University of Arkansas Press

Representationalism and Perceptual OrganizationAuthor(s): E. J. Green

Source: Philosophical Topics , Vol. 44, No. 2, New Directions in the Philosophy of
Perception (FALL 2016), pp. 121-148
Published by: University of Arkansas Press

Stable URL: [Link]

REFERENCES
Linked references are available on JSTOR for this article:
[Link]
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@[Link].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
[Link]

University of Arkansas Press is collaborating with JSTOR to digitize, preserve and extend
access to Philosophical Topics

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
philosophical topics
vol. 44, no. 2, Fall 2016

Representationalism and Perceptual

Organization

E. J. Green
New York University

ABSTRACT. Some philosophers have suggested that certain shifts in per-

ceptual organization are counterexamples to representationalism about
phenomenal character. Representationalism about phenomenal character
is, roughly, the view that there can be no difference in the phenomenal char-
acter of experience without a difference in the representational content of
experience. In this paper, I examine three of these alleged counterexamples:
the dot array (Peacocke 1983), the intersecting lines (Speaks 2010), and the
3 X 3 grid (Nickel 2007). I identify the two features of their phenomenol-
ogy that call for explanation: grouping and prominence. I then argue that
representationalists can adequately account for both of these features. I also
critique some previous treatments of grouping and prominence.

1. Introduction

Perceptual organization is present in visual phenomenology. For example, the

image in figure 1 can appear myriad different ways, depending on how you group
its constituent elements. Everyday cases illustrate the same point. Certain clusters
of items such as flocks of geese, trails of ants, or schools of fish seem to constitute
“visual units.” This paper concerns the phenomenology of perceptual organization.

121

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
More specifically, it is about how representationalists about phenomenal character
should account for this phenomenology.

Figure 1. Reproduced from Feldman (2003), with permission from Elsevier.

I’ll construe representationalism about phenomenal character as, roughly, the

view that there can be no difference in the phenomenal character of experience
without a corresponding difference in the representational content of experience.
In other words, if two experiences differ in what they are like for the subject, then
they also differ in how they represent the world as being. Several philosophers
have argued that certain cases of perceptual organization experience pose chal-
lenges to representationalism (see Peacocke 1983; Macpherson 2006; Nickel 2007;
Speaks 2010). In what follows, I’ll consider and try to meet these challenges. I’ll
argue that shifts of perceptual organization are plausibly accompanied by changes
in representational content.1
The plan is as follows. In section 2, I introduce three cases of perceptual organi-
zation that have been held to raise problems for representationalism: the dot array
(Peacocke 1983), the intersecting lines (Speaks 2010), and the 3 X 3 grid (Nickel
2007). I’ll identify two types of phenomenal shifts involved in these examples:
shifts of grouping and shifts of prominence. In section 3, I’ll propose an account of
grouping shifts, which I call the cue-­based composition account, and I’ll apply this
account to the dot array. In section 4, I’ll propose an account of prominence shifts,
and apply this (along with the cue-­based composition account) to the intersecting
lines and 3 X 3 grid. I’ll also critique some previous treatments of these examples.

1. I won’t consider all of the perceptual-­organization-­based challenges to representationalism.

Rather, I’ll only discuss those that I believe have not yet received adequate responses. Macpherson
(2006), for example, argues that ambiguous figures like Mach’s square/diamond pose a serious
difficulty for representationalists. I won’t consider this argument here, but for responses to it,
see Jagnow (2011), Orlandi (2011), and Raftopoulos (2011). I am inclined to endorse some ver-
sion of Jagnow’s or Raftopoulos’s account, on which the shifts in phenomenology associated with
Macpherson’s figures are accompanied by changes in perceptual reference frame.

122

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 2. Representationalism and Counterexamples

2.1. Representationalism

I’ll assume for the purposes of this paper that visual experiences have content. By
this, I mean that visual experiences represent the world as being some way or other.
As a result, your current visual experience can be either accurate or i­ naccurate
depending on whether the world is the way the experience represents it as being.
The key view I’ll be concerned with is the following, which we can call the rep-
resentationalist supervenience thesis (RST): Necessarily, for any pair of experiences
E and E*, if E and E* differ in their phenomenal character, then E and E* differ in
their representational content. In other words, differences in phenomenal charac-
ter are necessarily accompanied by differences in content.
RST is endorsed by a number of theorists (e.g., Byrne 2001; Dretske 1995;
Harman 1990; Tye 1995, 2002). The thesis is entailed by any view on which phe-
nomenal properties are identical with, constituted by, or wholly determined by
representational properties.2 The claims about perceptual organization that I’ll
develop below are consistent with any of these particular positions.
Note, however, that some philosophers who self-­identify as representational-
ists would not endorse RST. For example, Lycan (1998) holds that functional prop-
erties, in addition to representational properties, play a role in determining overall
phenomenal character. On this view, two experiences might have the same content
but nevertheless differ in their phenomenal character by virtue of having different
functional properties. Proponents of this type of view may attempt to account for
perceptual organization shifts by appeal to changes in an experience’s functional
properties. However, they may also adopt the proposals I’ll develop.
RST would also be false according to an “intramodal” version of representa-
tionalism. Intramodal representationalism claims only that if two experiences of
the same sense modality differ in their phenomenal properties, then they differ
in their representational properties.3 The view allows, for example, that a visual
and tactile experience might have the same content but still differ phenomenally.
Intramodal representationalism is a weaker thesis than RST. Nevertheless, the
counterexamples considered below also raise challenges for intramodal represen-
tationalism, given that they seem (at least prima facie) to concern visual experience
alone. Thus, intramodal representationalists are also in need of a viable account of
perceptual organization experience, and they may wish to adopt the proposals I’ll
recommend.

2. A phenomenal property is a property of having a certain phenomenal character. A representa-

tional property is a property of having a certain content.
3. Byrne (2001, 205) draws the distinction between inter-­and intramodal representationalism (which
he calls “intentionalism”). Intramodal representationalism is also a version of what Chalmers
(2004) calls “impure representationalism.”

123

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 I want to highlight one further feature of the claims I’ll defend below. My posi­
tive proposals will be consistent with reductive representationalism. By reductive
representationalism, I mean the view that the contents of perceptual experience
can be wholly understood without appeal to phenomenal properties.4 Reductive
representationalism rules out accounts on which perceptual experience represents
phenomenal properties (e.g., the property of having phenomenal character F). It
also rules out accounts on which perceptual experience represents relations to
experiences with certain phenomenal properties (e.g., the property of being dis-
posed to cause experiences with phenomenal character F).5 I am neutral concern-
ing whether visual experiences ever represent properties of this sort.6 However, I
believe that we do not need to appeal to such properties in specifying the content
of perceptual organization experience.

2.2. The Examples

The following three cases have been urged as examples where phenomenal change
obtains in the absence of representational change. As such, they threaten RST. I’ll first
introduce the cases, and then I’ll identify the two aspects of their phenomenology—­
grouping shifts and prominence shifts—­that representationalists need to explain.
The dot array in figure 2 can perceptually appear two different ways. Either
you see the display as organized into separate “rows” of dots, or you see the dis-
play as organized into separate “columns” of dots. The columns display greater
proximity. Each element is closer to its nearest column neighbors than to its near-
est row neighbors. The rows, on the other hand, display greater color similarity.
When a certain collection of dots is grouped, the dots in the collection seem to “go
together” in your experience.

Figure 2

4. See Chalmers (2004) for this characterization.

5. For the latter sort of view, see Shoemaker (2001).
6. I am, however, inclined to endorse some version of the transparency thesis for visual experience.
In other words, I believe that when we introspect our visual experiences it always seems to us
as though we are simply inspecting the scene in front of our eyes (e.g., Harman 1990). As such, I
believe that if we ever have visual experiences that represent the phenomenal properties of our
visual experiences, it does not seem to us as though this is the case.

124

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Peacocke (1983) holds that the difference in phenomenal character between
these two grouping experiences cannot plausibly be explained by appeal to a dif-
ference in content.7
The intersecting lines in 3 can perceptually appear at least four different ways.
Even without shifting your gaze, you can focus your attention on the intersection
of any of the four vertical lines with the horizontal line. With each shift of attention,
different pairs of lines can seem to “stand out” in your visual experience. Moreover,
when a given pair stands out, the lines in the pair also seem to go together in your
experience, forming a cross-­shape.

