Cognitive Psychology and Its Implications, Ch. 4
4
Mental Imagery
Try answering these two questions:
• How many windows are in your house? • How many nouns are in the American Pledge of Allegiance?
Most people who answer these questions have the same experience. For the first
question they imagine themselves walking around their house and counting windows.
For the second question, if they do not actually say the Pledge of Alliance out loud,
they imagine themselves saying the Pledge of Allegiance. In both cases they are creating
mental images of what they would have perceived had they actually walked around
the house or said the Pledge of Allegiance.
Use of visual imagery is particularly important. As a result of our primate heritage,
a large portion of our brain functions to process visual information. Therefore, we use
these brain structures as much as we can, even in the absence of a visual signal from
the outside world, by creating mental images in our heads. Some of humankind’s most
creative acts involve visual imagery. For instance, Einstein claimed he discovered the
theory of relativity by imagining himself traveling beside a beam of light.
A major debate in this field of research has been the degree to which the processes
behind visual imagery are the same as the perceptual and attentional processes that we
considered in the previous two chapters. Some researchers (e.g., Pylyshyn, 1973, in an
article sarcastically titled “What the mind’s eye tells the mind’s brain”) have argued that
the perceptual experience that we have while doing an activity such as picturing the
windows in our house is an epiphenomenon; that is, it is a mental experience that does
not have any functional role in information processing. The philosopher Daniel Dennett
(1969) also argued that mental images are epiphenomenal—that is, that the perceptual
components of mental images are not really functional in any way:
Consider the Tiger and his Stripes. I can dream, imagine or see a striped tiger, but
must the tiger I experience have a particular number of stripes? If seeing or imagining
is having a mental image, then the image of the tiger must—obeying the rules of
images in general—reveal a definite number of stripes showing, and one should be
able to pin this down with such questions as “more than ten?”, “less than twenty?”
(p. 136)
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 92
Verbal Imagery versus Visual Imagery | 93
Dennett’s argument is that if we are actually seeing a tiger in a mental image, we
should be able to count its stripes just like we could if we actually saw a tiger.
Because we cannot count the stripes in a mental image of a tiger, we are not having
a real perceptual experience. This argument is not considered decisive, but it does
illustrate the discomfort some people have with the claim that mental images are
actually perceptual in character.
This chapter will review some of the experimental evidence showing the ways that
mental imagery does play a role in information processing. We will define mental
imagery broadly as the processing of perceptual-like information in the absence of an
external source for the perceptual information. We will consider the following questions: • How do we process the information in a mental image? • How is imaginal processing related to perceptual processing? • What brain areas are involved in mental imagery? • How do we develop mental images of our environment and use these
to navigate through the environment?
•Verbal Imagery versus Visual Imagery
There is increasing evidence from cognitive neuroscience that several different
brain regions are involved in imagery. This evidence has come both from studies
of patients suffering damage to various brain regions and from studies of the
brain activation of normal individuals as they engage in various imagery tasks.
In one of the early studies of brain activation patterns during imagery, Roland
and Friberg (1985) identified many of the brain regions that have been investigated
in subsequent research. They had participants either mentally rehearse a
word jingle or mentally rehearse finding their way around streets in their neighborhoods.
The investigators measured changes in blood flow in various parts of
the cortex. Figure 4.1 illustrates the principal areas they identified.When participants
engaged in the verbal jingle task, there was activation in the prefrontal cortex
near Broca’s area and in the parietal-temporal region of the posterior cortex
FIGURE 4.1 Results from
Roland and Friberg’s (1985)
study of brain activation
patterns during mental imagery.
Regions of the left cortex
showed increased blood flow
when participants imagined
a verbal jingle (J) or a spatial
route (R).
Brain Structures
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 93
near Wernicke’s area. As discussed in Chapter 1, patients with damage to these
regions show deficits in language processing. When participants engaged in the
visual task, there was activation in the parietal cortex, occipital cortex, and temporal
cortex. All these areas are involved in visual perception and attention, as
we saw in Chapters 2 and 3.When people process imagery of language or visual
information, some of the same areas are active as when they process actual
speech or visual information.
An experiment by Santa (1977) demonstrated the functional consequence
of representing information in a visual image versus representing it in a verbal
image. The two conditions of Santa’s experiment are shown in Figure 4.2. In
the geometric condition (Figure 4.2a), participants studied an array of three
geometric objects, arranged with one object centered below the other two.
This array had a facelike property—without much effort, we can see eyes and a
mouth. After participants studied the array, it was removed, and they had to
hold the information in their minds. They were presented with one of several
different test arrays. The participants’ task was to verify that the test array contained
the same elements as the study array, although not necessarily in the same
94 | Mental Imagery
Study
array
arrays
Test
Test
arrays
Study
array
Identical,
same configuration
Same elements,
linear configuration
Different elements,
same configuration
Different elements,
linear configuration
Triangle Circle
Square
Triangle Circle
Square
Triangle Circle Square
Triangle Circle
Arrow
Triangle Circle Arrow
Identical,
same configuration
Same word,
linear configuration
Different words,
same configuration
Different words,
linear configuration
(a) Geometric condition
(b) Verbal condition
FIGURE 4.2 The procedure followed in Santa’s (1977) experiment demonstrating that visual
and verbal information is represented differently in mental images. Participants studied an initial
array of objects or words and then had to decide whether a test array contained the same
elements. Geometric shapes were used in (a), words for the shapes in (b).
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 94
spatial configuration. Thus, participants should
have responded positively to the first two test
arrays and negatively to the last two. Santa was
interested in the contrast between the two positive
test arrays. The first was identical to the
study array (same-configuration condition). In
the second array, the elements were displayed
in a line (linear-configuration condition). Santa
predicted that participants would make a positive
identification more quickly in the first case,
where the configuration was identical—because,
he hypothesized, the mental image for the study
stimulus would preserve spatial information. The
results for the geometric condition are shown in
Figure 4.3. As you can see, Santa’s predictions were confirmed. Participants were
faster in their judgments when the geometric test array preserved the configuration
information in the study array.
The results from the geometric condition are more impressive when contrasted
with the results from the verbal condition, illustrated in Figure 4.2b.
Here, participants studied words arranged exactly as the objects in the geometric
condition were arranged. Because it involved words, however, the study stimulus
did not suggest a face or have any pictorial properties. Santa speculated that participants
would read the array left to right and top down and encode a verbal
image with the information. So, given the study array, participants would encode
it as “triangle, circle, square.” After they studied the initial array, one of the test
arrays was presented. Participants had to judge whether the words were identical.
