Journal of Experimental Psychology:
Learning, Memory, and Cognition
1988, Vol. 14, No. 3, 468476
Copyright 1988 by the American Psychological Association, Inc.
0278-7393/88/$00.75
Picture Memory: Recognizing Added and Deleted Details
Kathy Pezdek
Ruth Maki
The Claremont Graduate School
North Dakota State University
Debra Valencia-Laver, Tony Whetstone, Janet Stoeckert, and Tom Dougherty
The Claremont Graduate School
When people are presented simple and complex pictures and then tested in a same-changed
recognition test with a simple or complex form of each, d' is greater for the simple than the
complex picture (Pezdek & Chen, 1982). The results of three experiments confirm the robustness
of this "asymmetric confusability effect" and test a model of the processes underlying this effect.
According to the model, pictures are schematically encoded such that the memory representation
of both simple and complex pictures is similar to the simple form of each. In Experiment 1, a
sentence was presented that described the central schema in the picture prior to subjects' viewing
each picture. This manipulation exaggerated the asymmetric confusability effect; schematic
processing thus underlies the effect. Results of Experiment 2 refute the hypothesis that the effect
results from subjects erroneously anticipating a recall test rather than a recognition test. Furthermore, although some of the nonschematic elaborative information in complex pictures is stored
in memory, it is difficult to retrieve to verify that something is missing when complex presentation
pictures are changed to simple test pictures (Experiment 3). Thus, although people are able to
distinguish large sets of old pictures from new distractor pictures, their ability to detect missing
elaborative visual details is more limited.
Our ability to recognize pictures that we have seen before
is remarkable. Researchers have demonstrated that individuals can distinguish large sets of old pictures from new distractor pictures at high levels of accuracy (Nickerson, 1965;
Shepard, 1967; Standing, Conezio, & Haber, 1970). However,
few investigators have examined memory for the elaborative
visual information in pictures that are remembered. This is
the focus of the present study.
The conceptual origins of the present study are two lines of
research that suggest that picture-recognition memory is positively correlated with the a m o u n t of information encoded
from pictures. The first of these lines of research includes
several studies that have reported that recognition memory
for pictures increases with the duration of the exposure (Loftus
& Bell, 1975; Potter, 1976; Potter & Levy, 1969; Tversky &
Sherman, 1975). The interpretation of these results has been
that increased exposure duration leads to better memory
because more information about the details o f the picture is
encoded and retained at the longer intervals. This interpretation is supported by findings that with more exposure time to
view pictures subjects make more eye fixations (Loftus, 1972)
and that eye fixations generally occur more often and for
longer durations in regions of high information density
(Antes, 1974; Yarbus, 1967).
The second line of research that suggests that picturerecognition memory is positively correlated with the amount
of information encoded from each picture is the finding that
pictures are recognized better than are comparable verbal
stimuli such as words or sentences (cf. Shepard, 1967). It has
been hypothesized that this picture superiority effect is due to
the additional amount of detailed information available in
pictures as compared with verbal material (Bevan & Steger,
1971; Evertson & Wicker, 1974; Reese, 1970). Presumably, if
the picture superiority effect can be accounted for by the
amount of detail in pictures, then more detailed pictures
should be better recognized.
To test this hypothesis Nelson, Metzler, and Reed (1974)
used a two-alternative forced-choice recognition test with old
and new test pictures. However, they found no differences in
recognition among items presented as (a) black-and-white
photographs, (b) complex, embellished line drawings, or (c)
simple, unembellished line drawings. The three versions of
each picture contained the same central information, but extra
contextual details, shading, and embellishment were added in
the complex pictures. Although recognition memory did not
differ among these three types of pictures, memory for all
three types of pictures was superior to memory for sentences
that had previously been derived from the pictures. Similarly,
Dirks and Neisser (1977) compared memory for added, deleted, and moved objects in real displays with memory for
photographs of the same displays. Although the real displays
included more detail than did the photographs of each, no
significant differences resulted between these two types of
visual materials with either recall or recognition measures.
This research was conducted while Kathy Pezdek was supported
by a grant from the James Irvine Foundation. Some of the data in
Experiment 1 were a portion of the master's thesis submitted by
Hsaun-Chih Chen.
Kirk Reynolds and Nusha Askari provided assistance with Experiments 2 and 3. We thank Vera Dunwoody, Rob Simon, Shawn
Burn, and Patricia Winter for volunteering their classes for this study.
We also thank Jane Gray for her help in preparing the manuscript.
Correspondence concerning this article should be sent to Kathy
Pezdek, Department of Psychology, The Claremont Graduate School,
Claremont, California 91711-6175.
468
469
RECOGNITION MEMORY FOR PICTURES
On the other hand, using Nelson et al.'s materials (1974)
and brief exposure durations (60 ms to 500 ms), Lofius and
Bell (1975) reported that the ability to distinguish old from
new pictures was greater for photographs than for simple and
complex line drawings; however, memory did not significantly
differ between simple and complex line drawings. Similarly,
Park, Puglisi, and Smith (1986) had subjects view pictures
with low, medium, and high levels of visual detail and found
that the ability to distinguish old from new pictures was
greater for relatively more detailed pictures. Thus, although
several lines of research have suggested that increasing the
amount of detailed information in a picture leads to improved
recognition memory, the support for this effect is mixed.
Furthermore--and pivotal, to the purpose of the present
study--each of the above studies testing the role of detail in
picture-recognition memory used an old-new recognition
tasks in which the distractor pictures were completely different
pictures. Thus, these previous studies tested the ability to
distinguish familiar pictures that activate old schemata from
unfamiliar pictures that activate new schemata. The present
program of research uses a same-changed recognition test
and thereby assesses memory for more specific details in
pictures, that is, memory for subschematic information. The
issue addressed in the present program of research is thus
different from the issue addressed in previous studies on the
role of detail in picture-recognition memory.
