Applied Cognitive Psychology, Appl. Cognit. Psychol.

26: 301–307 (2012)

Published online 15 December 2011 in Wiley Online Library ([Link]) DOI: 10.1002/acp.1825

Brief Exposure to Misinformation Can Lead to Long-Term False Memories

BI ZHU1,2, CHUANSHENG CHEN2, ELIZABETH F. LOFTUS2*, QINGHUA HE1,

CHUNHUI CHEN1, XUEMEI LEI1, CHONGDE LIN1 and QI DONG1*
1
State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, China
2
Department of Psychology and Social Behavior, University of California, Irvine, USA

Abstract: Do false memories last? And do they last as long as true ones? This study investigated whether experimentally created
false memories would persist for an extended period (one and a half years). A large number of subjects (N = 342) participated in a
standard three-stage misinformation procedure (saw the event slides, read the narrations with misinformation, and then took the
memory tests). The initial tests showed that misinformation led to a significant amount of false memory. One and a half years later,
the participants were tested again. About half of the misinformation false memory persisted, which was the same rate as for true
memory. These results strongly suggest that brief exposure to misinformation can lead to long-term false memory and that the
strength of memory trace was similar for true and false memories. Copyright © 2011 John Wiley & Sons, Ltd.

INTRODUCTION the false memories at the initial testing session, the rates of
long-term decay of such false memories would be the same
Human memory is not perfect. True memory fades, whereas as those for true (veridical) memories.
false memory appears (Brainerd & Reyna, 2005; Loftus, Thus far, only a few studies have examined the long-term
2003). Much is known about the longevity of true (or verid- (>1 year) persistence of false memories created in experi-
ical) memory (e.g. Ebbinghaus’ (1885/1964) classic ‘forget- mental settings with the misinformation paradigm. More
ting curve’). Much less is known, however, about the than a half century ago, Davis and Sinha (1950) conducted
longevity or persistence of false memories. How long do an experimental study of long-term false memory arising
false memories last? What types of false memories persist from misinformation. They asked the subjects to read a story
and why? Intuitively, false memories that are recounted once and three days later showed the subjects a picture that
in a while (e.g. repeatedly giving eyewitness accounts of an purportedly depicted the story. One group of the subjects
event with some false information) will last at least as long was shown a picture that included some information that
as they are being recounted. What about false memory that was inconsistent with the original story. Of relevance to the
is induced briefly in an experimental setting and is unlikely current study, a total of 16 subjects (nine in the misinforma-
to be recounted? Answers to this question should be of great tion group) were asked to recall the story after 1 year. Results
significance to our understanding of the enduring power of showed that details from the misinformation picture
false memories. ‘intruded’ into later recall of the story. Although limited in
Researchers have used several paradigms to study false scale and rigor (a very small sample size after much attrition,
memory [e.g. the Deese–Roediger–McDermott paradigm only one story with limited measures, no control items, no
(DRM), the misinformation paradigm]. For example, the statistical tests), this study presented initial empirical evi-
misinformation paradigm includes three standard stages— dence that false memory induced by experimentally pre-
experiencing an event, receiving misinformation about the sented misinformation can last for at least a year.
event, and being tested for memory of the event (Loftus, Since then, three other studies have examined long-term
2003)—with some variations in the specific design of each (≥1 year) maintenance of false memories created by misin-
stage (e.g. Bekerian & Bowers, 1983; McCloskey & Zaragoza, formation in children. Huffman, Crossman, and Ceci
1985). Studies have shown that the subjects’ memory of the (1997) explored whether source misattributions that had
original event was affected by the misinformation they been reported in an earlier study (by Ceci, Huffman, Smith
received, and false memories were hence created (see Loftus, & Loftus, 1994) persisted over time. In the earlier study
2003 for a review). The nature of such false memories, how- (Ceci et al., 1994), children had been asked to imagine a
ever, has been under much debate. Researchers have proposed false event (e.g. riding a hot-air balloon). For the following
various theories such as the source-monitoring framework, the 10 weeks, these children were interviewed weekly, during
activation-monitoring account, the fuzzy trace theory, and the which they were asked to ‘think real hard’ about whether
distinctiveness heuristic (see Brainerd & Reyna, 2005 for a the events had occurred (no matter whether true or false)
description of these theories). The source-monitoring frame- and try to ‘recollect’ them. At the end of these interviews
work, for example, proposed that the false memories occur (i.e. misinformation sessions), children assented to 22% of
because of misattribution or confusion of the sources of the false events. Two years later, Huffman et al. (1997)
information. Based on the source-monitoring framework, we followed up 22 children (6–7 years old) from the original
predicted that once the subjects misattributed the source of Ceci et al. (1994) study. Half of the children still assented
to at least one of the false events, and on average, the
*Correspondence to: Elizabeth F. Loftus, University of California, Irvine, 2393 children assented to 13% of all false events.
Social Ecology II, Irvine, CA, USA 92697; Qi Dong, State Key Laboratory of
Cognitive Neuroscience and Learning, Beijing Normal University, China. More recently, London, Bruck and Melnyk (2009) did a
E-mail: Elizabeth F. Loftus, eloftus@[Link]; Qi Dong, dongqi@[Link] follow-up interview with 45 children (4–6 years old) who

