Flashbulb memories in social groups: A comparativetest±retest study of the memory of French PresidentMitterrand’s death in a French and a Belgian group

Antonietta Curci

University of Bari, Italy

Olivier Luminet

Belgian National Fund for Scientific Research and University of Louvain at Louvain-la-Neuve, Belgium

Catrin Finkenauer

Utrecht University, The Netherlands

Lydia Gisle

University of Louvain at Louvain-la-Neuve, Belgium

Flashbulb memories are vivid and long-lasting memories for the reception context of an important publicevent (Brown & Kulik, 1977). They are assumed to be triggered by emotional factors (i.e., intensity ofemotional feeling, appraisal of the original event) and by social factors (i.e., social sharing of the news,following media debate about the event). The present study investigated the memory for the death of theformer President of France F. Mitterrand in two social groups, i.e., French and Belgian people. This studytests whether the flashbulb memory attributes, the memory for the original event, and the impact of theemotional and social determinants of flashbulb memory differed across groups. The results indicated thatthe flashbulb memory for Mitterrand’s death is affected by group provenance, as French people showedhigher levels of recall for the flashbulb memory attributes and their determinants than Belgian people.Time impaired recollections in both groups, so that flashbulb memories appear prone to decay and sharethe same destiny as ordinary memories. The theoretical construct of concern—as the most basic ante-cedent of emotional experiences and its related appraisal (Frijda, 1994)—is discussed in order to explainthe differences in memory of the two social groups.

MEMORY, 2001, 9 (2), 81–101

# 2001 Psychology Press Ltdhttp://www.tandf.co.uk/journals/pp/09658211.html DOI:10.1080/09658210042000120

Requests for reprints should be sent to Antonietta Curci, University of Bari, Department of Psychology, Piazza Umberto I, 1,70100 Bari, Italy. Email: a.curci@psico.uniba.it

Antonietta Curci is doing a PhD on flashbulb memory at the Department of Psychology, University of Bari, Italy. Olivier Luminetis post-doctoral researcher at the Belgian National Fund for Scientific Research. His contribution to this study was supported by grantFRFC 2.4546.97 and grant FNRS 1.5.124.00 from the Belgian National Fund for Scientific Research. Catrin Finkenauer is associateprofessor at the Department of Child and Adolescent Studies, Utrecht University, The Netherlands. Lydia Gisle is doing a PhD in thefield of Health Psychology at the Department of Psychology, University of Louvain, Louvain-la-Neuve, Belgium.

The authors gratefully acknowledge the help of Guglielmo Bellelli and Bernard Rime for their comments on earlier drafts of themanuscript, and Martin A. Conway and Daniel B. Wright for their review of the manuscript. They thank Jorg Blasius, MichaelGreenacre, and Guy Lories for their helpful suggestions on the statistical analyses. The authors also thank Jean-Francois and NadineBotermans, Patrick Chambres, Jacques Curie, Abdessadek El-Ahmadi, Marie-Claire Gay, Michel Huteau, Tracy Mayne, EmmanuelMerlin, Natascha Rainis, Jean-Pierre Rolland, Arielle Syssau, Nady Van Broeck, Martial van der Linden, and Francoise Van Duurenfor their help in collecting the data.

About twenty years ago, Brown and Kulik (1977)defined flashbulb memories as vivid, detailed, andlong-lasting memories for the circumstances inwhich people learned about a shocking publicevent (for instance, the assassinations of John F.Kennedy, Malcolm X, or Martin Luther King).According to Brown and Kulik, people retain thememory for the reception context in which theyfirst heard about the public event. In other words,people remember vividly and for a long time notjust the original event, but features of the recep-tion context, such as when they first learned aboutthe event, where they were, what they were doingand with whom, the informant, and the aftermathof the event (Bohannon, 1988; Brown & Kulik,1977; Conway et al., 1994; Larsen, 1992). Flash-bulb memory studies are usually for importantpublic events, although there are some studiesshowing that people have very vivid memoriesrelated to personal events (Pillemer, Goldsmith,Panter, & White, 1988; Rubin & Kozin, 1984).

According to Brown and Kulik (1977), the twomain determinants of flashbulb memories aresurprise and importance–consequentiality of theoriginal event. If an event triggers both a greatlevel of surprise and is rated as consequential, it issupposed to be remembered for a long time. Toillustrate, in the United States, Brown and Kulik(1977) investigated flashbulb memories of Afri-can-American and Caucasian participants aboutthe deaths of Martin Luther King and Malcolm X,two events for which the level of consequentialitywas substantially different in the two groups con-sidered. The results showed that African-American participants had significantly moreflashbulb memories for these events than Cauca-sian participants (Brown & Kulik, 1977). Brownand Kulik (1977) explained these results by pos-tulating a special encoding mechanism in the brainthat is triggered by the original event and that isassumed to make the memories vivid and long-lasting (Livingston, 1967). Other authors accountfor the special encoding hypothesis by emphasis-ing the role of surprise, importance–con-sequentiality, and emotional feeling states ingeneral (Conway, 1995; Pillemer, 1984).

The special encoding hypothesis has beencriticised by authors stressing the importance ofreconstructive post-encoding factors. Their theo-retical perspective is also known as constructivist(Christianson, 1989; McCloskey, Wible, & Cohen,1988; Neisser, 1982; Wright, 1993). According tothese authors, the rehearsal of the originalexperience plays an important role in yielding

flashbulb memories, both in their maintenanceand in their formation. Flashbulb memories areviewed as not particularly vivid and long-lasting.Instead, they are still inaccurate and prone todecay, as rehearsal processes make them con-tinuously modified. As with any ordinary memo-ries, these memories get distorted and biased overtime (Christianson, 1989; Schmolck, Buffalo, &Squire, 2000; Wright, 1993).