Figure 3

Speaks (2010) argues that the difference between these experiences cannot be
explained by appeal to a difference in representational content. Rather, he suggests
that the difference is due to a nonrepresentational change in the sui generis phenome­
nology of attention, which he contends is distinct from visual phenomenology.
Finally, the 3 X 3 grid in figure 4a can appear several different ways, depending
on which groups of squares seem to stand out in your experience. However, two
experiences predominate. Either the squares labeled 1, 3, 5, 7, and 9 in figure 4b
seem to stand out, or the squares labeled 2, 4, 6, and 8 seem to stand out. Again, as
with the intersecting lines, when a set of squares stands out, they also seem to go
together into a larger unit.

Figure 4a Figure 4b

Nickel (2007) argues that these two experiences have the very same contents,
and thus that the difference in their phenomenal characters cannot be explained
by appeal to a difference in content.

7. Peacocke (1992) later abandons this view. I will consider Peacocke’s more recent position below.

125

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 If these diagnoses of the examples are correct, then RST is false, because
there are at least some changes in phenomenal character that are unaccompanied
by changes in content. However, the difficulty posed for representationalism is, I
think, even worse than this. As we have seen, there are two kinds of experiential
shifts involved in these examples. First, there are shifts of grouping. That is, there
are changes in which items seem to “go together” in perceptual experience. Second,
there are shifts of prominence. That is, there are changes in which items seem to
“stand out” relative to other items. The dot array clearly involves shifts of grouping,
while the other two examples clearly involve shifts of both grouping and promi­
nence. The reason the challenge is so serious is that, if representationalists can’t
explain these aspects of visual phenomenology, then they can’t explain features that
are present in almost every visual experience. Let me briefly expand on this point.
While the foregoing stimuli are striking because of their multistability (they
can appear different ways at different times), grouping and prominence are present
whenever we perceive a sufficiently complex scene. As the Gestalt psychologists
recognized long ago, this is the case even when our perceptual interpretation of
a scene is stable. The significance of multistability is just that it makes the role of
perceptual organization in visual phenomenology particularly obvious.
Consider perceptual grouping. As Max Wertheimer noted: “If a number of
stimuli are presented together, a correspondingly large number of separate “givens”
do not generally occur for the human; rather there are more comprehensive giv-
ens, in particular segregation, combination, separation” (Wertheimer 1923/1958).
In line with this, the environment almost always seems to decompose into separate
units or objects, and when this occurs, the parts of an object are experienced as
“going together,” such as when a person’s arms, torso, and legs seem to cohere as a
unit. Furthermore, physical connectedness is not a necessary condition for percep-
tual grouping—­think, once again, of flocks of geese or trails of ants.8
Similar remarks apply to prominence. Certain parts of the environment almost
always seem more prominent than others. A cat is normally seen as more salient
than its background, and I currently see my coffee mug as more salient than the
table on which it rests. In fact, Wolfgang Köhler observed decades ago that shifts of
attention readily induce shifts in prominence (Köhler 1947, 170–71, 182), antici­
pating the more recent emphasis on this point (see section 4 below).9
Thus, if representationalists cannot provide adequate accounts of grouping and
prominence, then the view does not merely founder on a few isolated examples.
Rather, it is unable to explain two of the most commonplace features of perceptual

8. While grouping has received by far the most attention in the visual case, it is also ubiquitous in
audition. Temporally extended streams of sound (e.g., speech or musical melodies) appear to
break down into separate “chunks” or “units,” and sounds of different pitches heard together can
be grouped together into a unit, such as when one hears a musical chord (Bregman 1990; Matthen
2010; O’Callaghan 2008).
9. Some of Köhler’s displays (e.g., Köhler 1947, 182) are in fact quite similar to the intersecting lines.

126

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
phenomenology. I’ll argue in what follows that representationalists can give plau-
sible accounts of these phenomena, and moreover that these accounts generalize
to defuse each of the foregoing counterexamples.

3. Grouping

In this section I’ll examine the experience of perceptual grouping. I’ll first consider
some previous treatments of the dot array, and argue that they are inadequate. I’ll
then offer a general account of grouping in section 3.2, and apply this to the dot
array in section 3.3.

3.1. Treatments of the Dot Array

The dot array is usually thought to be unproblematic for representationalism. Thus,

Nickel (2007) suggests that representationalists can adequately account for this
example, even though they cannot account for the 3 X 3 grid. Moreover, Peacocke
(1992) amends his earlier view of the dot array, suggesting that there is plausibly
a difference in content between row-­grouping experiences and column-­grouping
experiences.10 But what is the difference supposed to be? As far as I am aware, two
answers have been offered, neither of which seems fully satisfactory.
Michael Tye’s (2002) answer is perhaps the most straightforward. Tye suggests
that one’s experience attributes different properties to the array in the two cases.
When you experience row grouping, your experience attributes to the array the
property of containing rows, and when you experience column grouping, your
experience attributes to the array the property of containing columns. Since these
are (ostensibly) different properties, the two experiences have different contents.
Thus RST is safe.11
The difficulty for this approach, I contend, is that it fails to capture the phe-
nomenology of certain temporally extended perceptual grouping experiences. Note
that whether a given collection of dots counts as a row or a column depends on its
orientation with respect to the perceiver. No set of dots is intrinsically a row rather
than a column. Thus, for an array to be experienced as containing rows, it must be
experienced as containing collections of dots oriented roughly horizontally with
respect to the perceiver, while for an array to be experienced as containing col-
umns, it must be experienced as containing collections of dots oriented roughly
vertically with respect to the perceiver.
Suppose, then, that you view the dot array in figure 2 and perceptually group
the dots according to color similarity. On the current approach, this simply involves

10. See also Block (2010, 29–30).

11. Nickel (2007) also recommends this answer.

127

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
representing the array as containing collections of dots oriented horizontally. But
now imagine that you slowly rotate the dot array 90º, so that the sets of dots group-
able by color similarity become oriented vertically, while the sets of dots groupable
by proximity become oriented horizontally. Suppose, further, that you maintain
grouping by color similarity across the rotation. In this case, it seems that your
experience continues to organize the array in the same way throughout. The very
same elements of the array seem to “go together” before and after the rotation, and
the groups that they form visually seem to persist throughout the rotation.12 But
this raises a problem. According to the foregoing account, the array should be
organized differently before and after the rotation, because you initially see the array
as organized row-­wise, and later as organized column-­wise. The account thus fails to
accommodate the fact that we can experience perceptual grouping as remaining
stable across changes in viewer-­centered orientation.
Peacocke (1992, 79) offers a different view of the dot array. Peacocke suggests
that the difference between row grouping and column grouping is accompanied by
a difference in which quadruples of dots are represented as collinear. In particular,
when one perceives row grouping, the quadruples of dots in each row are repre-
sented as collinear, and when one perceives column grouping, the quadruples of
dots in each column are represented as collinear. Note that this proposal has the
advantage of explaining how a single organization of the array can be maintained
across a 90º rotation. The same quadruples of dots might be represented as collin-
ear both before and after the rotation.
Unfortunately, this view also faces difficulties. Most importantly, it is unclear
why simply representing some dots as collinear should suffice for experiencing
them as going together into a unit. Collinearity is a kind of spatial relation among
points or line segments. But notice that perceiving some items as standing in a
particular spatial relation does not always suffice for grouping those items. Indeed,
some spatial relations seem to make grouping less likely. For example, when you
perceive one surface as partially occluding another, you are likely not to group the
two as a single entity (e.g., Nakayama et al. 1995). Plausibly, then, experiencing a
plurality of elements as forming a unified group involves something beyond the
representation of a spatial relation among those elements.
Furthermore, note that perceptual grouping can obviously occur among non-­
collinear elements, as in figures 1, 3, and 4a. But all of these figures (and of course
myriad other examples) also give rise to the characteristic experience of items
“going together.” Optimally, we would like to explain what is common to group-
ing phenomenology across these various cases, even though they involve different
particular spatial relations among the grouped elements. But the current proposal
supplies no such explanation.

12. As I’ll discuss below, there is also empirical evidence that perceptual groups can be perceptually
reidentified over time, even as they move with respect to the perceiver.