Copyright © 2011 John Wiley & Sons, Ltd.

302 B. Zhu et al.

had participated in an earlier study (by Bruck, Melnyk & Ornstein, 2001; Goodman, Batterman-Faunce, Schaaf, &
Ceci, 2000). In the earlier study, Bruck et al. (2000) intro- Kenney, 2002). These studies used misleading questions at
duced false memories by giving children both true and false the time of follow-up testing or used misleading information
reminders about a magic show they had previously watched. shortly before the follow-up testing to investigate whether
These reminders were presented twice (16 and 28 days after children became more susceptible to suggestions after some
the magic show). The children were tested twice (12 days time had passed since a key event occurred (up to 4 years,
after the second reminder and 15 months after that, the latter Goodman et al., 2002). In general, results of such studies
of which was summarized in London et al. (2009)). The rates confirmed their hypothesis that children were more suscepti-
of false memory (inaccurate assents for false items) were ble to the misinformation effect when the original event was
78% at the first test and 63% at the second test. These rates more distant in the past. These studies, however, did not
of false memory were much higher than those for the control address whether false memories from the original session
items (i.e. false details about the show that were not men- of misinformation/misleading questions persisted. Instead,
tioned during the reminding sessions), which were 17% they showed that false memories were easier to create when
and 37% for the first and second test, respectively. Results true memories of the original event were degraded.
of these two studies suggested that misinformation could Another related but distinct line of research on ‘long-term’
result in sustained false memory in children. false memory is that using the DRM paradigm (Seamon
A third study, however, did not find long-term mainte- et al., 2002). In contrast to the long-term maintenance of
nance of false memories. Peterson, Parson and Dean (2004) the specific implanted false memory via the misinformation
studied children (3–13 years old) who were injured in acci- paradigm, the longitudinal studies (up to a few months) of
dents and being treated in the hospital. A year later, children the false memory in the DRM paradigm dealt with consisten-
in the experimental group were given both false and true cies in spontaneous false memories that were recreated at
information about their accidents, whereas the control group every testing session. In other words, these consistencies
did not receive such information. An example of the misin- are likely to reflect the strength of associations between the
formation is ‘You hurt yourself on glass, what did the doctor stimuli presented and the critical ‘lure’, rather than any trace
do to make you feel better?’ ‘Glass’ was the misinformation of the ‘lure’ from the original testing session.
because the child did not hurt himself on glass. Misinforma- The current study is the first large-sample (N = 342 adults)
tion used in the questions was different for each child so as to study of long-term (1.5 years) maintenance of false memory
match his/her accident. One year after the misinformation from brief exposure to misinformation in an experimental
session (2 years after the accident), experimental group setting. To reiterate, based on the source-monitoring hypoth-
children (N = 36) were asked to recall the accident. Results esis of misinformation false memory, we expected that false
showed that virtually, none of the misleading information and true memories would decay at the same rate over the
was incorporated into long-term recall. 1.5 year period. We reasoned that once a person has a
There are various plausible explanations of the discrepan- memory, then this memory lasts regardless of the source of
cies between this study and the other studies because they the memory.
differed in many aspects of the study design (e.g. nature of
the original events, misinformation procedure, testing proce-
dure, length of delay). One of the most likely explanations is METHOD
that both Huffman et al. (1997) and London et al. (2009)
studies relied on multiple sessions of misinformation, Participants
whereas Peterson et al. (2004) relied on one misinformation
(misleading questions) session. This is a one-and-a-half year follow-up of a previous study
In addition to the previously mentioned long-term longitu- on misinformation false memory (Zhu et al., 2010). Of the
dinal studies of false memories induced by misinformation, original 437 participants (mean age = 20  1 years; 189
researchers have also examined the persistence of errors in males and 248 females), 342 (78.3%) took part in this
flashbulb memory. For example, based on the results of the long-term follow-up (mean age = 20  1 years; 152 males
44 college students in the 2.5 years follow-up memory study, and 190 females). To assess potential attrition biases, we
Neisser and Harsch (1992) found that false flashbulb memo- compared the followed-up sample with those who were not
ries persisted about 2.5 years even when participants were followed-up in terms of their demographic characteristics
confronted with their own original writings that they had (i.e. gender, age) and their scores on major study variables
done on the day after the original event. such as scores on the misinformation false memory tests.
This review of studies that examined the long-term main- Results showed no significant differences between these
tenance of false memories should be distinguished from 342 subjects and the 95 subjects who did not participate in
another line of research that focused on increased suscepti- the follow-up; t′ s ranged from 1.60 to 1.86, p′ s > .05.
bility to misinformation when a long as opposed to short
time interval has occurred between an event and the expo-
Procedure
sure to misinformation. Adult subjects were more suscepti-
ble to misinformation presented a week after an event than At Time 1, the classic misinformation test was given. It in-
to misinformation presented shortly afterward (Loftus, volved three standard stages (events, narrations, and then
Miller & Burns, 1978). Similar results have been found with the recognition and source-monitoring tests). See Figure 1
children (e.g. Burgwyn-Bailes, Baker-Ward, Gordon, & for details.

