Content uploaded by Nurhan Er
Author content
All content in this area was uploaded by Nurhan Er on Dec 12, 2018
Content may be subject to copyright.
APPLIED COGNITIVE PSYCHOLOGY
Appl. Cognit. Psychol. 17: 503–517 (2003)
Published online in Wiley InterScience 26 February 2003
(www.interscience.wiley.com) DOI: 10.1002/acp.870
A New Flashbulb Memory Model Applied to the
Marmara Earthquake
NURHAN ER*
Department of Psychology, Ankara University, Sihhiye, Ankara, Turkey
SUMMARY
This study was based on the recollections of people who experienced the Marmara earthquake and
those who had no direct experience with it but only heard it on the news. Four models of flashbulb
memory (the photographic model, the comprehensive model, the emotional-integrative model and
the importance-driven emotional reactions model) were compared in the study. Findings indicated
that the importance-driven emotional reactions model provided a better fit to the data than the others
for victim and comparison groups. In order to verify the accuracy and consistency of flashbulb
memories, a small sub-sample of participants from both groups was tested twice after the first
anniversary of the quake. The results indicated that the Marmara earthquake was accurately recalled
and flashbulb memories about the event are consistent after a delay of one year. Copyright #2003
John Wiley & Sons, Ltd.
Flashbulb memory, which was first coined by Brown and Kulik (1977), has been used for
unexpected, traumatic, vivid and important personal or national events. In many cases,
such emotionally shocking events may be well preserved in people’s memories for years.
According to Brown and Kulik, these vivid recollections remain unchanged in the memory
over time. Recently, contradictory findings have been obtained from studies of flashbulb
memory. While some research has found people reporting remarkably accurate and stable
memories as stated by Brown and Kulik (e.g. Conway et al., 1994; Neisser et al., 1996;
Pillemer, 1984), others have found flashbulb recollections as inaccurate as other kinds of
memories (e.g. Neisser, 1982; McCloskey et al., 1988). In recent years, flashbulb memory
studies have focused on assassinations or assassination attempts (see Christianson, 1989;
Pillemer, 1984; Winogrod and Killinger, 1983) and disasters (see Bohannon, 1988;
Wright, 1993; Christianson and Engelberg, 1999). The target events in flashbulb memory
studies have almost always been negative events (except for a few studies: e.g. Tekcan,
2001 compared a negative and a positive event).
In one study Wright et al. (1998), investigated people’s flashbulb memories for the
resignation of Margaret Thatcher and the Hillsborough football tragedy. They found that
the flashbulb memories of the subjects differed depending on their social class, age and
gender. In this study, males reported having clearer flashbulb memories, compared with
females, of the Hillsborough disaster even though they felt it was less important and less
Copyright #2003 John Wiley & Sons, Ltd.
Correspondence to: Dr Nurhan Er, Department of Psychology, Ankara University, 06100 Sihhiye, Ankara,
Turkey.
E-mail: ner@humanity.ankara.edu.tr
emotional for them. Results also showed that the Hillsborough disaster was rated as more
emotional and important than Thatcher’s resignation even though it was not recalled as
vividly as the latter. In another study, Christianson and Engelberg (1999) investigated
flashbulb memories of the Estonia ferry disaster on the Baltic Sea. The emotional reactions
of the subjects were measured shortly after the event and a year later. Results showed that
personal circumstances when hearing the news of the Estonia ferry disaster were well
retained, although far from perfect. Neisser and Harsch (1992) examined participants’
memories of the Challenger Space Shuttle explosion and found that although participants
were highly confident about their memories for this event, just three years after the event
their memories were not very accurate.
Shum (1998) pointed out that a more reasonable experiment would elicit individuals’
own flashbulb memories, rather then viewing them in terms of public occurrences. The
findings mentioned here raise the question of whether these public events are the main
types of events about which people have flashbulb memories. Although it is expected that
flashbulb information is not as rapidly diminished as normal information, the results of
some investigations showed that flashbulb information decreases in time. According to
Brown and Kulik (1977), the special flashbulb mechanisms may operate only in situations
involving a strong emotional reaction.
Rubin and Kozin (1984) have demonstrated flashbulb memories formed by personally
important life-events rather than national events. They asked undergraduate students to
describe their three clearest flashbulb memories and to rate the national importance,
personal importance, level of surprise, consequentiality, vividness, and emotionality
associated with each event on a 7-point Likert type scale. Most of the events reported
by the students were the ones that they had experienced directly (e.g. car accident, sexual
encounters). The resulting 174 memories were almost all rated to be high in personal
importance, but low in national importance. Students had difficulty recalling national
events such as assassinations. Like Pillemer (1984), Rubin and Kozin found no effect of
consequentiality or number of rehearsals. Unlike Pillemer’s findings there was no major
effect of emotions. In fact, they found that the flashbulb memories of the subjects were
mostly affected by personal importance and vividness. Furthermore, many previous
studies had suggested that importance, emotional reaction, and number of rehearsals
played a key role in the maintenance of vivid memories (e.g. Bohannon, 1988; Bohannon
and Symons, 1992; Conway et al., 1994; Finkenauer et al., 1998; Wright et al., 1998;
Wright and Nunn, 2000).