All the test stimuli involved words, but otherwise they presented the same
possibilities as the test stimuli in the geometric condition. The two positive stimuli
exemplify the same-configuration condition and the linear-configuration
condition. Note that the order of words in the linear array was the same as it
was in the study stimulus. Santa predicted that, unlike the geometric condition,
because participants had encoded the words into a linearly ordered verbal image,
they would be fastest when the test array was linear. As Figure 4.3 illustrates,
his predictions were again confirmed.
Different parts of the brain are involved in verbal and visual imagery,
and they represent and process information differently.
•Visual Imagery
Most of the research on mental imagery has involved visual imagery, and this
will be the principal focus of this chapter. One function of mental imagery is to
anticipate how objects will look from different perspectives. People often have
the impression that they rotate objects mentally to achieve perspective. Roger
Shepard and his colleagues have been involved in a long series of experiments
Visual Imagery | 95
Geometric
Verbal
Reaction time (s)
1.25
1.15
Same
configuration
Linear
configuration
FIGURE 4.3 Results from
Santa’s (1977) experiment. The
data confirmed two of Santa’s
hypotheses: (1) In the geometric
condition, participants would
make a positive identification
more quickly when the configuration
was identical than when
it was linear, because the visual
image of the study stimulus
would preserve spatial information.
(2) In the verbal condition,
participants would make a
positive identification more
quickly when the configuration
was linear than when it was
identical, because participants
had encoded the words from
the study array linearly, in
accordance with normal reading
order in English.
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 95
on mental rotation. Their research was among the first to study the functional
properties of mental images, and it has been very influential. It is interesting to
note that this research was inspired by a dream (Shepard, 1967): Shepard awoke
one day and remembered having visualized a 3-D structure turning in space.
He convinced Jackie Metzler, a first-year graduate student at Stanford, to study
mental rotation, and the rest is history.
Their first experiment was reported in the journal Science (Shepard &
Metzler, 1971). Participants were presented with pairs of 2-D representations
of 3-D objects, like those in Figure 4.4. Their task was to determine whether the
objects were identical except for orientation. The two objects in Figure 4.4a
are identical, as are the two objects in Figure 4.4b, but in both cases the pairs
are presented at different orientations. Participants reported that to match the
two shapes, they rotated one of the objects in each pair mentally until it was
congruent with the other object. There is no way to rotate one of the objects in
Figure 4.4c so that it is identical with the other.
The graphs in Figure 4.5 show the times required for participants to decide
that the members of pairs were identical. The reaction times are plotted as a
function of the angular disparity between the two objects presented. The angular
disparity is the amount one object would have to be rotated to match the other
object in orientation. Note that the relationship is linear—for every increment
in amount of rotation, there is an equal increment in reaction time. Reaction
time is plotted for two different kinds of rotation. One is for 2-D rotations
(Figure 4.4a), which can be performed in the picture plane (i.e., by rotating the
page); the other is for depth rotations (Figure 4.4b), which require the participant
to rotate the object into the page. Note that the two functions are very
similar. Processing an object in depth (in three dimensions) does not appear
to have taken longer than processing an object in the picture plane. Hence,
participants must have been operating on 3-D representations of the objects in
both the picture-plane and depth conditions.
These data might seem to indicate that participants rotated the object in a
3-D space within their heads. The greater the angle of disparity between the two
objects, the longer participants took to complete the rotation. Though the
participants were obviously not actually rotating a real object in their heads,
the mental process appears to be analogous to physical rotation.
96 | Mental Imagery
(a) (b) (c)
FIGURE 4.4 Stimuli in the Shepard and Metzler (1971) study on mental rotation. (a) The
objects differ by an 80° rotation in the picture plane (two dimensions). (b) The objects differ
by an 80° rotation in depth (three dimensions). (c) The objects cannot be rotated into
congruence. (From Metzler & Shepard, 1974. Reprinted by permission of the publisher. © 1974 by Erlbaum.)
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 96
There has been a great deal of subsequent research examining the mental
rotation of all sorts of different objects. The typical finding is that the time
required to complete a rotation does vary with the angle of disparity. In recent
years, there have been a number of brain-imaging studies that looked at what
regions are active during mental rotation. Consistently, the parietal region
(roughly the region labeled R at the upper back of the brain in Figure 4.1) has
been activated across a range of tasks. This finding corresponds with the results
we reviewed in Chapter 3 showing that the parietal region is important in spatial
attention. Some tasks involve activation of other areas. For instance, Kosslyn,
DiGirolamo, Thompson, and Alpert (1998) found that imagining the rotation of
one’s hand produced activation in themotor cortex.
Neural recordings of monkeys have provided some evidence about neural
representation during mental rotation involving hand movement. Georgopoulos,
Lurito, Petrides, Schwartz, and Massey (1989) had monkeys perform a task in
which they moved a handle at a specific angle in response to a given stimulus. In
the base condition, monkeys just moved the handle to the position of the stimulus.
Georgopoulos et al. found cells that fired for particular positions. So, for
instance, there were cells that fired most strongly when the monkey was moving to
the 9 o’clock position and other cells that responded most stronglywhen the monkey
moved to the 12 o’clock position. In the rotation condition, the monkeys had
to move the handle to a position rotated some number of degrees from the stimulus.
For instance, if the monkeys had to move the handle 90° counterclockwise and
the stimulus appeared at the 12 o’clock position, they would have to move the
handle to 9 o’clock. If the stimulus appeared at the 6 o’clock position, they would
have to move to 3 o’clock. The greater the angle, the longer it took the monkeys
Visual Imagery | 97
Angle of rotation (degrees)
(a) (b)
0 40 80 120 160
Mean reaction (s)
Mean reaction (s)
0 40 80 120 160
FIGURE 4.5 Results of the Shepard and Metzler (1971) study on mental rotation. The mean
time required to determine that two objects have the same 3-D shape is plotted as a function
of the angular difference in their portrayed orientations. (a) Plot for pairs differing by a rotation
in the picture plane (two dimensions). (b) Plot for pairs differing by a rotation in depth (three
dimensions). (From Metzler & Shepard, 1974. Reprinted by permission of the publisher. © 1974 by Erlbaum.)
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 97
to initiate the movement, suggesting that this task involved a mental rotation
process to achieve the transformation. In this rotation condition, Georgopoulos
et al. found that various cells fired at different times during the transformation.At
the beginning of a transformation trial, when the stimulus was presented, the cells
that fired most were associated with a move in the direction of the stimulus.By the
end of a transformation trial, when the monkey actually moved the handle, maximumactivity
occurred in cells associated with the movement. Between the beginning
and the end of the trial, cells representing intermediate directions were most
active. These results suggest that mental rotation involves gradual shifts of firing
from cells that encode the initial stimulus to cells that encode the transformed
stimulus or, in this case, the transformed response.