Rehearsal encompasses different processessuch as following event-related media debates,talking about the event with others, and thinkingof the original event. In other words, rehearsaldeals with media communication, social sharingand rumination about the experience. The con-structivist approach relies considerably on societalfactors to explain the formation and maintenanceof flashbulb memories (Wright & Gaskell, 1995).However, this constructivist perspective alsoacknowledges the direct impact of emotion, sur-prise, and importance–consequentiality (i.e., theencoding variables) as concurrent explanatoryfactors that act together with rehearsal in theformation of flashbulb memories. In this per-spective, emotion keeps its effect on memory afterthe original event has occurred through the socialsharing of and rumination about the event(Finkenauer et al., 1998). Studies about socialsharing and rumination have shown that theserehearsal factors are associated with the intensityof the emotional feeling state felt after an event(Luminet et al., 2000; Luminet, Zech, Rime, &Wagner, 2000; Rime et al., 1998; Rime, Philippot,Boca, & Mesquita, 1992). It follows that emotionsalways represent an important determinant offlashbulb memories. People remember a givenexperience for a long time, because they feltemotionally involved when it happened andrehearsed it as time passed. Emotion is supposedto be related both directly and indirectly toflashbulb memories and to exert its effect both atthe time the original event is encoded and duringits rehearsal.

EMOTIONAL DETERMINANTS OFFLASHBULB MEMORIES

Many studies on the formation and maintenanceof flashbulb memories stress the link betweenemotion and memory (Brown & Kulik, 1977;Conway et al., 1994; Pillemer, 1984; Ruiz-Vargas,1993). An important issue is to explain why peopleretain different memories for the same emotional

82 CURCI ET AL.

experience. As proposed by Brown and Kulik(1977), people are differently affected by an event,because of the different level of consequentialityascribed to it. In the Brown and Kulik study, theconcept of consequentiality is an implementationof the concept of biological significance proposedby Livingston (1967). In an evolutionary per-spective, some events can be more effective forthe survival of the species, others can be lesseffective; some events can enhance the safety ofhuman beings, others can be harmful. People reactdifferently in consequence of their evaluation ofthe potential enhancement or harm carried by anevent. In every context, flashbulb memories arisewhen an event is evaluated as highly con-sequential or, in other terms, as highly significantfor the safety of people living in that context.Thus, the high degree of consequentiality pro-motes a different encoding, which leads to a morevivid and long-lasting memory for the event(Brown & Kulik, 1977; Guy & Cahill, 1999).

The impact of people’s evaluations in elicitingemotional experiences has been taken intoaccount by the cognitive theories of emotion.Appraisal theorists suggest that emotions are dif-ferentiated by the cognitive evaluations of theoriginal event, that is, the so-called appraisals.These appraisals are automatic processings of agiven stimulus that determine the onset of differ-entiated subjective feelings (Frijda, 1987; Frijda,Kuipers, & ter Schure, 1989; Scherer, 1984, 1997;Smith & Ellsworth, 1985). People appraise theoriginal event differently, which then gives rise todifferent emotional feelings for the same event.

Recently, a new model has been tested whichemphasises the role of cognitive appraisals in eli-citing flashbulb memories (Finkenauer et al.,1998). The authors focus on the novelty andimportance–consequentiality appraisals. Theynotice that, since Brown and Kulik, the appraisalof novelty has been neglected by flashbulb mem-ory scholars. Brown and Kulik (1977) alsoassessed the role of consequentiality, even thoughthey did not clearly discriminate between thisappraisal and the emotion itself. According tothem, flashbulb memories are elicited becausepeople perceive that the occurring event has someconsequences for their life (Brown & Kulik, 1977).Finkenauer et al. (1998) tested a model wherenovelty is the direct determinant of surprise, whileimportance–consequentiality yields emotionalfeeling states, and thereby modulates the rehear-sal of the event. In this model, emotional feelingstates and their cognitive appraisals are struc-

turally linked to yield flashbulb memories. How-ever, the impact of appraisals and emotionoperate mainly through the rehearsal of the event.By rehearsing the event, people also maintain thememory for the reception context (i.e., the flash-bulb memory attributes) (Finkenauer et al., 1998).

SOCIAL DETERMINANTS OFFLASHBULB MEMORIES

Besides the approach just outlined relying mainlyon intra-individual processes to explain flashbulbmemories, there is a different perspective stress-ing the impact of societal factors (Wright & Gas-kell, 1995). The literature on collective memorysuggests that flashbulb memories are in part theoutcomes of shared experiences happening insocial contexts (Bellelli, 1999; Pennebaker, Paez,& Rime, 1997). The social sharing and therepeated thinking about a public event shape thecontent and features of the memory. The moreemotional and important an event is for the socialgroup, the more likely people will rehearse it(Finkenauer, Gisle, & Luminet, 1997; Pennebakeret al., 1997). Thus, people discuss and think abouta public experience in conformity with the inter-ests and emotional involvement of the group towhich they belong (Jodelet, 1998). In a broadersense, people discuss and rehearse especially whatis relevant for their country or social group(Robinson, 1996). Gaskell and Wright (1997) findin Tajfel’s social identity theory (1981) and inTurner’s social categorisation theory (1987) thefoundation for the functional analysis of flashbulbmemory with reference to the self-constructionprocess. According to the social identity theory,people strive to construct a positive self-concept.This process encompasses two aspects, their per-sonal identity and their social identity. The per-sonal identity is built of cognitions about theindividual herself or himself, while the socialidentity is built of cognitions about the groups towhich the individual belongs. Gaskell and Wright(1997) point out that the memory for politicalevents contributes to identify each person as anindividual and as a member of a social group.Memory for political events shapes individuals’personal and social identity. Therefore, memoriesremain vivid and long-lasting, such as flashbulbmemories, ‘‘because they contribute positively topersonal and social identity and thus serve tomaintain or enhance self-esteem’’ (Gaskell &Wright, 1997, p. 180).

FLASHBULB MEMORIES IN SOCIAL GROUPS 83

From these approaches, it becomes clear thatflashbulb memories are not just the outcomes ofan intra-individual remembering process. Ratherthey represent truly social experiences and areshaped by the constraints of the context in whichthe original events occur. In others words, flash-bulb memories should largely depend on theemotion experienced by social groups. As pre-viously shown, rehearsal is a social process whichcontributes to the elaboration and maintenanceof flashbulb memories for a specific public event.The way rehearsal shapes the memory is two-fold. First, people share the event and think of itin different ways, according to their group’shabits and requirements. Second, the way theevent is shared and ruminated on will alsodepend on the availability of mass media infor-mation and the way TV broadcasts, radio chan-nels, and newspapers build up the content ofpublic debates about the news. Again, thesereflect the group’s habits and requirements (Bel-lelli, 1999; Bellelli, Leone, & Curci, 1998).