One of the most interesting recent flashbulb memory studies was reported by Neisser
et al. (1996) and based on subjects who had direct experience with the 1989 California
earthquake. Three groups of subjects were compared. The first group was from Berkeley,
which had experienced a moderate earthquake, the second group was from the University
of California in Santa Cruz, which had experienced much more severe conditions. For the
informants in the Atlanta group, the earthquake was just something that happened
thousands of miles away. Results showed that informants from Berkeley and Santa
Cruz, who had experienced the quake, remembered much more than the Atlanta group.
In addition, the California groups remembered their direct experience of the quake better
than they remembered learning about the event. The most important finding of this
study was that personal involvement in the event led to greatly improved recall. Moreover,
this study revealed that individuals who had directly experienced the California earth-
quake in 1989 recalled their experiences accurately and confidently after a delay of a year
and half.
504 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
FLASHBULB MEMORY MODELS
Several different models have been suggested to explain flashbulb memory. Brown and
Kulik (1977), for example, who conducted the first formal study of flashbulb memory,
proposed the photographic model of flashbulb memory formation. The authors chose 10
major events (mostly assassinations) to test this model. Individuals were asked whether
they could recall the circumstances in which they first heard about the events which had
taken place 10 to 30 years before. Conway et al. (1994) proposed another model to account
for the formation of flashbulb memories. In this model, called a comprehensive model by
Finkenauer et al. (1998), flashbulb memories are formed by three encoding factors (prior
knowledge, importance and effect) and one post-encoding (rehearsal-elaboration) factor.
They conducted a large test–retest study to examine the subjects’ flashbulb memories
concerning the resignation of the British Prime Minister Margaret Thatcher. In the study,
subjects from different nations (UK, non-UK) completed a flashbulb memory question-
naire within 14 days of Margaret Thatcher’s resignation. A sub-sample of these subjects
was tested a second time approximately 11 months later. Finkenauer et al. (1998) have
formulated another flashbulb memory model (an emotional-integrative model of flashbulb
memory) on the basis of Belgian citizens’ recollections about King Baudouin’s death.
They investigated subjects’ recollections about the context in which they first heard about
the event from news reports.
Summarized similarities and differences between the three models of flashbulb memory
are shown in Figure 1. All three models agree that flashbulb memories are influenced by
the same three variables: surprise, importance or consequentiality, and emotional state or
affective reaction. Also, each model provides evidence for the crucial role of the
unexpected situations (novelty and surprise) for the formation of flashbulb memories.
FLASHBULB MEMORIES FOR THE MARMARA EARTHQUAKE
The flashbulb memory studies present various problems. First, in the studies of flashbulb
memory, the sample consisted of subjects who had heard about the event in the news rather
than those who had experienced the event directly —except Neisser et al.’s (1996) study.
The resulting problem is that one cannot determine whether the subjects are observing
flashbulb or normal memory processes if, as a consequence of the shocking event, people
do not experience emotional or affective states, or if there is no personal involvement in
the event. According to Wright and Gaskell (1995), the sampling problem is critical in all
naturalistic cognition and flashbulb memory studies. An event might be of flashbulb
calibre for a small well-defined sub-set of the population, but be insignificant to most
people. Second, many studies employed only small groups of subjects. According to
Conway et al. (1994), the use of small sample sizes may be problematic, because it
becomes very difficult to use multivariate analyses, which are essential if latent constructs
such as flashbulb memories, effect, consequentiality/importance, rehearsal, and the
relations between such constructs are to be assessed.
The aim of the present study is to test the various models of flashbulb memory by
sampling large groups of subjects who experienced the Marmara earthquake directly or
who only heard it in the news. On 17 August 1999, Turkey experienced a devastating
earthquake. Many buildings collapsed in Istanbul, Kocaeli, Sakarya, Bolu, Yalova and
Bursa provinces. The earthquake was became a traumatic experience and nightmare for
A new flashbulb memory model 505
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
people all over the country, as well as for the millions living in the residential areas of the
affected towns and city centers. More than 1000 relief workers from 19 countries joined the
frantic search for bodies— dead or alive— the day after the quake. According to the Crisis
Center in Ankara, the death toll was 15,851 and the reported number of injuries was 43,953.