The present study follows from a previous study in this
program of research by Pezdek and Chen (1982). In this
previous study, a subset of the line drawings used by Nelson
et al. (1974) was used to test subjects' memory for the elaborative detailed information in pictures. Each of 44 pictures
was available in both a simple and a complex form, such as
the examples in Figure I. As with the complete set of materials
used by Nelson et al. (1974), the simple and complex versions
of each picture were both "best described" by the same
sentence; however, extra shading, details, and elaboration
were added to the figure and the background in the complex
version of each picture. Subjects from three age groups were
presented 22 simple and 22 complex line drawings. During
the testing period, pictures were presented one at a time in a
same-different recognition test. Half of the simple and complex pictures were tested in the same form in which they had
been presented. The other half of the test items were changed
pictures; that is, pictures previously presented in a simple
form were tested with the complex form of the same pictures,
and pictures previously presented in a complex form were
tested with the simple form of the same pictures. In the
presentation phase subjects were instructed to study the pictures carefully. During the test phase they viewed pictures one
at a time and responded "same" or "changed" to each.
The results of the adults tested by Pezdek and Chen (1982)
are presented in Table 1. As can be seen, recognition sensitivity was significantly higher for simple (d' = 2.03) than for
complex (d' = 1.18) presentation pictures. This effect is
referred to as the asymmetric confusability effect. Subjects
more accurately distinguished copies of simple input pictures
from more complex distractor pictures than they distinguished copies of complex input pictures from simple distrac-
COMPLEX
SIMPLE
1
T
_._____
Figure 1. Examples of pictures in both simple and complex forms.
tor pictures. The asymmetric confusability effect was replicated in a second experiment by Pezdek and Chen (1982) and
has also been reported by Pezdek (1987) with first graders,
third graders, young adults, and adults older than 68.
The purpose of the present study was to test a particular
model for the processes underlying the asymmetric confusability effect first reported by Pezdek and Chen (1982). According to this model, pictures are schematically encoded
such that the memory representation of both simple and
complex pictures is similar to the simple version of each
picture. Furthermore, although at least some of the nonschematic elaborative information in complex presentation pictures is stored in memory, this elaborative information is
difficult to retrieve to verify changed (simple) test versions of
complex presentation pictures. That is, elaborative information is easy to recognize in same complex test pictures but is
Table 1
Mean Percentage Correct and d' Values in Each
Experimental Condition Reported by Pezdek and Chen
(1982): Adult Subjects in Experiment I
Simple
presentation
d' --- 2.03
Measure
Percentage correct
Same
test
71.0
Complex
presentation
d' -- 1.18
Changed S a m e Changed
test
test
test
83.8
68.7
66.5
470
PEZDEK, MAKI, VALENCIA-LAVER, WHETSTONE, STOECKERT, DOUGHERTY
difficult to retrieve when complex presentation pictures are
changed to simple test pictures.
The notion that pictures are schematically encoded comes
from several previous studies. Goodman (1980) found that
pictures were represented in terms of the action schema
depicted in each picture and that high schema-relevant items
were accurately recalled but poorly recognized, and low
schema-relevant items were accurately recognized but poorly
recalled. Similarly, Friedman (1979) reported that objects in
pictures that have the highest likelihood of being represented
in a picture's frame required less visual analysis for identification, and that transformations of these objects were less
likely to be noticed than were transformations of less expected
objects. Findings by Hock, Romanski, Galie, and Williams
(1978), Mandler and Johnson (1976), Mandler and Ritchey
(1977), and Pezdek (1978) provide further evidence that people use a scene schema to encode and represent information
presented in pictures as well as in real world scenes (Brewer
& Treyens, 1981).
The model tested in the present study is summarized in
Table 2 with the predicted pattern of results. Consider processing of the simple presentation pictures first. The model
predicts that when a simple picture is presented, subjects store
the visual information that is relevant for communicating the
schema of the picture. It should be noted that when the simple
pictures were developed by Nelson et al. (1974), an artist was
shown each photograph used by Nickerson (1965) and the
sentence that had been normed to best describe the photograph. The artist was instructed to sketch a line drawing of
the photograph that included the least amount of visual detail
in order to present the information in the photograph that
was necessary for communicating the one-sentence descriptor.
Thus, most of the visual information in each simple picture
is predicted to be retained in the schematic representation of
the picture in memory. Consequently, with a same (simple)
test picture, the predicted response would be "same" because
a match in memory would be made. With a changed (complex) test picture, the predicted response would be "changed"
because the extra details in the complex version of the simple
picture are not recognized as being part of the memory
representation of the simple presentation picture. The model
thus predicts high hit rates and high correct rejection rates in
the simple presentation conditions.
On the other hand, the complex pictures were defined by
Nelson et al. (1974) as being replicas of the simple form of
Table 2
Predictions of the Proposed Model
Simple presentation
Complex
presentation
Changed
Changed
Same test
test
Same test
test
Predicted
(simple)
(complex) (complex) (simple)
Response
"Same . . . . Changed . . . . Same . . . . Same"
Results
High hit
High CR
High hit
Low CR
rate
rate
rate
rate
Note. CR = correct rejection.
each picture with extra shading, details, and elaboration
added. Thus, when a complex picture is presented, the model
predicts that the picture will be schematically processed such
that the visual information that is relevant for communicating
the schema in the picture is retained in memory. This schemarelevant visual information in a complex picture is primarily
the information that is in the simple form of the complex
picture. Thus, the memory representation for complex presentation pictures would retain the information in the simple
form of the picture.
