ArticlePDF Available

Abstract

Five experiments with 304 undergraduates investigated the effects of imagery type (bizarre or common) on memory of word triplets. In Exp I bizarre imagery increased recall when imagery type was manipulated in a within-list design but not when imagery type was manipulated in a between-list design. Results of Exp II show that this effect occurred with imagery processing instructions and not with semantic processing instructions. Exp III indicated that bizarre imagery facilitated recall in a within-list design for both self-paced and experimenter-paced presentations of the stimuli, and Exp IV, the pattern of effects of bizarre imagery on memory (in a within-list design) did not parallel the effects of presentation rate on memory. Data from Exps I through IV are inconsistent with an attentional explanation of bizarre imagery effects. In Exp V, when several types of additional learning were interpolated between initial imaginal processing and testing, bizarre imagery produced better recall with a between-list manipulation, but only when the additional learning involved common imagery. In addition to delineating the conditions under which bizarre imagery improves recall, overall findings suggest that distinctiveness may underlie the bizarreness effect. (29 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Journal of Experimental Psychology:
Learning, Memory, and Cognition
1986, Vol. 12. No. 1, 54-65
Copyright 1986 by the American Psychological Association, Inc.
0278-7393/86/SO0.75
Bizarre Imagery as an Effective Memory Aid:
The Importance of Distinctiveness
Mark A. McDaniel
University of Notre DameGilles Q Einstein
Furman University
Five experiments investigated the effects of imagery type (bizarre or common) on memory. In Ex-
periment
1
bizarre imagery increased recall when imagery type was manipulated in a within-list design
but not when imagery type was manipulated in a between-list design. In Experiment 2 it was shown
that this effect occurs with imagery processing instructions and not with semantic processing instruc-
tions.
In Experiment 3 bizarre imagery facilitated recall in a within-list design for both self-paced and
experimenter-paced presentations of the stimuli, and in Experiment 4 the pattern of effects of bizarre
imagery on memory (in a within-list design) did not parallel the effects of presentation rate on memory.
The results of these experiments were inconsistent with an attentional explanation of bizarre imagery
effects. In Experiment 5 several types of additional learning were interpolated between initial imaginal
processing and testing. Bizarre imagery produced better recall with a between-list manipulation, but
only when the additional learning involved common imagery. In addition to delineating the conditions
under which bizarre imagery improves recall, the results of1 these experiments suggest that distinctiveness
may underlie the bizarreness effect.
Mnemonic devices have been used for centuries for the purpose
of enhancing learning and reducing forgetting. Although various
techniques have been developed over time, it is only within the
last 20 years that the important components of memory devices
have been systematically investigated. In popular memory train-
ing books (e.g., Cermak, 1975; Lorayne & Lucas, 1974; Yates,
1966),
the formation of interactive imagery has been described
as an important process in many widely used mnemonic sys-
tems (e.g., method of loci and pegword systems). Also, the claim
has been made that the images should be ridiculous, illogical,
and/or bizarre. Although research has clearly shown that using
interactive imagery facilitates memory relative to many verbal
processing strategies and to noninteractive images (Bower, 1970,
1972),
the importance of bizarre imagery for memory has not
been established. Many studies have addressed the issue of the
effects of bizarre imagery on memory, yet few of these have shown
any mnemonic advantage for bizarre imagery. The results of
most of the studies indicate either that bizarre and common
imagery produce equal levels of recall (Cox & Wollen, 1981;
Order of authorship was determined by the toss of a coin, reflecting
the equal contribution by the authors. Experiment
1
was reported at the
annual meeting of the Psychonomic Society in San Diego, November
1983,
and Experiments 3, 4, and 5 were reported at the International
Imagery Conference in Wales, April 1985,
We gratefully acknowledge the assistance of Gail Tomasheski in de-
veloping the materials and collecting the data for Experiments 1 and 2,
students in Furman University's Experimental and Statistical Methods
class for collecting the data for Experiment 3, Laura Dougherty and Ruth
Kaiser for their help in collecting the data for Experiment 4, and Jennifer
Bribois, Ron Riebschleger, Lydia Roper, and Donna Schwartz for their
help in collecting the data for Experiment 5.
Correspondence concerning this article should be addressed to Mark
McDaniel, Department of Psychology, University of Notre Dame, Notre
Dame, Indiana 46556.
Hauck, Walsh, & Kroll, 1976; Senter & Hoffman, 1976; Wollen,
Weber, & Lowry, 1972; Wood, 1967) or that bizarre imagery
negatively affects recall (Collyer, Jonides, & Bevan, 1972; Em-
merich & Ackerman,
1979;
Wollen &
Cox,
1981b). These findings
have occurred with such regularity that Postman (1975) wrote,
"It is also possible to report that despite popular misconceptions,
bizarre images and pictures are no better as memory aids than
common ones" (p. 322). Recently, however, the mnemonic ben-
efits of bizarre imagery have been demonstrated in several studies
(Merry, 1980; Merry & Graham,
1978;
Webber & Marshall, 1978;
Wollen & Cox, 1981a, 1981b). Given these inconsistent and
equivocal results, the general purposes of the present experiments
were (a) to determine under what conditions bizarre imagery
facilitates memory and (b) to explore the mechanisms through
which bizarreness affects memory.
A common characteristic of the studies that have shown bi-
zarreness to be effective is that bizarreness was manipulated as
a within-list (mixed list) variable. In all of the other studies listed
earlier, bizarreness was manipulated as a between-list (unmixed
list) variable. This pattern of results suggests that bizarre imagery
may be effective only when both bizarre and common images
are formed (Wollen & Cox, 1981b). This conclusion, however,
is based on comparisons across different experiments. Because
many variables were not controlled across these studies, the con-
clusion is not compelling. The purpose, then, of the first exper-
iment was to compare directly within- and between-list manip-
ulations of the bizarreness variable. Erlebacher's (1977, 1978)
method of comparison was used, which allows a direct test of
the interaction between the bizarreness variable and the design
type.
At least several hypotheses can be formulated that predict that
bizarre imagery will be effective only in the context of other,
common, images. Wollen and Cox (1981b) suggest that the su-
perior recall of bizarre images in mixed list designs is due to the
54
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
BIZARRE IMAGERY AND DISTINCTIVENESS55
manner in which the materials were constructed in the studies
that have demonstrated the effect (Merry, 1980; Merry & Gra-
ham, 1978; Wollen & Cox, 1981b). In these experiments, the
bizarreness of the images was controlled by manipulating the
sentence frame in which noun pairs appeared. The bizarre sen-
tences were created by interchanging the nouns that appeared
in the common sentences. For example, the common sentences
"The man smoked the cigar" and "The hen pecked the worm"
were rearranged to form the bizarre sentences "The man pecked
the worm" and "The hen smoked the cigar." Wollen and Cox
point out that with these materials, the possibility of intersentence
cuing exists for the bizarre materials but not for the common
materials. That
is,
recall of the bizarre sentence, "The hen smoked
the cigar" might suggest associates that appear in other bizarre
sentences such as
pecked,
worm, or man. This cuing of infor-
mation contained in other sentences is less likely with the com-
mon sentences. Consequently, Wollen and Cox argue that the
higher recall of bizarre images is due to an artifact of the nature
of the material, wherein the intersentence cuing is the factor
underlying high memory of bizarre images. In the present studies,
subjects in all conditions were presented with the same set of 10
noun triplets for learning, and bizarreness was manipulated by
varying the context in which the words appeared. Thus, the in-
tersentence cuing bias was avoided and unconfounded compar-
isons of bizarre and common imagery were possible.