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
 Long-term false memory 303

Figure 1. Diagram depicting the procedure of the experimental manipulation and tests

Events each event (i.e. ‘you saw the picture slides and read the
Two separate events were shown to the subjects, each con- narrations, please try your best to answer the following ques-
sisting of 50 digital color slides (taken from the research of tions based on what you saw in the picture slides’). There
Okado and Stark, 2005). All the subjects saw both of the was no explicit ‘warning’ that narrations included misinfor-
events in order to generate reliable results. The order of the mation. For the recognition test of each event at Time 1,
two events was counterbalanced across the participants. the questions were presented in random order (i.e. not
Each picture was shown for 3500 ms with an interslide inter- following the chronology of events depicted in the slides).
val of 500 ms. One of the events depicts a man breaking into Each question had three possible choices as answers.
a car and stealing things from it, and the other depicts a girl’s Choices were either a detail presented in the picture (‘origi-
wallet being stolen by a seemingly nice man. Of the 50 slides nal item’) or a detail presented in the narrations with misin-
that comprised each story, 12 were critical slides that would formation (‘misinformation item’) or a new foil detail (‘foil
be inaccurately described in the subsequent narrations item’). For example, the subjects might see in the pictures
(which are described). To attain a balanced design, two of a man hiding behind a door after stealing a girl’s wallet
different images of each critical slide (one for the first stage and would then read the narration that he was hiding behind
and the other for the second/misinformation stage) were a tree. For the critical question ‘Where was the man hiding
generated. They were counterbalanced across the partici- after stealing the girl’s wallet?’, the choices were ‘behind
pants. As presented in Figure 1, one subject may see a man the tree’ (misinformation item), ‘behind the door’ (original
put the wallet in his jacket’s outside pocket and would be item), and ‘behind the car’ (foil). The endorsement rates of
misinformed at the second stage that he put the wallet in the original, misinformation, and foil items represented the
his pants’ pocket, whereas another subject may see the man ‘overall true memory’, ‘overall false memory’ and ‘overall
put the wallet in his pants’ pocket and would be misinformed foil’, respectively.
at the second stage that he put the wallet in his jacket’s Immediately after the recognition test, the subjects took
outside pocket. the source-monitoring test. The participants were asked from
what presentation source they remembered the answers they
Narrations had given on the recognition test. Five options were pro-
After 30 minutes of filler tasks, the story narrations were vided as follows: ‘saw it in the picture only’, ‘read it in the
presented. The narratives consisted of one sentence for each narrations only’, ‘saw it in both and they were the same’,
slide image describing the scene depicted in the image. All ‘saw it in both and they conflicted with each other’, and
the subjects read two story narrations, one pertaining to each ‘guessed’. Critical details in the pictures that were accurately
of the two events they had previously seen. For each event, recognized and further endorsed on the source memory test
50 sentences were presented (including 12 inaccurate as ‘saw it in the picture only’ or ‘saw it in both and they con-
descriptions (misinformation) and 38 accurate descriptions flicted with each other’ were considered to be ‘robust true
(i.e. consistent with the original picture slides), and each memories’(RTM). Misinformation items that were further
sentence was shown for 3500 ms with an interval of 500 ms endorsed on the source memory test as ‘saw it in the picture
between sentences. The subjects were told that they were only’ or ‘saw it in both and they were the same’ were consid-
to read narrations made by an eyewitness to those events. ered ‘robust false memories’(RFM). Foil items that were
The subjects were not warned about potential discrepancies further endorsed on the source memory test as ‘saw it in
between the picture slides and the narrations. the picture only’, ‘read it in the narrations only’, and ‘saw
it in both and they were the same’ were considered ‘robust
Tests at time 1 foil’. In other words, the source-monitoring test gave the
After 10 minutes of a filler task, the subjects took the recog- subjects a chance to rethink about the source of their mem-
nition test and then the source-monitoring test (adapted from ory. Based on the source-monitoring test, we could affirm
Okado & Stark, 2005). For the recognition test, 12 critical whether the subjects really believed that they had seen the
questions (pertaining to the critical slides) were asked for misinformation in the original event.

