It Test
Resume:
JOURNAL OF MEMORY AND LANGUAGE
0014
ARTICLE NO.
Repeated Exposure to Suggestion and False Memory:
The Role of Contextual Variability
KAREN J. MITCHELL MARIA S. ZARAGOZA
AND
Kent State University
Although it has been well established that a single exposure to suggestion can result in the
creation of false memories for suggested events, little is known about the effects of repeated
exposure to suggestion. Zaragoza and Mitchell (in press) demonstrated that repeated exposure
to postevent suggestion increased subjects tendency to misremember witnessing the suggested
information. The experiments presented here examined the possibility that increasing contextual
variability between the repeated exposures would exacerbate this effect by impairing subjects
ability to discriminate accurately the precise source of the suggested items. Results from two
experiments show that increasing variability by changing surface features (i.e., modality) exagger-
ated the deleterious effects of repeated exposure to suggestion. Increasing the spacing between
exposures (Experiment 2), however, did not have the same effect. 1996 Academic Press, Inc.
One of the obstacles people face when they ing experienced events that were never actu-
attempt to remember a speci c event is dis- ally experienced. It is now well established
criminating between memories of the target that even single exposures to misinformation
event and other related information in mem- can result in genuine false memories for sug-
ory. For example, eyewitnesses who are called gested events (e.g., Ackil & Zaragoza, 1995;
to testify in a court of law must separate their Belli, Lindsay, Gales, & McCarthy, 1994;
memories of the event they witnessed from Lindsay, 1990; Zaragoza & Lane, 1994; Zara-
their own ruminations about the event as well goza & Mitchell, in press). This paper, how-
as any pertinent information they may have ever, is concerned with illusory memories that
acquired from other sources (e.g., newspaper result when people have been repeatedly ex-
accounts of the event, new information posed to misleading suggestions about an
gleaned from conversations with others). Oc- event they witnessed.
casionally, people s attempts to distinguish Interest in illusory memories is certainly not
between related memories fail. A potential limited to the domain of eyewitness memory.
consequence of such discrimination failures is For example, the possibility that people can
an illusory recollection a memory for hav- be led to create memories of events that never
happened has been attracting considerable in-
terest of late because of its relevance to the
These experiments composed a thesis submitted by
current recovered/false memory debate. At the
Karen J. Mitchell in partial ful llment of the requirements
heart of this debate is the claim that the recent
for the Master of Arts degree, Department of Psychology,
Kent State University. This research was funded by Na- rash of alleged recovered memories of child-
tional Institute of Mental Health Grant MH47858 to Maria
hood abuse are in fact false memories induced
S. Zaragoza. We thank Peggy Dombrowski, Sean Lane,
by certain therapeutic practices (e.g., see Lind-
John Mitchell, and Don Wolf for their patient help with
say & Read, 1994; Loftus & Ketcham, 1994,
stimuli preparation. We are grateful to Michelle Gilmour,
for recent discussions of this claim). Like the
Jeanine Raye, and Emily Roper for their assistance in
collecting data. Address correspondence and reprint re- situation in which a witness is misled, the con-
quests to Maria S. Zaragoza or Karen J. Mitchell, Depart-
cern here is that a suggestion introduced re-
ment of Psychology, Kent State University, Kent, OH
peatedly by a third party might result in a
44242-0001. E-mail: ********@*******.****.***. or
false memory. In fact, it is this opportunity
********@******.****.***.
0749-596X/96 $18.00 246
Copyright 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
AID JML 2464 / A002$$$201 04-11-96 10:32:49 jmla AP: JML
247
CONTEXTUAL VARIABILITY AND FALSE MEMORY
for repeated exposure to suggestion that is misleading information about the event on two
thought to make these types of situations espe- separate occasions. On a nal test (using open-
cially conducive to false memory creation. ended questions), the suggestibility of this
In spite of the recognition that repeated sug- group of subjects was compared to that of two
gestion may encourage the induction of false control groups, each of which was exposed to
memories, surprisingly little is known about misinformation on only one of the two occa-
repetition as a distinct factor in their develop- sions. Warren and Lane (1995) found that
ment. A review of the eyewitness suggestibil- adult subjects who had been exposed to misin-
ity literature reveals, for example, that the vast formation twice were no more likely to misre-
majority of these studies have assessed the member witnessing the suggested information
effects of only a single exposure to misinfor- than were subjects in either of the single expo-
mation (see Lindsay, 1994, for a review). sure groups. Thus, the results of this study
There are a few recent studies that have em- did not support the hypothesis that repeated
ployed repeated suggestion, and they report suggestion increases false memory.
rather striking examples of false memory in There is, however, one study which helps
which subjects claim to remember entire cti- establish a direct link between repeated expo-
tious events, such as getting lost in a mall as a sure to misinformation and the formation of
child (e.g., Ceci, Crotteau Huffman, Smith, & false memory (Zaragoza & Mitchell, in press).