Since the Marmara earthquake triggered strong emotional reactions and had a serious
impact on people from both groups, it is highly suitable to study flashbulb memory.
Directly experiencing an earthquake is more stressful and emotional than hearing about an
earthquake in the news. In fact, living through an earthquake is an emotionally and
personally significant event. Thus, the first assumption of this study is that the recollec-
tions of those who experienced the Marmara earthquake will be different from those of the
people who had no direct experience. It was assumed that a different flashbulb memory
model is needed depending one whether the earthquake was directly experienced or not.
Figure 1. Similarities and differences between the three models of flashbulb memory
Note: 1. The photographic model, 2. The comprehensive model, 3. The emotional-integrative model
506 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
An importance-driven emotional reactions model of flashbulb memory has been proposed.
In this model, personal consequences determine intensity of emotional reactions as critical
operator in the formation and maintenance of the real flashbulb memory.
METHOD
Participants
A total of 655 Turkish participants volunteered for the study. Of those, 335 (140 female,
195 male) were earthquake victims who experienced the Marmara earthquake directly.
The mean age of the earthquake victims was 34.4 (SD ¼14.25), ranging from 18 to 48
years. The other 320 participants (150 female, 170 male) were from Mersin, which had not
been effected by the Marmara earthquake. The comparison group was similar to the victim
group in terms of age, gender and education level. The mean age of the comparison group
was 30.22 (SD ¼11.19), ranging from 18 to 53 years.
Questionnaire
A modified version of Finkenauer et al.’s (1998) questionnaire was used. It consisted of
eight sets of questions investigating different sub-dimensions in the flashbulb memory
models: (1) flashbulb memory attributes, (2) original event memory, (3) overt rehearsal,
(4) emotional reactions, (5) surprise, (6) novelty, (7) importance/consequentiality, and (8)
affective attitude. The questionnaire contained a total of 32 items after the elimination of
problematic items in accordance with the results of exploratory factor analysis.
Flashbulb memory attributes
In the questionnaire used for the earthquake victim group, the subjects were asked five
questions related to the occurrence of the Marmara earthquake. Similar questions were
used for the comparison group, but different from the first group, this time participants
answered questions about when they first heard the news: (a) the exact time (they
experienced or heard about the quake; TIME ), (b) the place they were at (city, place,
location; WHERE ), (c) the people they were with (WHO), (d) the ongoing activity they
were involved in (WHAT ), (e) for each group, participants were asked to describe up to
five specific details of their personal context when they experienced the earthquake or they
first heard the news (DETAIL). Consequently, flashbulb memory attributes were assessed
in terms of five items for both of groups.
The first two items were scored from 0 to 3 according to the number of units subjects
recalled. For instance, in the first question, an answer was scored 3 when the exact date
was recalled (month, day, hour). The third and the fourth items were scored 1 when
respondents were able to remember the answer or 0 when they were not. For the last item,
each, of the five specific details was scored to elicit detailed aspects of environment and
the subjects’ interaction with the environment (see Finkenauer et al., 1998). The maximum
score for details was 5. Four independent raters individually examined each detail and
scored the last item. Inter-rater reliability was found to be as 0.92.
Original event memory
Original event memory was assessed by using six questions concerning general informa-
tion about the earthquake that was distributed by the media: (a) the exact time (date:
A new flashbulb memory model 507
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
month, day, hour; M.TIME), (b) the number of injured people, dead people, destroyed
buildings (NUMBER), (c) cities effected by the quake (CITIES), (d) reasons for the
occurrence (M.REASON), (e) help and lifeguard teams (HELP), and (f) on duty official
improvements (OFFICIAL).
All items were scored 0 to 3 according to the number of units, participants recalled. For
the last four items, five news reports were selected from the media and among these five,
participants were asked how many they recollected. For example, an answer scored 3
when at least 3 news reports were recalled.
Overt rehearsal
Overt rehearsal was assessed by six questions. Respondents were asked to rate how closely
they followed the media after the announcement of the Marmara earthquake. On this sub-
dimension, respondents were asked: (a) After the earthquake how often they listened the
media (FOLLOW), (b) how often they talked about the reasons for the occurrence
(T.REASON), (c) experiences during the quake (EXPERIENCE), (d) future plans
(FUTURE), (e) possible ways of protection from the effects of quakes (PROTECTION),
and finally, (f ) how often they participated in a meeting (conferences, seminars etc.;
MEETING). All items were rated on a 5 point scale (0 ¼never, 4 ¼very often).
Emotional reactions
Emotional reactions to the quake were assessed by six items. In this sub-dimension,
respondents were rated according to their level of fright, sadness, desperation, anxiety, and
anger. In the last item, subjects were asked to rate how long the emotional impact of the
earthquake lasted (DURATION of FRIGHT,SADNESS,DESPERATION,ANXIETY,
ANGER) on a 5-point scale (0 ¼not at all, 4 ¼very long).