Even without intersentence cuing, at least two other theories
can be used to predict the bizarreness effect in a within-list design.
According to the attentional or processing time hypothesis
(Merry, 1980; Wollen & Cox, 1981b), bizarre, novel, or unusual
sentences arouse attention and interest and therefore they receive
a greater amount of
processing.
This would be particularly true
in a within-list design where a context of common images pro-
vides the contrast necessary to focus attention upon the novel
or
bizarre.
This effect can also be predicted from notions focusing
on the mnemonic benefits of distinctive encodings (Lockhart,
Craik, & Jacoby, 1976). Distinctiveness is typically defined in a
relative manner, where the distinctiveness of an event is deter-
mined by its relation to other encoded events (Jacoby & Craik,
1979).
Therefore, bizarre materials should produce encodings
that are relatively more distinctive in the context of common
materials than in the context of bizarre materials. This framework
has been used to explain the positive effects of orthographic dis-
tinctiveness on memory. Hunt and Elliott (1980) found that or-
thographically distinct items were better recalled than ortho-
graphically common items in mixed lists but not in unmixed
lists.
They argued that unusual items become functionally more
distinct when presented in the context of common items. This
logic would also nicely describe any enhanced recall of bizarre
images in a mixed list design. Thus, both the attentional and
distinctiveness frameworks can be used to predict higher recall
of bizarre images when subjects form both bizarre and common
encodings.
Experiment 1
In the first experiment, different subjects were exposed to all
bizarre sentences, all common sentences, or half bizarre and
half common sentences; all subjects were asked to create images
for these sentences. Again, all subjects were presented with the
same combinations of noun triplets in all of the conditions, re-
sulting in a constant level of intersentence cuing across all con-
ditions. Thus, according to the intersentence cuing explanation
of the bizarreness effect, bizarre images should not be better
recalled than common images. If, however, the bizarreness effect
results from either distinctiveness or attentional considerations,
then bizarre images should be better recalled than common im-
ages in the within-list design. Several dependent measures were
collected to assess the different hypotheses.
Method
Subjects. The 60 subjects were introductory psychology students
who received course credit for their participation. Subjects were tested in
groups of two to four and the experimental sessions tasted approximately
30 min.
Design and materials. Twenty subjects were randomly assigned to
each of three groups that differed on the basis of the materials they were
presented with: (a) all bizarre material, (b) all common material, or (c)
half bizarre and half common material. These three groups were necessary
to generate the appropriate four cells in Erlebacher's (1977, 1978) design
for comparing the within- and between-subjects effects of an independent
variable. These four cells represent the factorial combination of two in-
dependent variables: design type (within-list and between-list) and type
of imagery (bizarre and common).
All subjects were presented with the same set of 10 word triplets to
image. Each triplet was presented as underlined words in a sentence, and
type of imagery was manipulated by varying the relation between the
words of the triplet. For example, the bizarre context for the list items
dog, bicycle, and street was "The dog rode the bicycle down the street,"
and the common context was "The dog chased the bicycle down the
street." A common and a bizarre sentence were created for each of the
10 word triplets, and two unmixed lists were constructed for subjects in
the between-subjects conditions. Subjects in the bizarre imagery group
were presented with the 10 bizarre sentences, and subjects in the common
imagery group were presented with the corresponding 10 common sen-
tences (these sentences are presented in the Appendix). In addition, two
mixed lists containing five bizarre and five common sentences were cre-
ated. Across these two lists, each word triplet appeared once in a bizarre
sentence and once in a common sentence. Each of these
lists was
presented
to half of the subjects in the mixed list group.
Each sentence was typed in the center of a 3 X 5 in. (7.62 X 12.70 cm)
index card, and subjects in the different conditions were presented with
their appropriate stack of 10 cards. A presentation order was randomly
determined and all subjects were presented the 10 noun triplets in the
same order. In the mixed list conditions, the order of bizarre and common
sentences was randomly determined with the restriction that no more
than two sentences of one type appear adjacently.
To assess our success in generating the bizarre materials, a group of
12 subjects was asked to rate on a 5-point scale the degree to which the
relation between the three underlined nouns in the sentence was "unusual
or atypical." As expected, subjects rated the bizarre sentences (M = 4.4)
to be significantly more unusual than the common sentences {M = 1.6),
t(l\) = 10.54.
Procedure. To ensure an incidental learning situation, subjects were
told that the purpose of the experiment was to investigate differences
between individuals in imaging ability. For each sentence, they were asked
to form an interactive mental image of the sentence that included the
three underlined nouns and to rate the vividness of that image on a 5-
point scale. Subjects were instructed to proceed through their stack of
cards one at a time. They were told to work at their own rate but they
were also informed that we were timing their performance with a stop-
watch (again, to look at individual differences in imaging ability). All
subjects were also asked to be certain to form a mental image before they
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
56
MARK
A.
McDANIEL
AND
GILLES
O.
EINSTEIN
rated
the
vividness
of the
image.
After imaging
and
rating
the
sentences,
subjects
were given
a 30-s
distractor
task
which required
them
to
count
backwards
by
threes
from
a
three-digit
number.
When
this
was
completed,
subjects
were
asked
to
recall
all of the
underlined
words
in the
sentences;
they
were
allowed
3 min for
this
free-recall
task.
Results and Discussion
For all analyses contrasting the between- and within-list effects
of the imagery variable, Erlebacher's (1977, 1978) analysis of
variance procedure was used. In this procedure the mean squares
of the main effects and the interaction are calculated in the nor-
mal fashion. The error terms for these effects are derived by
calculating three additional sources of
variation:
(a) the variation
due to subjects using only the data from the within-subjects (list)
portion of the experiment, (b) the variation due to the interaction
between subjects and treatments using only the data from the
within-subjects (list) portion of the experiment, and (c) the vari-
ation due to subjects within groups using only the data from the
between-subjects (list) portion of the experiment. As applied to
the present design, the appropriate error term for the main effect
of bizarreness and the interaction between bizarreness and type
of design is the weighted average of the mean squares of b and
c. The appropriate error term for the main effect of design type
is the weighted average of the mean squares of a and c. The alpha
level was set at .05.
Processing. An analysis of the vividness ratings indicated that
the images formed from the common sentences (M = 1.43) were
more vivid than the bizarre images (M - 2.84), ^(1,43) = 83.10,
MSC = .46. Neither the main effect of design type nor the inter-
action approached significance (Fs <
1).