128

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
3.2. Perceptual Grouping as Cue-­Based Composition

I’ll now offer a different account of grouping phenomenology, which I call the
cue-­based composition account. I’ll argue that this view explains grouping shifts in
the dot array, and also avoids the pitfalls of the previous approaches.
Cases of perceptual grouping exhibit the following pattern. The visual sys-
tem detects a property or relation (or multiple properties and relations) instanti-
ated by a plurality of elements in the scene, and as a result, treats the plurality as
forming a unified group. For example, row grouping in the dot array is facilitated
by the detection of both color uniformity and collinearity among elements of
the rows.
It has been known since the work of early Gestalt psychologists that only
certain perceptible properties and relations support grouping.13 These properties
and relations are called cues to grouping. The perceptual grouping principles state
generalizations linking these cues to grouping. Much recent work on perceptual
organization attempts to quantitatively specify the strengths of various grouping
cues, and to characterize how different cues trade off when placed in competition
with one another (e.g., Kubovy and van den Berg 2008).
Well-­known grouping cues include proximity, similarity, good continuation,
and common fate. Element connectedness is a more recent addition to this list
(Palmer and Rock 1994). Here are very rough statements of the grouping prin-
ciples corresponding to these cues. Applications are shown in figure 5:
Principle of proximity: Elements that are close together tend to be
grouped.
Principle of similarity: Elements that are similar (in color, shape, size,
orientation, etc.) tend to be grouped.
Principle of good continuation: Elements that fall along a smooth curve
tend to be grouped.
Principle of common fate: Elements that move together (e.g., along
equal and parallel motion trajectories) tend to be grouped.
Principle of element connectedness: Connected elements tend to be
grouped.

13. See Wertheimer (1923/1958). Nevertheless, it should also be noted that new grouping cues (and
corresponding principles) are still being discovered today (Wagemans et al. 2012a).

129

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Figure 5. Left column (top to bottom): Grouping by proximity, grouping by
common fate, grouping by element connectedness. Right column (top to bottom):
Grouping by good continuation, grouping by similarity.

A perceptual group is formed on the basis of perceptually representing the

items in the group as instantiating one or more grouping cues.14 Of course, a natu­
ral question concerns why we are disposed to group items when they instantiate
grouping cues. Why are our visual systems set up this way, and how is it advan-
tageous? This is a—­perhaps the—­fundamental question in Gestalt theory, and I
won’t attempt to answer it here.15 To rescue RST, however, I need only to systemat-
ically characterize the representational content of grouping experiences. I needn’t
explain why the visual system is disposed to produce such experiences.
I suggest that when elements are grouped together, they are experientially rep-
resented as composing a larger individual.16 They are seen as parts of a larger thing.
This claim is not quite as innocent as it may initially appear, for two reasons. First, it
implies that perceptual experiences represent mereological (part-­whole) relations.
As such, experiences represent more than shape, size, color, and spatial relations. A
fortiori, experiences represent more than the positions and local features of small
surface patches.17 Second, the view implies that perceptual experiences reify percep-
tual groups. Groups are represented as further things, over and above their mem-
bers. Both of these claims require support, and I’ll return to them shortly.
Suppose that you see four items, a, b, c, and d, as perceptually grouped by
virtue of their collinearity. I propose that your visual experience represents at least
five individuals—­the four grouped items along with a fifth individual, e, com-
posed of them. Moreover, the experience represents at least two kinds of relations:

14. I do not mean to suggest that this is a personal-­level inference. Rather, the transition is carried out
by the computational processing mechanisms of the visual system.
15. See Wagemans et al. (2012b) for an excellent overview of historical and contemporary theoretical
work on this topic.
16. This approach has precedent in the psychological literature (see Feldman 2003; Palmer 1977).
17. For an extended argument against this kind of “minimalist” view within the domain of shape
phenomenology, see Green (2015).

130

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
the grouping cues instantiated by the grouped items, and the part-­whole relations
between the grouped items and e. The content of the experience can thus be speci­
fied as in (1):
(1) Collinear(a, b, c, d) & Part-­of(a, e) & Part-­of(b, e) & Part-­of(c, e) &
Part-­of(d, e)

I call this the cue-­based composition account of perceptual grouping. Elements

are grouped by virtue of being represented as composing a larger entity, and the
(causal/computational) explanation of why elements come to be so represented is
that they are represented as instantiating certain grouping cues.
One immediate worry with the cue-­based composition account is that, as
stated, it casts perceptual content as singular or object-­involving. (The expres-
sions a, b, c, d, and e are singular terms that purport to refer to particulars.) There
are well-­known challenges to this approach, which I won’t rehearse here (see
Schellenberg [2010] and Soteriou [2000] for discussion). Fortunately, however, a
generalist about perceptual content could readily adopt a modification of the cue-­
based composition account. She could simply replace the singular terms in my
content specification with free variables and prefix the formula with quantifiers to
bind each of them. Thus, I will put the singularism-­generalism dispute aside for the
remainder of the paper.
I now return to the two key claims identified earlier. The first claim is that
perceptual grouping involves the representation of parthood, in addition to the
representation of grouping cues. I believe that introspection supplies two rea-
sons to believe this. First, as noted above, it seems likely that grouping experience
involves something beyond the representation of bare spatial relations (e.g., col-
linearity or proximity), since the representation of such relations is not plausibly
sufficient for grouping. I suggest that the representation of parthood offers a
compelling account of this extra experiential aspect. Second, by appeal to the
representation of parthood, we can offer a compelling account of the common
factor in various cases of perceptual grouping. Even though the specific grouping
cues differ from case to case, each one nonetheless involves the representation of
parthood or composition.
The second key claim is that perceptual groups figure in perceptual content as
individual objects or entities, alongside their constituents. One might be dubious
of the idea that perception reifies groups in this way. At the very least, we need
positive reasons to accept it. In fact, I think that there is good empirical evidence
that when items are grouped, the group is treated as an individual object.
Perhaps the strongest evidence that perception treats groups as individuals
in their own right is that we can perceptually reidentify a group despite changes
in the number of members that it contains. Indeed, we can even see a percep-
tual group as the continuation of a simple ordinary object (e.g., a closed polygon)
seen earlier. To see why this is relevant to the present issue, consider an analogy.
Suppose you are first shown a full six-­pack of beer, and then you see someone
remove three bottles from the pack. If you judge that the pack persists through this

131

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
change, then plausibly you are treating the pack as an individual, over and above
the particular bottles it contains.
Visual perception reidentifies objects over time. This is most clearly shown
by demonstrations such as apparent motion, the bouncing/streaming paradigm,
and the tunnel effect (see, e.g., Feldman and Tremoulet [2006] and Flombaum et
al. [2009] for discussion). For instance, suppose you see Frames 1 and 2 in fig-
ure 6 in rapid succession. You will almost certainly have one of the following two
­experiences: (i) you will see the square morph into the triangle and the circle
morph into the trapezoid, or (ii) you will see the square morph into the trapezoid
and the circle morph into the triangle. In both cases your experience reidentifies
objects (e.g., in (ii) the square is reidentified as the trapezoid). Moreover, there is
good evidence that object reidentification in apparent motion is genuinely per-
ceptual, since the rules governing the process are largely hardwired into the visual
system, and do not take into account high-­level knowledge or beliefs (see Carey
2009, 73; Dawson 1991; Flombaum et al. 2009; Scholl 2007).

Figure 6. Frame 1 (left) and Frame 2 (right)

In recent years, visual object reidentification has been extensively studied in

the multiple-­object tracking (MOT) paradigm. In the MOT task, participants are
asked to track a set of “target” objects as they move about the screen in the pres-
ence of moving distractors. We are quite good at performing this task for target set
sizes of up to four objects (Pylyshyn and Storm 1988). Critically, the task involves
reidentifying the original targets at the end of a trial, and this is thought to reflect
our capacity to perceptually reidentify targets from one moment to the next during
the trial (cf. Scholl and Pylyshyn 1999). Relevant for present purposes is a modi-
fication of the task studied by vanMarle and Scholl (2003). VanMarle and Scholl
showed participants a display in which the objects (both targets and distractors)
began as individual squares. However, as soon as a square started to move it would
break apart into pieces that moved together as a cluster, thus forming a common-­
fate-­based perceptual group. Critically, vanMarle and Scholl found that partici-
pants’ tracking capacities were the same in this condition as in the standard MOT
paradigm. They could reliably track about four groups. A plausible interpretation
of these results is that each of the initially seen squares visually seemed to persist
during a trial, but did not appear to persist as any individual, connected object.
Rather, they appeared to persist as groups.

132

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Consider, further, the following experiment due to Wynn, Bloom, and Chiang
(2002). Infants were repeatedly shown an array containing two moving groups of
three dots. After they habituated to this display, they were shown one of two test
displays. One of the test displays contained four moving groups of two dots, while
the other contained two moving groups of four dots. It was found that infants
looked significantly longer toward the former display than toward the latter dis-
play. In other words, they were more surprised to see a test display that had a
different number of groups than one that had the same number of groups, even
though the two test displays had the same number of individual dots. A plausible
explanation of the infants’ surprise is that they perceptually represented two things
in the habituation display, and then represented either two things or four things
in the test display. In support of this “perceptual” interpretation of infants’ enu­
meration capacities, many psychologists have argued that tracking and enumera-
tion in infancy draws on the very same mid-­level visual tracking system recruited
in apparent motion and multiple-­object tracking (see Carey 2009; Scholl and
Leslie 1999).
Given this evidence, I want to emphasize two points. First, it seems that we
can see a group as the continuation of an entity seen earlier without seeing any
members of the group as the continuation of that entity. Second, it seems that we
can see a group as persisting without seeing its members as persisting. These facts
strongly suggest that groups are perceptually represented as individual things, over
and above their members.18
I conclude, then, that the two key claims associated with the cue composition
account have both introspective and empirical support. Grouping involves the
representation of parthood, and groups are experienced as individual entities.