OVERVIEW AND HYPOTHESES

The first aim of the present study is to investi-gate flashbulb memories among two socialgroups. No model about the formation andmaintenance of flashbulb memories is beingtested, although we bear in mind that thesememories correspond to a well assessed patternof remembering and it is accounted for by somespecific encoding and rehearsal variables (Bel-lelli, 1999; Brown & Kulik, 1977; Conway et al.,1994; Finkenauer et al., 1998). As the literaturesuggests, we define flashbulb memories withreference to so-called canonical categories of thereception context, including time, location, otherpeople present, details of the reception context,ongoing activity, and changes in ongoing activity(Bohannon, 1988; Brown & Kulik, 1977; Conwayet al., 1994; Neisser & Harsch, 1992). We focusalso on the changes in ongoing activity becausethey are indicative of the disruptiveness of thesituation with respect to the ordinary scripts ofdaily life. In fact, previous research work(Bohannon, 1988; Winograd & Killinger, 1983;Wright & Gaskell, 1992) suggested that thememorability of some events could be partiallydue to the changes these events prompt in thecourse of ordinary life.

The original event of the present study was thedeath of the former French President, Francois

Mitterrand, on January 8, 1996.1 Mitterrand wasan important politician and he was very familiar toFrench citizens. Born in 1916, he became a min-ister immediately after the end of World War II.Leader of the Left since the beginning of the1970s, Mitterrand was elected as President ofFrance in 1981. He was the first Socialist Presidentin that country since the 1950s. He was then re-elected in 1988 and terminated his mandate in1995, only a few months before he died. He hadthen been the President of France for 14 years.Although F. Mitterrand suffered from prostatecancer for a couple of years and his cancer was inits terminal state, his death came as a big surpriseto French people. This was reflected in the enor-mous amount of news coverage following hisdeath. French citizens felt very upset by this event,independent of their personal opinion aboutMitterrand’s politics.

The memory for Mitterrand’s death wasexamined in two French-speaking groups, ofFrench and Belgian citizens, at two differenttimes, one at 2 months after the event and theother 1 year later. The present study aimed toexamine more systematically how flashbulbmemories vary across different social groups.Specifically, flashbulb memory attributes, con-fidence in flashbulb memory, memory for theoriginal event, emotional determinants, rehearsal,and previous knowledge about Mitterrand amongFrench and Belgian citizens were compared.

Mitterrand’s politics were a very relevantmatter for French society. French citizens directlyexperienced the effects of Mitterrand’s politics,and they had been directly concerned by hispolitical choices. For Belgian citizens, Mitter-rand’s politics were only a foreign matter withmuch less effect on their societal life. We expectedthat the memory for the original event would varyacross the land of origin of the groups, at bothtimes of measurement. We also hypothesised thatthe memory for the reception context would varyacross groups, in that French citizens wouldremember the features of the context in whichthey learned about the event better than Belgians.

1 Generally, the literature about flashbulb memories dealswith unexpected and shocking public events. It has been arguedthat people can also retain photographic memories for expec-ted events (Neisser, 1982). However the literature rarelyaddresses theoretical and empirical differences between auto-biographical memories for expected and unexpected events.The present study attempts to examine whether the char-acteristics of flashbulb memories for an unexpected event holdfor the memory for an expected event.

84 CURCI ET AL.

In other words, French citizens were expected toexperience more detailed flashbulb memories forMitterrand’s death than Belgians did. Confidenceis generally considered among the distinctivefeatures of flashbulb memories, because peopleusually feel very confident about their answerswhen questioned about this kind of memory.Confidence contributes so noticeably to supportthe recollection of relevant public events, that itwas even supposed to have some effects on thedevelopment of flashbulb memories (Bellelli,1999; Neisser & Harsch, 1992). Thus, we expectedthat French citizens would be more confidentabout their recollections than Belgians.

Although the event is appraised on the samedimensions, the extent to which the cognitiveappraisals are experienced may differ acrossgroups. In other words, we expected the sameappraisal dimensions (i.e., novelty and impor-tance–consequentiality) to have an impact, but weexpected them to have a stronger impact in theFrench than in the Belgian group.

Similarly to cognitive appraisals, we expectedthat the original news would yield the same pat-tern of negative emotional feeling states in the twogroups. We hypothesised, however, that theintensity of the emotion would be greater for theFrench group than for the Belgian group. We alsopredicted that each group would rehearse theexperience to a different degree, depending ontheir emotional involvement (Pennebaker et al.,1997). More precisely, the rehearsal would bemore repetitive for French than for Belgians,given the direct relationship between emotion andrehearsal of the event in the structural models offormation and maintenance of flashbulb mem-ories (Finkenauer et al., 1998). In the presentstudy the measure of the rehearsal of the originalevent comes from three variables, including socialsharing, rehearsal through the media, and mentalrumination. Furthermore, social sharing encom-passes global conversations about the originalevent, sharing specific facts of the original event,sharing emotional reactions, and sharing infor-mation about the reception context. We expectedthat the difference between French and Belgianswould be slightly higher for sharing emotionalreactions because of the greater emotional invol-vement of French citizens in the event.

Finally, the more people evaluate Mitter-rand’s politics as a prominent topic, the morethey are interested in it, and the more theyshould know about it. Therefore we hypothe-sised that prior knowledge and interest in Mit-

terrand and his politics would be stronger for theFrench group than for the Belgian group, but wedid not expect that the French would be globallymore favourable towards Mitterrand than theBelgians.

The second aim of the study was to examineconsistency of memory across time, as data col-lection at two different points in time allowed usto examine this question. Flashbulb memorieswere originally considered as long-lasting memo-ries that remain vivid and detailed over time(Brown & Kulik, 1977). According to Pillemer(1998), the public impact of a news item and themedia sensationalism contribute to the main-tenance and consistency of some memories forpublic events. Christianson and Engelberg (1999)argued that the consistency of flashbulb memoriesis related to the degree of involvement andimportance people experience for a given event.Based on these findings, we predicted that theconsistency of the flashbulb attributes and thememory for the event would be higher for Frenchcitizens than for Belgians.