The model further predicts that the elaborative visual information in complex presentation pictures that is not central
to communicating the schema in each picture is encoded and
stored differently than is the schema-relevant visual information. A consequence of this difference is that the elaborative visual details are more difficult to retrieve than are the
schema-relevant visual details. Thus, elaborative information
in complex presentation pictures is difficult to retrieve when
a changed (simple) test picture is viewed but can be recognized
when it is included in same (complex) test pictures. High hit
rates but low correct rejection rates are thus predicted for
complex presentation pictures. The difficulty in retrieving
nonschematic elaborative details is hypothesized to result
from one of two sources. The difficulty in retrieving nonschematic details may result from weaker associations in memory
between nonschematic details and schematic details than exist
among schematic details. Alternatively, Goodman (1980) suggested that this difficulty occurs because nonschematic details
are not retained in memory with the schematic information.
The present study was not designed specifically to distinguish
between these two interpretations.
Experiment 1
In Experiment 1, half of the subjects participated in the
same-changed recognition procedure used by Pezdek and
Chen (1982) with a 5-s presentation rate. The other half of
the subjects participated in this same procedure except that
they were presented a one-sentence description of each picture
just prior to viewing the picture. Subjects in the sentenceprompt condition were instructed to keep each sentence in
mind while studying the picture that followed. This procedure
was suggested by Friedman (1979) and Bartlett, Till, and Levy
(1980) as a method for encouraging thematic or schemadriven processing of each picture.
It was anticipated that although both simple and complex
pictures are schematically processed, presenting the sentence
prompt before each picture would serve to more clearly
differentiate the schema-relevant visual information from the
elaborative visual information. It was predicted that the sentence prompt would increase the probability that pictures are
encoded and retained in terms of their central schema, and
furthermore, would increase the probability of encoding and
storing nonschematic elaborative information in a form that
makes subsequent retrieval difficult. Thus, an interaction was
predicted between the sentence-prompt condition and presentation form; the asymmetric confusability effect was pre-
RECOGNITION MEMORY FOR PICTURES
471
dieted to be larger in the sentence-prompt condition than in
the n o - p r o m p t condition.
phase to ensure that subjects understood the task. The entire procedure lasted about 35 rain.
Method
Results
Subjects and design. The subjects were 40 undergraduates who
volunteered from classes at California State University, San Bernardino. Age and sex were not specifically controlled in this experiment
nor throughout the study. No students participated in more than one
experiment. Twenty subjects were randomly assigned to the sentenceprompt condition and 20 to the no-prompt condition. All subjects
viewed simple and complex line drawings in a presentation phase and
were tested with a same-changed recognition procedure.
Materials. The 44 pairs of pictures used by Pezdek and Chen
(1982) were used in each of the experiments in this study. These
pictures were a subset of the pictures used by Nelson et al. (1974)
and originally constructed by Nickerson (1965). Each of 44 basic
pictures had been drawn in both a simple, unembellished line drawing
form and a complex, embellished line drawing form, which yielded a
total of 88 pictures. All drawings were black and white. The simple
and complex forms of each picture were constructed by Nelson et al.
(1974) such that both were "best described" by the same sentence;
extra shading, details, and elaboration, however, were added to the
figure and the background in the complex version of each picture.
None of the details added in the complex form of each picture were
unexpected or implausible, and these details were not essential for
communicating the theme of the picture. Examples of stimulus
pictures are shown in Figure I.
The 44 test pictures consisted of 22 same pictures (11 simple and
11 complex) and 22 changed test pictures. There were two types of
changed test pictures---those that included additions and those that
included deletions. Additions involved simple pictures at presentation
and complex versions of those same pictures at test. Deletions involved complex pictures at presentation and simple versions of those
same pictures at test.
In addition to the pictorial stimuli, a typed one-sentence description of each picture was reproduced on each of 44 additional slides.
The sentences were those generated by Nelson et al. (1974); two
judges independently examined the pictures and created a one-sentence description of each. Differences between the two judges' descriptions were reconciled by a third judge.
Procedure. Subjects participated in groups of 3 to 5 subjects each.
The experiment consisted of a study phase followed immediately by
a 3-min maze-tracing task and then a test phase. The maze-tracing
task was inserted between the study and test phases to ensure that the
test assessed long-term memory. In the study phase, subjects were
presented 88 slides for 5 s each, with 44 representing pictures. The
projected image of each picture measured 64 x 97 cm. This size was
held constant for each experiment in this study. The sentence-prompt
condition thus included 22 simple pictures and 22 complex pictures
randomly sequenced, with each immediately preceded by the one
sentence that described the picture. In the no-prompt condition the
study phase included the same pictures presented in the same order
as in the sentence-prompt condition. However, a blank slide was
presented prior to each picture to equalize the intertrial interval in
the two conditions. The background brightness of the blank slides
was similar to that of the sentence slides. Subjects were instructed to
study each picture carefully, because this would be important in a
later part of the experiment. In addition, subjects in the sentenceprompt condition were instructed to keep each sentence in mind
while studying the picture that followed. During the test phase,
subjects looked at each test picture and circled on a response sheet
"SAME"or "CHANGED."Four sample items were shown before the test
The percentage correct and d' data were calculated across
subjects and are presented in Table 3. T h r o u g h o u t this study
the rejection region for all analyses was p < .05. A 2 (sentence
p r o m p t / n o prompt) x 2 (simple or c o m p l e x presentation
form) analysis o f variance (ANOVA)was first performed on the
d ' data. Recognition sensitivity was greater for picture presented in the simple (d' --- 1.65) than complex form (d' =
0.87), F ( I , 38) = 15.64, MSc --- .78. Although the presence of
a sentence p r o m p t did not produce an overall m a i n effect ( F
< 1.00), this variable did interact significantly with presentation form, F(1, 38) = 4.34, MSe = .78. The sentence p r o m p t
increased the recognition advantage o f simple over complex
pictures relative to the n o - p r o m p t condition.