Surprise
Surprise was assessed by three questions. The subjects were asked to rate how shocked
they were by the huge number of destroyed buildings (SURPRISE 1), the very large
number of people injured by the quake (SURPRISE 2), and the occurrence of the quake
(SURPRISE 3) on a 5 point scale (0 ¼not at all, 4 ¼a lot).
Novelty
Novelty was assessed by two questions. Participants rated their responses on a 5-point
scale. The first item was the extent to which quake was unexpected for them (UN-
EXPECTANCY)(0¼not at all, 4 ¼very much), the second was whether they had
experienced any quake before or not (BEFORE)(0¼never, 4 ¼three times or more).
Importance/consequentiality
This sub-dimension was assessed by two questions. Respondents were asked to rate how
important the quake was for them (IMPORTANT), and how much it changed their lives
(CHANGE) on a 5-point scale (0 ¼not at all, 4 ¼very much).
Affective attitude
Respondents rated their affective attitude towards the earthquake victims (T.VICTIMS) and
volunteer help officers and workers (T.HELP) on a 5-point scale (0 ¼not at all, 4 ¼strong
empathy, for the first item; 0 ¼not at all, 4 ¼very grateful, for the second item).
508 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
PROCEDURE
Questionnaires were applied by the experimenter to small groups in various temporary
camps for the earthquake victims, schools, hospitals etc. Participants were given written
instructions and asked to read them carefully. After reading the instructions, they were
asked if they had any questions. If they had, the experimenter provided further explana-
tion.
The collection of data took place 6 to 9 months after the Marmara earthquake.
Moreover, a small sub-sample of participants from the victim and comparison groups
which were given the original questionnaires six months after the event were tested again
after a six-month interval.
RESULTS
Four different statistical analyses were carried out to examine the data: (1) descriptive
statistics, (2) exploratory factor analysis (3) analysis of variance (ANOVA), and (4)
confirmatory factor analysis.
Descriptive analyses
Flashbulb memory
For flashbulb memory attributes, the percentage of personal recall by the subjects in both
groups, except on the specific details, was over 90. While 82% of the participants in the
victim group recalled all five specific details, the percentage was only 58 in the comparison
group. The difference between the two groups was significant (z¼7.83, p<0.001).
While many people in both groups could recall details of their personal context at the time
(whether they experienced the earthquake or heard it in a news reports) a large percentage
of the recollections of the participants in victim group were more vivid than the recol-
lection of those from the comparison group.
Original event memory
According to the percentage of correct recall, the earthquake victims generally recalled
more accurately than the comparison group in terms of original event memory, but
differences observed between the two groups were not significant. Thus, participants from
both of the groups had high correct recall of the earthquake-related information that was
distributed by the media. More specifically, 76–98% of the participants provided correct
answers to six questions concerning original event memory.
Secondary measures
The results showed that the Marmara earthquake had a strong impact on Turkish citizens.
Participants reported a high level of fright (M¼3.92, SD ¼0.22), sadness (M¼3.81,
SD ¼0.14), desperation (M¼3.94, SD ¼0.08), anxiety (M¼3.87, SD ¼0.14), and anger
(M¼3.90, SD ¼0.22). The results indicate that these negative emotions are still being
experienced by the participants (M¼3.69, SD ¼0.45). In addition, the participants
reported that the Marmara earthquake had a considerable effect on them (M¼3.75,
SD ¼0.19), and changed their lives (M¼3.81, SD ¼0.21). They rated the quake as
unexpected (‘unexpectancy’, M¼3.78, SD ¼0.23) and surprising (‘surprise 1’;
A new flashbulb memory model 509
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
destruction of the buildings, M¼3.39, SD ¼0.43, ‘surprise 2’; injuries to people,
M¼3.73, SD ¼0.56, ‘surprise 3’; occurrence of the quake, M¼3.56, SD ¼0.34). As
expected, the incident was very often rehearsed by the participants in different ways. It
was rehearsed by following the media (M¼3.92, SD ¼0.29) through conversations about
the reasons for the occurrence (M¼3.85, SD ¼0.11), discussion of future plans
(M¼3.23, SD ¼0.23), means of protection (M¼3.55, SD ¼0.18), and participation in
meetings (M¼3.02, SD ¼0.21).
Exploratory factor analysis
In order to evaluate the extent to which items were loaded on the expected sub-dimensions,
exploratory factor analysis was conducted. In general, it was observed that each sub-
dimension was represented by items with high loadings. It was found that problematic
items, which were identified based on item-total correlations, either had low factor loading
or were crossloaded under two or more factors and these problematic items were
eliminated. According to the results of the analysis, the questionnaires consisted of 32
items after the elimination of the problematic items. Internal consistency reliabilities for
each sub-dimension were also examined.