When people must transform the orientation of a mental image to make
a comparison, they rotate its representation through the intermediate
positions until it achieves the desired orientation.
Image Scanning
Something else we often do with mental images is to scan them looking for some
critical information. For instance, when people are asked how many windows
there are in their houses (the task described at the beginning of this chapter),
many report mentally going through the house visually as they count the
windows. Researchers have been interested in the degree to which people are
actually scanning perceptual representations in such tasks, as opposed to just
retrieving abstract information. For instance, are we really “seeing” each window
in the room or are we just remembering how many windows are in the room?
Brooks (1968) performed an important series of experiments on the scanning
of visual images. He had participants scan imagined diagrams such as the
one shown in Figure 4.6. For example, the participant was to scan around an
imagined block F from a prescribed starting point and in a prescribed direction,
categorizing each corner of the block as a point in the top or bottom (assigned a
yes response) or as a point in between (assigned a no response). In the example
(beginning with the starting corner), the correct sequence of responses is yes, yes,
yes, no, no, no, no, no, no, yes. For a nonvisual contrast task, Brooks also gave
participants sentences such as “A bird in the hand is not in the bush.” Participants
had to scan the sentence while holding it in memory, deciding whether
each word was a noun or not. A second experimental variable was how participants
made their responses. Participants responded in one of three ways:
(1) said yes or no; (2) tapped with the left hand for yes and with the right hand
for no; or (3) pointed to successive Y’s or N’s on a sheet of paper such as the
one shown in Figure 4.7. The two variables of stimulus material (diagram or
sentence) and output mode were crossed to yield six conditions.
Table 4.1 gives the results of Brooks’s experiment in terms of the mean
time spent in classifying the sentences or diagrams in each output condition.
The important result for our purposes is that participants took much longer
for diagrams in the pointing condition than in any other condition, but this
was not the case when participants were working with sentences. Apparently,
98 | Mental Imagery
FIGURE 4.6 An example of a
simple block diagram that Brooks
used to study the scanning of
mental images. The asterisk and
arrow show the starting point
and the direction for scanning
the image. (From Brooks, 1968.
Reprinted by permission of the publisher.
© 1968 by the Canadian Psychological
Association.)
Anderson7e_Chapter_04.qxd 8/21/09 6:58 PM Page 98
scanning a physical visual array conflicted with scanning a mental array.
This result strongly reinforces the conclusion that when people are scanning
a mental array, they are scanning a representation that is analogous
to a physical visual array. Requiring the person simultaneously to engage
in a conflicting scanning action on an external physical visual array disrupts
the mental scan.
One might think that Brooks’s result was due to the conflict between
engaging in a visual pointing task and scanning a visual image.
Subsequent research makes it clear, however, that the interference is not a
result of the visual character of the task per se. Rather, the problem is spatial
and not specifically visual; it arises from the conflicting directions in
which participants had to scan the physical visual array and the mental
image. For instance, in another experiment, Brooks found evidence of
similar interference when participants had their eyes closed and indicated
yes or no by scanning an array of raised Y’s and N’s with their fingers. In
this case, the actual stimuli were tactile, not visual. Thus, the conflict is
spatial, not specifically visual.
Baddeley and Lieberman (reported in Baddeley, 1976) performed an
experiment that further supports the view that the nature of the interference
in the Brooks task is spatial rather than visual. Participants were
required to perform two tasks simultaneously. All participants performed
the Brooks letter-image task. However, participants in one group simultaneously
monitored a series of stimuli of two possible brightnesses and had
to press a key whenever the brighter stimulus appeared. This task involved
the processing of visual but not spatial information. Participants in the
other condition were blindfolded and seated in
front of a swinging pendulum. The pendulum
emitted a tone and contained a photocell. Participants
were instructed to try to keep the beam of
a flashlight on the swinging pendulum. Whenever
they were on target, the photocell caused the
tone to change frequency, thus providing auditory
feedback. This test involved the processing
of spatial but not visual information. The spatial
auditory tracking task produced far greater
impairment in the image scanning task than did
the brightness judgment task. This result also
indicates that the nature of the impairment in
the Brooks task was spatial, not visual.
People suffer interference in scanning a mental image if they have to
simultaneously process a conflicting perceptual structure.
Visual Comparison of Magnitudes
A fair amount of research has focused on the way people judge the visual details
of objects in their mental images. One line of research has asked participants
Visual Imagery | 99
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
N
N
N
N
N
N
N
N
N
N
N
FIGURE 4.7 A sample output sheet of
the pointing condition in Brooks’s study
of mental image scanning. The letters
are staggered to force careful visual
monitoring of pointing. (From Brooks, 1968.
Reprinted by permission of the publisher. © 1968
by the Canadian Psychological Association.)
TABLE 4.1
Results of Brooks’s (1968) Experiment Showing Conflict
Between Mental Array and Visual Array Scanning
Mean Response Time (s)
by Output Mode
Stimulus Material Pointing Tapping Vocal
Diagrams 28.2 14.1 11.3
Sentences 9.8 7.8 13.8
From Brooks, 1968. Reprinted by permission of the publisher. © 1968 by the
Canadian Psychological Association.
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 99
to discriminate between objects based on some dimension such as size. This
research has shown that when participants try to discriminate between two
objects, the time it takes them to do so decreases continuously as the difference
in size between the two objects increases.
Moyer (1973) was interested in the speed with which participants could
judge the relative size of two animals from memory. For example, “Which is
larger, moose or roach?” and “Which is larger, wolf or lion?”Many people report
that in making these judgments, particularly for the items that are similar in
size, they experience images of the two objects and seem to compare the sizes
of the objects in their images.
Moyer also asked participants to estimate the absolute size of these animals.
He plotted the time required to compare the imagined sizes of two animals
as a function of the difference between the two animals’ estimated sizes.
Figure 4.8 reproduces these data. The individual points represent
comparisons between pairs of items. In general, the
judgment times decreased as the difference in estimated size
increased. The graph shows that judgment time decreases linearly
with increases in differences in the size of the animals.
Note, however, that the differences have been plotted logarithmically.
A log-difference scale makes the distance between small
differences large relative to the same distances between large
differences. Thus, the linear relationship in the graph means
that increasing the size difference has a diminishing effect on
reaction time.