Despite the confidence people usually exhibit,flashbulb memories are not immune to forgetting(Christianson, 1989; Christianson & Engelberg,1999; Weaver, 1993). Time has been found todecrease the accuracy of flashbulb attributes,while the ‘‘memory for the central event isenhanced by the impressiveness of the news’’(Larsen, 1992, p. 61). As a consequence, time isexpected to impair the memory for flashbulbattributes but not the memory for the event andthe level of related confidence. We examined theimpact of time on the ratings of emotion andimportance, and expected that they would remainstable over time, because these estimates maydepend on the perceived quality of the relatedmemories (Christianson & Engelberg, 1999;Neisser, 1982). Similarly, we examined the impactof time on the other encoding factors (i.e., surpriseand novelty) and on the variables correspondingto the ratings of social sharing, rehearsal by themedia, rumination, and personal knowledge aboutMitterrand.

Finally, in line with the idea that flashbulbmemory does not imply the existence of a specialmemory mechanism (see Christianson, 1989;Conway, 1995; McCloskey et al., 1988) we pre-dicted that, as with ordinary memories, flashbulbmemories would decrease over time in a similarway for both the groups. Thus, no interactioneffects of time by provenance were predicted onthe flashbulb attributes and the memory for the

FLASHBULB MEMORIES IN SOCIAL GROUPS 85

event scores. Similarly, we did not expect anyinteraction effect on the encoding and rehearsalfactors.

METHOD

Design

Participants were French-speaking French andBelgian citizens. The independent variables werethe provenance of the participants (i.e., French vsBelgian), and the phase of the data collection (i.e.,the test–retest factor). We tested the memory ofparticipants twice: (1) 1–2 months after Mitter-rand’s death, and (2) 1 year after the first datacollection. The dependent variables are themeasures of the flashbulb memory attributes.These attributes correspond to the canonicalcategories usually defining flashbulb memories,such as time, location, other people present,details of the reception context, ongoing activity,changes in ongoing activity, and other receptioncontext details (Bohannon, 1988; Brown & Kulik,1977; Conway et al., 1994; Finkenauer et al., 1998;Neisser & Harsch, 1992). We also considered asdependent variables measures of the memory forthe original event (i.e., Mitterrand’s death) andfactors that are supposed to affect the flashbulbmemory formation, such as the appraisals ofnovelty and importance–consequentiality, sur-prise, negative emotional feeling state, andrehearsal (Finkenauer et al., 1998).

Participants

A total of 362 French participants (54.7% of thetotal sample, M age = 29.4; SD = 12.3) and 302Belgian participants (45.3% of the total sample,M age = 35.4; SD = 14.7) volunteered to partici-pate in a study on memories and reactionsrelated to Mitterrand’s death. The first data col-lection took part 1–2 months after that event.One year after the first data collection, 34.5% ofthe participants (N = 229), completed the ques-tionnaire a second time. A total of 124 partici-pants of the retest-group (54.1%) were French(M age = 28.4; SD = 11.2), and 105 (45.8%) wereBelgians (M age = 34.5; SD = 14.2). In thepresent paper, we will only consider those parti-cipants who took part in both data collections(i.e., N = 229 participants). Table 1 reports thecharacteristics of the sample, including genderfrequencies.

Measures

The questionnaire is very similar to the one usedby Finkenauer et al. (1998). One difference,however, is that it aimed at investigating theimpact of surprise more closely by using a largerset of items. Some questions were also addedconcerning rumination on the experience andprior knowledge about Mitterrand and his politics.The questionnaire was composed of different setsof items: (1) flashbulb memory attributes, (2)confidence about the flashbulb memory attri-butes, (3) memory for the original event, (4)negative emotional feeling state, (5) surprise, (6)appraisal of novelty, (7) appraisal of importance,(8) rehearsal and rumination, (9) prior knowledge,(10) personal interest, and (11) attitudes.

1. Flashbulb memory attributes. The flashbulbmemory measures investigated the recall of thecircumstances in which participants first learnedabout the event. The questions dealt with theexact time participants heard the news (date, dayof week, and hour), the source of information(family, friends, colleagues, media), the place theywere (country, city, room, or other kind of loca-tion, i.e. in the car), the other people they werewith, their ongoing activity, and the degree ofchanges in their ongoing activity following theannouncement. These questions correspond to thecanonical categories of flashbulb memories(Bohannon, 1988; Brown & Kulik, 1977; Conwayet al., 1994; Finkenauer et al., 1998).

For all the questions, answers scored 1 whenmentioned, 0 when missing. For the date question,the answer scored 2 when it included the year, themonth, and the day the respondent heard aboutthe news. It scored 1 when it included only theyear and the month, and 0 when no answer was

TABLE 1

Sample characteristics as a function of group, gender, andtime of data collection

GroupsOne to two months

after the eventn

Retest groupn

FrenchMales 106 40Females 256 84

BelgiansMales 146 52Females 156 53

86 CURCI ET AL.

provided. For the hour question, the answerscored 2 when an hour was mentioned, 1 whenonly the part of the day was mentioned, and 0when no answer was provided.

Respondents were asked to list additionaldetails (with a maximum of 5) of the personalreception context in which they first heard thenews. In order to rule out details just inferred onthe basis of participants’ familiarity with theenvironment, each listed detail was scored 1 if itmet one of the following criteria, 0 otherwise. Thecriteria used to score details were: (1) to include achangeable aspect of the environment (e.g., ‘‘itwas raining’’ = 1; ‘‘a green carpet in living room’’ =0); (2) to refer to the respondent him/herself orhis/her position in the environment (e.g., ‘‘I wassitting on a chair near to the radio’’) (see alsoFinkenauer et al., 1998).

The variable assessing the source of informationwas not included among the flashbulb indicators asa ceiling effect was found for its distribution, that is,almost everybody remembered how she/he heardthe news. We therefore collapsed the other sixindicators into a single measure of flashbulbmemory attributes (see later).