Confirmatory factor analysis
The proposed structural model in the present study was tested in two stages by utilizing
LISREL 8.30 (Jo
¨reskog and So
¨rbom, 1996). In the first stage, a measurement model was
separately tested to see how well the indicators serve as measurement instruments for the
underlying latent variables and to examine the correlations among the latent variables in
both of the groups. On the basis of modifications for both victims and comparison groups,
substantial improvements were observed to fit the data. The final model presented much
more satisfactory data for both groups (
2
(94, N¼337) ¼245.535, p<0.001,
RMS ¼0.05, AGFI ¼0.92, NNF ¼0.94, CFI ¼0.95 for the victim group,
2
(96,
N¼322) ¼292.560, p<0.001, RMS ¼0.05, AGFI ¼0.89, NNF ¼0.91, CFI ¼0.92 for
the comparison group). In the next stage, the photographic model, the comprehensive
model, the emotional-integrative model, and the model proposed in the present study
(importance-driven emotional reactions model of flashbulb memory) were compared
through the application of structural equation modeling. The fit statistics and comparative
fit indices for the victim group are presented in Tables 1 and 2 for the comparison group for
all models. To decide how appropriately the model fits the data, the following indicators
were used: The lowest ratio of
2
to df, the values of the GFI and CFI greater than 0.90 and
Table 1. A summary of fit indices and estimations of the four models in the victim group
The importance- The emotional- The The
driven-emotional integrative photographic comprehensive
reactions model model model model
2
(SD) 976.25 (336) 1425 (356) 974 (256) 985 (225)
GFI 0.945 0.901 0.881 0.882
AGFI 0.951 0.872 0.852 0.804
CFI 0.967 0.894 0.874 0.865
RMSEA 0.014 0.045 0.063 0.071
510 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
the RMSEA value of approximately 0.05 or less (Anderson and Gerbing, 1988; Bollen,
1989).
In this study, a structural diagram for each model was generated but only proposed
model’s diagram was presented both of the groups (Figure 3 for the earthquake victim
group, Figure 4 for the comparison group). Each structural diagram includes relationships
between latent (sub-dimensions or constructs) variables and observed (measured) vari-
ables. In Figures 3 and 4, latent variables are shown in the ovals and observed variables are
in the squares.
As shown in Table 1, the proposed model provided a better fit to the data than the others
for the victim group and all the estimated parameters were significant. In this model, the
ratio of
2
to df and the RMSEA value were lowest and the GFI and CFI values were
greatest. Based on the consideration of parsimonious fit, the proposed model seems to be
preferable to the others. In Figure 3, according to the structural diagram of the proposed
model, flashbulb memory and original event memory are collected under the same
construct or factor. In other words, flashbulb memory is the original event memory.
Similarly, novelty and surprise are collected under one construct.
According to fit indices for the four models in the comparison group, the proposed
model provides an acceptable fit to the data (Table 2). All the item loadings and all of the
causal path linking factors are significant. In Table 2, the results were less favourable for
the emotional-integrative model, the photographic and the comprehensive models of
flashbulb memory compared to the proposed model. As seen in Figure 4, flashbulb
memory is different from original event memory in the comparison group, but novelty and
surprise are under the same construct like the victim group. On the basis of these analyses,
the proposed models (importance-driven emotional reactions models) for the two groups,
seem to be an suitable fit to the data.
Consistency of flashbulb memories
Two procedures were applied to verify the accuracy and consistency of flashbulb memories.
First, 96 participants from the victim group were married couples who had experienced the
quake together. After completing the questionnaires, these couples rated each other’s
recollections to evaluate their accuracy, noting each specific detail in the last item on a 3-
point scale (1¼inaccurate, 0 ¼I couldn’t remember it, 1 ¼accurate). The results
indicated 85% of the couples assessed their partner’s memories as accurate, 1% of them
assessed it as inaccurate and 10% of them couldn’t be sure about their partner’s answer.
Second, a small sub-sample of participants from the victim and comparison groups were
tested twice after the first anniversary of the Marmara earthquake. Eighty participants, 40
Table 2. A summary of fit indices and estimations of the four models in the comparison group
The importance- The emotional- The The
driven emotional integrative photographic comprehensive
reactions model model model model
2
(SD) 1086 (436) 1255 (456) 985 (325) 1075 (236)
GFI 0.962 0.931 0.903 0.822
AGFI 0.958 0.930 0.891 0.850
CFI 0.984 0.936 0.866 0.864
RMSEA 0.041 0.062 0.089 0.063
A new flashbulb memory model 511
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
(20 female, 20 male) of them from the victim group and 40 (20 female, 20 male) of them
from the comparison group participated in the retest. They were selected from participants
who had taken the original questionnaires six months after the event. Thus, participants in
the sub-sample were retested after a six-month interval. In the retest procedure,
participants answered questions only about the flashbulb memory attributes of the
questionnaire. It was thought that the answers to the questions from other sub-dimensions
might have been influenced by the media and by the lapse of time. To score the retest
data, a procedure originally developed by Neisser and Harsh (1992) was used (this
procedure was also used in various other studies: e.g. Cohen et al., 1994; Conway et al.,
1994; Neisser et al., 1996). All items were scored from 0 to 2 (2 ¼exactly consistent
answer, 1 ¼not exactly consistent answer, 0 ¼not completed or inconsistent answer).