Significantly, very similar results are obtained when people
visually compare physical size. For instance, D. M. Johnson
(1939) asked participants to judge which of two simultaneously
presented lines was longer. Figure 4.9 plots participant judgment
time as a function of the log difference in line length. Again, a
linear relation is obtained. It is reasonable to expect that the
100 | Mental Imagery
1.4
Mean reaction time (s)
Difference in line length (mm)
2.3
3.2
2.9
1.7
2.0
2.6
1 2 3 4 6 8
1.5
1.4
1.3
1.2
1.1
1.0
0.9
0.8
0.10 1.10 2.10
Estimated difference in animal size
Mean reaction time (s)
FIGURE 4.8 Results from Moyer’s experiment demonstrating that
when people try to discriminate between two objects, the time it
takes them to do so decreases as the difference in size between
the two objects increases. Participants were asked to compare the
imagined sizes of two animals. The mean time required to judge
which of two animals is larger is plotted as a function of the estimated
difference in size of the two animals. The difference measure
is plotted on the abscissa in a logarithmic scale. (From Moyer, 1973.
Reprinted by permission of the publisher. © 1973 by Perception and Psychophysics.)
FIGURE 4.9 Results from the
D. M. Johnson (1939) study
demonstrating that making
mental comparisons involves
difficulties of discrimination
similar to those involved in
making perceptual comparisons.
Participants compared the
lengths of two lines. The mean
time required to judge which
line was longer is plotted as a
function of the difference in line
length. The difference measure
is plotted on the abscissa in
a logarithmic scale.
Anderson7e_Chapter_04.qxd 8/20/09 9:42 AM Page 100
more similar the lengths being compared are, the longer perceptual judgments
will take, because telling them apart is more difficult under such circumstances.
The fact that similar functions are obtained when mental objects are compared
indicates that making mental comparisons involves the same process as those
involved in perceptual comparisons.
People experience greater difficulty in judging the relative size of two pictures
or of two mental images that are similar in size.
Are Visual Images Like Visual Perception?
Research has been done to determine to what degree visual imagery is like
perception. In one experiment, Finke, Pinker, and Farah (1989) set out to
determine the degree to which people were able to recognize the images they
constructed in the same way they would be able to recognize objects they saw.
These researchers asked participants to create mental images and then engage
in a series of transformations of those images. Here are two examples of the
problems that they read to their participants:
• Imagine a capital letter N. Connect a diagonal line from the top right
corner to the bottom left corner. Now rotate the figure 90° to the right.
What do you see? • Imagine a capital letter D. Rotate the figure 90° to the left. Now place a
capital letter J at the bottom.What do you see?
Participants closed their eyes and tried to imagine these transformations as
they were read to them. The participants were able to recognize their composite
images just as if they had been presented with them on a screen. In the first
example, they saw an hourglass; in the second, an umbrella. The ability to
perform such tasks illustrates an important function of imagery: It enables us
to construct new objects in our minds and inspect them. It is just this sort of
visual synthesis that structural engineers or architects must perform as they
design new bridges or buildings.
The Finke et al. (1989) study shows that we can make judgments about
imagined objects and come to the same conclusions as we would about seen objects.
But are these imagined objects really the same as seen objects? Researchers
have been doing studies to see whether subtle properties associated with visual
perception are also reproduced in mental images. For instance, Wallace (1984)
did an experiment to study the Ponzo illusion (Berbaum & Chung,
1981) illustrated in part (a) of Figure 4.10: Even though the two horizontal
lines have the same length, the upper one seems longer. In the
experiment, participants were first shown the converging lines in
part (b) and then shown the two horizontal lines in part (c) and
asked to imagine them together so that the top horizontal line
touched the two diagonals. They were asked to compare the length
of the two horizontal lines in their mental images. Participants rated
as high in imagery ability reported as strong an illusion for their
mental image as for the actual stimulus in part (a). Thus, it appears
Visual Imagery | 101
(a) (b) (c)
FIGURE 4.10 The Ponzo illusion
is illustrated in part (a): Even
though they have the same
length, the upper horizontal
line seems longer. Parts (b) and
(c) illustrate the material Wallace
(1984) used to study the illusion
in imagery. Participants imagined
the parts superimposed so
that the diagonal lines in
(b) just touched the horizontal
lines in (c). (From Wallace, 1984.
Reprinted by permission of the publisher.
© 1984 by the Psychonomic Society.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 101
that the imagery system can reproduce a detailed visual illusion, supporting
the apparent equivalence between imagery and perception.
Chambers and Reisberg (1985) reported a study that seemed to
indicate differences between a mental image and visual perception of the
real object. Their research involved the processing of reversible figures,
such as the duck-rabbit shown in Figure 4.11. Participants were briefly
shown the figure and asked to form an image of it. They had only
enough time to form one interpretation of the picture before it was removed,
but they were asked to try to find a second interpretation. Participants
were not able to do this. Then they were asked to draw the image
on paper to see whether they could reinterpret it. In this circumstance,
they were successful. This result suggests that visual images differ from pictures
in that one can interpret visual images only in one way, and it is not possible to
find an alternative interpretation of the image.
Subsequently, Peterson, Kihlstrom, Rose, and Gilsky (1992) were able to get
participants to reverse mental images by giving them more explicit instructions.
For instance, participants might be told how to reverse another figure or be
given the instruction to consider the back of the head of the animal in their
mental image as the front of the head of another animal. Thus, it seems apparent
that although it may be more difficult to reverse an image than a picture,
both can be reversed. In general, it seems harder to process an image than the
actual stimulus. Given a choice, people will almost always choose to process an
actual picture rather than imagine it. For instance, players of Tetris prefer to
rotate shapes on the screen to find an appropriate orientation rather than rotate
them mentally (Kirsh & Maglio, 1994).
It is possible to make many of the same kinds of detailed judgments about
mental images that we make about things we actually see, though it is more
difficult.
Visual Imagery and Brain Areas
Brain imaging studies indicate that the same regions are involved in perception
as in mental imagery. As already noted, the parietal regions that are involved in
attending to locations and objects (last chapter) are also involved in mental
rotation. A study by O’Craven and Kanwisher (2000) illustrates how closely the
brain areas activated by imagery correspond to the brain areas activated by
perception. Recall from Chapter 2 that the fusiform region of the temporal cortex
responds preferentially to faces. There is another region of the temporal
cortex, the parahippocampal place area (PPA), that responds preferentially to
pictures of locations. O’Craven and Kanwisher asked participants either to view
faces and scenes or to imagine faces and scenes. The same areas were active
when the participants were seeing as when they were imagining. Figure 4.12
compares activation in the fusiform face area (FFA) with activation in the PPA.