To examine the amount
of time necessary to form bizarre and common images, a single-
factor analysis of variance
(ANOVA)
was performed on the total
processing time for the three different groups. This processing
time measure represents the total time it took subjects to image
and rate all 10 sentences. Because of experimenter error, pro-
cessing times were not available for 11 subjects and, consequently,
the analysis was performed on the remaining 15-17 subjects in
each group. The analysis indicated that processing time differed
among the groups, F\2, 46) = 3.39, MSe = 664. A Newman-
Keuls analysis revealed that subjects in the common condition
{M = 88.3 s) took significantly less time to form and rate images
than subjects in the bizarre (M
=
117.4 s) and within (M
=
111.6
s) conditions, and these latter two groups did not differ reliably.
Hence, it appears that bizarre images require more processing
time.
Recall. Recall was initially examined by calculating the pro-
portion of the underlined items recalled, and these data are pre-
sented in Table 1 as a function of type of imagery and design.
The
ANOVA
indicated that there was no main effect of design
type {F < 1). Items from the bizarre images were better recalled
than items from the common images, F(l, 39) = 7.84, MSe =
.03,
but this effect was qualified by a significant interaction be-
tween design type and type of
imagery,
F(l, 39) = 13.32, MSe =
.03.
A planned comparison of this interaction indicated that
bizarre images were better recalled than common images only
for the within-list manipulation, F\\,39) = 20.80, MSK = .03.
There was no evidence for higher recall of bizarre images when
bizarreness was manipulated between lists (F < \).
To localize the effects of bizarreness, the recall protocols were
Table 1
Experiment I: Mean Proportion of Items Recalled, Mean
Proportion of Sentences Recalled, and Mean Number
of Items Per Sentence Recalled as a Function of Type
of Imagery and Design
Design type
Between-list
Recall
Sentences
Items per sentence
Within-list
Recall
Sentences
Items per sentence
Bizarre
.52
.63
2.54
.62
.75
2.50
Common
.55
.60
2.77
.38
.47
2.42
analyzed for the proportion of sentences recalled and the number
of items per sentence recalled, and these data are also presented
in Table 1. For the former measure, a sentence was scored as
recalled if at least one word from that sentence was recalled. As
is evident in Table 1, the results of this measure paralleled that
of the overall recall measure. That is, there was a main effect of
imagery, F(l, 37) = 13.25, MSt = .04, no main effect of design
type,
F <
1,
and a reliable interaction between these two variables,
F(\, 37) = 8.62, MSe = .04. Furthermore, planned comparisons
of the interaction indicated that more bizarre sentences were
accessed than common sentences in the within-list design, F(\,
37) = 21.62, MSC = .04, but not in the between-list design, F <
1.
An
ANOVA
performed on the number of items per sentence
recalled revealed no reliable effects of imagery, design type, or
their interaction, F < 1, F\U 31) = 2.36, F\U 45) = 3.24, re-
spectively. Hence, these analyses indicated that the mnemonic
advantage of bizarre imagery in the within-list design was due
to enhanced accessibility to the sentences or images and not to
increased accessibility to the items within the sentences.
The primary result of this experiment was the direct dem-
onstration of the interaction between type of imagery and design
type.
Bizarre imagery produced higher recall than common im-
agery in a mixed list but not in an unmixed list design. Further,
the greater recall resulting from bizarre imagery was due entirely
to superior access to the images, indicating enhanced retrieva-
bility of bizarre images. Once accessed, however, the bizarreness
of the image does not seem to affect how well the individual
items within the image are recovered. This view is consistent
with previous work using mixed lists. Higher recall of bizarre
images has typically been shown in studies using free recall
(Merry, 1980; Merry & Graham, 1978; Wollen & Cox, 1981a,
198
lb) but not in studies using cued recall (Senter & Hoffman,
1976;
Wollen & Cox, 1981a, 1981b; Wollen, Weber, & Lowry,
1972).
The present findings are important because they run counter
to existing claims that bizarreness is an irrelevant variable in
mnemonic imagery (Kroll, Schepeler, & Angin, 1986, this issue;
Postman, 1975; Senter & Hoffman, 1976; Wollen et al., 1972).
These results also provide some information on the important
factors underlying the bizarreness effect. As outlined in the In-
troduction, Wollen and Cox (1981b) have suggested that the bi-
zarreness effect obtained in several recent mixed list designs is
not due to the nature of the design, but rather to the particular
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
BIZARRE
IMAGERY
AND
DISTINCTIVENESS
57
method of constructing the materials in these studies. Essentially,
they argue that when bizarre noun pairs are formed by inter-
changing highly associated common noun pairs, an intersentence
cuing bias is created that favors the bizarre materials. In the
present study, the same word triplets were presented to all sub-
jects,
resulting in a constant level of intersentence cuing across
all conditions. Therefore, the intersentence cuing explanation is
untenable for our data. Another possible explanation that can
be rejected is Kroll et al.'s (1986) suggestion that previous dem-
onstrations of the bizarreness effect resulted from failures to
control the degree of interaction among the items within the
images. Although we did not directly measure the degree of in-
teraction in the sentences used in the study, Kroll et al.'s (1986)
view does not predict the obtained interaction between type of
imagery and design type. If there were a greater degree of inter-
action among the items in the bizarre images, then the bizarreness
effect should have emerged in both the within- and between-list
manipulations.
The results of this study are consistent with both the attentional
and distinctiveness views. Although the processing time data in-
dicate that subjects spend more time forming bizarre images, it
is not clear that this is a necessary condition for obtaining the
bizarreness effect. Experiments 3-5 directly address these views,
as neither can be unambiguously supported or rejected at this
point.
Experiment
2
In the first experiment, subjects were presented with sentences
(rather than pictures of interacting objects), and it was assumed
that the vividness rating task forced subjects to form images of
the underlined words in the sentences. Although similar rating
procedures have been used in past research investigating bizarre
imagery (e.g., Wollen & Cox, 1981b), it is possible that our task
was ineffective in getting subjects to form images. Subjects may
have ignored our explicit instructions to form images before rat-
ing the sentences and instead may have semantically processed
the items. If this were the case, then semantic and imagery ori-
enting activities should produce similar results. To explore this
possibility, subjects in this experiment were asked to perform
either semantic or imaginal processing on the mixed list used in
Experiment 1.
Method
In
this
2x2
mixed
factorial
design,
type
of
materials
(bizarre
and
common)
was
varied
within
subjects
and
type
of
processing
(semantic
and
imaginal)
was
varied
between
subjects.
The
vividness
rating task
from
Experiment
I was
used
to
encourage imaginal
processing,
and se-
mantic
processing
was
induced
by
having
subjects rate
the
bizarreness
of the
sentences.
For
this
latter
task,
subjects
were
asked
to
rate
the
degree
to
which
the
relation
among
the
three underlined
words
in
each
sentence
was
unusual.
In
both
rating
tasks,
subjects
indicated
their
decisions
on a
5-point
scale.