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
304 B. Zhu et al.

Test at time 2 endorsement rates of the original items and increases in endorse-
The follow-up test occurred 1.5 years after the initial study. ment rates of misinformation and foil items across two time
At Time 2, the subjects saw the same two 50-slide events points).
again (with the same event presentation procedure as at The average endorsement rates reported showed persis-
Time 1), except that the presentation of these slides stopped tence of false as well as true memories, but it was not clear
right before each critical slide. The critical slides were not how much of the false memory at Time 2 was exactly the
presented. Instead, the participants were asked what had same as that of Time 1. After all, endorsement of misinfor-
happened in this missing critical slide based on what they mation items at Time 2 could be due to three sources [persis-
had seen ‘in the picture slides’ one and a half years earlier. tence of false memory from Time 1, ‘new’ false memory due
For each event, the content and format (i.e. a recognition test to original misinformation or exposure of misinformation
with three alternative choices) of these 12 critical questions item in the test (i.e. sleeper effect or test effect), and random
were exactly the same as the original recognition test used guesses]. Random guesses were controlled for by the use of
at Time 1. The only procedural difference was that at Time 2, foil items.1 To separate persistence of false memory from
the questions were presented as part of the story, thus the sleeper effect of misinformation, we compared the items
following the chronology of events depicted in the slides, endorsed at both time points. As shown in Table 1 and
whereas at Time 1, the order of presentations of the recogni- Figure 2, we calculated the conditional probabilities of the
tion test was random. The order of events was not counterba- consistent choices for items on the misinformation test
lanced at Time 2, because we did not find the order effect at {e.g. conditional probability of the misinformation items =
Time 1. The noncritical items were not assessed at Time 2. [the number of misinformation items endorsed at both Times
We used this embedded procedure because the 1.5 years 1 and 2 consistently] divided by [the number of misinforma-
long-term memory of events might be better retrieved under tion items endorsed at Time 1], Brainerd, Reyna, & Brandse
the context of events presentation, possibly eliminating floor (1995)}. Of the endorsed misinformation items at Time 1
effects. All of these tests were self-paced and administered (i.e. overall false memory), 53% were endorsed at Time 2.
on computers. Similarly, of the endorsed original items at Time 1 (i.e. over-
The participants were debriefed at the end of the follow-up all true memory), 53% were endorsed at Time 2. In contrast,
test. It should be noted that at the end of the original test, the only 28% of the foil items endorsed at Time 1 were
participants were not told that they would be tested again re-endorsed at Time 2. In other words, there was a strong
1.5 years later. Instead, as part of a larger project, these sub- evidence of persistence of false as well as true memory, sig-
jects were asked to complete several questionnaires and nificantly higher than the 33% baseline, t(340) = 17.23, p < .001,
tested with many instruments during the first phase and were d = .93, for false memory, and t(340) = 26.55, p < .001, d = 1.44,