Loftus, 1994; Ceci, Loftus, Leichtman, & We describe this study in detail because it
Bruck, 1994; Hyman & Billings, 1995; Hy- serves as the springboard for the experiments
man, Husband, & Billings, 1995; Hyman & reported here. Zaragoza and Mitchell (in
Pentland, this issue; Loftus & Ketcham, press) employed a procedure in which subjects
1994). Given the apparent facility with which viewed a videotape depicting a home burglary
false memories have been obtained in these and later answered questions about it. Subjects
studies, it is tempting to conclude that repeti- were questioned about the events of the bur-
tion is a powerful means of inducing false glary in chronological order, and this process
memory. However, closer examination of the was repeated three times in succession. Each
methodology employed in these studies re- time through, subjects were asked about the
veals that it is not at all clear to what extent same set of events, although they were asked
repetition might have caused these effects. Be- about different aspects of these events each
cause these studies were designed to mimic time. Some of the questions contained mis-
the complexity of real-world suggestive inter- leading suggestions. For example, when sub-
view techniques, they employed procedures in jects were questioned about a scene in which
which subjects were repeatedly pressed, the thief leaves the house, some subjects were
across multiple sessions, to describe events asked the following question, which presup-
from their childhood that never actually oc- poses the misleading suggestion, gun : As
curred. The locus of the interpretive dif culty the thief was leaving the house, he put his
is that, in all these studies, number of repeated hand on the gun at his waist, looked both ways
suggestions was confounded with several and walked out the door. Did he step out onto
other variables including the passage of time a porch? This question was misleading be-
and demand (cf., Zaragoza & Mitchell, in cause the thief in the video did not have a
press). Thus, the role repetition played in the weapon of any sort. The critical manipulation
creation of these memories is dif cult to dis- was that for each subject some suggestions
cern. occurred once in the course of questioning and
Warren and Lane (1995) employed a very others occurred three times. So, for example,
different procedure one that permits clearer although all subjects answered questions
inferences about the effects of repeated expo- about the thief leaving the house on three sep-
sure to suggestion. In this study, subjects rst arate occasions, for some subjects the mis-
viewed an event and then were exposed to leading suggestion gun appeared in only
AID JML 2464 / A002$$$202 04-11-96 10:32:49 jmla AP: JML
248 MITCHELL AND ZARAGOZA
TABLE 1 in the video, ruling out the possibility that
repetition merely increased subjects belief
MEAN PROPORTION OF RECOGNIZED SUGGESTIONS AT-
that the suggested events transpired (e.g.,
VIDEO (SOURCE MISATTRIBUTION ER-
TRIBUTED TO THE
RORS) AS A FUNCTION OF NUMBER OF EXPOSURES (ZARA- Arkes, Hackett, & Boehm, 1989; Bacon,
GOZA & MITCHELL, IN PRESS)
1979; Begg, Anas, & Farinacci, 1992; Begg &
Armour, 1991; Hasher, Goldstein, & Toppino,
No. of
1977; Schwartz, 1982).
exposures Repetition
effect Why did repeated exposure to suggestion
Condition 1 3 (3 minus 1)
increase source misattribution errors? In at-
tempting to answer this question, it is useful
Experiment 1
to rst consider what is known about source
Immediate Group .21 .40 .19
monitoring processes in general. According to
48-h Delay Group .43 .57 .14
1 Week Delay Group .41 .59 .18 the source monitoring framework (Johnson,
Experiment 2 .15 .28 .13
Hashtroudi, & Lindsay, 1993), memory for
source is an attribution that results from evalu-
Note. Source misattribution errors were de ned as the
ating the characteristics of the underlying
proportion of de nitely yes (Experiment 1) or re-
member (Experiment 2) responses to the In Video? memory representation. These characteristics
probe for recognized suggestions.
re ect the conditions under which the memory
was acquired and include contextual informa-
tion (i.e., spatial/temporal information), sen-
one of the three questions about the thief leav- sory/perceptual detail, and any record of the
ing, whereas for others gun appeared in all
cognitive processes engaged in during encod-
three questions about the scene.
ing. Judgements about source are made by
In order to assess whether subjects had
evaluating the quantity and nature of these
come to misremember witnessing the sug-
characteristics. Errors in source memory may
gested information in the video, in the last
occur for several reasons, including: (1) a
phase of the experiment subjects were tested
memory has characteristics typical of another
on their memory for the source of the sug-
source, (2) there is an absence of characteris-
gested items (e.g., that the thief had a gun).
