1976, p.145).
[xiii] Repleteness simply means that more features "matter." Diagrams,
e.g., are less replete than pictures because they refer independently of
having a specific line thickness or color (ibid. pp. 229–230).
[xiv] Ibid. p. 229.
[xv] See Goodman (1976). For a summary of Goodman's central thesis as well
as various criticisms see Chasid (2004). Notice that Goodman's idea of
unlimited differences in pictorial mapping carries serious flaws, and his
thesis may even be inconsistent with basic pictorial phenomena. See ibid.
[xvi] Goodman (1976), p. 35.
[xvii] Walton (1990), p. 328.
[xviii] Goodman (1984), p. 130.
[xix] Goodman's claim reflects a shift in his own view. Cf. Goodman (1976),
pp. 36–38) and (1984), p. 126 ff.
[xx] See Walton's theory on perceiving photographs (1984) and Zemach
(1999).
[xxi] Sèrullaz (1978), p. 20.
[xxii] Besides familiarity, there are other content features that tend to
reduce the number of vacant features. Walton's idea of realism as vivacity
refers to one of them. See Walton (1990) p. 296 ff.
[xxiii] Goodman 1976, p. 38.
[xxiv] Lopes (1995), p. 279.
[xxv] It simply seems unlikely that familiarity with the rules enhances an
intersystemic realism degree. This is possibly the reason behind Lopes's
attempt to elaborate on Goodman's idea (see ibid.)