Subjects were
told
to
work
at
their
own rate, and
their
speed
of
rating
was not
timed.
All
other
procedures
were identical
to
those
used
in the
first
experiment,
and 14
subjects
were
assigned
to
each
of the two
groups.
Results and Discussion
Processing. As in Experiment 1, an analysis of the vividness
ratings for the subjects in the image rating group indicated that
Table 2
Experiment 2: Mean Proportion of Items Recalled, Mean
Proportion of Sentences Recalled, and Mean Number of Items
per Sentence Recalled as a Function of
Type
of Materials and
Type of Processing
Type
of
processing
Imaginal
Recall
Sentences
Items
per
sentence
Semantic
Recall
Sentences
Items
per
sentence
Type
of
material
Bizarre
.63
.74
2.51
.42
.52
2.43
Common
.40
.48
2.50
.51
.57
2.72
the common images (M
=
1.26) were more vivid than the bizarre
images (M = 2.95), F\i, 12) = 28.44, MSt = .71. The subjects
in the semantic processing group rated the bizarre sentences (M -
4.4) as more unusual than the common sentences (M = 1.6),
F\l, 12) = 437.56, MSe = .13. As in the first experiment, two
mixed lists were used, and across the two lists each word triplet
appeared once in a bizarre context and once in a common con-
text. In each of the analyses just mentioned, list was included as
a factor and there were no reliable main effects or interactions
involving the list variable.
Recall. Analyses were performed on the proportion of
underlined words recalled, the proportion of sentences accessed
(a sentence was scored as accessed if one or more words were
recalled from that sentence), and the mean number of words per
sentence recalled. These recall measures are presented in Table
2.
Each of these measures was included in a 2 X 2 mixed
ANOVA
and the alpha level was set at .05.
An analysis of the proportion of underlined nouns recalled
revealed that the main effects of type of materials and type of
processing were not reliable, F(
1,
26) = 2.68 and F < 1, respec-
tively. There was, however, a significant interaction between these
two variables, F(\, 26) = 14.34, MSt = .02. As shown in Table
2,
subjects who performed the vividness rating task recalled a
greater proportion of items from the bizarre sentences than from
the common sentences, F(\y 26) = 14.80, MSe = .02. With the
semantic rating task, a greater proportion of items was recalled
from the common sentences than from the bizarre sentences,
but this difference was not reliable, F(\, 26) = 2.37.
The analysis of the proportion of sentences recalled revealed
a main effect of type of materials, F{\, 26) = 5.98, MSe = .03,
and a significant interaction between type of materials and type
of processing, F(\t 26) - 12.48, MSC = .03. The main effect of
type of processing was not reliable (F < 1). As can be seen in
Table 2, subjects who engaged in vividness rating recalled more
bizarre sentences than common sentences, F\l, 26) = 17.88,
MSe = .03. This result confirms the one obtained in Experiment
1,
and again indicates that bizarre images are more accessible
than common images. For subjects who performed the semantic
processing task, there was no difference between the recall of
bizarre and common sentences (F <
1).
The analysis of the num-
ber of items recalled per sentence produced no main effect of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
58MARK A. MCDANIEL AND GILLES O. EINSTEIN
type of processing, a reliable effect of type of material, and a
significant interaction between these two variables, F <
\,F{\,
26) = 4.49, MS, = .06, and F(\, 26) = 5.43, MSe = .06, respec-
tively. For vividness rating subjects, the number of items per
sentence recalled was nearly equal for bizarre and common sen-
tences (F <
1).
Semantic processing subjects, however, recovered
more items from the common sentences than from the bizarre
sentences, F(\? 26) = 9.93, MS9 = .06.
The two major results of this study were (a) the replication of
the bizarreness effect obtained in Experiment
1
with the vividness
rating task and (b) the finding that different results occurred with
the semantic and imagery processing tasks. Thus, it appears that
the higher recall of bizarre items in a within-list design depends
on imagery instructions. It seems that vividness rating instruc-
tions lead subjects to engage in something other than semantic
processing, and there is no reason to suspect that this processing
is not imagina! in nature. Our results also suggest that some of
the past failures to obtain the bizarreness effects in within-list
designs may be related to the type of orienting activities used
during learning. For example, in the first experiment reported
by Kroll et al. (1986, this issue), subjects were asked to form
images of presented sentences and to rate the images for vividness.
Additionally, subjects were asked to rate the bizarreness of the
sentences and to rate the degree of interaction among the nouns
in the sentences. In their second experiment, subjects were asked
to form images and then perform the interaction rating task.
Although speculative, it seems that the bizarreness and interac-
tion tasks primarily encouraged semantic processing. Thus, Kroll
et al.'s failure to find a recall advantage for bizarre sentences is
not surprising. Our results suggest that the nature of the orienting
activities can influence the relative recall of bizarre and common
materials and that an emphasis on imagery is essential for pro-
ducing the bizarreness effect.
Experiment 3
Given our consistent findings that bizarre images are better
recalled than common images in a mixed list design, the purpose
of this experiment was to test predictions derived from the at-
tentional and distinctiveness explanations of this effect. In this
experiment, all subjects were presented with a mixed list of sen-
tences and later they were asked to recall the underlined words
from the sentences. All subjects were told to form images of
these sentences and to rate the vividness of the images. Half of
the subjects processed the sentences at their own rate (self-paced),
and the other subjects were forced to process the sentences at a
fixed rate (experimenter-paced). According to the attentional
theory, bizarre materials should be recalled better than common
materials only when they receive extra processing or attention
(i.e.,
the self-paced condition). On the other hand, according to
the distinctiveness hypothesis, recall should be higher for bizarre
materials under both the self-paced and the experimenter-paced
conditions. That is, there should be a memory advantage for
bizarre materials even when the bizarre and common images
receive equal attention (i.e., the experimenter-paced condition).
Method
Subjects. The 80 subjects were volunteers from the general student
population at Furman University. Subjects were tested individually and
the experimental sessions lasted about 30 min.
Design. The design of this study was a 2 X 2 mixed factorial, including
the variables of type of imagery (bizarre, common) and type of presen-
tation (subject-paced, experimenter-paced). The type of imagery variable
was manipulated within subjects, and the type of presentation variable
was varied between subjects. Subjects were randomly assigned to the
self-
and experimenter-paced conditions.
Materials and procedure. All subjects were presented with a mixed
list of six common and six bizarre sentences. Each sentence contained
three underlined words. The materials were the same as those used in
the mixed list portion of Experiment 1 (with the addition of a common
and a bizarre sentence for a total of 12 sentences; these are presented in
the Appendix). Two mixed lists were constructed, and across these two
lists each word triplet appeared once in a bizarre sentence and once in
a common sentence. Each of the mixed lists was presented to half of the
subjects in each condition. A presentation order was randomly determined
(with the restriction that no more than two sentences of one type appear
adjacently), and all subjects were presented with the word triplets in the
same order.