tics that uniquely specify the item s source,
That is, subjects were asked both (1) whether
and (3) the subject fails to engage in reasoning
they remembered the suggested item from the
that would prevent an error (i.e., utilize gen-
video and (2) whether they remembered the
eral knowledge).
suggested item from the questions. In Experi-
When subjects attempt to answer mis-
ment 1, subjects (who were tested at one of
leading questions about witnessed events they
three retention intervals) were asked to indi-
are likely to think about and imagine the infor-
cate their con dence in their source judge-
mation (both accurate and suggested) de-
ments. Experiment 2 employed a more direct
scribed in the questions (Zaragoza & Lane,
measure of subjects phenomenal experience
1994). At the time of test, subjects might mis-
by asking them to discriminate between items
attribute their imagined representation of the
they speci cally remembered from the video
suggested item to the video because it contains
and those they believed were in the video (cf.
sensory/perceptual characteristics similar to
Tulving, 1985). Table 1 shows that, at all re-
memories of perceived events (Johnson et al.,
tention intervals, subjects were more likely
1993). With repetition, the image that the sub-
to claim that they de nitely remembered
ject creates of the suggested event may be-
seeing the suggested items in the video when
come increasingly elaborate, detailed, and
they had been exposed to them three times
seemingly real, increasing the likelihood of
than when they had been exposed only once.
misattribution (see Suengas & Johnson, 1988,
Similarly, they were more likely to claim that
they remembered seeing the suggestions for evidence that rehearsing imagined events
AID JML 2464 / A002$$$202 04-11-96 10:32:49 jmla AP: JML
249
CONTEXTUAL VARIABILITY AND FALSE MEMORY
serves to preserve and embellish them; Zara- bile). The effects of contextual variability are
goza & Mitchell, in press). also analogous to the effects of retention inter-
Although the role of imaginal processing in vals exhibited in many paradigms. Both can
the effect of repetition remains unclear, there produce a loss of distinct contextual/source
is at least one other factor that may have con- attributes which in turn leads to generalization
tributed to this effect. Note that each repetition of responding (see Riccio, Ackil, & Burch-
of the suggested information occurred in a Vernon, 1992 for a review; cf., Rovee-Collier,
somewhat different context (i.e., in different 1991).
questions and after several intervening items) Given that contextual variability is an inher-
and this variability may have reduced sub- ent characteristic of repetition, the goal of the
jects ability to discriminate between the video present study was to manipulate that variabil-
and the questions as the source of the sug- ity in order to assess whether it plays a role in
gested information. In other words, it is likely the errors that result from repeated suggestion.
that as a function of having encountered the Speci cally, the present experiments were de-
suggestions in a variety of contexts, the sug- signed to investigate whether increasing the
gested information was highly familiar yet contextual variability of the repeated expo-
lacking in discrete information regarding the sures to postevent suggestions would increase
item s source, leading subjects to overgenera- subjects suggestibility, as measured by their
lize when judging the suggested item s origin. tendency to make source misattribution errors.
Interestingly, William James made a similar The rst experiment employed the same gen-
observation about the consequences of repeti- eral three-phase procedure used by Zaragoza
tion many years ago: and Mitchell (in press). In phase 1, subjects
viewed a brief video of a house burglary. During
phase 2, misleading suggestions were intro-
If a phenomenon is met with, however, too often,
and with too great a variety of contexts, although duced in questions about the video. Within these
its image is retained and reproduced with corre-
questions each subject received some sugges-
spondingly great facility, it fails to come up with
tions once and others three times, while still
any one particular setting, and the projection of it
other suggestions served as never presented con-
backwards to a particular past date consequently
trol items. The innovation introduced here in-
does not come about. (James, 1890/1918, p. 673)
volved varying the context in which the thrice
In fact, empirical evidence that contextual repeated suggestions were encountered by pre-
variability can affect memory via the process senting them in different modalities. For subjects
of generalization is provided by the work of in the Single Modality Group, the three presen-
Rovee-Collier and her colleagues (see Rovee- tations of the suggestions were all in the same
Collier, 1991, for a review of this work). modality (i.e., all in print, all on videotape, or
These studies utilized a contingency proce- all on audiotape), a procedure analogous to that
dure in which infants foot-kicking responses previously used by Zaragoza and Mitchell (in
during training sessions moved an overhead press). However, for subjects in the Mixed Mo-
crib mobile to provide reinforcement. Mem- dality Group, the suggestions were presented
ory for the circumstances of learning were once in each of three different modalities (print,
measured by the infants response to a test audiotape, and videotape), thus increasing the
mobile presented sometime later. The studies contextual variability of the repeated exposures.