told only that they would be contacted again in the future for for true memory; but the persistence of foil was slightly lower
more data collection. than the 33% baseline, t(340) = 2.60, p = .01, d = .14.
These results were confirmed when we limited our analy-
ses to robust memories [e.g. conditional probability of the
RESULTS RFM = (the number of item endorsed as RFM at Time 1
and then also endorsed as misinformation item at Time 2
The endorsement rates for the true and false items are shown in consistently) divided by (the number of item endorsed as
Table 1. At Time 1, the subjects endorsed 61% of the original RFM at Time 1)]. Of the RFM at Time 1, 47% were
items [overall true (veridical) memory], 31% of the misinforma- endorsed at Time 2. Similarly, of the RTM at Time 1, 52%
tion items (overall false memory), but only 8% of the foil items were endorsed at Time 2. In contrast, only 18% of the robust
(overall foil). Results showed that the subjects had good mem- foil items endorsed at Time 1 were re-endorsed at Time 2.
ory of the original items (original versus foil), t(341) = 56.40, To further understand the decay of long-term true and
p < .001, Cohen’s d = 4.35, but also a strong misinformation false memories, we examined the true and false memories
effect (misinformation versus foil) t(341) = 22.70, p < .001, that did not persist from Time 1 to Time 2, which accounted
d = 1.85. It should be noted that these results are only for those for about half (47%) of the true and misinformation false
individuals who completed the tests at both Times 1 and 2. At memories at Time 1. These changed memories showed sim-
Time 2 (1.5 years later), the average endorsement rate of the ilar probabilities of going from endorsing the original items
original items was 45%, that for misinformation items was (true memory) at Time 1 to endorsing misinformation items
39%, and that for foil items was 17%. Memory decay was (new false memory) at Time 2 (32%) as from endorsing mis-
evident as shown by the decreased endorsement rate of the orig- information items (false memory) at Time 1 to endorsing the
inal items, t(341) = 17.38, p < .001, d = 1.22, as well as by the original items (‘true’ memory) at Time 2 (31%). These
increased endorsement of the foils, t(341) = 18.76, p < .001,
d = 1.41. The endorsement of misinformation items also signifi- 1
The foil items were selected by Okado and Stark (2005) to be comparable
cantly increased, t(341) = 7.76, p < .001, d = .55. More impor- to the true and misinformation items. To further ensure that foil items served
tantly, the misinformation effect was just as strong at Time 2 as a good baseline in our study, a separate sample (N = 27) was recruited and
given the recognition test and to guess the responses for each critical slide
as at Time 1, (misinformation versus foil), t(341) = 29.13, with no prior exposure to the events. Results showed that the foil items were
p < .001, d = 2.64. There was a significant interaction between endorsed 29% of the time, which was not significantly different from ran-
time (two time points) and item types (original, misinformation, dom guess (33% of the three alternatives). In other words, endorsement rates
of the foil items served as a good baseline. Finally, this conclusion was fur-
and foil), F(2, 340) = 283.18, p < .001, partial eta squared = .63, ther confirmed by the random pattern of endorsement among the three
reflecting the opposite trends by item type (decreases in options at Time 2 for those who endorsed foils at Time 1 (see text).