Consistency scores were also computed for thetime, location, ongoing activity, other peoplepresent, details, and changes in ongoing activitymeasures. For each item, the responses at the firstdata collection and at the retest were compared.The value 2 was assigned if respondents providedexactly the same answer at the first data collectionand at retest. The value 1 was assigned if theanswers were substantially but not entirely iden-tical, and 0 if the answers were totally different ormissing at the retest.

2. Confidence. For each of the flashbulbmemory attributes, participants rated the degreeof confidence about their recollection on 7-pointscales (1 = not at all, 7 = very much).

3. Memory for the original event. The memoryfor the event was assessed by three items con-cerning the exact time of the Mitterrand’s death(date, day of week, and hour), the place where ithappened, and the cause of death. Each answer tothis set of questions was scored 2 if it was com-pletely right, 1 if it was basically but not entirelycorrect (e.g. the answer about the cause of thedeath scored 2 for ‘‘prostate cancer’’, 1 for ‘‘can-cer’’, 0 if it was wrong or missing).

Consistency scores were also computed forthese measures. For each item the responses at the

first data collection and at retest were compared: 2was assigned if respondents provided exactly thesame answer at both times of measurement, nomatter whether this answer was right or wrong; 1was assigned if the answers were substantially butnot entirely identical, no matter whether theywere right or wrong; 0 was assigned if the answerswere totally different or missing at the retest.

4. Negative emotional feeling state. This com-ponent was assessed by asking respondents to rateon a 7-point scale (1 = not at all; 7 = very much)the extent to which they were upset by the news.Additionally, participants rated on three 7-pointscales the impact of the news by assessing howmuch they felt (1) shaken (1 = not shaken at all; 7= very shaken), (2) affected (1 = not affected at all;7 = very affected), and (3) indifferent (1 = notindifferent at all; 7 = very indifferent). This lastscale was inverted. Finally, respondents rated on7-point scales (1 = not at all; 7 = very much) thedegree to which they experienced eight basicnegative emotions (anger, sadness, guilt, fear,anxiety, disgust, contempt, and shame).

5. Surprise. Respondents rated on five 7-pointscales (1 = not at all; 7 = very much) to what extentthey had experienced the news as (1) surprising,(2) expected, (3) foreseeable, (4) unbelieveable,and (5) astonishing. The items corresponding tothe labels ‘‘expected’’ and ‘‘foreseeable’’ wereinverted.

6. Appraisal of novelty. Respondents rated oneight 7-point scales (1 = not at all; 7 = very much)how (1) usual, (2) common, (3) current, (4) habi-tual, (5) unusual, (6) special, (7) uncommon, and(8) unique the event was for them. The itemscorresponding to the labels ‘‘usual’’, ‘‘common’’,‘‘current’’, and ‘‘habitual’’ were inverted.

7. Appraisal of importance–consequentiality.Respondents rated the extent to which the eventwas important for themselves and their life on a 7-point scale (1 = not important at all, 7 = veryimportant). Additionally, they rated to extent towhich the original event had consequences forthem on a 7-point scale (1 = no consequences atall, 7 = many consequences).

8. Rehearsal. Ten items assessed overtrehearsal. First, six items assessed the socialsharing of the event with others. One item con-cerned the global frequency of conversations

FLASHBULB MEMORIES IN SOCIAL GROUPS 87

about the original event. The next three itemsassessed the extent to which conversations con-cerned (1) specific facts about the original event,(2) personal reactions and feelings related to theevent, and (3) information about the receptioncontext. These items were all rated from 1 (never)to 5 (very often). Also participants indicated thenumber of people they shared the event with (1 =nobody, 6 = more than six persons), and the timeelapsed until they shared for the first time on a 6-point scale (1 = immediately after, 6 = never).These items, except the one measuring the delayin sharing, were used to assess the social sharing ofthe experience. Second, three items referred tothe frequency of rehearsal by the media. Partici-pants rated how often they had followed the news(1) on TV, (2) on the radio, and (3) in the news-papers (1 = never, 5 = more than 10 times).Finally, the frequency of rumination was assessedby one item. Respondents rated on a 5-point scale(1 = never; 5 = more than six times) how manytimes they had had spontaneous thoughts orimages about the event during the days followingthe event.

9. Prior knowledge. Respondents were ques-tioned about the exact date of Mitterrand’s elec-tion, the names of his seven Prime Ministers, andthe years he was a candidate for the presidency.Specifically, with reference to the election date,participants were asked to indicate the exact day,month, and year Mitterrand became President ofFrance. For each part of the date, the answersscored 1 when people remembered exactly, and 0when the answers were incorrect or missing. Thescores for this question ranged from 0 to 3. Withreference to the question about the seven PrimeMinisters, each correctly reported name scored 1,so the scores for this question ranged from 0 to 7.For the previous candidatures question, threeyears were listed and respondents had to indicatewhether or not Mitterrand ran for the presidencyin each of these years. Thus, the scores for thisquestion ranged from 0 to 3.

10. Personal interest. Respondents rated theirpersonal interest in French politics, and the extentto which they usually follow TV programmes andnewspaper articles about French politics on three7-point scales (1 = not at all, 7 = very much).

11. Attitudes. Participants rated on three 7-point scales (1 = not at all, 7 = very much) to whatextent they (1) had sympathy for Mitterrand, (2)

admired him, and (3) were favourable to hispolitics.

Procedure

Questionnaires were distributed to the partici-pants 1–2 months after Mitterrand’s death andagain 1 year after the first data collection. Ques-tionnaires distributed at the retest were identicalto those distributed at the first data collection.Respondents were asked to recall their experienceand to evaluate their emotional states and therelated appraisals when the event took place. Forthe rehearsal measures, they were asked to give anestimate of the extent to which they shared andruminated about the event in the weeks thatimmediately followed it. Finally, general measuresof interest, attitudes, and knowledge about Mit-terrand’s politics were collected again.

Respondents were recruited among under-graduate psychology students, researchers frompsychology departments, and experimenters’acquaintances. A preliminary short text explainedthat the questionnaire dealt with memories andpersonal reactions towards Mitterrand’s death.Participants were told that there were no right andwrong answers to any of the questions, the studybeing concerned with the memory for mass mediaevents.