demonstrated that infants who were trained In the nal phase, a source memory test was
using variable training stimuli (e.g., different administered. The measure of primary interest
colored crib mobiles at each training session) was subjects tendency to misattribute the sug-
were more likely to generalize the foot-kick- gested items to the originally witnessed event. It
ing behavior to a novel test mobile than were was hypothesized that increasing the variability
infants who had been repeatedly exposed to between repeated exposures would increase
the same stimuli (e.g., same colored crib mo- false memory for the repeated suggestions, as
AID JML 2464 / A002$$$202 04-11-96 10:32:49 jmla AP: JML
250 MITCHELL AND ZARAGOZA
TABLE 2
evidenced by more source misattributions to re-
peated suggestions in the Mixed Modality EXAMPLES FROM THE POSTEVENT QUESTIONNAIRE
Group than in the Single Modality Group. (SUGGESTION: THE THIEF HAD A GUN )
3-Exposure level
EXPERIMENT 1
Later, as he was leaving the house the thief, putting his
Method hand on the gun at his waist, looked both ways and
went out the door. Did he slam the door behind
Subjects and Design
himself?
Before leaving the house the thief checked the gun at
One hundred eighty undergraduates were
his waist and looked both ways to see if anyone was
randomly assigned to either the Mixed Modal-
watching. After he got out the door, did he begin to
ity Group (n 90) or the Single Modality run?
Group (n 90). A 2 (Mixed vs Single Modal- As the thief was leaving the house, he put his hand on
ity) 1 3 (0, 1, or 3 exposures) mixed design the gun at his waist, looked both ways and walked
out the door. Did he step out onto a porch?
was utilized, with modality group as a be-
tween-subjects factor and number of expo-
1-Exposure level
sures to postevent suggestion as a within sub-
jects variable. Subjects participated in partial Later, as he was leaving the house the thief looked both
ways and went out the door. Did he slam the door
ful llment of a course requirement.
behind himself?
Before leaving the house the thief looked both ways to
Materials and Procedure see if anyone was watching. After he got out the
door, did he begin to run?
Subjects were run in small groups (i.e.,
As the thief was leaving the house, he put his hand on
10). They were told that they would be the gun at his waist, looked both ways and walked
participating in an experiment designed to out the door. Did he step out onto a porch?
study memory for complex events. They
Note. These questions represent the sixth question in
were further informed they would view a
each of the three subsets for each exposure level. For the
brief videotape and then be asked some
0-exposure level (control), there was no reference to the
questions about it. gun in any of the questions.
Phase 1 The eyewitness event. Subjects
rst viewed a 5-min videotaped scene taken
the question (see Table 2 for examples). The
from a police training lm. It depicted a bur-
questions of each subset were ordered according
glary of a home by two youths and an ensuing
to the chronology of the video. Thus, ordering
police car chase. The clip was rich in action
was identical for each subset.
and dialogue.
All subjects answered all 36 questions, and,
Phase 2 Misleading postevent questioning.
by inserting suggested items in speci c ques-
Immediately after seeing the video, subjects an-
tions as necessary, number of exposures to
swered the same 36 postevent questions used
suggestions was manipulated within subjects
by Zaragoza and Mitchell (in press). The set of
(see Table 2 for examples). For each subject,
questions was actually composed of three 12-
four suggestions were presented in all three
question subsets. Each question in a subset re-
subsets of questions (3-exposure level), four
ferred to one of 12 unique events in the video
suggestions were presented in only the last
(e.g., the thief entering the home). For each of
subset of questions1 (1-exposure level), and
these events a misleading postevent suggestion
was constructed (see Appendix A). The sugges-
1
In a previous experiment (Zaragoza & Mitchell, in
tions were composed of details, action, and dia-
press) conducted with these same materials we showed
logue which supplemented, rather than contra-
that placement of the single-exposure suggestions (i.e.,
dicted, the information in the scene. The sug- whether in the rst, second, or third subset of questions)
gested items were inserted in the prefatory text had no effect on any of the dependent measures. For this
of the questions and were never the focus of reason, to keep the number of counterbalancing groups
AID JML 2464 / A002$$$203 04-11-96 10:32:49 jmla AP: JML
251
CONTEXTUAL VARIABILITY AND FALSE MEMORY
nally, four suggestions served as never pre- (Time to complete the questions was approxi-
sented control items. Counterbalancing as- mately equal for all groups, about 5 min.) Sub-
sured that all items served at all exposure lev- jects were told that they should answer each
els equally often across subjects. It is im- and every question even if they had to guess.