Every time the participants viewed or imagined a face, there was increased activation
in the FFA, and this activation went away when they processed places.
Conversely, when they viewed or imagined scenes, there was activation in the
102 | Mental Imagery
FIGURE 4.11 The ambiguous
duck-rabbit figure used in
Chambers and Reisberg’s study
of the processing of reversible
figures. (From Chambers & Reisberg,
1985. Reprinted by permission of the
publisher. © 1985 by the American
Psychological Association.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 102
PPA that went away when they processed faces. The responses during imagery
were very similar to the responses during perception, although a little weaker.
The fact that the response was weaker during imagery is consistent with the
behavioral evidence we have viewed suggesting that it is more difficult to process
an image than a real perception.
There are many studies like these that show the involvement of cortical
regions that are further along in the visual processing stream, but the evidence is
less clear about activation in the primary visual cortex (areas 17 and 18) where
visual information first reaches the brain. The O’Craven and Kanwisher study
found activation in the primary visual cortex during imagery. Such results are
important because they suggest that visual imagery includes relatively low-level
perceptual processes. However, activation has not always been found in the
primary visual cortex. For instance, the Roland and Friberg study illustrated
in Figure 4.1 did not find activation in this region (see also Roland, Eriksson,
Stone-Elander, & Widen, 1987). Kosslyn and Thompson (2003) reviewed
59 brain-imaging studies that looked for activation in early visual areas. About
half of these studies find activation in early visual areas and half do not.
Visual Imagery | 103
−0.5
0.0
Signal change (%) Signal change (%)
0.5
1.0
1.5
2.0
−1.0
−0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Perception Imagery Perception Imagery
FFA
PPA
FIGURE 4.12 Results from the O’Craven and Kanwisher study showing that visual images are
processed in the same way as actual perceptions and by many of the same neural structures.
Although participants alternately perceived (or imagined) faces and places, brain activation
was seen in the fusiform face area (FFA, upper panel) and the parahippocampal place area
(PPA, lower panel). (From O’Craven & Kanwisher, 2000. Reprinted by permission of the publisher. © 2000 by the
Journal of Cognitive Neuroscience.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 103
Their analysis suggests that the studies that find activation in these early visual
areas tend to emphasize high-resolution details of the images and tend to focus
on shape judgments. As an instance of one of the positive studies, Kosslyn et al.
(1993) did find activation in area 17 in a study where participants were asked
to imagine block letters. In one of their experiments, participants were asked to
imagine large versus small letters. In the small-letter condition, activity in the
visual cortex occurred in a more posterior region closer to where the center of
the visual field is represented (recall that the visual field is topographically
organized—see Figure 2.6). This would make sense because a small image
would be more concentrated at the center of the visual field.
Imaging studies like these show that perceptual regions of the brain are
active when participants are engaged in mental imagery, but they do not
establish whether these regions are actually critical to imagery. To return to the
epiphenomenon critique at the beginning of the chapter, it could be that the
activation plays no role in the actual tasks being performed. A number of experiments
now have used transcranial magnetic stimulation (TMS—see Figure
1.13) to investigate the causal role of these regions in the performance of
the underlying task. For instance, Kosslyn et al. (1999) presented participants
with 4-quadrant arrays like those in Figure 4.13 and asked them to form a mental
image of the array. Then, with the array removed, participants had to use
their image to answer questions like “Which has longer stripes: Quadrant 1 or
Quadrant 2?” or “Which has more stripes: Quadrant 1 or Quadrant 4?” Application
of TMS to primary visual area 17 significantly increased the time they
took to answer these questions. Thus, it seems that these visual regions do play
a causal role in mental imagery, and temporarily deactivating them results in
impaired information processing.
When performing a visual imFagery task, there is activation in brain regions
involved in visual perception, and disruption of these regions results in
disruption of the imagery task.
104 | Mental Imagery
FIGURE 4.13 Illustration of stimuli used in Kosslyn et al. (1999).
The numbers 1, 2, 3, and 4 were used to label the four quadrants,
each of which contained a set of stripes. After memorizing the
display, the participants closed their eyes, visualized the entire
display, heard the names of two quadrants, and then heard
the name of a comparison term (for example, “length”); the
participants then decided whether the stripes in the first-named
quadrant had more of the named property than those in the
second.
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 104
Imagery Involves Both Spatial and Visual Components
We have reviewed the Brooks (1968) research indicating that people are sensitive
to the spatial structure of their mental images and the Moyer (1973) research
indicating that people process visual details such as size in their mental images.
The distinction between the spatial and visual attributes of imagery has proved
important in recent research.We can encode the position of objects in space by
seeing where they are, by feeling where they are, or by hearing where they are.
Such encodings use a common spatial representation that integrates information
that comes in from any sensory modality. On the other hand, certain aspects of
visual experience, such as color, are unique to the visual modality and seem separate
from spatial information. Thus, imagery involves both spatial and visual
components. In the discussion of the visual system in Chapter 2, we reviewed
the evidence that there is a “where” pathway for processing spatial information
and a “what” pathway for processing object information (see Figure 2.1). Corresponding
to this distinction, there is evidence that the parietal regions support
the spatial component of visual imagery, whereas the temporal lobe supports the
visual aspects (see Figure 4.1). We have already noted that mental rotation, a
spatial task, tends to produce activation in the parietal cortex. Similarly, temporal
structures are activated when people imagine visual properties of objects
(Thompson & Kosslyn, 2000).
Patient studies of imagery also support this association of spatial imagery
with parietal areas of the brain and visual imagery with temporal areas.
Levine, Warach, and Farah (1985) compared two patients, one who suffered
bilateral parietal-occipital damage and the other who suffered bilateral inferior
temporal damage. The patient with parietal damage could not describe the
locations of familiar objects or landmarks from memory, but he could
describe the appearance of objects. The patient with temporal damage had
an impaired ability to describe the appearance of objects but could describe
their locations.
Farah, Hammond, Levine, and Calvanio (1988) carried out more detailed
testing of the patient with temporal damage. They compared his performance
on a wide variety of imagery tasks to that of normal participants. They found
that he showed deficits on only a subset of these tasks: ones in which he had to
judge color (“What is the color of a football?”), sizes (“Which is bigger, a popsicle
or a pack of cigarettes?”), the lengths of animals’ tails (“Does a kangaroo
have a long tail?”), and whether two states in the United States had similar
shapes. In contrast, he did not show any deficit in performing tasks that seemed
to involve a substantial amount of spatial processing: mental rotation, image
scanning, letter scanning (as in Figure 4.7), or judgments of where one U.S.
state was relative to another state. Thus, temporal damage seemed to affect only
those imagery tasks that required access to visual detail, not those that required
spatial judgments.