Statistical analysis

In a first step, descriptive analyses were run on themeasures corresponding to the memory for theoriginal event, flashbulb memory attributes, andmemory consistency. In a second step, measure-ment issues were considered for all variables usedin the present study. Finally, structural analyseswere carried through mixed-design analyses ofvariance models.

To deal with the measurement issue and inorder to summarise each of the listed sets ofindicators in single composite scores, principalcomponents analyses of mixed measurement leveldata were performed by means of SPSSPRINCALS2 routine. PRINCALS is an optimalscaling procedure (Young, 1981) aimed at lookingfor new scales as closely correlated as possible to aset of observed scores (Greenacre, 1993). This

2 Principal non-linear component analysis by means ofalternating least squares.

88 CURCI ET AL.

procedure was chosen because it allowed us todeal at the same time with nominal, ordinal, andnumerical variables (Van de Geer, 1993; Young,1981; Young, Takane, & De Leeuw, 1978). Forinstance, the variable assessing changes in ongoingactivities is a nominal one with two levels (0 = noanswer and 1 = mentioned activity), while theemotion indicators are numerical variables rang-ing from 1 to 7. For the purpose of consistency, weapplied the same procedure to all variables in thequestionnaire, even in the case of variables forwhich we had to summarise only numerical indi-cators to obtain a single composite measure.

Separate analyses were run on each set ofindicators, and, for each set, only the solutions ofunidimensional analyses were retained. We madethis choice for two reasons. First, our purpose wasto aggregate the observed data in syntheticalmeasures. Second, for each set of variables, wefound that the unidimensional solution wasalready a good approximation of the observeddata.

The eigenvalues listed in Table 2 are measuresof fit (contribution to the inertia) of the uni-dimensional solutions. Eigenvalues range from 0to 1, where 0 means no fit and 1 the best possiblefit. A fit of 1 means that the items considered areperfectly correlated and can be replaced by asingle composite score (Greenacre, 1993, 1994). Ifan eigenvalue is 1, the composite score resultingfrom the scaling procedure is a perfect combina-

tion of the corresponding set of indicators. In asmuch as an eigenvalue is lower than 1, as Table 2shows, the scaling is a slightly imperfect combi-nation of the corresponding set of indicators in theanalysis (Van de Geer, 1993). For each set ofvariables, we checked whether the contribution ofthe second dimension to the inertia of the solu-tions was more or less redundant with respect tothe first one. For all sets this never happened, andonly the unidimensional solutions were con-sidered.

PRINCALS starts from the contingency matrixof the raw data which contains the frequencies ofthe observation categories for each variable in theset (Greenacre, 1984; Weller & Romney, 1990).Optimal scores are mathematically equivalent tothe correlations of the observed scores to thedimensional solutions and they are related to theobserved frequency of the categories of the vari-ables (Greenacre, 1993). Comparable to ordinarycorrelation coefficients, optimal scores can bepositive or negative in conformity with the direc-tion of the correlation of the correspondingobserved scores with the dimensional solution.For each set of indicators, an optimal score isassociated to each respondent. To illustrate, apositive optimal score obtained by a respondenton the flashbulb memory attributes means thatthis respondent has a better memory than arespondent achieving a negative score. A negativeoptimal score on emotion means that the respon-dent experienced a lower emotional state than arespondent scoring positively on that variable.The saved optimal scores resulting from the ana-lyses were used as the final measures for each setof variables. For rumination, we directly enteredthe raw scores in subsequent analyses, becauseonly one item was used in the questionnaire toassess respondents’ rumination about theirexperience.

The subsequent analyses were performed inorder to test whether flashbulb memory attributesthat referred to Mitterrand’s death and theirencoding and rehearsal determinants were affec-ted by the provenance of the subgroups and bychanges over time. The optimal scores for each setof variables were first analysed in a 2 £ 2 mixed-design analysis of variance, where provenance(i.e., French vs Belgians) was the between-subjects factor, and test–retest was the within-subject factor (i.e., 1–2 months after the event vs13–14 months after the event). Then, optimalscores resulting from the consistency measureswere analysed by two separate one-way analyses

TABLE 2

PRINCALS eigenvalues

Composite variables Eigenvalues

Flashbulb memory attributes .500Confidence .466Memory for the original event .450Flashbulb memory consistency .421Memory for the original event consistency .412Negative emotional feeling state .708Surprise .561Novelty .459Importance .897Rehearsal (social sharing) .519Rehearsal (by media) .582Prior knowledge .607Personal interest .826Attitude .832

Eigenvalues are measures of fit (contribution to the inertia)of the unidimensional solutions. Eigenvalues range from 0 to 1,where 0 means no fit and 1 the best possible fit. A fit of 1 meansthat the items considered are perfectly correlated and can bereplaced by a single composite score (Greenacre, 1993, 1994).

FLASHBULB MEMORIES IN SOCIAL GROUPS 89

of variance, where provenance (i.e., French vsBelgians) was the independent variable. Respon-dents’ gender and age were included as covariatesin all comparisons, as the two groups of Frenchand Belgian respondents appeared different withrespect to these characteristics. The correspond-ing effects are reported only if significant. Other-wise, the covariates were dropped from theanalysis and only the main effects and interactioneffects were considered.

Given the high number of multiple compar-isons on the scores and in order to reduce thepossibility of type I errors, we used Bonferroni’sadjustment by setting the significance level atvalue of p = .001.3 As a consequence, comparisonssignificant at p >.001 level were not considered.Effect sizes were also computed for all the com-parisons.

RESULTS

Descriptive analyses

Figure 1 shows the percentages of participantswho were able to answer the questions aboutmemory for the original event. These percentages

are very high (about 94% of respondents) formemory for the cause of Mitterrand’s death, bothat the first data collection and at the retest. Figure1 also shows the percentages of people answeringthe questions on the flashbulb memory attributes.For all attributes, percentages exceed 60% ofrespondents both at the first data collection and atthe retest.

Consistency scores were computed for memoryfor the original event and for the flashbulb mem-ory attributes. As Figure 2 shows, percentages arevery high especially for memory for the time andcause of the event (respectively, 72.3% and 88.6%of respondents), and for the flashbulb memoryattributes of time, place, and other people present(respectively, 74.2%, 97.4%, and 68.6% ofrespondents).