portant to note that some actually perceived After subjects inquiries were addressed,
items were also manipulated in the questions the rst 12 questions were delivered. When
so that they were presented either once or all subjects were nished, instructions were
three times. Therefore, number of repetitions given for the next subset of questions. Again,
in the questions was not correlated with the these varied depending on how the questions
accuracy of the information. These items, were to be delivered, but were similar to those
however, were not counterbalanced. given for the rst subset of questions. As a
To allow implementation of the modality ruse for the additional questions, subjects were
manipulation, all question subsets were indi- told either: (1) we were interested in investi-
vidually prepared in three modalities, namely, gating the effect of a different modality (it
printed, read on videotape by a male, and, read was named) on people s recall of information
on audiotape by a female. Equal numbers of about a movie (Mixed Modality) or (2) we
subjects in the Single Modality Group were were interested in investigating the effect of
randomly assigned to receive all 36 of their answering additional questions on people s re-
postevent questions in either printed, video- call of information about a movie (Single Mo-
taped, or audiotaped form (n s 30 each). dality). The same procedure followed for the
Subjects in the Mixed Modality Group got nal set of 12 questions. Note that procedure
one 12-question subset in each of the three and instructions were parallel for all groups,
modalities (e.g., 12 questions in print, 12 on thus keeping time lag between question sub-
videotape, 12 on audiotape). Order of modal- sets relatively equal.
ities was counterbalanced so that all modal- When all 36 questions had been delivered
ities occurred equally often at all ordinal posi- subjects engaged in a 7-min ller task.
tions. In all then, counterbalancing assured Phase 3 Source memory test. A surprise
that across subjects, all suggestions occurred source memory test followed the 7-min lled
equally often at all exposure levels in all mo- interval. The probes, which were presented in
dalities and that all modalities were repre- the same randomized order to all subjects,
sented equally at all ordinal positions. were 32 statements read on a cassette recorder
Instructions varied somewhat for each in a male voice, so as to be as different from
group depending on how the rst 12-question all other stimuli exposures as possible. The
subset was to be administered. However, they interitem interval was 8 s. Twelve of these
were constructed to be equated for length and statements contained the critical items (e.g.,
of parallel semantic structure. Depending on The thief had a gun. ). For any one subject,
the group, subjects were told either that they 4 of these were novel items (i.e., 0 exposures)
would hear an audiotape of someone reading and 8 were suggested items that appeared in
them questions about the video they had just the postevent questions only (4 each at 1 and
seen, see and hear someone reading them 3 exposures). The remaining 20 test probes
questions on videotape, or, be given a sheet were ller items that were included solely for
with questions printed on it. Subjects who purposes of ensuring that the test list contained
were to receive the questions via videotape or equal numbers of test probes from each of
audiotape were told they would have 8 s in the four possible source categories (i.e., video
which to answer each question. Written ques- only, questions only, both video and ques-
tions were answered at the subjects own pace. tions, and neither video nor questions). Spe-
ci cally, because the critical items contributed
8 probes from the questions only category
manageable, we embedded all single-exposure sugges-
and 4 probes that were new to the test list,
tions in the last set of questions.
AID JML 2464 / A002$$$203 04-11-96 10:32:49 jmla AP: JML
252 MITCHELL AND ZARAGOZA
the ller items consisted of 8 video only questions here referred to any of the questions
items, 8 both items, and 4 additional they had answered about the burglary video.
new items. It should be noted that ller To ensure that subjects responses accurately
items were chosen with the goal of giving re ected their memory for the source of the
subjects a clear benchmark against which to test items, the experimenter went through each
evaluate their memories of the suggested of the four possible source categories (i.e.,
items, rather than with the goal of assessing video only, questions only, both, neither), and
subjects general source monitoring ability. indicated what the appropriate response pat-
Therefore, these items were selected to be tern should be. For example, subjects were
very obvious members of their source cate- told that if they remembered the information
gory. For example, we chose as video only contained in the test statement from the video
ller items highly salient objects that had been only, they should circle yes in the
visible throughout large segments of the video Video column and no in the Ques-
but were never mentioned in the questions. tions column. They were cautioned that they
Subjects were given both written and verbal must make two responses to each test item.
instructions for the source memory test. They The question of primary interest in this
were told that they would hear 32 statements study was whether subjects memory for the
read to them on a tape recorder at 8-s intervals. source of the repeated suggestions varied as
All subjects were explicitly informed that of a function of contextual variability, and for
the 32 statements they would hear, some con- this reason we present the results for the sug-
gested items only.2 A yes response to a
tained information that was only in the video
of the burglary scene, some contained infor- suggested item in the Video column indi-
mation that was not in the video of the bur- cated a source misattribution error, while a
glary scene but was contained in the questions yes response to a suggestion in the Ques-
they answered, some contained information tions column indicated a correct source
that was in both the video and the questions, judgement. Note that these are not mutually
and nally, some of the test statements con- exclusive judgements. A subject could both
tained information that was in neither the misattribute a suggested item to the video and
video of the burglary nor the questions. In- correctly attribute it to the questions.