Three independent judges scored all items and inter-rater reliability was found to
be 0.94.
In order to investigate the consistency of flashbulb memories on the sub-sample parti-
cipants, a mixed model analysis of variance (ANOVA) was conducted on the scores assigned
to each of the attributes of memory. In this analysis, the group was a between-subjects
variable with two levels (victim and comparison) and flashbulb memory attributes were
treated as a repeated measure with five levels (TIME,WHERE,WHO,WHAT,DETAIL).
Results revealed that the main effect of group was the only significant effect
F(1, 178) ¼20.12, p<0.001). In Figure 2, the mean consistency scores for each of the
five items in the flashbulb memories sub-dimension are shown for both victim and
comparison groups.
According to this result, memory attributes for the comparison group were less accurate
than for the victim group. In other words, participants of the victim group were exactly
correct on all memory attributes. This finding is particularly important because it shows
that victim group recollections were highly consistent in the two testing sessions.
Figure 2. Accuracy scores of flashbulb memory attributes for the victim and comparison groups
512 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
DISCUSSION
The present study was based on the recollections of 335 people who had experienced the
Marmara earthquake and 320 people who had had no direct experience of it but had only
heard about it on the news. The main findings obtained from this study indicated that the
Turkish people had been strongly affected by the event. The findings secondary findings
suggested that this disaster generated high levels of surprise, consequentiality and
emotional reaction for both groups of people who had experienced or heard about the
quake. Participants from both groups rated their emotions at high levels and reported that
the Marmara earthquake had a serious effect them.
A comparison has been made between the pattern of accuracy scores for the memory
attributes within the victim and comparison groups. The results of the memory attributes data
indicate that the Marmara earthquake was accurately recalled and flashbulb memories about
the event were consistent after a year. It was reasoned that a different flashbulb memory
model was needed, depending on whether an earthquake had been experienced or not. In
order to test this consideration, three models previously proposed in the literature and a new
model proposed in the present study, called ‘importance-driven emotional reactions model of
flashbulb memory’, were compared through the application of confirmatory factor analyses.
According to results of this analysis, two important similarities were found between the
importance-driven emotional reactions model for both groups and the other three models
proposed in the literature. First, all four models agreed that importance or consequentiality
determined the intensity of the emotional states. Second, rehearsal affected flashbulb
memory directly in all models. In the following section, the focus is on the specificities of
the importance-driven emotional reactions model for both groups and the implications of
this study for the understanding of the nature of flashbulb memory.
IMPORTANCE-DRIVEN EMOTIONAL REACTIONS OF FLASHBULB
MEMORY MODEL FOR THE VICTIM GROUP
On the basis of structural equation modeling, the proposed model is more appropriate than
the other three models. As seen in Figure 3, all paths linking factors are significant and
flashbulb memory is the original event memory. This result is different from the models
previously proposed in the literature.
According to this finding, the memories of people who experienced the Marmara
earthquake were preserved as a whole and unchanged. Results of the retest show that the
long-term memories of the victim group are more complete, more durable and more
consistent than those of the comparison group. The findings provide support for the
flashbulb and autobiographical memory studies that personal importance is a critical
predictor of maintaining clear memories (e.g. Rubin and Kozin, 1984; Wright and Nunn,
2000). Indeed, in the present study, structural equation modelling revealed that importance
fires strong emotional reactions and emotional reactions determine flashbulb memory.
There is a similarity between the proposed model in this study and the other two models
proposed in the literature. The emotional-integrative model (Finkenauer et al., 1998) and
the comprehensive model (Conway et al., 1994) assume that personal importance or
consequentiality determine the intensity of the emotional state (see also Conway, 1995;
Conway et al., 1996). The importance-driven emotional reactions model proposed for the
victim group also showed that importance directly affected flashbulb memory formation.