The percentage correct data were analyzed in a 2 (sentence
p r o m p t / n o prompt) x 2 (simple or complex presentation
form) x 2 (same or changed test form) ANOVA. Subjects were
significantly m o r e accurate in recognizing pictures presented
in the simple (74.2%) than complex (64.5%) form, F ( I , 38)
= 12.61, MSe = 2.92, and m o r e accurate in recognizing same
(hit rate = 73.5%) than changed (correct rejection rate =
65.0%) test pictures, F ( I , 38) = 23.71, MS~ = 2.28. The
Presentation F o r m x Test F o r m interaction was also significant, F(1, 38) = 5.17, MSc = 5.54. Although the hit rate for
simple (72.5%) and c o m p l e x presentation pictures (74.6%)
did not differ, the correct rejection rate was significantly less
for complex (54.4%) than for simple presentation pictures
(75.6%). This first-order interaction did not significantly vary
as a function o f the sentence-prompt manipulation, nor did
the sentence-prompt manipulation produce a significant m a i n
effect on the percentage correct measure.
F o u r planned comparisons were p e r f o r m e d to c o m p a r e
performance in the sentence-prompt versus n o - p r o m p t conditions for hit rates and correct rejection rates with simple
and c o m p l e x pictures. As seen in Table 3, comparisons were
m a d e on the pairs o f means in each o f the four columns in
the top half o f the table. The only difference between the
sentence-prompt and n o - p r o m p t conditions was in the correct
rejections o f c o m p l e x pictures (47.3% vs. 61.5%), t(38) =
2.10. The presence o f the sentence p r o m p t reduced subjects'
recognition accuracy for changed versions of pictures that had
been presented in their c o m p l e x form. This difference conTable 3
Mean Percentage Correct and d' Values in Each
Experimental Condition in Experiment I
Condition
Sentence-prompt
No-prompt
Simple presentation
Complex presentation
Same Changed
test:
test:
%
%
correct correct d'
Same Changed
test:
test:
%
%
correct correct d'
77.0
67.9
74.2
77.0
1.95
1.34
74.3
74.8
47.3
61.5
0.76
0.97
472
PEZDEK, MAKI, VALENCIA-LAVER, WHETSTONE, STOECKERT, DOUGHERTY
tributed to the significant interaction of the sentence-prompt
condition and presentation form on d'.
The results of Experiment 1 are consistent with the predictions of the model being tested. The patterns of results in the
sentence-prompt and no-prompt conditions were similar,
with the asymmetric confusability effect exaggerated in the
sentence-prompt condition. It appears that subjects in both
conditions were schematically encoding the pictures and subsequently using schematic cues to guide retrieval of visual
information. However, the sentence prompt caused subjects
to do this more effectively.
In addition, if subjects are schematically processing pictures
as predicted by the model, then the elaborative detail that is
less essential for communicating the schema of each complex
picture would more likely be encoded and stored in a form
that makes subsequent retrieval difficult. Thus, the correct
rejection rate for changed complex presentation pictures
would be lower than for changed simple presentation pictures.
This occurred in Experiment 1. Furthermore, exaggerating
schematic processing of pictures by the use of a sentence
prompt would be expected to decrease the correct rejection
rate for complex presentation pictures relative to simple presentation pictures more in the sentence-prompt condition than
in the no-prompt condition. Again, this occurred in Experiment 1.
Experiment 2
Experiment 2 tested whether subjects used schema-guided
or thematic encoding of the pictures in the previous experiments because they erroneously anticipated that the memory
test would be a recall test rather than a recognition test. This
interpretation follows from Nickerson and Adam's conclusion
(1979) that "the visual details of an object, even a very familiar
object, are typically available from memory only to the extent
that they are useful in everyday life" (p. 287). This suggests
that subjects in the present study may have been encoding
the type of information that they anticipated would be "useful" for the subsequent memory task.
Furthermore, it has been demonstrated that subjects encode
different types of information from pictures as a function of
whether they are anticipating a recall- versus a recognitionmemory test (Frost, 1972; Tversky, 1973). In particular, in at
least one study (Goodman, 1980), it was reported that although high thematically relevant details are better recalled
than are low thematically relevant details, low thematically
relevant details are better recognized than are high thematically relevant details. Thus, one interpretation of the results
of the present study is that subjects were using thematic
encoding of the pictures because they were anticipating a
recall-memory test.
Experiment 2 is a replication of the procedure used by
Pezdek and Chen (1982), with the added feature that half of
the subjects were told prior to the presentation phase that the
test that followed would be a same-changed recognition test,
and half were told to anticipate a recall test. The recognitioninstruction group was shown eight sample items demonstrat-
ing the type of changes they would see during the test. The
recall-instruction group was given eight sample pictures and
a practice recall test before the presentation sequence. Both
groups in fact received the identical same-changed recognition-memory test. If the memory advantage for simple over
complex pictures is due to subjects erroneously anticipating a
recall test, then the difference between recognition of simple
and complex pictures should be less in the recognition-instruction condition than in the recall-instruction condition in
Experiment 2.