Taken together, these results indicate that, ingeneral, a high number of respondents retained atleast some details of the event and flashbulbmemory attributes. Furthermore, many memoryattributes showed temporal consistency. As aconsequence, the following step was to checkwhether these memory attributes and the impact ofthe encoding and rehearsal variables would differwith respect to the provenance of the groups.

3 Bonferroni’s adjustment led to the threshold of p = .003, as the significance level of .05 is divided by the number of the multiplecomparisons on the PRINCALS scores (i.e. 13 + 2, see Tables 3 and 4). Nevertheless we preferred to use a more conservativeapproach and we lowered the p level to .001.

Figure 1. Percentages of responses for memory for the original event and for flashbulb memory attributes.

90 CURCI ET AL.

Analyses of emotional feeling state

A MANOVA was run on the emotional feelingstate raw scores, with type of emotion (eight levels)and test–retest factor (two levels) as within-subjectfactors and provenance (two levels) as between-subjects factor. Figure 3 provides a visual display ofthe main effect of the type of emotion on theintensity of the emotional feeling state, both for thefirst data collection and the retest, F (7, 1449) =66.31, p < .001. Neither provenance nor test–retest

effects were found on the negative emotionalfeeling states. In other words, French and Belgians’emotional feelings about the death of Mitterrandwere really close and did not differ over time.Finally, all interactions were non significant—except for the interaction effect of provenance bytype of emotion, F(7, 1449) = 4.27, p < .001. Sadnesswas the most intense emotion felt by participants,both at the first data collection and at the retest,and French were more affected by this emotionalstate than Belgians, F (1, 207) = 6.24, p < .02.

Figure 2. Percentages of consistency scores for memory for the original event and for flashbulb memory attributes.

Figure 3. Negative emotional states as a function of group provenance and time of data collection.

FLASHBULB MEMORIES IN SOCIAL GROUPS 91

Analyses by provenance

Flashbulb memory attributes, confidence, andmemory for the event. Table 3 shows the resultsfrom the analyses on the PRINCALS optimalscores. As can be seen, the provenance of thesubgroups had an effect on a large number of thedependent variables investigated.

Figure 4 displays the mean differences betweenFrench and Belgian respondents for all PRIN-CALS scores entered in the MANOVAs. Meandifferences for flashbulb memory attributes,memory for the original event, and confidencewere significantly higher for French respondents.Thus, confirming our predictions, the flashbulbmemory attributes were better remembered bythe French group than by the Belgians. Asexpected, the former group was also more con-fident about its recollections and exhibited a bet-ter memory for the original event.

Emotion, surprise, and appraisal of novelty andimportance. As Figure 4 shows, the highest meandifferences between French and Belgian respon-dents were found for negative emotional statesand appraisal of importance. As predicted, theFrench group, compared to the Belgian group,was significantly more emotionally involved onhearing the news, and rated the event as sig-nificantly more important. However, no sig-nificant effects were found for surprise, except for

age of the respondents, which was found to posi-tively influence the respondents’ level of surprise,­ = .340, t (227) = 5.376, p < .001. Thus, olderpeople experienced a greater level of surprise onhearing the news than younger people, but thiseffect of age did not interact with provenance. Thedifference between French and Belgians was notfound to be significant even after controlling forthe respondents’ age.

Finally, for novelty appraisal the effect of pro-venance just approached the fixed significancelevel, F (1, 222) = 10.30, p = .002, r = .211, with theFrench scoring higher on this variable.

Rehearsal, prior knowledge, and attitudes.French participants shared the experience to agreater extent than Belgian participants (see alsoFigure 4). The sharing scores aggregated differentaspects of the phenomenon, i.e. the globalfrequency of conversations about the originalevent, the sharing of specific facts about the ori-ginal event, the sharing of personal reactionsrelated to the event, and the sharing of informa-tion about the reception context. The two groupswere expected to differ greatly on the sharing ofpersonal reactions, because of the greater emo-tional involvement of French respondents. Tocompare across groups each aspect of sharing, aMANOVA was run separately on the four rawmeasures of sharing, with provenance (two levels)as between-subjects factor and the test–retest

Figure 4. Mean differences between French and Belgian participants for all PRINCALS scores entered in the MANOVAs. Pointsrepresent the differences between the means of the French respondents and the means of the Belgian respondents; vertical linesdepict confidence intervals of the mean differences. A positive mean difference indicates that the French scored higher than theBelgians on that measure; on the contrary, a negative mean difference indicates that Belgians scored higher than French respondentson that measure.

92 CURCI ET AL.

TA

BL

E3

MA

NO

VA

so

nth

eP

RIN

CA

LS

optim

alsc

ore

s(p

rove

nan

ceb

yte

st±r

ete

stfa

ctor

)

Pro

vena

nce

Tes

t–re

test

Inte

ract

ion

FB

Mfe

atur

esF (df)

M(F

r)(S

D)

M(B

e)(S

D)

effe

ctsi

ze(r

)F (df)

M(t

1)(S

D)

M(t

2)(S

D)

effe

ctsi

ze(r

)F (df)

M(F

r-t1

)(S

D)

M(B

e-t1

)(S

D)

M(F

r-t2

)(S

D)

M(B

e-t2

)(S

D)

Fla

shbu

lbat

trib

utes

28.0

8*(1

,227

).2

69(.

693)

7.3

18(1

.198

).3

3253

.22*

(1,2

27)

.205

(.79

7)7

.205

(1.1

35)

.234

.57

(1,2

27)

.454

(.44

8)7

.090

(.99

7).0

84(.

834)

7.5

46(1

.336

)

Con

fide

nce

25.7

3*(1

,227

).2

42(.

921)

7.2

85(1

.019

).3

1944

.34*

(1,2

27)

.233

(.94

5)7

.233

(1.0

04)

.404

.98

(1,2

27)

.505

(.86

6)7

.089

(.93

6)7

.023

(.90

2)7

481

(1.0

64)

Mem

ory

for

the

even

t32

.61*

(1,2

27)

.281

(1.0

69)

7.3

32(.

796)

.354

1.04

(1,2

27)

.039

(1.0

27)

7.0

39(.