As in Experiment I, subjects were told that the goal of the experiment
was to study individual differences in imagery ability. The sentences were
presented to subjects on 3 X 5 in. (7.62 X 12.70 cm) index cards and
each subject was asked to form an interactive image of each sentence
and to rate the vividness of each image on a
5-point
scale. Subjects in
the self-paced condition were told to proceed through the index cards,
one at a time and at their own rate. The processing time for each sentence
was recorded and this time included the time to form the image and to
rate the vividness of the image. Subjects in the experimenter-paced con-
dition were allowed a total of 8.5 s to form an image and to rate the
vividness of the image for each sentence. This 8.5-s interval was signaled
to subjects with a beep from a tape recorder and subjects were given
several practice trials to familiarize them with the length of this interval.
After rating the sentences, all subjects were given a
2-min
distractor task
which involved solving simple math problems. Subjects were then given
4 min to free recall the underlined words in the sentences.
Results and Discussion
The total number of underlined words recalled, the number
of sentences recalled (as in Experiments
1
and 2, a sentence was
scored as recalled if at least one word from that sentence was
recalled), and the number of items per sentence were tabulated
for each subject.1 These measures are presented in Table 3, and
each of these measures was subjected to a 2 X 2 mixed
ANOVA.
The alpha level for all statistical tests was set at .05.
The analysis of the total number of words recalled revealed a
main effect of imagery type, F(\, 78) = 13.89, MSe = 8.44, no
main effect of type of presentation, F < 1, and no reliable in-
teraction between these two variables, F([, 78) = 1.01. As shown
in Table 3, items from the bizarre sentences were better recalled
than items from the common sentences in both the self-paced
and experimenter-paced conditions.
As in the earlier experiments, further analyses indicated that
the bizarreness effect was due entirely to the number of sentences
accessed. As shown in Table 3, this measure produced the same
pattern of results as that obtained with the total recall measure.
That
is,
there was a main effect of type of imagery which indicated
that more bizarre sentences were recalled than common sen-
1 The analyses in Experiments 3-5 were performed on the total number
of words recalled and the number of sentences recalled. Given that the
number of items presented varied across experiments, these measures
are presented as proportions in all of the tables.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
BIZARRE IMAGERY AND DISTINCT1VENESS59
Table 3
Experiment 3: Mean Proportion of Items Recalled, Mean
Proportion of Sentences Recalled and Mean Number
of hems per Sentence Recalled as a Function of
Type
of Imagery and Type of Presentation
Type of presentation
Self-paced
Recall
Sentences
Items per sentence
Experimenter-paced
Recall
Sentences
Items per sentence
Type of imagery
Bizarre
.54
.67
2.4
.49
.67
2.3
Common
.42
.51
2.3
.42
.51
2.4
tences, F(\, 78) = 20.97, MSC = 1.42. Neither the main effect
of type of presentation nor the interaction between type of pre-
sentation and type of imagery was significant (both Fs < 1).
Another analysis was performed on the number of items per
sentence recalled. As is evident in Table 3, this measure was
unaffected by variations in type of imagery (F < 1) and type of
presentation (F < 1). Also, there was no significant interaction
between these two variables, F\l, 78) = 1.77.
The processing times of subjects in the self-paced conditions
were analyzed to determine if these subjects gave more attention
to the bizarre sentences. Because of experimenter error, pro-
cessing times were not available for two subjects and, therefore,
the analysis was performed on the remaining 38 subjects in the
self-paced condition. The analysis indicated that subjects devoted
reliably more processing time to the bizarre sentences (M =
13.8 s) relative to the common sentences (M = 11.1 s), /(37) =
2.39. It should be noted that the mean processing times for the
bizarre and common sentences were considerably higher than
the experimenter-paced rate of 8.5 s. Thus, it appears unlikely
that experimenter-paced subjects were able to do anything other
than process the presented sentences during these 8.5-s intervals.
In addition to providing further support for the mnemonic
effectiveness of bizarre imagery in mixed lists designs, the present
experiment provided data that are difficult to incorporate within
the attention framework. According to this framework, there
should be a recall advantage of bizarre over common imagery
only when the bizarre images receive more processing time. In
this experiment, bizarre imagery produced higher recall than
common imagery under both self-paced and experimenter-paced
conditions. Under self-paced presentation conditions, subjects
devoted more processing time to bizarre, relative to common,
images and these results are in accord with the attentional
framework. The results obtained with experimenter pacing,
however, are inconsistent with the attentional explanation. The
bizarreness effect was found despite equal processing time for
the bizarre and common images. Thus, it appears that greater
processing time for the bizarre images is not a necessary require-
ment for producing the bizarreness effect. Given the important
theoretical ramifications of this result, Experiment 4 was per-
formed to provide additional information on the importance of
differential processing time for the bizarreness effect.
Experiment 4
In this experiment processing time for the mixed list of sen-
tences was always experimenter-paced. To assess directly the ef-
fects of additional processing time on our dependent measures,
two presentation rates were utilized (7 s and 14
s).
A dissociation
between the pattern of the effects of bizarre imagery and the
pattern of the effects of presentation rate would further damage
the processing time explanation of bizarre imagery effects. Also,
a positive effect of bizarre imagery at the 7-s presentation rate
would be difficult to reconcile with an attentional explanation
because our previous work with the sentences used in this study
indicates that subjects use on the average approximately 6.5 s to
form an image of the common sentences (McDaniel & Einstein,
1983).2
With a 7-s presentation rate it appears that subjects would
have little or no extra time to use for reviewing previously pre-
sented bizarre sentences. Finally, in this experiment we included
an additional memory measure: a yes-no recognition test of the
target words in each sentence.
Method
Subjects. Subjects were 28 introductory psychology students partic-
ipating for extra course credit. Half of the subjects were assigned a 7 s
per sentence presentation rate, and the other half of the subjects were
assigned a 14 s per sentence presentation rate. Imagery type was manip-
ulated within subjects.
Procedure. A microcomputer controlled the presentation of the in-
structions and the sentences, and it was also used to collect the imagery
ratings. Each subject was seated in front of a CRT and an accompanying
keyboard. After reading the instructions (which included a description
of the presentation rate), the subjects initiated presentation of the set of
sentences by pressing the space
bar.
The sentences were the same as those
used in Experiment 3, and the two mixed lists were used equally often
in each presentation rate condition. Upon the appearance of a sentence
on the screen, subjects were asked to form a mental image of the sentence's
intent. After the allotted time (7 or 14 s), the sentence was replaced by
a 5-point rating scale. The subject then rated the vividness of his or her
mental image by entering a number from
1
(clear, vivid image) to 5 {not
very clear image). Subjects were given 4 s to enter the rating, after which
the next sentence was presented. The order in which sentences were pre-
sented was randomly determined for each subject. After all the sentences
had been presented, subjects were required to work on a sheet of math
problems for 2 min. Subjects were then administered a free-recall test of
the target items. They were given 4 min for the recall task. Next, subjects
were presented with a yes-no recognition test. For this test, the 36 target
words from the sentences and 56 distractor nouns semantically related
to the targets were randomly ordered and typed on a sheet of paper. For
each subject, 28 of the distractors were semantically related to items from
the bizarre sentences, and 28 of the distractors were semantically related
to items from the common sentences (e.g., the distractor SHRIMP was
considered to be a semantic associate of the target item
LOBSTER;
the
distractor
CARPET
was considered to be a semantic associate of the target
FLOOR).