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
 Long-term false memory 305

Table 1. Average endorsement rate (% and SD) at Time 1, consistent choice rate (% and SD), and conditional probabilities (% and SD) on the
misinformation tests
Consistent choices Conditional probability
Time 1 Time 2 after 1.5 years rate after 1.5 years of consistent choices

Overall false memory 31 (17) 39 (9) 16 (10) 53 (21)

Overall true memory 61 (17) 45 (9) 32 (11) 53 (14)
Overall foil 8 (5) 17 (7) 2 (3) 28 (35)
Robust false memories 13 (9) — — 7 (6) 47 (35)
Robust true memories 50 (17) — — 26 (11) 52 (18)
Robust foil 4 (4) — — 1 (2) 18 (35)

for the correlation between long-term false memory and

overall false memory at Time 1; r = .16, p < .01 for the
correlation between long-term false memory and RFM at
Time 1. To put these correlations in context, comparable cor-
relations for true memories were .18 between long-term true
memory and overall true memory at Time 1 and .16 between
long-term true memory and RTM at Time 1 (see Table 2),
indicating similar stability in individual differences in true
and false memories.
Finally, although tangential to our study, we found modest
Figure 2. Average conditional probabilities (and standard errors) of gender differences for the true and false memories at Time 1.
the consistent choices for items on the misinformation test
Specifically, males had higher scores than females in false mem-
ories (i.e. overall false memory: F (1, 341) = 4.99, p < .05,
changes were more than twice as likely to occur as those d = 0.24; RFM: F (1, 341) = 7.87, p < .01, d = 0.31). In contrast,
from either true memory to endorsing foils (15%) or from males had lower scores than females for true memories (i.e.
false memory to foils (15%). These results suggested that overall true memory: F (1, 341) = 5.98, p < .05, d = 0.27;
the traces of true and false memories were similarly strong RTM: F (1, 341) = 7.88, p < .01, d = 0.31). No gender dif-
and both were stronger than that of the foils. Also illuminat- ference was found in endorsement rates of the foil items at
ing were the changes for the foil items. The conditional prob- Time 1 (i.e. overall foil: F (1, 341) = .21, p > .05; robust foil:
abilities were similar (about one-third) for the three choices F (1, 341) = .53, p > .05). At Time 2, there were no significant
(i.e. 28% from foils at Time 1 to foils at Time 2, 29% from gender differences in long-term memories. (i.e. overall false
foils at Time 1 to the original items at Time 2, and 33% from memory: F (1, 341) = 2.25, p > .05; overall true memory:
foils at Time 1 to the misinformation items at Time 2). These F (1, 341) = 2.38, p > .05; overall foil: F (1, 341) = .00, p > .05).
results suggest that the foil items had a random pattern of
changes as would have been expected from well-chosen foils
(also see footnote 1). DISCUSSION
Another approach for detecting persistence of false mem-
ory is to use the individual differences approach. Correla- The current study showed that with a brief exposure to misin-
tional analysis showed modest but significant correlations formation, false memories were produced and persisted for an
for false memory between Times 1 and 2: r = .15, p < .01 extended period. First, the participants endorsed false memory

Table 2. Bivariate correlations among indices of false memories

LTFM LTTM LT foil OFM RFM OTM RTM O foil
Time 2
LTFM
LTTM .68***
LT foil .37*** .36***
Time 1
OFM .15** .14** .02
RFM .16** .12* .05 .56***
OTM .16** .18** .00 .94*** .52***
RTM .12* .16** .03 .83*** .42*** .87***
O foil .01 .06 .06 .26*** .18*** .05 .02
R foil .04 .05 .02 .09 .10 .26*** .22*** .53***
LTFM, long-term false memory; LTTM, long-term true memory; OFM, overall false memory; OTM, overall true memory; RFM, robust false memory; RTM,
robust true memory; LT,Long-term; O,Overall; R,Robust; FM,false memory; TM,true memory.
*p < .05
**p < .01
***p < .001

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
306 B. Zhu et al.