cluding this explicit warning about misinfor-
Results and Discussion
mation allows one to rule out the possibility
that subjects would merely believe that the Item recognition. A distinction can be made
misinformation was contained in the video be- between subjects ability to recognize a sug-
cause the experimenter said it was so. Rather gested item as old and their ability to iden-
they were told they were to base their source tify the suggested item s source. We measured
judgements on their memory of the events. item recognition as subjects ability to identify
They were instructed that for each statement a suggestion as being from either the video
they heard they were to answer two questions. and/or the questions. As can be seen in the
Did they remember the information contained top half of Table 3, overall, item recognition
in the test probe: (1) from the video of a bur- was better for suggestions presented three
glary they saw? and (2) from the postevent times than for those presented once (F(1,178)
52.31, MSE .02, p .0001). Of greater
questions?
For each test probe, subjects circled Yes interest is the nding that item recognition did
or No for each of these two questions on not vary between the Single and the Mixed
their answer sheet, which contained two col-
umns, one labeled Video? and one labeled 2
Note also that analysis of the ller item data is not
Questions? Care was taken to differentiate informative given that these items were selected so that
for the subjects what was meant by video and subjects could easily identify their source and perfor-
questions. It was made clear to subjects that mance thus tended toward ceiling.
AID JML 2464 / A002$$$203 04-11-96 10:32:49 jmla AP: JML
253
CONTEXTUAL VARIABILITY AND FALSE MEMORY
TABLE 3 peated exposure to suggestion increased
source misattribution errors as evidenced by
MEAN PROPORTION OF SUGGESTIONS RECOGNIZED AS
more errors in the 3-exposure condition than
FUNCTION OF GROUP AND NUMBER OF EXPOSURES IN
A
EXPERIMENTS 1 AND 2 in the 1-exposure condition (F(1,178)
39.98, MSE .05, p .0001). This result
No. of
replicates the repetition effect reported by Zar-
Exposures
agoza and Mitchell (in press). More impor-
tantly, Fig. 1 clearly shows that although the
Condition 1 3
Mixed Modality Group committed more er-
Experiment 1 rors overall (F(1,178) 4.00, MSE .25,
Single Modality .86 .97
p .05), the detrimental effect of repeated
Mixed Modality .84 .96
exposure was greater for the Mixed than the
Experiment 2
Single Modality Group, as evidenced by a sig-
Single Modality Consecutive .86 .97
Single Modality Spaced .81 .97 ni cant Group X Exposures interaction
Mixed Modality Consecutive .81 .94 (F(1,178) 5.82, MSE .05, p .05).
Mixed Modality Spaced .81 .96
Planned comparisons con rmed that there was
a signi cant repetition effect in both groups
(F(1,89) 8.64, MSE .04, p .04, for
the Single Modality Group; F(1,89) 34.23,
Modality groups (F(1,178) .72, MSE .03,
MSE .05, p .0001, for the Mixed Modal-
p .10), in spite of the fact that contextual
ity Group) and that the locus of the interaction
variability has previously been shown to im-
was a difference between the groups at the 3-
prove item memory (see Greene, 1992, for
exposure level (F(1,178) 7.58, MSE .15,
a review). However, ceiling effects in the 3-
p .01) but not at the 1-exposure level
exposure case may have precluded nding dif-
(F(1,178) .85, MSE .15, p .10). This
ferences between the groups.
nding helps rule out a global confusion ac-
Obviously, one cannot attribute a source to
count of this contextual variability effect. In
an item that one does not remember. There-
other words, it does not appear that Mixed
fore, to control for the item recognition differ-
Modality subjects were more confused in gen-
ences between exposure conditions, the source
eral, but rather the confusion was limited to
data reported below (and throughout) were
those items to which they were actually ex-
conditionalized on item recognition. That is
posed in all three modalities.
to say, for each source (video or questions),
we report the proportion of recognized sug-
gestions that were attributed to that source.