Neuropsychological evidence suggests that imagery of spatial information
is supported by parietal structures, and imagery of objects and their visual
properties is supported by temporal structures.
Visual Imagery | 105
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 105
Cognitive Maps
Another important function of visual imagery is to help us figure out and
remember the spatial structure of our environment. Our imaginal representations
of the world are often referred to as cognitive maps. The connection
between imagery and action is particularly apparent in cognitive maps. We
often find ourselves imagining our environment as we plan how we will get
from one location to another.
An important distinction can be made betweeen routemaps and survey maps
(Hart &Moore, 1973).A route map is a path that indicates specific places but contains
no 2-D information. It can even be a verbal description of a path (“Straight
until the light, then turn left, two blocks later at the intersection . . .”). Thus,with a
pure route map, if your route from location 1 to location 2 were blocked, you
would have no general idea of where location 2 was, and so youwould be unable to
construct a detour.Also, if you knew (in the sense of a route map) two routes from
a location, you would have no idea whether these routes formed a 90° angle or a
120° angle with respect to each other. A survey map, in contrast, contains this information,
and is basically a spatial image of the environment.When you ask for
directions from typical online mapping services, they will provide both a route
map and a survey map to support both mental representations of space.
Thorndyke and Hayes-Roth (1982) investigated workers’ knowledge of the
Rand Corporation Building (Figure 4.14), a large, maze-like building in Santa
106 | Mental Imagery
Northwest
lobby
Computer center
Administrative
conference
room
East lobby
Cashier
Snack bar
Common
room
South
lobby
First-floor building plan
The Rand Corporation
Supply
room
Feet
0 50 100
N
S
W E
FIGURE 4.14 The floor plan for part of the Rand Corporation Building in Santa Monica, California.
Thorndyke and Hayes-Roth studied the ability of secretaries to find their way around the building.
(From Thorndyke and Hayes-Roth, 1982. Reprinted by permission of the publisher. © 1982 by Cognitive Psychology.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 106
Monica, California. People in the Rand Building quickly acquire the ability to
find their way from one specific place in the building to another—for example,
from the supply room to the cashier. This knowledge represents a route map.
Typically, though, workers had to have years of experience in the building before
they could make such survey-map determinations as the direction of the lunchroom
from the photocopy room.
Hartley, Maguire, Spiers, and Burgess (2003) looked at the different brain
activity when people used these two types of representations. They had participants
navigate virtual reality towns under one of two conditions. In one condition,
called route-following, participants learned to follow a fixed path through
the town. In the other condition, called way-finding, participants first freely
explored the town and then had to find their way between locations. The fMRI
activation was compared when engaged in the route-following or way-finding
tasks. The results on the experiment are illustrated in Color Plate 4.1. In the
way-finding task, participants showed greater activation in a number of regions
found in other studies of visual imagery including the parietal cortex. Also,
they found greater activation in the hippocampus (see Figure 1.7), a region that
has been implicated in navigation in many species. In contrast, in the routefollowing
task participants showed greater activation in more anterior regions
motor regions. It would seem that the survey map is more like a visual image
and the route map is more like an action plan. This is a distinction that is supported
in other fMRI studies of route maps versus survey maps (e.g., Shelton &
Gabrieli, 2002).
Our knowledge of our environment can be represented in either survey maps
that emphasize spatial information or route maps that emphasize action
information.
Egocentric and Allocentric Representations of Space
Navigation becomes difficult when we must tie together multiple different
representations of space. In particular, we often need to relate the way space
appears as we perceive it to some other representation of space, such as a cognitive
map. The representation of “space as we see it” is referred to as an egocentric
representation. Figure 4.15 illustrates an egocentric representation that one might
have upon entering Camp Snoopy at the Mall of America in Bloomington,
Minnesota. Even young children have little difficulty understanding how to navigate
in space as they see it—if they see an object they want, they go for it. Problems
arise when one wants to relate what one sees to such representations of the
space as cognitive maps, be they route maps or survey maps. Similar problems
arise when one wants to deal with physical maps, such as the map of Camp
Snoopy shown in Figure 4.16. This kind of map is referred to as an allocentric
representation because it is not specific to a particular viewpoint, though, as is
true of most maps, north is oriented to the top of the image. Try to identify which
building on the map corresponds to the circled building in Figure 4.15, assuming
Visual Imagery | 107
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 107
108 | Mental Imagery
FIGURE 4.16 An allocentric
representation of Camp Snoopy.
(Courtesy of Camp Snoopy/Bob Cole.)
that the viewpoint is the one of the stick figure illustrated
in Figure 4.16.When people try to make such judgments,
the degree to which the map is rotated from their actual
viewpoint has a large effect. Indeed, people will often
rotate a physical map so that it is oriented to correspond
to their point of view. The map in Figure 4.16 would have
to be rotated approximately 90º counterclockwise to be
oriented with the representation shown in Figure 4.15.
When it is not possible to rotate a map physically,
people show an effect of the degree of misorientation that
is much like the effect we see for mental rotation (e.g.,
Boer, 1991; Easton & Sholl, 1995; Gugerty, deBoom,
Jenkins, & Morley, 2000; Hintzman, O’Dell, & Arndt,
1981). Figure 4.17 shows a recent result from a study by Gunzelmann and
Anderson (2002), who looked at the time required to find an object on a standard
map (i.e., north oriented to the top) as a function of the viewer’s location.
When the viewer is located to the south, looking north, it is easier to find the
object than when the viewer is north looking south, just the opposite of the
map orientation. Some people describe imagining themselves moving around
the map, others talk about rotating what they see, and still others report using
verbal descriptions (“to the left of Snoopy, before the Ferris wheel”). The fact
that the angle of disparity in this task has as great an effect as it does in mental
rotation has led many researchers to believe that the processes and representations
involved in such navigational tasks are similar to the processes and representations
involved in mental imagery.
Physical maps seem to differ from cognitive maps in one important way:
Physical maps show the effects of orientation, and cognitive maps do not. For
FIGURE 4.15 An egocentric
representation of Camp Snoopy
at the Mall of America. (Courtesy
of Camp Snoopy/Bob Cole.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 108
example, imagine yourself standing against various walls of your bedroom, and
point to the location of the front door of your home or apartment. Most people
can do this equally well no matter which position they take. In contrast, when
given a map like the one shown in Figure 4.16, people find it much easier to point
to various objects on the map if they are oriented in the same way the map is.