3.3. The Dot Array

Let’s now apply the cue-­based composition account to the dot array. I’ve labeled
the dots in the top row and left column in figure 7.

Figure 7

18. Cf. Koffka’s (1935, 176) famous claim that a perceptual group is “other than” the sum of its parts.

133

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 As noted above, the columns of dots are groupable because they instantiate
the proximity cue: Each element is closer to its nearest column neighbors than
it is to its nearest row neighbors.19 To make this more precise, let’s introduce the
notion of a proximity chain: An ordered set of entities (o1, . . . , on) is a proximity
chain (PC) just in case (a) o1 is at least as near to o2 as it is to anything else, (b) on is
at least as near to on-­1 as it is to anything else, and (c) for all oi where 1 < i < n, the
two nearest neighbors of oi are oi -­ 1 and oi +1. Thus, when a subject perceives column
grouping, her experience represents the column on the left as in (2):
(2) PC(d1, d5, d6 , d7) & Part-­of(d1, a) & Part-­of(d5, a) & Part-­of(d6 , a) &
Part-­of(d7 , a)

Likewise for the other columns in the array.

A similar treatment can be given for row grouping. Here, the relevant group-
ing cue is color uniformity. Thus, when a perceiver experiences row grouping, her
experience represents the top row as in (3):
(3) Same-­color(d1, d2 , d3 , d4 ) & Part-­of(d1, a) & Part-­of(d2 , a) & Part-­
of(d3 , a) & Part-­of(d4 , a)

The cue-­based composition account accommodates the intuition that percep-

tual grouping involves both perceiving elements as instantiating some perceptible
relation (e.g., collinearity, proximity, or color similarity) and perceiving the ele-
ments as going together. To perceive some elements as going together just is to
perceive them as parts of a larger individual. Further, note that the account can
capture the stability of perceptual grouping across a 90º rotation. The very same
dots are represented as composed throughout the rotation, and they are repre-
sented as instantiating the very same grouping cue. By virtue of this, you experi-
ence the array as grouped in the same way throughout the rotation.
I conclude, then, that the dot array poses no problem for RST. Shifts in group-
ing phenomenology in this display (and in general) are plausibly accompanied by
changes in content.

4. Prominence

In this section I’ll consider the experience of prominence. I’ll first examine some
available treatments of prominence, and argue that they are inadequate. I’ll then
offer my own account in section 4.2, and apply this to the intersecting lines (sec-
tion 4.3) and the 3 X 3 grid (section 4.4).

19. For quantitative analyses of grouping by proximity, see Kubovy et al. (1998) and Kubovy and van
den Berg (2008). Of particular importance is Kubovy and colleagues’ “pure distance law,” which
states that the likelihood of grouping a dot array along a given direction v is a decreasing expo-
nential function of the ratio of the distance between neighboring dots along v and the shortest
inter-­dot distance in the array.

134

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
4.1. Prominence: Existing Accounts

Several philosophers have tried to explain the experience of prominence in the

intersecting lines and 3 X 3 grid examples. Note that visual attention seems to play
an important role in bringing about these shifts. For instance, when you attend to
a certain group of squares in the grid, that group tends to become prominent. It
is thus natural to try to explain these experiential changes by identifying an effect
that attention exerts on perceptual content. In developing my own account below
I’ll appeal to one such influence of attention on perceptual content. However, I will
first briefly evaluate some alternative accounts of this type.
Nanay (2010) proposes that when you allocate attention to a particular object
or location, your visual experience represents more determinate properties of that
object, or more determinate properties exemplified at that location:
Attention makes the attended property more determinate. If I am attend-
ing to the color of my office telephone, I attribute very determinate
(arguably super-­determinate) properties to it. If, as it is more often the
case, I am not attending to the color of my office telephone, I attribute
only determinable properties to it (of, say, being light-­colored or maybe
just being colored). (Nanay 2010, 266)20

Thus, objects outside the window of attention are represented only as having deter-
minable properties (e.g., being red, being quadrilateral), while objects inside the
window of attention are represented as having more determinate properties (e.g.,
being red17, being square).
Nanay suggests that this view offers an explanation of the changes in con-
tent that accompany changes in phenomenal character in the grid example, but
unfortunately he does not offer precise details about how the account is supposed
to work. He claims, for instance, that when squares 1, 3, 5, 7, and 9 of the grid are
attended, they are represented as having more determinate properties than the
other squares (Nanay 2010, 267). But in which respects are the attended squares
supposed to appear more determinate?
One possibility is that Nanay believes that the locations of the edges of the
attended squares are represented more determinately—­or perhaps with higher
spatial resolution (cf. Carrasco and Barbot 2015; although see Bronner 2015). This
might in turn suffice for making the squares bounded by those edges more expe-
rientially prominent. However, if this is the idea, then, as Nickel (2007) notes, it
is problematic. Square 5 shares all of its edges with squares 2, 4, 6, and 8. As such,
if squares 2, 4, 6, and 8 stand out simply because their edges are represented with
higher resolution, then, for the same reason, square 5 should stand out too. But

20. Although Nanay says that attention “makes the attended property more determinate,” what he
really means is that attention leads one’s perceptual experience to represent more determinate
properties. Attention presumably does not affect the represented properties themselves, but only
which properties are perceptually represented.

135

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
since it is possible to experience squares 2, 4, 6, and 8 as prominent without expe-
riencing square 5 as prominent, this account does not work.
A different proposal is that the attended squares appear more determinate
in respect of their brightness or luminance properties. Thus, the attended squares
may be experienced as having a very specific luminance value, while the un-
attended squares may only be represented as having luminance values within some
broader range. Consistent with this, there is evidence that attention decreases
the variability of subjects’ judgments concerning the brightness of a stimulus
(Prinzmetal et al. 1997). One explanation for this (though by no means the only
one) is that perceptual content becomes more determinate vis-­à-­vis brightness
under conditions of attention. However, even if this is so, it does not strike me as
a plausible account of experiential prominence. First, the proposed connection
between prominence and the determinacy of perceived luminance is somewhat
mysterious. It is unclear why a figure should appear more salient simply by virtue
of being attributed a more determinate luminance value. Second, even if attention
does lead to changes in the determinacy of perceived brightness, these changes
are not especially striking upon introspection. It is not immediately obvious that
the attended squares appear to have more determinate luminance values. On the
other hand, shifts in prominence are immediately striking. This leads to a puzzle. If
shifts in prominence are due to shifts in the determinacy of perceived brightness,
then why aren’t the latter shifts immediately obvious to us whenever we undergo
prominence shifts?
I think that a more compelling account of experiential prominence in this
case is rather that when certain squares are attended, their perceived luminance
increases in intensity. The attended squares, on this proposal, appear brighter than
the unattended squares. In this vein, Ganson and Bronner (2013) and Wu (2014,
133) both suggest that attention-­induced changes in apparent brightness may
account for changes in the appearance of the 3 X 3 grid. Moreover, this account
has prima facie plausibility. Brighter things do sometimes seem to “stand out” from
darker things. They are, in some sense, more salient in one’s phenomenology.
However, it is an open question whether the sense of “prominence” in which
brighter things are more prominent than darker things is the same sense in which
the attended squares of the grid are more prominent than the remaining squares.
In fact, there are reasons to doubt that certain squares are more prominent simply
by virtue of appearing brighter. Consider figures 8a and 8b. In figure 8a, squares
1, 3, 5, 7, and 9 are slightly darker than the remaining squares, and the reverse is
true in figure 8b. If all it is for a particular group of squares to be prominent is for
the members of that group to appear slightly brighter than the remaining squares,
then we should be unable to see squares 1, 3, 5, 7, and 9 as standing out in figure
8a, and similarly for squares 2, 4, 6, and 8 in figure 8b. But this does not seem cor-
rect. These figures seem just as bistable as the original grid. And critically, when
the darker squares appear prominent, they also appear darker than the remain-
ing squares. Thus, it does not seem that when certain squares are experienced as
promi­nent, this is because they appear brighter than the other squares.