Method
Subjects and design. Forty-five students in an introductory psychology class at California State University, Los Angeleswere subjects
in the recognition-instruction condition. Forty students in an introductory psychologyclass at Chaffey CommunityCollege were subjects
in the recall-instrnction condition. The students participated in class
as part of a class exercise. Classes were randomly assigned to each of
the two conditions. However, because students participated in class,
this precluded random assignment of subjects to conditions and thus
stretches the definition of a "true experiment." All subjects viewed
simple and complex line drawings in a presentation phase and were
tested with same and changed test pictures in a same-changed recognition procedure.
Materials and procedure. The 44 pictures from Experiment 1 were
used in this experiment. In the study phase, 44 pictures were randomly
arranged and presented for 5 s each. This phase was followed by a 3min maze-tracing task and then a test phase. In the test phase 44 test
pictures were randomly arranged and presented for 8 s each in a same
or changed form. Subjects responded "SAME"or "CHANGED"on the
response sheet provided. Four sample items were shown before the
test phase to ensure that subjects understood the task. The entire
procedure took about 30 min.
Before viewing the presentation pictures, subjects in the recognition-instructions condition were instructed to study each picture
carefully and told that they would be tested with a recognition
memory test afterward. They were then shown eight demonstration
pairs, each consisting of a presentation picture followed by a test
picture. The eight demonstration pairs included two each of the four
presentation form/test form combinations (i.e., present simple-test
same, present simple-test changed, present complex-test same, present
complex-test changed). With each demonstration test picture, the
experimenter indicated the correct response for that item and described the reason for the correct response in terms of the picture
being (a) the same as the previous picture, (b) a changed version of
the previous picture because of specific added elaborative detail, or
(c) a changed version of the previous picture because of specific
deleted elaborative detail.
Before viewing the presentation pictures, subjects in the recallinstructioncondition were told to study each picture carefully because
afterward they would be given a recall test. A recall test was defined
as a test in which they would be asked to write down a one-sentence
description of each of the pictures they could remember from the
original sequence. They were then shown eight sample pictures followed by a practice test in which they were instructed to recall in
writing as many of the pictures as they could remember.
Results
The percentage correct and d' data were calculated across
subjects and are presented in Table 4. A 2 (recognition or
RECOGNITION MEMORY FOR PICTURES
Table 4
Mean Percentage Correct and d' Values in Each
Experimental Condition in Experiment 2
Simple presentation
Condition
Recognition
instructions
Recall
instructions
Same Changed
test:
test:
%
%
correct correct d'
Complex presentation
Same Changed
test:
test:
%
%
correct correct d'
73.3
78.0
1.64
67.8
66.4
1.03
77.3
84.0
1.91
73.9
66.8
1.32
recall instructions) x 2 (simple or complex presentation form)
ANOVA was first performed on the d" data. Recognition sensitivity was greater for pictures presented in their simple form
(d' = 1.76) than complex form ( d ' = 1.16), F(1, 83) = 15.72,
MSe = .96. The effect of prior instructions did not significantly
affect d', F(1, 83) = 3.36, MSe = 1.41, nor did the instruction
variable interact with presentation form, F < 1.00.
The percentage correct data were analyzed in a 2 (recognition or recall instructions) x 2 (simple or complex presentation form) x 2 (same or changed test form) ANOVA. The only
significant effects were the main effect of presentation form,
F ( l , 83) = 37.67, MSe = 1.99, and the Presentation Form x
Test Form interaction, F ( l , 83) = 12.04, MSe = 1.74. In line
with the asymmetric confusability effect, subjects were more
accurate recognizing simple (78.2%) than complex pictures
(68.7%), and the difference in accuracy between simple and
complex items was significant for correct rejections (81.0%
vs. 66.6%) but not for hits (75.3% vs. 70.9%). The manipulation of recognition versus recall instructions did not produce
a significant main effect, F ( l , 83) = 3.31, MSe = 4.29, nor
were any of the interactions involving instructions significant.
Experiment 2 tested if the recognition-memory advantage
for simple over complex pictures is due to subjects' erroneously anticipating a recall test rather than a recognition
test. According to this interpretation, the difference between
recognition of simple and complex pictures should be less
with recognition instructions than with recall instructions.
This did not occur in Experiment 2. The results in both
conditions in Experiment 2 replicated the results of Experiment 1 as well as results o f Pezdek and Chen (1982); primarily,
subjects discriminated original from changed simple presentation pictures significantly better than they discriminated
original from changed complex presentation pictures. The
asymmetric confusability effect is thus a robust effect that is
not diminished by showing subjects exactly what type of
same-changed recognition test to anticipate. The interpretation that subjects use schema-guided or thematic encoding of
the pictures in this task because they erroneously anticipated
a recall rather than a recognition memory test is thus rejected.
Experiment 3
Experiment 3 tested an additional assumption in the model
tested in Experiment 1. This assumption is that the nonsche-
473
matic elaborative information in complex presentation pictures that is retained in memory is difficult to retrieve. Thus,
the nonschematic elaborative information in complex pictures can be recognized when it is available in a same test
picture (i.e., high hit rates for complex pictures in the previous
experiments), but this information is difficult to retrieve from
memory when it is not available in a changed test picture
(i.e., low correct rejection rate for complex pictures). According to this prediction, the lower correct rejection rate for
complex than simple presentation pictures is due to greater
difficulty verifying that something is missing from changed
complex presentation pictures than from verifying that something is added to changed simple presentation pictures.
To test this hypothesis, subjects in the experimental condition in Experiment 3 viewed the simple and complex pictures as they had been presented in the previous experiment,
but they were tested with only changed test pictures. The task
of the subjects was to look at each changed test picture and
decide if elaborative details were added to or deleted from the
previously seen version of the picture. They were then asked
to identify what elaborative details had been added to or
deleted from the previously seen picture. It was predicted that
subjects would be significantly more accurate detecting additions made to changed simple pictures than detecting deletions made from changed complex pictures. Furthermore, it
was predicted that subjects would more accurately identify
the specific elaborative details that were added to changed
simple pictures than they would be identifying those that were
deleted from changed complex pictures.