Four minutes were also given for this test. After completing
the recognition test, subjects were asked if they expected to be tested on
their memory for the words in the sentences, and if
so,
if they had done
2 The times discussed refer to the number of seconds taken to process
and form an image of each sentence; they do not include the time involved
in recording the vividness rating. In Experiment 3 the processing times
included time to record the vividness rating; this accounts for the several
additional seconds allowed in Experiment 3.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
60
MARK
A.
MCDANIEL
AND
GILLES
O.
EINSTEIN
anything
in
addition
to the
imagery task
to
prepare
for the
test.
One
subject
in the 14-s
group
and 2
subjects
in the 7-s
group
indicated
that
they
had
expected
a
memory
test
and had
tried
to
prepare
for it, and
these
subjects
were
replaced.
Results and Discussion
For each dependent measure, a two-factor mixed
ANOVA
was
performed, with imagery type as the within-subjects variable and
presentation rate as the between-subjects variable. For all analyses
the rejection level was set at .05.
Vividness
ratings.
No subject failed to enter a vividness rating
for any sentence within the allotted time interval. As in Exper-
iment 1, bizarre images were rated as significantly less vivid than
common images, F\i, 26) = 14.09, MSC - .85 (average ratings
were 2.53 and 1.61, respectively). Also, the images tended to be
rated as more vivid with the 14-s presentation rate than with the
7-s rate, F(l, 26) = 3.33, p < .10 (means were 1.87 and 2.27,
respectively), and there was no significant interaction between
presentation rate and imagery type (F < 1).
Recall. The mean proportion of target words recalled, the
mean proportion of sentences recalled (scored as in Experiments
I and 2), and the mean number of items per sentence recalled
as a function of imagery type and presentation rate are displayed
in Table 4. Items from the bizarre sentences were recalled sig-
nificantly better than items from the common sentences, F(l,
26) = 26.99, MS, = 9.53, and items presented at the 14-s rate
were better recalled than items presented at the 7-s rate, F(\,
26) = 9.01, MSe = 9.71. There was no interaction between pre-
sentation rate and imagery type (F < 1).
Consistent with the previous experiments, the bizarreness effect
was totally due to the number of sentences accessed. Significantly
more bizarre sentences were accessed than common sentences,
F\\, 26) =
30.33,
MSC = 1.41, but there was no significant effect
of imagery type for the number of words per sentence recalled
(F < 1). The presentation rate effect, however, was due to sig-
nificant increases at the 14-s rate for both the number of sentences
accessed and the number of words per sentence recalled, F\l,
26) = 4.87, MSe = .83, and F(l, 26) = 7.17, MSe = .66, respec-
tively. Paralleling the total recall analysis, presentation rate and
imagery type did not significantly interact for either the sentence
access or words per sentence scores.
These results, along with those of Experiment 3, strongly sug-
gest that the mnemonic benefits of bizarre imagery in a within-
list design are not due to increased processing time afforded the
bizarre material. The mnemonic effects of bizarre imagery in
all of the experimenter-paced conditions were strikingly similar
to the effects of bizarre imagery in the self-paced conditions. The
mnemonic effects of presentation rate, on the other hand, did
not parallel the effects of bizarre imagery (e.g., effects were ob-
tained for the words per sentence recalled for presentation rate
but not for imagery type). This dissociation (between processing
time and imagery effects) weakens the potential argument that,
despite nominally equivalent presentation rates for bizarre and
common sentences, subjects spent time processing the bizarre
sentences while the common sentences were being presented.
This possibility is also unlikely because subjects would seem to
have had neither the time to continue processing a previous sen-
tence (at least for the 7-s presentation condition) nor the incli-
Table 4
Experiment 4: Mean Proportion of Items, Mean Proportion of
Sentences, Mean Number of Items per Sentence Recalled,
Mean Recognition Score, and Mean Proportion of Hits
and False Alarms (FA) as a Function of
Imagery Type and Presentation Rate
Presentation rate"
7sRecall
Sentences
Items per sentence
Recognition1"
Hits
False alarms
14 s
Recall
Sentences
Items per sentences
Recognition
Hits
False alarms
Type of imagery
Bizarre
.45
.58
2.3
.65
.66
.04
.57
.63
2.7
.73
.74
.02
IHit) - iXFA)
Common
.19
.25
2.0
.61
.63
.06
.35
.38
2.8
.62
.63
.02
1 - P(FA)
nation to continue processing a previous sentence (subjects were
not forewarned about the memory test).
Recognition. The mean proportions of hits and false alarms
are shown in Table 4. Also included in Table 4 is a recognition
memory score calculated from the hits and false alarms that
corrects for guessing (Zechmeister & Nyberg, 1982, p. 204).3
False alarms for the bizarre items were calculated using the 28
distractors that were semantically related to the bizarre items,
and the false alarms for the common items were calculated using
the remaining 28 distractors that were semantically related to
the common items. Note that the relatedness of the distractors
to the targets did not systematically vary as a function of imagery
type because the particular targets (and, thus, the particular dis-
tractors) assigned to each imagery condition were counterbal-
anced across subjects. Imagery type did not significantly affect
the number of hits, number of false alarms, or the recognition
score. Presentation rate also did not significantly affect the num-
ber of hits or the recognition score, but there were significantly
more false alarms at the 7-s rate than at the 14-s rate, F[l, 26) =
4.81,
MSe = 2.32. No significant interactions between imagery
type and presentation rate were found. Similar to the recall data,
then, the effects of imagery type and presentation rate did not
parallel one another.
Thus,
the results indicate differential effects of increased pro-
cessing time (i.e., the presentation rate variable) and bizarre im-
agery. Increased processing time appears to affect several aspects
of memory: retrievability (sentence access scores) and accessibility
of individual items within an event (words per sentence scores
! Recognition Score = PjBit) - f(False Alarm)
1 - /"(False Alarm)
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
BIZARRE IMAGERY AND DISTINCT! VENESS
61
and
false alarms). The mnemonic effects of
bizarre
imagery,
on
the
other
hand,
appear to be limited to
retrievability
of
the
image;
there
seem to be no effects on
accessibility and/or discriminability
of individual
items within the
image.
An
attentional explanation
of bizarre
imagery effects, therefore, appears to be neither suf-
ficient
nor
necessary.
The
remaining position, a
distinctiveness
view, is not incon-
sistent
with the present findings, but direct support for this view
has
not yet been provided. According to the distinctiveness
framework,
the critical feature for
enhanced
memory
of bizarre
images
is not a within-list design per se, but rather that some
contrast
is provided that enhances the
distinctiveness
of
the
bi-
zarre
materials.