items more than foils 1.5 years after the original exposure. Sec- In that case, theoretically, we could better answer the ques-
ond, about half of the false memory from Time 1 was main- tion about whether both true and false memory last regardless
tained at Time 2, at the same rate as true memory. Third, we of the source of memory. From the fuzzy trace theoretical
found significant correlations between false memories at the perspective, the verbatim traces fade more quickly than the
time of the initial misinformation session (Time 1) and those gist trace. But we found that the conditional probabilities of
1.5 years later. In other words, those participants who had more consistent choices for memory towards original items and
false memory at Time 2 were the ones who also had more false post-event misinformation items are the same. Therefore,
memories earlier. These correlations, though modest, were we believe the memory for original items and post-event mis-
similar to those for true memories. information items are behaving like verbatim traces. So the
Before we discuss the implications of these results for our fuzzy trace theory might be a better explanation for the spon-
understanding of false memories, it is worth pointing out that taneous DRM false memory, rather than for the false memory
our design may have underestimated the degree to which induced by post-event misinformation.
false memories persist. First, the foils used at Time 2 were Our results also have implications for the forensic world.
the same as those used at Time 1. Previous research showed Although we used the embedded memory test at Time 2
that mere memory testing can create false memory because for several reasons, primarily to reduce floor effects in mem-
of exposure to alternative choices in the prior memory test ory, it may actually resemble the kind of testing that does go
(Brainerd & Reyna, 1996). It is conceivable that this phe- on in some real-world settings. In modern real life, the media
nomenon may be partially responsible for the increase of [e.g. television news reports or video caught by the closed-
endorsement rate of foils, which had been seen by the sub- circuit television (CCTV)] might give the eyewitness some
jects once when they were tested at Time 1. visual information about the original events but not all of
Second, the recognition test at Time 2 was embedded in the details, which is similar to the ‘re-experiencing’ events
the picture slides show (a modification to the original recog- presentation we provided to the subjects at Time 2. The po-
nition test). Based on the source-monitoring theory, this con- lice might ask the eyewitness to watch the video caught by
text should facilitate the memory toward the original picture CCTV, and the police might pause the video presentation
slides rather than the memory of misinformation from the at the critical points and ask questions about the event details
narrations (Johnson, Hashtroudi, & Lindsay, 1993; Smith which were not caught by CCTV. Based on our findings, we
& Vela, 2010). It is thus conceivable that if the context for believed, in testimony of victims or witness, it is not appro-
the recognition test had been the narrations, the endorsement priate to make assumptions about differences between true
rate for misinformation items (which were presented in nar- and false memory in terms of how long they last, because
rations) would have been higher. Also consistent with this they seem to have the same strength of memory trace. Our
conjecture, previous studies have found that chronologically long-term false memory study further illustrates the impor-
ordered questions led to fewer misinformation errors than tance of minimizing post-event misinformation or at least
did questions presented in random order (Bekerian & considering the power of such misinformation in creating
Bowers, 1983). Future research might profitably use both long-lasting false memories.
contexts as well as other formats (e.g. context-free recall or
recognition) to enhance our understanding of the conditions
under which false memories persist. Of course, the use ACKNOWLEDGEMENTS
of multiple formats and/or multiple contexts has its own
drawbacks (e.g. interference for the within-subject design, This study was supported by the 111 Project from the Minis-
needing a large sample for between-subject design). try of Education of China (B07008). The authors thank Yoko
Future studies should also include one or more (if feasible) Okado and Craig Stark for sharing their slides with us and
control groups which would not receive the post-event mis- for their valuable input. We also want to thank the editor
information, or would receive accurate post-event informa- and Professor Ira Hyman and an anonymous reviewer for
tion, or would simply participate in testing at Time 2 without their extremely helpful suggestions.
taking the memory tests at Time 1, because testing per se
might enhance later retention (Roediger & Karpicke,
2006). Other variables that can be examined include the REFERENCES
format of studied materials (e.g. sentences for the original
event but pictures for the misinformation session as in Davis Bekerian, D. A., & Bowers, J. M. (1983). Eyewitness testimony: were we
and Sinha (1950) versus the reverse in our study or the same misled? Journal of Experimental Psychology. Learning, Memory, and
format for both sessions) and different lengths of delay in Cognition, 9, 139–145.
Brainerd, C. J., & Reyna, V. F. (1996). Mere memory testing creates false
follow-up.
memories in children. Developmental Psychology, 32(3), 467–478.
Taken together, our study presented evidence that the false Brainerd, C. J., & Reyna, V. F. (2005). The Science of False Memory. New
memory briefly introduced in an experimental setting seemed York, NY: Oxford University Press.
to have similar strength of memory trace as true memory. Brainerd, C. J., Reyna, V. F., & Brandse, E. (1995). Are children’s false
These results are consistent with the source-monitoring memories more persistent than their true memories? Psychological
framework. In other words, once the initial misinformation Science, 6, 359–364.
Bruck, M., Melnyk, L., & Ceci, S. J. (2000). Draw it again Sam: the effects
was misattributed, its resultant false memory may be similar of repeated drawing on the accuracy of children’s reports and source
in strength to true memory strength. Further studies might monitoring attributions. Journal of Experimental Child Psychology, 77,
benefit from including a source-monitoring test at Time 2. 169–196.