Recordings from single cells in the hippocampal region (inside the temporal
lobe) of rats suggest that the hippocampus plays an important role in maintaining
an allocentric representation of the world. There are place cells in the
hippocampus that fire maximally when the animal is in a particular location in
its environment (O’Keefe & Dostrovsky, 1971). Similar cells have been found in
recordings from human patients in a procedure to map out the brain before surgery
to control epilepsy (Ekstrom et al., 2003). Brain-imaging studies have shown
that when humans are navigating their environment, there is high hippocampal
activation (Maguire et al., 1998). Another study (Maguire et al., 2000) showed
that the hippocampal volume of London taxi drivers was greater than that of
people who didn’t drive taxis and increased with the length of time they had been
taxi drivers. It takes about 3 years of hard training to gain a knowledge of London
streets adequate to become a successful taxi driver, and this training appears to
have an impact on the structure of the brain. The amount of activation in hippocampal
structures has also been shown to correlate with age-related differences
in navigation skills (Pine et al., 2002) and may relate to gender differences in
navigational ability (Gron, Wunderlich, Spitzer, Tomczak, and Riepe, 2000).
These regions include the PPA, which O’Craven and Kanwisher (2000) found
active when participants imagined places (see Figure 4.12).
Although it appears that the hippocampus is important in supporting
allocentric representations, it seems that the parietal cortex is particularly
important in supporting egocentric representations (Burgess, 2006). In one
fMRI imaging study comparing egocentric and allocentric spatial processing
(Zaehle et al., 2007), participants where asked to make judgments that emphasized
either an allocentric or an egocentric perspective. In the allocentric conditions,
participants would read a description like “The blue triangle is to the left
of the green square. The green square is above the yellow triangle. The yellow
Visual Imagery | 109
S SW W NW N NE E SE S
0.0
0.5
1.5
1.0
2.0
2.5
3.0
3.5
Time to find object on map (s)
Direction in which viewer is looking
FIGURE 4.17 Results from
Gunzelmann and Anderson’s
study to determine how much
effect the angle of disparity
between a standard map
(looking north) and the viewer’s
viewpoint has on people’s
ability to find an object on
the map. The time required for
participants to identify the
object is plotted as a function
of the difference in orientation
between the map and the
egocentric viewpoint. (After
Gunzelmann & Anderson, 2002. Adapted
by permission of the publisher. © 2002
by Erlbaum.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 109
triangle is to the right of the red circle.” Then they would be asked a question
like “Is the blue triangle above the red circle?” In the egocentric condition, they
would read a description like “The blue circle is in front of you. The yellow
circle is to your right. The yellow square is to the right of the yellow circle.”
They would then be asked a question like “Is the yellow square to your right?”
There was greater hippocampal activation when participants were answering
questions in the allocentric condition than in the egocentric condition.
Although there was considerable parietal activation in both conditions, the egocentric
condition was distinguished by greater parietal activation.
Our representation of space includes both allocentric representations of
where objects are in the world and egocentric representations of where they
are relative to ourselves.
Map Distortions
Our mental maps often have a hierarchical structure in which smaller regions
are organized within larger regions. For instance, the structure of my bedroom
is organized within the structure of my house, which is organized within the
structure of my neighborhood, which is organized within the structure of Pittsburgh.
Consider your mental map of the United States. It is probably divided
into regions, and these regions into states, and cities are presumably pinpointed
within the states. It turns out that certain systematic distortions arise because
of the hierarchical structure of these mental maps. Stevens and Coupe (1978)
documented a set of common misconceptions about North American geography.
Consider the following questions taken from their research:
• Which is farther east: San Diego or Reno? • Which is farther north: Seattle or Montreal? • Which is farther west: the Atlantic or the Pacific entrance to the Panama
Canal?
The first choice is the correct answer in each case, but most people hold the
wrong opinion. Reno seems to be farther east because Nevada is east of California,
but this reasoning does not account for the westward curve in California’s
coastline. Montreal seems to be north of Seattle because Canada is north of the
United States, but the border dips south in the east. And the Atlantic is certainly
east of the Pacific—but consult a map if you need to be convinced about the
location of the entrances to the Panama Canal. The geography of North America
is quite complex, and people resort to abstract facts about relative locations of
large physical bodies (e.g., California and Nevada) to make judgments about
smaller locations (e.g., San Diego and Reno).
Stevens and Coupe were able to demonstrate such confusions with
experimenter-created maps. Different groups of participants learned the maps
illustrated in Figure 4.18. The important feature of the incongruent maps is that
the relative locations of the Alpha and Beta counties are inconsistent with the
locations of the X and Y cities. After learning the maps, participants were asked a
series of questions about the locations of cities, including “Is X east or west of Y?”
110 | Mental Imagery
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 110
Visual Imagery | 111
for the left-hand maps and “Is X north or south of Y?” for the right-hand maps.
On these questions, participants were in error 18% of the time for the congruent
maps and 15% percent of the time for the homogeneous maps—but 45% of the
time for the incongruent maps. Participants were using information about the
locations of the counties to help them remember the city locations. This reliance
Dimension tested
Alpha
County
Beta
County
Congruent
Horizontal
Vertical
Alpha
County
Beta
County
X Y
Alpha
County
Beta
County
Alpha
County
Beta
County
Homogeneous
Incongruent
FIGURE 4.18 Maps studied by participants in the experiments of Stevens and Coupe, which
demonstrated the effects of higher order information (location of county lines) on participants’ recall
of city locations. (From Stevens & Coupe, 1978. Reprinted by permission of the publisher. © 1978 by Cognitive Psychology.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 111
on higher order information led them to make errors, just as similar reasoning
can lead to errors in answering questions about North American geography.
When people have to work out the relative positions of two locations, they
will often reason in terms of the relative positions of larger areas that contain
the two locations.
Translating Words to Images
Another important use of spatial cognition involves the creation of a mental
image when we hear a word description of a spatial structure. This happens, for
instance, when we process directions, read a description of an event, or hear a
sportscast. The British psychologist Alan Baddeley reported that he was not able
to drive while trying to follow an American football game on the radio because
of the conflicting imagery. To get an idea of how difficult it can be to construct
a spatial image from words, try reading the description in Table 4.2.
Franklin and Tversky (1990) presented participants with such stories. After
the participants read the stories, they were asked to reorient themselves based
on descriptions such as
As you remain where you are on the balcony, you turn your body 90 degrees
to your right, and you now face the lamp. You look again at the short, rigid
pole by which it is fixed to the wall. Perhaps this is a precautionary feature in
case of earthquake.
Then they were asked to judge what was in a specific direction. For instance,
on different trials they would be asked what is on the right, on the left, above,
below, behind, or in front of them.