If this
is the
case,
it should be
possible
to produce
better
memory after bizarre imagery in a between-list design if
some
part of the learning session provides a contrast for the
bizarre
encodings.
Experiment 5 was performed to test this pre-
diction.
Experiment
5
In
this experiment we manipulated type of
imagery
as a be-
tween-list
variable,
and we also
manipulated
the context in which
the
imaged material was to be
remembered.
Some subjects were
required
to
image
and
recall
an
additional list of bizarre
sentences,
some
subjects were required to image and recall an additional
list
of
common
sentences, and some subjects were required to
solve
math problems and recall the solutions. If
distinctiveness
underlies
the effects of
bizarre
imagery,
then the
bizarreness
effect
should
be found when the intervening activity involves learning
common
material.
We
assumed that
encoding intervening images
that were common
would
provide
the
necessary contrast by which
bizarrely
encoded sentences would become
distinctive,
but that
intervening bizarre images
and
intervening encoding of numerical
information
would not
provide
the
contrast necessary
to
enhance
the distinctiveness
of
originally
encoded bizarre
sentences.
One
other
feature
of this
experiment
deserves
comment.
Thus far, all
of
our results rest on one set of
materials.
For this experiment,
we
constructed another set of
sentences
to be used in addition
to
our previous sentence set.
Method
Subjects. The 108 subjects were introductory psychology students
participating for course credit. Eight of the 18 subjects in each condition
were from Furman University and the remaining subjects were from the
University of Notre Dame.
Design. The design of this experiment was a 2 X 3 factorial in which
the variables of type of imagery on the first list presented (bizarre, com-
mon) and type of intervening material (bizarre sentences, common sen-
tences,
math problems) were manipulated. Both of these variables were
varied between subjects.
All subjects performed a series of four tasks in a specified sequence.
First, they formed interactive images for a set of sentences and rated the
vividness of each image (hereinafter labeled the target list). Second, they
performed an intervening task that involved either intentionally studying
a new set of sentences or solving and studying math problems. Third,
they recalled the material from the intervening task. Finally, they recalled
the capitalized words (the word triplets) from the target list. Thus, the
design enabled us to evaluate the effects of both proactive interference
and retroactive interference on the recall of bizarre and common sentences.
Materials. Four lists of 12 sentences were used in this experiment,
with each sentence containing three capitalized words. Two of the lists
were the unmixed lists used in Experiment 1, with the addition of two
sentences. These two lists contained sentences composed from the same
word triplets, with the word triplets embedded in bizarre sentences for
one list and common sentences for the other list. The other two lists of
sentences were derived from a different set of 12 noun triplets. Again,
an unmixed list of bizarre sentences and an unmixed list of common
sentences were developed from these triplets. The items in this latter set
of word triplets were judged by the experimenters to be semantically
related to the items in the other set of word triplets. All four lists of
sentences are presented in the Appendix. Subjects who were presented
with two lists of sentences (one for vividness rating and one for intentional
learning) received a list of sentences containing a different set of word
triplets on the second presentation. Otherwise, the sets of sentences were
completely counterbalanced across experimental conditions. That is, each
set was used equally often in the image rating and in the sentence inter-
vening task portions of the experiment.
Procedure. A microcomputer was used to present the instructions
and the sentences and also to collect the imagery ratings and processing
times for each sentence. Subjects were seated at a desk, in front of a CRT
and keyboard. They were initially presented with the imagery rating in-
structions used in Experiment 1. After they formed a mental image of
the sentence, they pressed a key which caused the 5-point vividness rating
scale (used in the previous experiments) to appear on the screen. After
entering the appropriate number, the next sentence was presented. Thus,
the presentation of this initial list was self-paced. After all of the sentences
were presented, subjects were given instructions appropriate to their par-
ticular intervening task. Subjects in the math intervening learning con-
ditions were given a sheet containing simple math problems and told to
solve and study these for 3 min. After 3 min, they were given 2 min to
recall as many solutions to the math problems as possible. The other
subjects received a different set of 12 bizarre or common sentences during
intervening learning. The instructions and materials for this segment of
the experiment were also presented on a microcomputer. Each of the 12
sentences was presented for 15 s (a total time of 3 min), and subjects
were told to study for a later recall test by forming an image of each
sentence. When the presentation of the sentences was completed, these
subjects were given 2 min to recall the capitalized words from the inten-
tional learning list. Finally, all subjects were given as much time as needed
to recall the capitalized words from the initial list.
Results
and Discussion
The
recall and processing time results are presented first for
the
target list, and then recall for the
intervening list
is
presented.
The
rejection level for inferring
statistical
significance was set at
.05.
Preliminary
ANOVAS
performed on the recall and processing
time
measures from the target list included sentence set as a
variable.
These
analyses
indicated that there were no main effects
or
interactions due to sentence set; that is, effects of type of
imagery
and type of
intervening
material were
generally
observed
regardless of the particular sets of sentences
used.
Consequently,
the
reported analyses are collapsed over the two sentence sets.
Target
list. Consistent with the previous
experiments,
com-
mon
images (M = 2.02) were rated as more vivid than bizarre
images
(M = 2.56), F\i, 102) = 7.39, MSe = 1.05. There were
no
other reliable effects from the vividness rating
analysis
(both
Fs
< 1). To examine processing times, the mean number of
seconds
taken to form images of the sentences was recorded.
(These
times do not include the amount of time taken to rate
the
vividness of the images.) Processing times were nominally
longer
for
bizarre images
(M = 10.65)
relative
to common images
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
62MARK A. McDANIEL AND GILLES O. EINSTEIN
(M = 9.60), but this difference was not reliable (F < 1). Also,
the intervening task variable had no effect on processing times
(F < 1), and there was no interaction between the intervening
task and type of imagery variables {F < 1).
Using the scoring procedures described in Experiment I, the
mean number of target words, sentences, and items per sentence
recalled were tabulated for the target sentences (i.e., the first list
presented). These measures are presented in Table 5. Each of
these measures was included in a 2 X 3
ANOVA,
with type of
imagery and type of intervening task as the variables. For the
total number of target words recalled, there was a main effect of
intervening
list,
F{2,102) = 32.72, MSt = 29.38, and a Newman-
Keuls analysis indicated reliable differences among all three in-
tervening conditions. As shown in Table 5, a bizarre sentence
intervening task produced more interference than a common
sentence intervening task, and recall was highest following the
math intervening task. There was also a marginally significant
(.10 > p > .05) main effect of type of imagery, F{\y 102) = 3.15,
MSK - 29.38, and more important, a marginally significant in-
teraction between the type of imagery and type of intervening
task variables, F{2, 102) = 2.99, MSt = 29.38. A Newman-
Keuls analysis revealed that bizarre imagery produced higher
recall than common imagery only after intervening learning with
common materials. When subjects performed intervening tasks
with bizarre sentences or math problems, there was no significant
difference between the recall of bizarre and common images. In
fact, after the bizarre sentence intervening task, recall for items
from common sentences was nominally higher than recall of
items from bizarre sentences.