An additional control condition was included in Experiment 3 to test whether subjects who had never seen the
presentation pictures would be able to indicate whether the
correct response should be "added" or "deleted" for each test
picture. This may have been possible if simple test pictures
look relatively more empty than do complex test pictures. To
test this interpretation, a separate group of subjects in Experiment 3 was tested with the changed test pictures, but this
second group was not shown the presentation pictures. If such
a guessing effect were operating, it would be expected that
subjects who did not see the presentation pictures would be
more likely to respond "added" to changed (complex) versions
of simple presentation pictures than to respond "deleted" to
changed (simple) versions of complex presentation pictures.
Method
Subjects and materials. Fifteen students in a cognitive psychology
class at California State University, FuUerton participated in the
experimental condition that included presentation pictures. They
were presented the same 44 simple and complex pictures as were
used in the previous experiments. Nineteen students in a social
psychology class at the same university participated in the no-presentation control condition. As in Experiment 2, classes were randomly assigned to the two conditions. However, because students
participated in classes, they were not randomly assigned to conditions.
Procedure and design. Subjects in the presentation condition
viewed the 22 simple and 22 complex presentation pictures for 5 s
each, followed by a 3-min maze-tracing task, and then the 44 test
474
PEZDEK, MAKI, VALENCIA-LAVER, WHETSTONE, STOECKERT, DOUGHERTY
pictures. In the presentation phase, these subjects were instructed to
study each picture carefully, because it would be important in a later
part of the experiment.
The test consisted of 44 changed versions of the pictures from the
presentation phase; there were no same test pictures. Subjects were
instructed that they would view 44 changed test pictures. For each
test picture, the task was first, to detect whether the change in the test
picture involved details that were added to or deleted from the original
picture and to circle the corresponding word "ADDED"or "DELETED"
on a response sheet. This was the detection test. Second, they were
asked to identify in the space provided what elaborative details had
been added to or deleted from the original picture. This was the
identification test. If at least one of the added or deleted details was
identified for each test picture, this was scored as an accurate identification response. Subjects were shown four practice test items in
advance to ensure that they understood the task.
Subjects in the no-presentation condition were told that a different
group of subjects had been shown a series of 44 pictures and then a
test on these pictures. They were told that they would be given the
same test as this previous group. They would not see, however, the
original presentation pictures. Their task was thus to look at each
picture and guess whether elaborative details had been added to or
deleted from the original version of the picture seen by the other
group. Then they were to indicate what elaborative details had been
added to or deleted from the original picture. Subjects in both groups
used the same response sheet. Subjects in the no-presentation condition were shown examples of 20 pictures, similar to those seen during
the presentation phase by the previous group, to give them a sample
of the range in the amount of visual detail typical with this kind of
line drawings. These 20 sample pictures were from the original set
used by Nelson et al. (1974) but were not included in the 44 pictures
used throughout the current study. Subjects in the no-presentation
condition were also shown the four practice test items to demonstrate
the types of changed pictures in the test.
Results
The accuracy in correctly detecting additions in changed
simple presentation pictures and deletions in changed complex presentation pictures was calculated and averaged for
subjects in both conditions. A 2 (simple vs. complex presentation form) x 2 (presentation vs. no-presentation group)
ANOVA was performed on the percentage correct detection
data. Subjects in the condition with presentation pictures
(84.8% correct) were significantly more accurate than were
subjects in the no-presentation condition (69. 1% correct), F(1,
32) = 8.42, MSe = 4.90, and these group differences interacted
with presentation condition, F(1, 3 2 ) = 7.11, MSe = 3.20.
Planned comparisons indicated that although subjects in the
presentation condition were significantly more accurate in
detecting changes in simple pictures (92.6%) than in complex
pictures (76.9%), t(14) = 7.8 l, in the no-presentation condition, detection of changes in simple (65.3%) and complex
pictures (72.9%) did not differ, t(18) = 1.25. The results of
the no-presentation group suggest that better accuracy at
detecting additions than deletions by the presentation group
does not result from a greater tendency to guess "added" than
"deleted."
Additional analyses were performed on subjects' second
response for each item; that is, the mean percentage accurate
identification given correct detection in each condition. In
the presentation condition, subjects were significantly more
accurate in identifying the details that were added to changed
simple presentation pictures, given that additions were detected (90.2%), than they were in identifying the details that
were deleted from changed complex presentation pictures,
given that deletions were detected (54.2%), t(14) - 7.81.
Similarly, in the no-presentation condition, subjects were
significantly more accurate in identifying the details that were
added to changed simple pictures, given that additions were
guessed correctly (55.1%), than they were in identifying the
details that were deleted from changed complex pictures,
given that deletions were correctly guessed (12.8%), t(18) =
6.37. Thus, not only are subjects better able to detect that
something is added to changed simple pictures than that
something is deleted from changed complex pictures; they are
also better able to identify the details that were added than
those that were deleted.
Together, these results verify the predictions of the model.
Although the nonschematic detail information in complex
presentation pictures can be recognized when it is included
in same (complex) test pictures (i.e., the high hit rate for
complex presentation pictures), it is difficult to retrieve when
it is deleted from changed (simple) test pictures (i.e., the low
correct rejection rate for complex presentation pictures).