136

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Figure 8a Figure 8b

One might suggest that it is not changes in perceived brightness intensity that
underlie shifts in appearance of the grid, but rather changes in perceived bright-
ness contrast. Evidence suggests that both involuntary and voluntary attention can
increase perceived contrast (Carrasco et al. 2004; Liu et al. 2009).21 Given this, a
natural proposal is that although squares 1, 3, 5, 7, and 9 do not appear brighter
than the others in figure 8a, attention boosts the perceived difference in luminance
between these squares and their surroundings.22 Perhaps, for example, the effect of
attention is flexible. Attention may render squares apparently brighter in the origi­
nal grid, but (e.g.) render squares 1, 3, 5, 7, and 9 apparently darker in figure 8a.
Consistent with this proposal, in Peter Tse’s (2005) attention-­induced bright-
ness illusion (discussed in more detail below), attention makes items presented
against a white background appear darker, while it makes items presented against
a black background appear brighter. However, it is unclear whether the factors
responsible for Tse’s illusion are at work in the grid. The attended items in Tse’s
example also come to appear transparent, while this does not seem to occur for
the attended squares in the grid. I’ll return to Tse (2005) below when I offer my
alternative proposal.
Despite the scientific evidence for the claim that attention alters apparent con-
trast, I think it is unlikely that this phenomenon underlies prominence shifts in
the grid. Suppose that one sees squares 2, 4, 6, and 8 as prominent. Then, on the
current proposal, the difference between the perceived luminance of these squares
and the perceived luminance of their surroundings should be boosted. However,
if this is the case, then presumably the perceived contrast of square 5 should also
be boosted, because it is bordered by the four attended squares. Thus, if boosts
in perceived contrast are what underlie the relevant phenomenal shifts, then it

21. Block (2010) argues that these effects of attention on apparent contrast cannot be explained in
terms of changes in content. Block’s arguments are independent of the putative counterexamples
considered here, and I will not address them in the current paper. However, for potential replies
to Block, see Hill (manuscript) and Watzl (forthcoming).
22. A different possibility is that attention boosts apparent contrast internal to attended objects (e.g.,
differences in apparent brightness between light and dark regions of the attended object), rather
than between the attended object and its surroundings. Indeed, attention does have this effect on
Gabor patches (Carrasco et al. 2004). However, given that each square in the grid has uniform
brightness, it is unclear how this strategy would apply here.

137

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
is unclear why we are able to see squares 2, 4, 6, and 8 as prominent without also
seeing square 5 as prominent. Unless the account is supplemented with further
details, this remains unexplained.23
I conclude that shifts in apparent determinacy, brightness, or contrast are
unlikely to account for the grid’s shifts in prominence. I have not explicitly con-
sidered the intersecting lines figure, but I think that similar worries for these pro-
posals arise in that case as well. Representationalists need a different account of
prominence shifts in these examples.

4.2. Prominence as Perceived Protuberance

I now offer an alternative explanation of the type of prominence involved in our

experiences of the intersecting lines and the 3 X 3 grid. My proposal is quite straight-
forward. When an item is experienced as more prominent than other items in these
displays, it is represented as closer to the perceiver than those other items. In other
words, the type of experiential prominence involved in these displays is explained by
the experiential representation of protuberance. This also offers a plausible explana-
tion of prominence in familiar cases. It is, for example, a plausible account of why
a cat is experienced as more prominent than its background, or my coffee mug is
experienced as more prominent than the table on which it rests. I contend, further,
that attention can induce changes in perceived depth, and this leads to prominence
shifts in multistable displays such as the intersecting lines and the grid.
In line with this view, there is a host of evidence indicating that attention
shifts are associated with changes in perceived depth and occlusion. For example,
when observers are instructed to allocate attention to one face of the Necker cube,
that face is generally perceived as being in front (Inui et al. 2000; Kawabata 1986;
Meng and Tong 2004). It is possible to produce a parallel depth reversal through
attention shifts even when the two faces are not seen as surfaces of a single 3-­D
object (see figure 9). Moreover, attention can also induce reversals of relative depth
in other well-­known bistable stimuli, such as the Schroeder staircase (Leopold and
Logothetis 1999).24

23. In response, one might propose that for a group of squares to be seen as prominent, they must all
have apparent contrast in the same direction. In other words, while it is sometimes the case that the
prominent squares appear brighter than most of their surroundings, and other times the case that the
prominent squares appear darker than most of their surroundings, it can never be the case that some
of the prominent squares appear brighter while the others appear darker. However, this account also
faces a difficulty. Suppose that you see squares 2, 4, 6, and 8 as darker than most of their surroundings.
(The “most of” is critical here, because these squares touch one another.) Then what makes it the case
that they stand out as a group, as opposed to square 5 standing out on its own? For, square 5 would
in this case appear brighter than most of its surroundings, and it has already been conceded that
this is at least sometimes sufficient for appearing prominent. So why is it not sufficient in this case?
24. Reflection on examples like the Necker cube suggests, I believe, that either overt or covert atten-
tion shifts can alter depth perception. (Overt attention shifts involve corresponding changes in
fixation or gaze direction, while covert attention shifts do not.) However, in presenting their

138

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Figure 9

Peter Tse’s (2005) illusion also seems to illustrate the effect of attention on per-
ceived depth. When you stare at any of the fixation dots in figure 10 but covertly
attend to one of the overlapping discs, the disc to which you attend comes to appear
darker than the other discs. (One can produce the opposite effect—­viz., attention-­
induced brightening—­with discs placed on a black background.) This example is
usually used to demonstrate the effect of attention on perceived brightness or con-
trast. However, it is also obvious that when you attend to one of the discs, it comes
to appear in front of the others—­it looks like a transparent figure partly occluding
them. Note, moreover, that this attentional influence on depth perception seems to
be critical to producing the brightness effect. Tse reports that when one of the discs
is specified as being behind the other two via binocular disparity, the brightness
illusion can no longer be elicited for the backmost figure (Tse 2005, 1097).

Figure 10. A version of Tse’s brightness illusion.

Thus, attention plausibly can influence depth perception. I suggest that this
occurs in both the intersecting lines (figure 3) and the 3 X 3 grid (figure 4a).

­challenges to representationalism, Nickel and Speaks focus solely on covert attention shifts. This
is because changes in phenomenology associated with overt attention shifts may have a number of
plausible explanations. For example, if you shift your fixation, then the areas of your field of vision
toward which you fixate will plausibly come to be represented with greater spatial resolution. As
such, for the representationalist to adequately account for the alleged counterexamples, she needs
only to identify an effect of covert attention on perceptual content. Nevertheless, the proposal
that attention influences depth perception may also help us to understand certain changes in
phenomenal character associated with overt attention shifts.

139

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
4.3. The Intersecting Lines

Return to the intersecting lines figure. Two aspects of the phenomenology of this
figure are noteworthy. First, depending on one’s allocation of attention, a certain
pair of lines may seem more prominent than the rest. Second, the lines that seem
prominent also seem to go together as a cross-­shaped group. We are now in a posi-
tion to understand both of these shifts in terms of changes in content.
When you attend to the intersection between the horizontal line and the ver-
tical line second from the left, this pair of lines appears nearer to you than the
remaining lines, and also appears to occlude the remaining lines. This is the reason
that they appear prominent. Furthermore, this plausibly has an influence on per-
ceptual grouping. Because the pair of attended lines is represented as occupying
a nearer depth plane than the remaining lines, they are also seen as disconnected
from the remaining lines, and so are unlikely to be grouped with them. However,
the two attended lines are represented as connected to each other. Because they are
represented as connected, they can be perceptually grouped under the principle of
element connectedness (Palmer and Rock 1994). This is why they seem to form a
cross-­shape.25
Thus, I propose that the reason why the attended lines are experienced as a
cross that is more prominent than the rest of the lines is that (i) they are repre-
sented as nearer to the observer, (ii) they are represented as physically connected,
and, as a result, (iii) they are represented as composing a larger individual.
The proposal that attention affects depth perception also helps us understand
a similar example, discussed by Chalmers (2004). Suppose that you see two small
red lights against a black background, and covertly shift attention from one to the
other. There is clearly a phenomenal change accompanying these attention shifts,
but it is not initially clear what representational change might be involved. Again,
I suggest that the light to which you attend appears closer than the light to which
you do not attend. Of course, this is compatible with there being other attention-­
induced differences as well (say, in the determinacy or intensity of their repre-
sented brightness).
I conclude that the intersecting lines figure poses no problem for RST. The
shifts in its appearance are plausibly accompanied by changes in content.