Source misattribution errors. The measure
of primary concern was the extent to which
subjects misattributed the suggested items to
the video.3 Figure 1 shows that, overall, re-
3
Base rates of false alarms (i.e., yes responses to
Video? for items at the 0-exposure level) were very
low (M s .09 and .07 for the Single and Mixed Modality
Groups, respectively) and did not vary between groups
( p .05). Nor did performance on the 0-exposure items
vary on any other dependent variable in either experiment
of the present study (all p s .05). Therefore, since the
primary interest was in determining whether the effects
of repetition vary differentially as a function of contextual FIG. 1. Mean proportion of recognized suggestions attrib-
variability, we focus on the 3-exposure vs 1-exposure uted to the video (source misattribution errors) as a function
level comparison. of group and number of exposures in Experiment 1.
AID JML 2464 / A002$$$203 04-11-96 10:32:49 jmla AP: JML
254 MITCHELL AND ZARAGOZA
TABLE 4 come to affect source judgements in this repe-
tition paradigm, two changes were made in
MEAN PROPORTION OF RECOGNIZED SUGGESTIONS AT-
the methodology. First, we utilized a measure
POSTEVENT QUESTIONS (ACTUAL
TRIBUTED TO THE
SOURCE MEMORY) AS A FUNCTION OF GROUP AND NUM- of source memory that allowed subjects to rate
BER OF EXPOSURES IN EXPERIMENTS 1 AND 2
their con dence in their source memory and
included unsure as a response option (Zar-
No. of
agoza & Lane, 1994, Experiment 5). Because
Exposures
subjects can choose unsure, this measure
Condition 1 3 allows one to partial out the possible contribu-
tion of guessing to performance. Second, we
Experiment 1
sought to assess whether the results of Experi-
Single Modality .91 .99
ment 1 would generalize to a form of contex-
Mixed Modality .85 .95
tual variation which occurs naturally in real-
Experiment 2
Single Modality Consecutive .93 .97 world repetition situations, namely, spacing of
Single Modality Spaced .89 .94 exposures. This was accomplished by intro-
Mixed Modality Consecutive .89 .98
ducing a 5-min lled interval between post-
Mixed Modality Spaced .87 .97
event question subsets. We reasoned that the
lled interval would increase contextual vari-
ability, most notably because subjects re-
Memory for actual source. The top half of turning to the postevent questions after a dis-
Table 4 shows the proportion of recognized tractor task should be in a different frame of
suggestions that were correctly attributed to mind upon return than should subjects who
the postevent questions. Overall, repeated ex- work on the question subsets consecutively.
posure to suggestion improved memory for Four groups were formed by orthogonally
the suggested items actual source (F(1,178) varying modality and spacing. In other words,
34.72, MSE .02, p .0001), a nding there were two groups which replicated those
consistent with those of Zaragoza and Mitch- in Experiment 1 (Single and Mixed Modality
ell (in press). Of greater interest is the nding Consecutive groups) and two which repre-
that Mixed Modality subjects made fewer cor- sented an extension (Single and Mixed Modal-
rect source attributions than Single Modality ity Spaced groups).
subjects overall (F(1,178) 5.90, MSE .04, Method
p .05) (see the top half of Table 4), although
Subjects and Design
the Group X Exposures interaction was not
Subjects were 396 undergraduates who par-
signi cant (F(1,178) .11, MSE .02, p
ticipated in partial ful llment of a course re-
.10). The results imply that subjects who were
quirement. They were randomly assigned to
exposed to suggestions in more varied con-
one of four groups, Single Modality Consecu-
texts had greater dif culty remembering the
tive (n 108), Mixed Modality Consecutive
actual source of all the suggested items, not
(n 108), Single Modality Spaced (n 90),
just those that were repeated in varied con-
or Mixed Modality Spaced (n 90), as de-
texts.
scribed below. This resulted in a 2 (Modality:
Taken together these results support the
Single vs Mixed) X 2 (Spacing: Consecutive
conclusion that contextual variability weakens
vs Spaced) X 3 (Exposure Level: 0, 1, or 3
subjects ability to make accurate source attri-
exposures) mixed design, with Modality and
butions.
Spacing as between-subject factors and Expo-
sure Level as a within-subject variable.
EXPERIMENT 2
Materials and Procedure
Experiment 2 was designed to replicate and
extend the ndings of Experiment 1. To better With the exception of two bogus ller tasks
understand how contextual variability may (described below) and the nature of the re-
AID JML 2464 / A002$$$203 04-11-96 10:32:49 jmla AP: JML
255
CONTEXTUAL VARIABILITY AND FALSE MEMORY
sponse options on the source memory test (de- tasks by using instructions and answer sheets
scribed below), the materials and procedure of comparable complexity and tone as those
were the same as those used in Experiment 1. used in the actual experimental tasks. The
Subjects were informed that they were par- tasks were administered in the same order (vi-
ticipating in an experiment designed to inves- olence rating, word association) to all sub-
tigate various cognitive abilities including jects.