General Discussion
These three experiments testify to the robustness of the
"asymmetric confusability effect" initially reported by Pezdek
and Chen (1982). When subjects in Experiment 1 were given
a sentence that described the central schema in each picture
prior to viewing each picture, the asymmetric confusability
effect was exaggerated. This suggests that schematic processing
of pictures underlies the effect. A model was tested in which
it was proposed that when pictures are schematically processed
the visual information that communicates the schema is more
likely to be encoded and retained in memory than is the
elaborative information not necessary for communicating the
schema. The memory representation for both simple and
complex pictures is thus similar to the simple version of each
picture. Elaborative details less essential for communicating
the central schema of a picture are represented in memory in
a manner that makes them difficult to retrieve. Thus elaborative visual information is easy to recognize in same complex
test pictures, but is difficult to retrieve to verify that "something is missing" when complex presentation pictures are
changed to simple test pictures. The results of Experiment 3
bear out this prediction of the model.
The finding that subjects are better able to notice what is
added to changed simple presentation pictures than what is
deleted from changed complex presentation pictures has been
observed in several other studies. Agostinelli, Sherman, Fazio,
and Hearst (1986) reported that subjects were more accurate
at detecting and identifying additions to line drawings than
deletions from line drawings of objects. Similarly, Healy
(1981) reported that in the process of proofreading, subjects
are less likely to notice missing features of a letter (e.g., when
"students" is misspelled as "studcnts") than they are to notice
added features of a letter (e.g., when "factors" is misspelled as
RECOGNITION MEMORY FOR PICTURES
"factors"). The asymmetric confusability effect thus appears
to be a fairly general effect.
The asymmetric confusability effect could be described
simply as an example of Weber's Law. According to this
description, people have more difficulty perceiving changes
in complex than simple pictures because the proportion of
change for a stimulus with more information is less than the
proportion of change for a stimulus with less information.
However, Weber's Law does not explain the effect, nor would
it be likely to predict it. Furthermore, Weber's Law would
not have predicted the result of Experiment 1, that schematic
processing of the presentation pictures exaggerates the asymmetric confusability effect. The proposed model goes beyond
the perceptual effect described by Weber's Law and specifically predicts the obtained results.
How do these results relate to the findings from a number
of studies that memory for pictures increases as the exposure
time per picture increases (Lofius & Bell, 1975; Potter, 1976;
Potter & Levy, 1969; Tversky & Sherman, 1975)? The interpretation of these findings has been that increased study time
leads to better memory because more information about the
details of the pictures is encoded and retained at the longer
intervals. However, only one of the studies reporting increased
picture-recognition memory with study time (Tversky & Sherman, 1975) required subjects to distinguish same from
changed test pictures, and in this study the recognition data
were not analyzed separately for hits and correct rejections.
Also, in an experiment by Pezdek (1987) in which Pezdek
and Chen's (1982) procedure was used, although increasing
the presentation duration from 5 s to 15 s increased overall
accuracy, it had no effect on subjects' ability to correctly reject
changed complex pictures. This finding suggests that study
time and complexity have independent effects on the ability
to distinguish same from changed test pictures. Thus, although
increasing study time does lead to better picture memory,
there is no support for the hypothesis that this is simply
because quantitatively more details are encoded and retained
at the longer intervals.
In line with the schematic processing notion outlined above,
one interpretation of the finding that pictures are better
recognized at longer study intervals is that this is due to
qualitative differences rather than quantitative differences in
encoding and storage of detail information. If we view picture
memory as a schema-driven process, then with longer ontime subjects would be able to better abstract the central
schema of each picture and enrich the memory representation
of the schema by incorporating into it more of the schemarelevant information in the picture. This would result in better
schemata in memory--not merely the storage of more details
from the pictures.
There is an apparent conflict between the results of the
present study and those of Friedman (1979) and Goodman
(1980). These two other studies have concluded that subjects
are more likely to recognize changes in unexpected, or low
thematically relevant items in a picture than changes in
expected, or high thematically relevant items. This conclusion
appears at odds with the findings of the present study that
subjects do not accurately retain the elaborative detail infor-
475
mation that differentiates complex from simple pictures, that
is, detail information that is not essential for depicting the
central schema in the picture. Although the present study
appears similar to the studies of Friedman (1979) and Goodman (1980), there are significant differences in the materials
used in these studies.
In Friedman's (1979) study, subjected viewed line drawings
of scenes with 25 to 34 objects in each scene. They studied
each picture for 30 s. Objects had been previously rated as
expected (e.g., a coffee pot in the kitchen) or unexpected (e.g.,
a plant in the kitchen). Changes in the unexpected items were
better noticed than were changes in the expected items. Differences between the results of this study and the current
study can be accounted for by the fact that the schemata
presented in the pictures used in the current study were more
unusual than were those used in Friedman's pictures. Also,
none of the detail information added to or deleted from
pictures in the present study involved unexpected items,
inasmuch as all of the information in both the simple and
complex versions of each picture had existed in the photograph of the real-world scene from which each drawing had
been derived.
In Goodman's (1980) study, subjects viewed line drawings
depicting action schemata such as a girl reading a book at a
desk. Each picture was studied for 10 s followed by a recognition test for objects that were schema-relevant (a bookcase)
or schema-irrelevant (a house plant). Although the materials
used by Goodman (1980) appear similar to those used in the
present study, her study was set up to compare recognition
memory for physical changes in schema-relevant versus
schema-irrelevant information. On the other hand, the present study compared subjects' ability to discriminate same
from changed simple versus complex pictures, and information added to changed simple presentation pictures was detected better than was information deleted from changed
complex presentation pictures. Detecting the addition or deletion of information in pictures apparently involves different
processes than detecting changed objects. Combining Goodman's (1980) findings with those of the present study, one
might predict that although information deleted from pictures
is detected less accurately than is information added to pictures, the size of this memory difference would be greater for
schema irrelevant items than for schema relevant items. However, a test of this prediction remains to be done.