4.4. The 3 X 3 Grid

I think the 3 X 3 grid poses the most difficult challenge out of our three examples,
because it is the most complex. I’ll begin with what I take to be the most promising
existing account of this figure, due to Jagnow (2011). I’ll argue that while Jagnow’s
account may provide a satisfactory explanation of the shifts in appearance a­ ssociated

25 It is also possible that the perpendicularity of the [Link] lines facilitates grouping. Indeed,
some authors count perpendicularity as a grouping cue (Feldman 1999; Leordeanu et al. 2009).

140

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
with Nickel’s figure, it does not generalize to explain shifts in appearance of a simi­
lar figure. Then I’ll argue that the cue-­based composition account coupled with
the proposal that attention affects depth perception offers a unified explanation
of both cases.
Jagnow suggests that changes in the appearance of the grid are underpinned
by shifts in figure-­ground organization. And indeed, the 3 X 3 grid indeed bears
some similarity to classic reversible figure-­ground displays, such as the face/vase
image. In both cases, certain regions of the display can appear to stand out while
other regions intuitively appear to “recede” into the background. Thus, it is natural
to seek a common explanation of the experiential shifts in both cases.
As Jagnow notes, there are several features distinctive of figure-­ground phe-
nomenology. For present purposes, I will highlight two (cf. Nakayama et al. 1989;
Baylis and Driver 1995; Palmer and Brooks 2008). First, and most obviously, the
figure appears closer than the ground. For instance, in Rubin’s famous face/vase
display, figure-­ground shifts are associated with shifts in which region appears to
be closer to the perceiver. Second, the edge between the regions is assigned to the
figure, rather than to the ground. The second feature is often called “one-­directional
edge assignment.” Intuitively, when the vase is seen as figural, the edges marking its
boundaries with the face-­shaped regions seem to “go with” the vase-­shaped region,
and not with the face-­shaped regions.26
In the 3 X 3 grid, Jagnow proposes that when a certain square looks promi-
nent while its neighbor does not, the prominent square is treated as figural while
the nonprominent square is treated as background. Thus, the prominent square
appears closer, and the shared edge is assigned to the former and not to the lat-
ter. Moreover, because perceived edge ownership is one-­directional, we have an
explanation of why certain experiences of the grid predominate: Generally, either
squares 2, 4, 6, and 8 are treated as figural, or squares 1, 3, 5, 7, and 9 are treated
as figural. These experiences are consistent with one-­directional edge assignment.
Other experiences, however, would require violating that requirement, and thus
are unlikely. For example, experiencing squares 1, 3, 5, 6, and 8 as figural would
require assigning the edge between squares 5 and 6 to both regions simultaneously.
The claim that shifts in prominence in the 3 X 3 grid are due to figure-­ground
reversals would not be helpful to representationalists unless it could also be shown
that figure-­ground reversals can be explained by appeal to changes in content.
Jagnow argues that representationalists can account for figure-­ground phenome-
nology. He writes:
[I]t is not sufficient to characterize the contents of perceptual expe-
riences exclusively in terms of objects and their properties. Rather,

26. Unfortunately, theorists are rarely explicit about what it means for an edge to be “assigned to” a
region. I’ll return to this issue in the appendix.

141

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 perceptual contents also have a figure/ground structure, that is, they
attribute the properties being a figure (or object) and being a ground to
parts of the scene in front of the perceiver. (2011, 332–33)

Thus, because prominence shifts are associated with changes in which parts of the
scene are represented as figure and which are represented as ground, RST is safe.
While I think that it is, overall, quite attractive, Jagnow’s account has some
puzzling features. For instance, one might wonder about the properties being figure
and being ground. What is it for something to be a figure, and how does this differ
from being a ground? Moreover, why are these properties not attributed to objects?
In the appendix, I’ll develop a different account of figure-­ground phenomenology.
I’ll argue that it can be explained by appeal to the cue-­based composition account,
without bringing in the properties being figure and being ground. However, I’ll put
aside these details for now.
I agree with Jagnow that figure-­ground reversals occur in the 3 X 3 grid, and
will retain this part of the view. However, I also believe that we can improve on
Jagnow’s account. I will argue in what follows that by combining the cue-­based
composition account with the claim that attention influences perceived protuber-
ance, we can retain the spirit of Jagnow’s explanation of the original 3 X 3 grid,
but we can also extend the resulting account to explain other examples that his
account does not explain.
Consider the array of squares in figure 11a, numbered as in figure 11b. Just
as with the original 3 X 3 grid, it is possible to see certain groups of squares (e.g.,
squares 1, 3, 5, 7, and 9) stand out in this display. Moreover, the shifts in appearance
of figure 11a seem qualitatively similar to the shifts in appearance of the origi-
nal grid (though they are, I submit, less striking 27). However, because the squares
do not share edges with their neighbors, one-­directional edge assignment does
not apply in this case. Figure-­ground processes should presumably treat all of the
squares as figural.

Figure 11a Figure 11b

27. I believe that the difference in salience between the phenomenal shifts in the original grid and
the phenomenal shifts in figure 11a is due to the fact that the original grid involves shifts in edge
assignment while figure 11a does not. I’ll return to this in the appendix.

142

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 We can offer a unified account of the experiential shifts in both the original
grid and in figure 11a. Certain subsets of the squares in both grids (e.g., squares
1, 3, 5, 7, and 9, or squares 2, 4, 6, and 8) form symmetric patterns. This serves as a
cue for grouping the squares together, and so for representing them as composing
a larger individual. Once some squares have been grouped into a single individ-
ual, the individual they form can be targeted by visual attention. However, at this
stage, various groups may compete for visual attention. For instance, the group
composed of squares 1, 3, 5, 7, and 9 may compete with the group composed of
squares 2, 4, 6, and 8. Finally, when one of these groups “wins out” and gets targeted
by visual attention, the squares in the group are seen as closer to the observer than
the remaining squares. According to my proposal, it is because these squares are
represented as closer that they are prominent in one’s experience.
There is empirical evidence supporting this proposal. First, note that symme-
try is a well-­known grouping cue (see Feldman 2007; Palmer 1999, 258–59). As
such, it is plausible that visually representing some squares as forming a symmetric
pattern is indeed a cue for grouping those squares. Second, there is evidence that
when elements are perceptually grouped, the resulting group attracts visual atten-
tion (see Kimchi 2009). It is natural, then, that multiple candidate groups may
compete for attention (cf. Vecera 2000, who suggests that when multiple objects or
groups have been segregated, they generally compete for visual attention). Finally,
there is evidence that attention exerts an influence on perceived depth ordering in
bistable displays (see section 4.2).
There is also strong evidence that attention influences figure-­ground percep-
tion. When a subject attends to a region (either voluntarily, or as the result of an
exogenous cue), she tends to see that region as figural (Baylis and Driver 1995;
Vecera et al. 2004). I suggest that this tendency reflects a more general influence of
attention on perceived depth. The effect of attention on figure-­ground organiza-
tion is, I propose, a result of the effect of attention on depth perception.
Before ending this discussion, I should address a challenge from Nickel (2007).
I’ve suggested that prominence shifts are explained by changes in perceived depth
or protuberance. However, Nickel briefly considers the idea that shifts in promi-
nence are associated with depth reversals, and argues that it should be rejected. He
writes the following:
This proposal will not do, because of general features of depth percep-
tion. If two objects subtend the same angle in a perceiver’s visual field
and one object appears closer than the other, the apparently closer one
also appears smaller. But in the case of the more prominent squares, we
do not observe a difference in apparent size. The prominent and non-­
prominent squares have the same apparent size. That strongly suggests
that the squares do not appear to be at different distances. (Nickel 2007)

The idea is that because all the squares in the grid both subtend the same visual
angle and also (Nickel contends) appear to have the same linear (distal) size, it is
implausible that some of the squares appear closer than the others. For if they did,

143

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
then the perceived depth of some of the squares would conflict with their per-
ceived linear size and visual angle (due to the laws of perspective). Nickel suggests
that this feature of the grid renders it a more serious counterexample to represen-
tationalism than familiar bistable figures such as the Necker cube, since shifts of
phenomenology are plausibly associated with depth reversals in the latter cases.
Does this argument undermine my proposal?
It does not. In fact, Nickel fails to notice that the Necker cube already offers
a forceful rebuttal to his argument. Note that in the standard Necker cube, the
reversible front and back square faces are drawn the same size, and thus subtend
the same visual angle. Furthermore, we can clearly see one face as closer than the
other. But even when this occurs, both faces nonetheless appear to have the same
linear size—­indeed, the figure appears cubical. The Necker cube nicely illustrates,
then, that perceived shape and linear size are not simple functions of perceived
distance and visual angle (for a related point, see Pizlo 2008, 154–55). As such,
Nickel’s argument against the proposal that prominence shifts are associated with
depth reversals is unpersuasive.
I conclude that the 3 X 3 grid poses no problem for RST. Shifts of grouping
and prominence in this display are readily explained by appeal to changes in rep-
resentational content.