judgement and memory. They were told that As in Experiment 1, a 7-min lled delay
one task would involve watching a short vid- followed the last subset of postevent questions
eotape and answering some questions about for all subjects, after which the surprise source
it, but that there would be a series of different memory test was administered.
tasks included in the experimental session. Source memory test. The test probes and
Except for the two additional ller tasks, procedure were the same as those used in Ex-
all subjects were exposed to the same general periment 1. The instructions were nearly iden-
three-phase procedure used in Experiment 1. tical, except for minute changes necessary to
Subjects in the two Consecutive groups per- facilitate the new answer options (e.g., rather
formed the two ller tasks rst, followed by than tell the subjects to circle yes in the
phase 1 and phase 2, as explained in Experi- Video column if they remembered the item
ment 1. For subjects in the Spaced groups one from the video they were told to . . . circle
ller task was completed after each of the rst one of the yes responses ).
two postevent question subsets. Subjects were given an answer sheet which
The ller tasks were as follows: contained two columns labeled Video and
Violence rating task. Ten 20-s segments of Questions. Each column contained 7-op-
popular music were recorded on cassette with tion Likert-type scales containing the follow-
8-s interitem intervals. Subjects were in- ing responses: de nitely yes, probably
formed they would hear a series of 20-s music yes, maybe yes, unsure, maybe no,
clips played on a cassette recorder. For each probably no, de nitely no. Subjects
clip their task was to: (1) make a yes/no were told to circle the response in each column
judgement about whether they recognized the that best described the nature of their memory
song, and, (2) rate the violence of the content about the source of the information contained
of the song on a 7-point Likert-type scale with in each of the 32 probes.
anchors at 1 (not at all violent) and 7 (very
Results and Discussion
violent). Subjects circled their responses on a
printed answer sheet. The analyses revealed that the spacing ma-
Word association task. Stimuli were 15 nipulation had no effect on any of the mea-
number word pairs, taken from MacLeod sures of interest (i.e., there were no reliable
(1988), presented individually at 20-s inter- main effects or interactions involving spacing,
vals via slide projector. The pairs were care- all p s .10; see the bottom half of Tables 3
fully selected so as not to overlap at all with and 4, and Table 5). Therefore, we will focus
the theme of the burglary video, nor the sug- on the effect of the Modality variable in our
gestions made in the postevent questions. Sub- discussion of the results. For ease of compari-
jects were told that as each number word pair son with the results of Experiment 1, we rst
was shown they were to: (1) write down the analyzed the data dichotomized as either
two-digit number contained on the slide, and yes or no responses (i.e., summed
(2) write down the rst three words that came across de nitely, probably, and maybe). Over-
to their mind when they thought about the all, the pattern of results closely replicated
word contained on the slide. those of Experiment 1.
Each test took approximately 5 min to com- Item Recognition. Once again, as can be
plete. Every attempt was made to convince seen in the bottom half of Table 3, item recog-
the subjects that these were real experimental nition was better for repeated items than those
AID JML 2464 / A002$$$204 04-11-96 10:32:49 jmla AP: JML
256 MITCHELL AND ZARAGOZA
TABLE 5
MEAN PROPORTION RECOGNIZED SUGGESTIONS ATTRIBUTED TO THE VIDEO (SOURCE MISATTRIBUTION ERRORS)
OF
FUNCTION OF GROUP AND NUMBER OF EXPOSURES IN EXPERIMENT 2
AS A
No. of exposures Repetition
effect
Condition 1 3 (3 minus 1)
Single Modality Consecutive .49 .60 .11
Single Modality Spaced .51 .61 .10
Mixed Modality Consecutive .54 .72 .18
Mixed Modality Spaced .45 .62 .17
suggested once (F(1, 392) 154.36, MSE evidenced by a reliable Exposure X Modality
.02, p .0001). As in Experiment 1, there interaction (F(1,392) 5.09, MSE .05, p
.05). Planned comparisons con rmed that
were no item recognition differences between
the Single Modality (M s .83 and .97 for there was a signi cant repetition effect in both
Modality Groups (F(1,197) 27.34, MSE .
one and three exposures, respectively) and
Mixed Modality subjects (M s .81 and .95 04, p .0001, and F(1,197) 52.57, MSE
.06, p .0001, for Single and Mixed Mo-
for one and three exposures, respectively).
Source misattribution errors. Figure 2 de- dality, respectively) and that the interaction
picts conditionalized source misattribution er- resulted because there was no difference in
rors collapsed across Spacing Group (although error rates for the two groups at the 1-exposure
level (F(1,394) .002, MSE .16, p 1.0),
individual group means are reported in Table
5). Inspection of the gure shows that the
Copyright 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.
Contact this candidate