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
 Long-term false memory 307

Burgwyn-Bailes, E., Baker-Ward, L., Gordon, B. N., & Ornstein, P. A. McCloskey, M., & Zaragoza, M. (1985). Misleading postevent information
(2001). Children’s memory for emergency medical treatment after one and memory for events: arguments and evidence against memory impair-
year: the impact of individual difference variables on recall and suggest- ment hypotheses. Journal of Experimental Psychology. General, 114,
ibility. Applied Cognitive Psychology, 15, 25–48. 1–16.
Ceci, S. J., Huffman, M., Smith, E., & Loftus, E. F. (1994). Repeatedly Neisser, U., & Harsch, N. (1992). Phantom flashbulbs: false recollections of
thinking about a non-event: source misattributions among preschoolers. hearing the news about Challenger. In E. Winograd, & U. Neisser
Consciousness and Cognition, 3, 388– 407. (Eds.), Affect and Accuracy in Recall: Studies of “Flashbulb Memories”
Davis, R. D., & Sinha, D. (1950). The effect of one experience upon the (pp. 9–31). Cambridge: Cambridge University Press.
recall of another. Quarterly Journal of Experimental Psychology, 2, 43–52. Okado, Y., & Stark, C. E. L. (2005). Neural activity during encoding pre-
Ebbinghaus, H. (1885/1964). Memory: A Contribution to Experimental dicts false memories created by misinformation. Learning and Memory,
Psychology. New York, NY: Dover. 12, 3–11.
Goodman, G. S., Batterman-Faunce, J. M., Schaaf, J. M., & Kenney, R. Peterson, C., Parson, T., & Dean, M. (2004). Providing misleading and rein-
(2002). Nearly 4 years after an event: children’s eyewitness memory statement information a year after it happened: effects on long-term mem-
and adults’ perceptions of children’s accuracy. Child Abuse & Neglect, ory. Memory, 12(1), 1–13.
26, 849–884. Roediger, H. L., & Karpicke, J. D. (2006). The power of testing memory:
Huffman, M., Crossman, A. M., & Ceci, S. J. (1997). “Are false memories basic research and implications for educational practice. Perspectives
permanent?”: an investigation of the long-term effects of source misattri- on Psychological Science, 1, 181–210.
butions. Consciousness and Cognition, 6(4), 482–490. Seamon, J. G., Luo, C. R., Kopecky, J. J., Price, C. A., Rothschild, L.,
Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source monitor- Fung, N. S., & Schwartz, M. A. (2002). Are false memories more
ing. Psychological Bulletin, 114(1), 3–28. difficult to forget than accurate memories? The effect of retention
Loftus, E. F. (2003). Make-believe memories. American Psychologist, 58, interval on recall and recognition. Memory and Cognition, 30(7),
864–873. 1054–1064.
Loftus, E. F., Miller, D. G., & Burns, H. J. (1978). Semantic integration of Smith, S. M., & Vela, E. (2010). Environmental context-dependent memory:
verbal information into a visual memory. Journal of Experimental a review and meta-analysis. Psychonomic Bulletin & Review, 8(2),
Psychology: Human Learning and Memory, 4, 19–31. 203–220.
London, K., Bruck, M., & Melnyk, L. (2009). Post-event information affects Zhu, B., Chen, C., Loftus, E. F., Lin, C., He, Q., Chen, C., et al. (2010).
children’s autobiographical memory after one year. Law and Human Individual differences in false memories from misinformation: cognitive
Behavior, 33, 344–355. factors. Memory, 18, 543–555.

Copyright © 2011 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 26: 301–307 (2012)
Copyright of Applied Cognitive Psychology is the property of John Wiley & Sons, Inc. and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written
permission. However, users may print, download, or email articles for individual use.