5. Conclusion

In this paper I’ve argued that representationalists can account for two ubiquitous
aspects of perceptual organization: grouping and prominence. Grouping can be
explained by appeal to the representation of composition, while prominence can
be explained by appeal to the representation of depth. I conclude that these fea-
tures pose no difficulty for representationalism. Indeed, they can be accommodated
in ways that are both introspectively plausible and empirically supported.

Appendix

Here I will show how to extend the cue-­based composition account to explain one-­
directional edge assignment in figure-­ground organization. This will complete my
account of the phenomenology of Nickel’s 3 X 3 grid, and will also show how
Jagnow’s account of the grid can be subsumed under my own. Moreover, it will
illuminate the extra representational change associated with shifts in prominence
in the original grid, and hence explain why these shifts are more salient than the
shifts in figure 11a.
Let’s first distinguish the interior regions of the grid’s squares from the contours
that bound and separate those regions. I’ve labeled these below in separate figures.
The interior regions are labeled in figure 12a, and the edges are labeled in figure 12b.

144

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
 Figure 12a Figure 12b

I’ve argued that when squares 1, 3, 5, 7, and 9 are prominent in one’s experience,
they are represented as being in front. Because of this, they are also experienced as
figural. So, given one-­directional edge assignment, edge e5 is “assigned” to region r1
rather than r2. (I’ll return shortly to the explanation of why depth perception has
this effect on edge assignment and hence figure-­ground organization.)
What is it for an edge to be “assigned” to a region? Intuitively, it involves expe-
riencing the edge as going with that region, rather than its neighbor. The fact that
this description is so natural suggests that figure-­ground organization may involve
perceptual grouping between edges and regions.28 Consistent with this, recent
research has uncovered a connection between these two processes. If one or more
grouping cues link a region with its bounding edge, then that region is more likely
to be treated as figural (see Palmer and Brooks 2008).29 If we suppose, then, that
figure-­ground organization involves perceptual grouping between regions and
edges, this permits a more precise characterization of what it is for an edge to be
assigned to a region. It is just for the region and the edge to be experientially repre-
sented as parts of a single individual—­a bounded surface. The edge is represented
as part of that surface (viz., its physical boundary).
Moreover, note that just as with perceptual grouping, figure-­ground organi-
zation is based on the visual detection of certain “special” properties, often called
figure-­ground cues. Thus, when a region is more convex, symmetric, or familiarly
shaped than the region it borders, it is more likely to be treated as figural (Rubin

28. I am using the term “edge” loosely here to apply both to contrast borders (i.e., locations where
color or luminance changes abruptly) and so-­called line-­edges. Line edges are the contours that
usually bound regions in artificial displays such as line drawings, but are very rarely found in
natural environments.
29. Suppose, for example, that a display contains two regions separated by a contrast border, and both
regions contain texture elements—­e.g., small dots. Suppose that the contrast border separating
the regions begins to move, and that the texture elements of one of the regions move with the
same speed and direction as the border. This gives rise to the experience that the region and the
edge are moving together (recall the grouping cue of common fate), and Palmer and Brooks have
found that it reliably leads the visual system to treat the region that moves along with the border
as figural, while treating the other (static) region as background. Palmer and Brooks found similar
effects on figural judgments when a region contained texture elements that could group with the
bounding edge by shared color, orientation, blur, or flicker timing.

145

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
1915/1958; Kanizsa and Gerbino 1976; Peterson 2003). I propose that we construe
these features as additional grouping cues. Thus, when a region is convex and bor-
dered by an edge, this serves as a cue for grouping the region and the edge as parts
of a single bounded surface. One reason why the grid is so unstable is that these
standard cues do not favor any of the squares over their neighbors.
To make this clearer, let’s contrast the content associated with seeing square
5 as figural with the content associated with seeing square 6 as figural. In the first
case, your experience represents the following:
Part-­of(e9 , a) & Part-­of(e13 , a) & Part-­of(e16 , a) & Part-­of(e12 , a) &
Part-­of(r5 , a)

In the second case, your experience represents the following:

Part-­of(e10 , a) & Part-­of(e14 , a) & Part-­of(e17 , a) & Part-­of(e13 , a) &
Part-­of(r6 , a)

The key difference is that the first experience fuses e13 with the edges and interior
region of square 5, while the second experience fuses e13 with the edges and inte-
rior region of square 6. The original grid, but not the modified grid in figure 11a,
involves changes in which edges are fused with which regions.
But why should changes in perceived depth be associated with changes in
edge assignment? Reflecting on the nature of occlusion, the answer is obvious.
When one opaque surface occludes another, the visible edge separating the visible
regions of the two surfaces is the physical boundary of the nearer (occluding) sur-
face, rather than the farther one. It is part of the nearer surface, and not part of the
farther one. Therefore, if the visual system interprets one region as being nearer
than an adjoining region, then the edge between them should be grouped with the
former region, not with the latter. It appears that this rule is built into the visual
system’s processing mechanisms (e.g., Nakayama et al. 1989).

Acknowledgments

I have benefited greatly from discussions with Brian McLaughlin, Jacob Feldman,
and an audience at the University of Nebraska-­Lincoln.

References

Baylis, G. C., and J. Driver. 1995. “One-­Sided Edge Assignment in Vision: 1. Figure-­Ground Segmentation
and Attention to Objects.” Current Directions in Psychological Science 4: 140–46.
Block, N. 2010. “Attention and Mental Paint.” Philosophical Issues 20: 23–63.
Bregman, A. S. 1990. Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, MA:
MIT Press.
Bronner, B. 2015. “Representationalism and the Determinacy of Visual Content. Philosophical Psychology
28(2): 227–39.
Byrne, A. 2001. “Intentionalism Defended.” Philosophical Review 110: 199–240.