To close on a general note, these results suggest that although people are very good at distinguishing large sets of
"old" pictures from "new" distractor pictures, their ability to
detect missing elaborative visual details is more limited. Studies that require people to distinguish old pictures from completely new distractor pictures only assess how well pictures
that draw on familiar schemata can be distinguished from
pictures that draw on unfamiliar schemata. Such studies do
not assess memory for the elaborative visual information in
pictures. This latter issue has been the focus of this article.
References
Agostinelli, G., Sherman, S. J., Fazio, R. H., & Hearst, E. S. (1986).
Detecting and identifyingchange:Additions versus deletions. Jour-
476
PEZDEK, MAKI, VALENCIA-LAVER, WHETSTONE, STOECKERT, DOUGHERTY
nal of Experimental Psychology: Human Perception and Performance, 12, 445-454.
Antes, J. R. (1974). The time course of picture viewing. Journal of
Experimental Psychology, 103, 62-70.
Bartlett, J. C., Till, R. E., & Levy, J. C. (1980). Retrieval characteristics
of complex pictures: Effects of verbal encoding. Journal of Verbal
Learning and Verbal Behavior, 19, 430-449.
Bevan, W., & Steger, J. A. (1971). Free recall and abstractness of
stimuli. Science, 172, 597-599.
Brewer, W. F., & Treyens, J. C. (1981). Role of schemata in memory
for places. Cognitive Psychology, 13, 207-230.
Dirks, J., & Neisser, U. (1977). Memory for objects in real scenes:
The development of recognition and recall. Journal of Experimental Child Psychology, 23, 315-328.
Evertson, C. M., & Wicker, F. W. (1974). Pictorial concreteness and
mode of elaboration in children's learning. Journal of Experimental
Child Psychology, 17, 264-270.
Friedman, A. (1979). Framing pictures: The role of knowledge in
automatized encoding and memory for gist. Journal of Experimental Psychology." General, 108, 316-355.
Frost, N. (1972). Encoding and retrieval in visual memory tasks.
Journal of Experimental Psychology, 95, 317-326.
Goodman, G. S. (1980). Picture memory: How the action schema
affects retention. Cognitive Psychology, 12, 473-495.
Healy, A. F. (1981). The effects of visual similarity on proofreading
for misspellings. Memory & Cognition, 9, 453-460.
Hock, H. S., Romanski, L., Galie, A., & Williams, C. S. (1978). Realworld schemata and scene recognition in adults and children.
Memory & Cognition, 6, 423-431.
Loftus, G. R. (1972). Eye fixations and recognition memory for
pictures. Cognitive Psychology, 3, 525-551.
Loftus, G. R., & Bell, S. M. (1975). Two types of information in
picture memory. Journal of Experimental Psychology: Human
Learning and Memory, 1, 103-113.
Mandler, J. M., & Johnson, N. S. (1976). Some of the thousand
words a picture is worth. Journal of Experimental Psychology:
Human Learning and Memory, 2, 529-540.
Mandler, J. M., & Ritchey, G. H. (1977). Long-term memory for
pictures. Journal of Experimental Psychology: Human Learning
and Memory, 3, 386-396.
Nelson, T. O., Metzler, J., & Reed, D. A. (1974). Role of details in
the long-term recognition of pictures and verbal descriptions. Jour-
nal of Experimental Psychology, 102, 184-186.
Nickerson, R. S. (1965). Short-term memory for complex meaningful
visual configurations: A demonstration of capacity. Canadian Journal of Psychology, 19, 155-160.
Nickerson, R. S., & Adams, M. J. (1979). Long-term memory for a
common object. Cognitive Psychology, 11, 287-307.
Park, D. C., Puglisi, J. T., & Smith, A. D. (1986). Memory for
pictures: Does an age-related decline exist? Journal of Psychology
andAging, 1, 11-17.
Pezdek, K. (1978). Recognition memory for related pictures. Memory
& Cognition, 6, 64-69.
Pezdek, K. (1987). Memory for pictures: A life-span study of the role
of visual detail. Child Development, 58, 807-815.
Pezdek, K., &Chen, H.-C. (1982). Developmental differences in the
role of detail in picture recognition memory. Journal of Experimental Child Psychology, 33, 207-215.
Potter, M. C. (1976). Short-term conceptual memory for pictures.
Journal of Experimental Psychology. Human Learning and Memory, 2, 509-522.
Potter, M. C., & Levy, E. I. (1969). Recognition memory for a rapid
sequence of pictures. Journal of Experimental Psychology, 81, 1015.
Reese, H. W. (1970). Imagery and contextual meaning. In H. W.
Reese (Chm.), Imagery in children's learning: A symposium. Psychological Bulletin, 73, 404-414.
Shepard, R. N. (1967). Recognition memory for words, sentences,
and pictures. Journal of Verbal Learning and Verbal Behavior, 6,
156-163.
Standing, L., Conezio, J., & Haber, R. N. (1970). Perception and
memory for pictures: Single-trial learning of 2500 visual stimuli.
Psychonomic Science, 19, 73-74.
Tversky, B. (1973). Encoding processes in recognition and recall.
Cognitive Psychology, 5, 275-287.
Tversky, B., & Sherman, T. (1975). Picture memory improves with
longer on time and off time. Journal of Experimental Psychology.
Human Learning and Memory, 44, 114-118.
Yarbus, A. L. (1967). Eye movements and vision. New York: Plenum.
Received July 11, 1986
Revision received J u n e 8, 1987
Accepted J u n e 22, 1987 •