112 | Mental Imagery
TABLE 4.2
Description Read by Participants in Franklin and Tversky’s (1990) Experiment
You are hob-nobbing at the opera. You came tonight to meet and chat with
interesting members of the upper class. At the moment, you are standing next to
the railing of a wide, elegant balcony overlooking the first floor. Directly behind you,
at your eye level, is an ornate lamp attached to the balcony wall. The base of the
lamp, which is attached to the wall, is gilded in gold. Straight ahead of you, mounted
on a nearby wall beyond the balcony, you see a large bronze plaque dedicated to
the architect who designed the theatre. A simple likeness of the architect, as well as
a few sentences about him, are raised slightly against the bronze background. Sitting
on the shelf directly to your right is a beautiful bouquet of flowers. You see that the
arrangement is largely composed of red roses and white carnations. Looking up,
you see that a large loudspeaker is mounted to the theatre’s ceiling about 20 feet
directly above you. From its orientation, you suppose that it is a private speaker for
the patrons who sit in this balcony. Leaning over the balcony’s railing and looking
down, you see that a marble sculpture stands on the first floor directly below you.
As you peer down toward it, you see that it is a young man and wonder whether it
is a reproduction of Michelangelo’s David.
From Franklin & Tversky (1990). Reprinted by permission of the publisher. © 1990 by Journal of Experimental
Psychology: General.
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 112
Figure 4.19 shows the times required for participants
to make their judgments in the various directions.
Participants made the above-below judgments
fastest and the right-left judgments slowest. Franklin
and Tversky point out that such results make sense if
we assume that participants are constructing a spatial
framework while reading the description. Both the
up-down vertical axis and the front-back horizontal
axis are relevant when we navigate in the real world. In
contrast, we suffer many left-right confusions because
of the bilateral symmetries of the body.We have similar
confusions about left and right in our imagery.
In another experiment, H. A. Taylor and Tversky
(1992) compared the relative effectiveness of three
types of spatial information: route descriptions,
survey descriptions, and actual maps. The route
descriptions were a mental tour through the environment,
whereas the survey descriptions gave a bird’seye
view of the environment. After studying the descriptions or the actual map,
participants had to verify route or survey statements about the environment.
A route statement might be
Driving from the Town Hall to the gas station, you pass Maple Street on your right.
A survey statement might be
Horseshoe Drive runs along the northern shore of Pigeon Lake.
Participants were asked to judge whether these sentences were true of the environment
they studied. They were equally fast at judging such questions whether
they had studied route descriptions, survey descriptions, or actual maps. From
this result, Taylor and Tversky concluded that people can construct cognitive
maps from verbal descriptions as effectively as from actual maps.
We can convert verbal descriptions into cognitive maps of our environment.
•Conclusions: Visual Perception and Visual Imagery
This chapter has also reviewed some of the evidence that the same brain regions
that are involved in visual perception are also involved in visual imagery. Such
research has presumably put to rest the issue raised at the beginning of the chapter
that the visual imagery experiences are purely epiphenomenal. However,
although it seems clear that perceptual processes are involved in visual imagery,
it remains an open question to what degree the mechanisms of visual imagery
are the same as the mechanisms of visual perception. Evidence for a substantial
overlap comes from neuropsychological patient studies (see Bartolomeo, 2002,
for a review). Many patients who have cortical damage leading to blindness
have corresponding deficits in visual imagery. As Behrmann (2000) notes, the
Conclusions: Visual Perception and Visual Imagery | 113
Response times (s)
2.5
1.5
2.0
1.0
0.5
Ahead Behind Right Left
Direction
Above Below
FIGURE 4.19 Results from
Franklin & Tversky’s experiment
demonstrating that we create
mental images from a word
description of a spatial
structure. Participants read
stories that required them to
orient themselves in an
imaginary environment, then
were asked to identify the
locations of objects in the
environment. The results are the
times required to identify objects
in various directions. (From
Franklin & Tversky, 1990. Reprinted by permission
of the publisher. © 1990 by Journal
of Experimental Psychology: General.)
Anderson7e_Chapter_04.qxd 8/20/09 9:43 AM Page 113
correspondences between perception and imagery can be quite striking. For
instance, there are patients who are only unable either to perceive or to image
faces and colors, but are otherwise unimpaired in either perception or imagery.
Nonetheless, there exist cases of patients who suffer perceptual problems but have
intact visual imagery and vice versa. Behrmann argues that visual perception and
visual imagery are best understood as two processes that overlap but are not identical,
as illustrated in Figure 4.20. To perceive a kangaroo requires low-level visual
information processing that is not required for visual imagery. Similarly, to form
a mental image of a kangaroo requires generation processes that are not required
by perception. Behrmann suggests that patients who suffer only perceptual losses
have damage to the low-level part of this system, and patients who only suffer
imagery losses have damage to the high-level part of this system.
114 | Mental Imagery
Low-level
visual analysis
High-level
generation
Intermediate
visual
processing
Perception Imagery
FIGURE 4.20 A representation
of the overlap in the processing
involved in visual perception and
visual imagery.
1. It has been hypothesized that our perceptual system
regularly uses mental rotation to recognize objects in
nonstandard orientations. In Chapter 2 we contrasted
template and feature models for object recognition.
Would mental rotation be more important to a template
or feature model?
2. Consider the following problem:
Imagine a wire-frame cube resting on a tabletop with
the front face directly in front of you and perpendicular
to your line of sight. Imagine the long diagonal that
goes from the bottom, front, left-hand corner to the
top, back, right-hand one. Now imagine that the cube is
reoriented so that this diagonal is vertical and the cube
is resting on one corner. Place one fingertip about a
foot above the tabletop and let this mark the position of
the top corner on the diagonal. The corner on which
the cube is resting is on the tabletop, vertically below
your fingertip.With your other hand, point to the
spatial locations of the other corners of the cube.
Hinton (1979) reports that almost no one is able to performthis
task successfully. In light of the successes we have
reviewed for mental imagery,why is this task so hard?
3. The chapter reviewed the evidence that many different
regions are activated in mental imagery tasks—parietal
and motor areas in mental rotation, temporal regions
in judgments of object attributes, and hippocampal
regions in reasoning about navigation.Why would
mental imagery involve so many regions?
4. Consider the map distortions such as the tendency to
believe San Diego is west of Reno. Are these distortions
in an egocentric representation, an allocentric representation,
or something else?
Questions for Thought
Key Terms
allocentric representation
cognitive maps
egocentric representation
epiphenomenon
fusiform face area
mental imagery
mental rotation
parahippocampal
place area
route maps
survey maps