146

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
Carey, S. 2009. The Origin of Concepts. Oxford: Oxford University Press.
Carrasco, M., S. Ling, and S. Read. 2004. “Attention Alters Appearance.” Nature Neuroscience 7: 308–13.
Carrasco, M., and A. Barbot. 2015. “How Attention Affects Spatial Resolution.” Cold Spring Harbor
Symposia on Quantitative Biology. DOI: [Link]
Chalmers, D. 2004. “The Representational Character of Experience.” In The Future for Philosophy,
ed. B. Leiter. Oxford: Oxford University Press.
Dawson, M. R. 1991. “The How and Why of What Went Where in Apparent Motion: Modeling
Solutions to the Motion Correspondence Problem.” Psychological Review 98: 569–603.
Dretske, F. 1995. Naturalizing the Mind. Cambridge, MA: MIT Press.
Feldman, J. 1999. “The Role of Objects in Perceptual Grouping.” Acta Psychologica 102: 137–63.
Feldman, J. 2003. “What Is a Visual Object?” Trends in Cognitive Sciences 7: 252–56.
Feldman, J. 2007. “Formation of Visual ‘Objects’ in the Early Computation of Spatial Relations.”
Perception & Psychophysics 69(5): 816–27.
Feldman, J., and P. D. Tremoulet. 2006. “Individuation of Visual Objects over Time.” Cognition 99:
131–65.
Flombaum, J. I., B. J. Scholl, and L. R. Santos. 2009. “Spatiotemporal Priority as a Fundamental Principle
of Object Persistence.” In The Origins of Object Knowledge, ed. B. Hood and L. Santos, 135–64.
Oxford: Oxford University Press.
Ganson T., and B. Bronner. 2013. “Visual Prominence and Representationalism.” Philosophical Studies
164(2): 405­18.
Green, E. J. 2015. “A Layered View of Shape Perception.” The British Journal for the Philosophy of Science.
DOI: 10.1093/bjps/axv042.
Harman, G. 1990. “The Intrinsic Quality of Experience.” Philosophical Perspectives 4: 31–52.
Hill, C. ms. “A Problem in Dretske’s Philosophy of Mind.”
Inui, T., S. Tanaka, T. Okada, S. Nishizawa, M. Katayama, and J. Konishi. 2000. “Neural Substrates for
Depth Perception of the Necker Cube: A Functional Magnetic Resonance Imaging Study in
Human Subjects.” Neuroscience Letters 282: 145–48.
Jagnow, R. 2011. “Ambiguous Figures and the Spatial Contents of Perceptual Experience: A Defense
of Representationalism.” Phenomenology and the Cognitive Sciences 10: 325–46.
Kanizsa, G., and W. Gerbino. 1976. “Convexity and Symmetry in Figure-­Ground Organization. In Vision
and Artifact, ed. M. Henle, 25–32. New York: Springer.
Kawabata, N. 1986. “Attention and Depth Perception.” Perception 15(5): 563–72.
Kimchi, R. 2009. “Perceptual Organization and Visual Attention.” Progress in Brain Research 176: 15–33.
Koffka, K. 1935. Principles of Gestalt Psychology. London: Lund Humphries.
Köhler, W. 1947. Gestalt Psychology. New York: Liveright Publishing.
Kubovy, M., A. O. Holcombe, and J. Wagemans. 1998. “On the Lawfulness of Grouping by Proximity.”
Cognitive Psychology 35: 71–98.
Kubovy, M., and M. van den Berg. 2008. “The Whole Is Equal to the Sum of Its Parts: A Probabilistic
Model of Grouping by Proximity and Similarity in Regular Patterns.” Psychological Review 115:
131–54.
Leopold, D. A., and N. K. Logothetis. 1999. “Multistable Phenomena: Changing Views in Perception.”
Trends in Cognitive Sciences 3(7): 254–64.
Leordeanu, M., M. Hebert, and R. Sukthankar. 2009. “Using Simple Features and Relations.” In Object
Categorization: Computer and Human Vision Perspectives, ed. S. Dickinson, A. Leonardis, B. Schiele,
and M. J. Tarr. Cambridge: Cambridge University Press.
Liu, T., J. Abrams, and M. Carrasco. 2009. “Voluntary Attention Enhances Contrast Appearance.”
Psychological Science 20(3): 354–62.
Lycan, W. G. 1998. “In Defense of the Representational Theory of Qualia (replies to Neander, Rey, and
Tye). Philosophical Perspectives 12: 479–87.
Macpherson, F. 2006. “Ambiguous Figures and the Content of Experience.” Noûs 40(1): 82–117.
Matthen, M. 2010. “On the Diversity of Auditory Objects.” Review of Philosophy and Psychology 1: 63–89.
Meng, M., and F. Tong. 2004. “Can Attention Selectively Bias Bistable Perception? Differences between
Binocular Rivalry and Ambiguous Figures.” Journal of Vision 4(7): 539–51.
Nakayama, K., Z. J. He, and S. Shimojo. 1995. “Visual Surface Representation: A Critical Link between
Lower-­Level and Higher-­Level Vision.” In Visual Cognition: An Invitation to Cognitive Science,
Volume 2, ed. S. M. Kosslyn and D. N. Osherson, 1–70. Cambridge, MA: MIT Press.
Nakayama, K., S. Shimojo, and G. H. Silverman. 1989. “Stereoscopic Depth: Its Relation to Image
Segmentation, Grouping, and the Recognition of Occluded Objects.” Perception 18: 55–68.
Nanay, B. 2010. “Attention and Perceptual Content.” Analysis 70: 263–70.

147

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]
Nickel, B. 2007. “Against Intentionalism.” Philosophical Studies 136(3): 279–304.
O’Callaghan, C. 2008. “Object Perception: Vision and Audition.” Philosophy Compass 3: 803–29.
Orlandi, N. 2011. “Ambiguous Figures and Representationalism.” Phenomenology and the Cognitive
Sciences 10(3): 307–23.
Palmer, S. E. 1977. “Hierarchical Structure in Perceptual Representation.” Cognitive Psychology 9:
441–74.
Palmer, S. E. 1999. Vision Science: Photons to Phenomenology. Cambridge, MA: MIT Press.
Palmer, S. E., and I. Rock. 1994. “Rethinking Perceptual Organization: The Role of Uniform
Connectedness.” Psychonomic Bulletin & Review 1: 29–55.
Palmer, S. E., and J. L. Brooks. 2008. “Edge-­Region Grouping in Figure-­Ground Organization and Depth
Perception.” Journal of Experimental Psychology 34(6): 1353–71.
Peacocke, C. 1983. Sense and Content. Oxford: Oxford University Press.
Peacocke, C. 1992. A Study of Concepts. Cambridge, MA: MIT Press.
Peterson, M. A. 2003. “On Figures, Grounds, and Varieties of Surface Completion.” In Perceptual
Organization in Vision: Behavioral and Neural Perspectives, ed. R. Kimchi, M. Behrmann, and
C. R. Olson, 87–116. Mahwah, NJ: Lawrence Erlbaum.
Pizlo, Z. 2008. 3D Shape: Its Unique Place in Visual Perception. Cambridge, MA: MIT Press.
Prinzmetal, W., I. Nwachuku, L. Bodanski, L. Blumenfeld, and N. Shimizu. 1997. “The Phenomenology
of Attention: 2. Brightness and Contrast.” Consciousness and Cognition 6: 372–412.
Pylyshyn, Z. W., and R. W. Storm. 1988. “Tracking Multiple Independent Targets: Evidence for a Parallel
Tracking Mechanism.” Spatial Vision 3: 179–97.
Raftopoulos, A. 2011. “Ambiguous Figures and Representationalism.” Synthese 181: 489–514.
Rubin, E. 1915/1958. “Figure and Ground.” In Readings in Perception, ed. D. C. Beardslee and
M. Wertheimer. Princeton, NJ: Van Nostrand.
Schellenberg, S. 2010. “The Particularity and Phenomenology of Perceptual Experience.” Philosophical
Studies 149(1): 19–48.
Scholl, B. J. 2007. “Object Persistence in Philosophy and Psychology.” Mind & Language 22: 563–91.
Scholl, B. J., and A. M. Leslie. 1999. “Explaining the Infant’s Object Concept: Beyond the Perception/
Cognition Dichotomy.” In What Is Cognitive Science?, ed. E. Lepore and Z. W. Pylyshyn, 26–73.
Oxford: Blackwell.
Scholl, B. J., and Z. W. Pylyshyn. 1999. “Tracking Multiple Items through Occlusion: Clues to Visual
Objecthood.” Cognitive Psychology 38(2): 259–90.
Shoemaker, S. 2001. “Introspection and Phenomenal Character.” Philosophical Topics 28(2): 247–73.
Soteriou, M. 2000. “The Particularity of Visual Experience.” European Journal of Philosophy 8: 173–89.
Speaks, J. 2010. “Attention and Intentionalism.” Philosophical Quarterly 60(239): 325–42.
Tse, P. U. 2005. “Voluntary Attention Modulates the Brightness of Overlapping Transparent Surfaces.”
Vision Research 45: 1095–98.
Tye, M. 1995. Ten Problems of Consciousness. Cambridge, MA: MIT Press.
Tye, M. 2002. “Visual Qualia and Visual Content Revisited.” In Philosophy of Mind: Classical and
Contemporary Readings, ed. D. Chalmers, 447–56. Oxford: Oxford University Press.
vanMarle, K., and B. J. Scholl. 2003. “Attentive Tracking of Objects versus Substances.” Psychological
Science 14: 498–504.
Vecera, S. P. 2000. “Toward a Biased Competition Account of Object-­Based Segregation and Attention.”
Brain and Mind 1: 353–84.
Vecera, S. P., A. V. Flevaris, and J. C. Filapek. 2004. “Exogenous Spatial Attention Influences Figure-­
Ground Assignment.” Psychological Science 15(1): 20–26.
Wagemans, J., J. H. Elder, M. Kubovy, S. E. Palmer, M. A. Peterson, M. Singh, and R. von der Heydt. 2012.
“A Century of Gestalt Psychology in Visual Perception: I. Perceptual Grouping and Figure-­Ground
Organization.” Psychological Bulletin 138: 1172–1217.
Wagemans, J., J. Feldman, S. Gepshtein, R. Kimchi, J. R. Pomerantz, P. A. van der Helm, and C. van Leeuwen.
2012. “A Century of Gestalt Psychology in Visual Perception: II. Conceptual and Theoretical
Foundations.” Psychological Bulletin 138: 1218–52.
Watzl, S. Forthcoming. “Can Intentionalism Explain How Attention Affects Appearances?” In Themes
from Block, ed. A. Pautz and D. Stoljar. Cambridge, MA: MIT Press.
Wertheimer, M. 1923/1958. “Principles of Perceptual Organization.” In Readings in Perception, ed.
D. C. Beardslee and M. Wertheimer. Princeton, NJ: Van Nostrand.
Wu, W. 2014. Attention. New York: Routledge.
Wynn, K., P. Bloom, and W.-­C. Chiang. 2002. “Enumeration of Collective Entities by 5-­Month-­Old
Infants.” Cognition 83: B55–B62.

148

This content downloaded from

[Link] on Wed, 01 Dec 2021 [Link] UTC
All use subject to [Link]