A new flashbulb memory model 513
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
The results show that novelty and surprise are critical factors operating together in the
formation of flashbulb memories. All four models agree that flashbulb memories are
directly determined by a new or unexpected event. Thus, novelty is a necessary condition
for flashbulb memory. This study indicate that novelty and surprise, which are important
variables for the flashbulb memory, are under the same construct, and that this construct is
directly affected by importance/ consequentiality. Interestingly, people rated the earth-
quake as something new and surprising even though they had been living in the quake area
for a year or more. In the present model, overt rehearsal directly contributes to flashbulb
memory formation and maintenance, but this construct is affected by emotional reactions
and the affective attitude. Emotional effects, especially in the victim group, have persisted
for a very long time. These negative feelings are repeated by way of social sharing (e.g.
following the media, conversing about the quake and participating in meetings). Overt
rehearsal may lead to accurate flashbulb memories for the victim group by reinforcing the
trace of original event memory. This process may be responsible for the finding that
flashbulb memory is the original event memory. In other words, since the earthquake is
regarded by the people as an important event in their lives, it continues to elicit intense
emotional reactions. These negative feelings repeatedly shared by others cause flashbulb
memories to remain vivid, detailed and accurate on the basis of the original event.
Participants from the victim group who were heavily affected by the results of the quake
talked about their experiences excessively.
In Neisser et al.’s (1996) study, which investigated memories of the 1989 California
earthquake, most Californian subjects who had directly experienced the earthquake
Figure 3. Relations between secondary variables for importance-driven emotional reactions model
in the victim group
Note: All of the items loadings and all of the causal paths linking factor are significant
514 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
reported their reactions and emotions on the low levels. For this reason, the authors
suggested that repeated narrative rehearsals may have played a more important role than
emotions and consequentiality. In the present study, participants from the victim group had
not only experienced the earthquake directly, but had also been greatly affected by the
results. They had talked a lot about all the experiences, just like Neisser et al.’s subjects.
They had rated their reactions and emotions on higher levels, and they reported that the
Marmara earthquake was more important for them than it was for the comparison group.
IMPORTANCE-DRIVEN EMOTIONAL REACTIONS OF FLASHBULB
MEMORY MODEL FOR THE COMPARISON GROUP
According to the results of the structural equation modelling, there are some differences
between the proposed models for the victim and comparison groups. First, in the model of
the comparison group, relationships between latent variables and observed variables,
although reliable, are comparatively lower than in the model of the victim group. Second,
in contrast to the victim group, flashbulb memory and original event memory are separate
constructs in the model of the comparison group. Recollections of subjects who did not
directly experience the quake are different from the original event but original event
memory has a strong effect on the flashbulb memory. This finding parallels the emotional-
integrative model. Third, importance/consequentiality does not directly affect flashbulb
memory. Rather, it triggers emotional reactions, which directly determine flashbulb
memory. It also affects overt rehearsal.
Finally, as seen from Figures 3 and 4 some processes are common to the formation and
maintenance of flashbulb memories for both groups. The critical difference between the
two groups hinges on whether the earthquake is directly experienced or not. Therefore, it
seems that the processes underlying importance/consequentiality have separate effects on
the two groups.
The Marmara earthquake was a nationally publicized event that seriously affected
people, yet the results of this event were personal, especially for the people who
experienced the quake directly. People have been shocked and surprised by the news of
the earthquake and also the event also led to personal consequences for themselves. The
present study illustrates that flashbulb memory is event related. If the event is assessed as
highly important in their own lives and is associated with a high reaction, then the
flashbulb memory is the original event memory. If the importance of the event in their own
lives does not reach a critical level, flashbulb memory formed is different from original
event memory. In this case, the observed differences between the two proposed models
may be due to differences in the importance of the quake in the lives of the people who
experienced it. For this reason, these two models have been given the same name:
importance-driven emotional reactions model of flashbulb memory.
Flashbulb memory studies often involve the recollections of public events in previous
investigations. Perhaps the reason for the decrease in accuracy of memories is the fact that
the subjects were not directly affected by the events dealt with in these studies (e.g.
Christianson, 1989; McCloskey et al., 1988; Neisser and Harsch, 1992; Pillemer, 1984).
Most flashbulb memories study a social psychological phenomenon, how people with little
if any personal consequences are affected by public tragedies. The Marmara earthquake,
especially for the victim group, is a real flashbulb memory event and has real con-
sequences. The present data all seems to support the conclusion that real personal
A new flashbulb memory model 515
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
consequences of the event are served to maintain real flashbulb memories. According to
the proposed model in this study, the higher the degree of importance and the more
emotional reactions are elicited, the more accurate and detailed the memory. This study
sheds some light on the controversy concerning the accuracy of flashbulb memory studies
and it also suggests the need for future reseach regarding the way in which people’s
recollections change depending on whether they experienced the event directly or not.
ACKNOWLEDGEMENTS
While the author was conducting this study she was a lecturer in the Department of
Psychology at Mersin University.