The analysis of the number of sentences (images) accessed
yielded the same pattern of results as that obtained with the total
recall measure. That is, there were reliable main effects of type
of imagery and type of interference and a reliable interaction
between these two variables, F{1, 102) = 4.50, MSe = 3.81, F{2,
102) = 34.22, MSC = 3.81, and F{2, 102) = 3.10, MSt = 3.81,
respectively. A Newman-Keuls analysis indicated that there were
reliable differences among all three interference conditions. As
shown in Table 5, sentence access was lowest following the bizarre
sentence intervening task and highest after the math intervening
task. In general, more bizarre sentences were accessed than com-
mon sentences. According to a Newman-Keuls analysis, however,
this effect was only reliable following an interfering task with
common sentences.
The items per sentence score is a conditional measure which
is dependent upon access to the sentences. Thus, subjects who
did not access any sentences were excluded from this analysis.
Although the effects of type of imagery and of the interaction
were not reliable (both Fs < 1), there was a significant main
effect of type of intervening task,
^1,
94) = 5.93, MSt = .30. A
Newman-Keuls analysis indicated that the number of items per
sentence recalled was lower following bizarre sentence interfer-
ence relative to common sentence and math interference. These
latter two types of intervening tasks did not differentially affect
access to items within the sentences.
These results are consistent with a distinctiveness framework.
By this view, bizarre imagery facilitates memory if it leads to a
distinctive encoding of an event, with distinctiveness referring
to "change against some background of commonality" (Jacoby
& Craik, 1979, p. 3). As predicted by this framework, bizarre
Table 5
Experiment 5: Mean Proportion of
Target
Items and Sentences
Recalled and Mean Number of hems per Sentence
Recalled From the Target List
Sentence type:
Target list
Bizarre
Recall
Sentences
Items per sentence"
Common
Recall
Sentences
Items per sentence15
Bizarre
.13
.18
2.0
.16
.19
2.3
Type of intervening task
Common
.36
.43
2.6
.22
.26
2.6
Math
.45
.53
2.6
.41
.48
2.5
Note.
Subjects
who
did not access any sentences were excluded from the
items per sentence analysis.
rts
by task are as follows: bizarre,
15;
common,
17;
and math, 18.
b /is by task are as follows: bizarre,
14;
common,
18;
and math, 18.
imagery enhanced recall only when intervening learning provided
a background context against which the bizarre images became
distinctive, in particular, a background context consisting of
common images. When intervening learning involved additional
bizarre images, there was no contrastive background by which
the originally encoded bizarre images could become distinctive,
and bizarreness failed to significantly improve recall. In fact, in
this instance the levels of recall for bizarre imagery did not reach
those observed for common imagery. When the intervening
learning involved remembering numerical information, the
background was very dissimilar to the imaginally encoded verbal
information, regardless of imagery type. Again bizarreness failed
to significantly improve recall, with recall being relatively high
for both bizarre and common imagery conditions.
Note that the present enhancement of recall for bizarre images
(when intervening learning consisted of common material) is the
only reported instance in which a bizarreness effect has been
found with a between-list design when using short (less than a
day) acquisition-test intervals. Given that this effect was predicted
by the distinctiveness framework, it provides nontrivial corrob-
oration for the framework. Further, because the pattern of effects
was obtained with two sets of materials, the pattern is probably
not due to any idiosyncratic characteristics of our materials. Nor
is
an attentional explanation of the pattern tenable because there
were no significant differences in processing times between the
common and bizarre sentences. Bizarre sentences did require
nominally more processing time (approximately 1 s) than com-
mon sentences, but the time differential was observed for con-
ditions in which there were no significant imagery effects (the
bizarre intervening list and the math intervening list conditions)
as well as in the condition for which bizarre imagery produced
significantly better recall (the common intervening list condition).
Recall of the intervening
list.
To examine the possible proac-
tive interfering effects of the target list on the recall of bizarre
and common sentences from the intervening lists, a 2 X 2
ANOVA
(including the variables of type of sentence in the intervening
list, type of imagery in the target list) was performed for each of
the dependent measures. These measures were the number of
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
BIZARRE
IMAGERY
AND
DISTINCTIVENESS
63
Table
6
Experiment 5: Mean Proportion of
Target
Items and Sentences
Recalled and Number of Items per Sentence Recalled
From the Intervening List
Type of
images
in the target list
Sentence type:
Intervening list
Bizarre
Recall
Sentences
Items per sentence*
Common
Recall
Sentences
Items per sentence*1
Bizarre
.36
.40
2.5
.54
.55
2.7
Common
.49
.57
2.8
.44
.48
2.7
Note.
Subjects who did not access any sentences were excluded from the
items per sentence analysis.
"
ns
by task are as follows: bizarre,
18;
and common, 18.
b
ns
by task are as follows: bizarre,
18;
and common, 17.
target words, sentences, and items per sentence recalled from
the intervening lists, and these scores are presented in Table 6.
The analysis of the total number of intervening list target words
recalled indicated that there was neither a main effect of type of
sentence, F(\, 68) = 2.13, nor a main effect of type of imagery
of the target list, F < 1. There was, however, a reliable interaction
between these two variables, F(\, 68) = 7.13, MS* = 49.27. A
Newman-Keuls analysis indicated that recall was higher for
common sentences than for bizarre sentences when subjects
formed bizarre images during the initial rating task (i.e., on the
target list). Although bizarre sentences were better recalled than
common sentences when subjects formed common images during
the initial rating task, this difference was not reliable. The analysis
of the number of sentences recalled produced identical results.
Although neither main effects were significant, Fs(I, 68) < 1.31,
there was a significant interaction between the type of sentence
and type of imagery of the target list, i^l, 68) = 7.29, MSe =
5.15. As shown in Table 6, the pattern of this interaction par-
alleled the one obtained with the total recall measure. An
ANOVA
performed on the number of items per sentence recalled produced
no reliable effects of type of
sentence,
images formed during the
initial task, or the interaction, i^l, 67) = 1.92, F < 1, and F(l,
67) - 2.40, respectively.
These results do not clearly resolve whether material learned
prior to the encoding of bizarre sentences will serve as a back-
ground context against which the bizarre sentences will become
distinctive. Perhaps the intentional nature of the intervening
learning task caused subjects to use learning strategies other than
the instructed imagery strategy (McDaniel & Kearney, 1984),
thus contaminating possible imagery effects. Or perhaps at-
tempted retrieval of the background material is necessary to
optimize the background material's contrastive effects. The results
of the studies presented herein indicate that significant mnemonic
enhancement of bizarre imagery is observed only when the con-
tents of common images have been previously recalled or are
concurrently being recalled. Clearly, unambiguous specification
of the influence of material learned prior to the target images
must await further research.
Another issue that deserves further exploration is the condi-
tions under which common images are better recalled than bi-
zarre images. The significant recall advantage for common sen-
tences when subjects were exposed to bizarre sentences in the
initial list,