A preliminary account of this study was presented at the XI National Congress of
Psychology held in Izmir in September 2000. I would like to thank my colleagues for their
helpful comments on earlier drafts of this paper at the XI National Congress of
Figure 4. Relations between secondary variables for importance-driven emotional reactions model
in the comparison group
Note: All of the items loadings and all of the causal paths linking factor are significant
516 N. Er
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
Psychology. I am grateful to the participants who kindly volunteered to participate in the
study. The reviewers’ useful suggestions are also much appreciated.
REFERENCES
Anderson JC, Gerbing DW. 1988. Structural equation modeling in practice: a review and
recommended two-step approach. Psychological Bulletin 103: 411–423.
Bohannon JN. 1988. Flashbulb memories for the space shuttle disaster: a tale of two theories.
Cognition 29: 179–196.
Bohannon JN, III, Symons VL. 1992. Flashbulb memories: confidence, consistency, and quantitiy. In
Affect and Accuracy in Recall: Studies of ‘Flashbulb Memories’, Winograd E, Neisser U (eds).
Cambridge University Press: Cambridge; 65–91.
Bollen K. 1989. Structural Equations with Latent Variables. Wiley: New York.
Brown R, Kulik J. 1977. Flashbulb memories. Cognition 5: 73–99.
Christianson S-A. 1989. Flashbulb memories: special, but not so special. Memory Cognition 17:
435–443.
Christianson S-A, Engelberg E. 1999. Memory and emotional consistency: the MS Estonia Ferry
Disaster. Memory 7: 471–482.
Cohen G, Conway MA, Maylor EA. 1994. Flashbulb memories in older adults. Psychology and
Aging 9: 454–463.
Conway MA. 1995. Flashbulb Memories. Erlbaum: Hove.
Conway MA, Anderson SJ, Larsen SF, Donelly CM, Mcdaniel MA, Mcclelland AGR, Rawles RE,
Logie RH. 1994. The formation of flashbulb memories. Memory & Cognition 22: 326–343.
Conway MA, Collins AF, Gathercole SE, Anderson SJ. 1996. Recollections of true and false
autobiographical memories. Journal of Experimental Psychology: General 125: 69–95.
Finkenauer C, Luminet O, Gisle L, El-Ahmadi A, Van Der Linden M, Philippot P. 1998. Flashbulb
memories and the underlying mechanisms of their formation: toward an emotional-integrative
model. Memory & Cognition 26: 516–531.
Joreskog K, Sorbom D. 1996. LISREL 8: User’s Reference Quide. Scientific Software International:
Chicago, IL.
McCloskey M, Wible CG, Cohen NJ. 1988. Is there a special flashbulb memory mechanism? Journal
of Experimental Psychology: General 177: 171–181.
Neisser U. 1982. Snapshots or benchmarks? In Memory Observed, Neisser U (ed.). Freeman:
San Francisco, CA; 43–48.
Neisser U, Harsch N. 1992. Phantom flashbulbs: flase recollections of hearing the news about
Challenger. In Affect and Accuracy in Recall: Studies of ‘Flashbulb Memories’, Winograd E,
Neisser U (eds). Cambridge University Press: Cambridge; 9–31.
Neisser U, Winograd E, Bergman ET, Schreiber CA, Palmer SE, Weldon MS. 1996. Remembering
the earthquake: direct experience vs. hearing the news. Memory 4: 337–357.
Pillemer DB. 1984. Flashbulb memories of the assassination attempt on President Reagan. Cognition
16: 63–80.
Rubin DC, Kozin M. 1984. Vivid memories. Cognition 16: 81–95.
Shum MS. 1998. The role of temporal landmarks in autobiographical memory processes. Psycho-
logical Bulletin 124: 423–442.
Tekcan A
_
I. 2001. Flashbulb memories for a negative and a positive event: news of desert storm and
acceptance to college. Psychological Reports 88: 323–331.
Winograd E, Killinger WA. 1983. Where were you when President Kennedy was assassinated?
Bulletin of the Psychonomic Society 11: 133–135.
Wright DB. 1993. Recall of the Hillsborough disaster over time: systematic biases of ‘flashbulb’
memories. Applied Cognitive Psychology 7: 129–138.
Wright DB, Gaskel GD. 1995. Flashbulb memories: conceptual and methodological issues. Memory
3: 67–80.
Wright DB, Gaskel GD, Muircheartaigh CA. 1998. Flashbulb memory assumptions: using national
surveys to explore cognitive phenomena. British Journal of Psychology 89: 103–122.
Wright DB, Nunn JA. 2000. Similarities within event clusters in autobiographical memory. Applied
Cognitive Psychology 14: 479–489.
A new flashbulb memory model 517
Copyright #2003 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 17: 503–517 (2003)