975)

.034

2.68

(1,2

27)

.370

(1.0

97)

7.3

53(.

775)

.192

(1.0

38)

7.3

12(.

818)

Neg

ativ

eem

otio

nalf

eelin

gst

ate

10.8

7*(1

,227

).1

90(1

.007

)7

.222

(1.0

16)

.214

2.95

(1,2

27)

7.0

40(1

.053

).0

42(1

.009

).0

57.3

7(1

,227

).1

63(1

.043

)7

.279

(1.0

19)

.217

(.97

3)7

.165

(1.0

15)

Surp

rise

12.

60(1

,226

).1

46(.

991)

7.1

82(.

993)

.107

1.46

(1,2

27)

0.34

(1.0

10)

.026

(1.0

00)

.040

1.13

(1,2

27)

7.1

42(1

.003

).2

43(.

984)

7.1

50(.

984)

.121

(1.0

03)

Nov

elty

10.3

0(1

,222

).1

33(1

.034

)7

.188

(1.1

26)

.211

.07

(1,2

22)

7.0

35(1

.123

)0.

11(1

.052

).0

092.

19(1

,222

).0

43(1

.166

)7

.188

(1.1

26)

.224

(.87

8)7

.254

(1.1

85)

Impo

rtan

ce12

.68*

(1,2

24)

.158

(1.1

44)

7.1

93(.

757)

.232

13.9

9*(1

,224

)7

.170

(.92

7).1

69(1

.047

).1

23.5

6(1

,224

)7

.038

(1.0

60)

7.3

26(.

713)

.355

(1.1

95)

7.0

57(.

780)

Reh

ears

al(S

hari

ng)

40.4

7*(1

,223

).2

87(.

938)

7.3

75(1

.045

).3

921.

77(1

,223

).0

24(1

.026

)7

.052

(1.0

56)

.044

.00

(1,2

23)

.341

(.87

6)7

.350

(1.0

68)

.233

(.99

6)7

.402

(1.0

25)

Reh

ears

al(b

ym

edia

)1.

39(1

,222

).0

65(1

.038

)7

.077

(.98

0).0

79.0

0(1

,222

)7

.010

(.99

5).0

12(1

.035

).0

00.3

7(1

,222

).0

40(1

.014

)7

.069

(.97

4).0

91(1

.066

)7

.085

(.99

2)

Rum

inat

ion2

7.85

(1,2

07)

2.65

6(1

.285

)2.

381

(1.1

84)

.191

.48

(1,2

07)

2.55

5(1

.267

)2.

502

(1.2

17)

.023

.012

(1,2

07)

2.67

9(1

.297

)2.

412

(1.2

23)

2.63

4(1

.266

)2.

351

(1.1

46)

Pri

orkn

owle

dge1

88.6

4*(1

,226

).4

37(.

772)

7.5

26(1

.001

).5

3111

.50*

(1,2

27)

.058

(.97

7)7

.067

(1.0

32)

.111

1.76

(1,2

27)

.476

(.72

9)7

.436

(1.0

04)

.397

(.81

4)7

.616

(.99

4)

Per

sona

lint

eres

t3.

18(1

,225

)7

.094

(.94

8).1

09(1

.056

).1

18.0

8(1

,225

)7

.012

(1.0

27)

.009

(.98

0).0

09.0

4(1

,225

)7

.097

(.97

9).0

89(1

.078

)7

.091

(.92

1).1

29(1

.038

)

Att

itud

e.3

3(1

,223

).0

27(1

.044

)7

.051

(.95

7).0

38.0

3(1

,223

)7

.010

(1.0

33)

7.0

07(.

979)

.006

.02

(1,2

23)

7.0

29(1

.069

)7

.056

(.99

3).0

25(1

.023

)7

.046

(.92

4)

*p

<.0

01.

Fr=

Fre

nch

grou

p,B

e=

Bel

gian

grou

p,t1

=1–

2m

onth

saft

erM

itte

rran

d’sd

eath

,t2

=1

year

afte

rfi

rstd

ata

colle

ctio

n,F

r-t1

=F

renc

hgr

oup

1–2

mon

thsa

fter

Mit

terr

and’

sde

ath,

Be-

t1=

Bel

gian

grou

p1–

2m

onth

saf

ter

Mit

terr

and’

sde

ath,

Fr-

t2=

Fre

nch

grou

p1

year

afte

rfi

rst

data

colle

ctio

n,B

e-t2

=B

elgi

angr

oup

1ye

araf

ter

firs

tda

taco

llect

ion.

1T

heco

mpa

riso

nw

asco

ntro

lled

for

resp

onde

nts’

age.

2R

awsc

ores

,wit

hra

nge

1–7.

93

factor (two levels) as within-subject factor. Theresults of this MANOVA are shown in Table 4.The French always had higher scores than theBelgians on all four measures of sharing, but forthe first measure of sharing (i.e., the frequency ofconversations about the original event) the maineffect of provenance was significant; for the otherthree measures the results just approached thesignificance level (p µ .006). This indicates that, insharing their experience, French respondents didnot seem to prefer certain topics of conversationover others. In the present study, social sharingseemed a rather homogeneous phenomenon moreintensely experienced by French people.

A shorter delay elapsed before French partici-pants spoke about the event as compared to Bel-gian ones, À2 (5, N = 442) = 17.93, p < .005).However, rehearsal by the media and mentalrumination about the event were not affected bythe provenance of the group.

Finally, French participants showed betterprior knowledge about Mitterrand and his poli-tical actions than Belgian participants. Controllingfor respondents’ age, this effect remained sig-nificant although prior knowledge was found to bepositively related to respondents’ age, ­ = .195, t(227) = 3.382, p = .001. Contrary to expectations,Belgians scored higher on personal interest inFrench politics than French respondents, eventhough there was no significant effect of proven-ance on this measure and on attitudes towardsMitterrand.

Analyses by test±retest factor

Flashbulb memory attributes, confidence, andmemory for the event. Memory for the flashbulbattributes was impaired over time, and, contraryto expectations, people’s confidence about theirrecollections also diminished as time passed. Aspredicted, there was no significant effect of thetest–retest factor on memory for the event.

Figure 5 displays the mean differences betweenthe two data collections for all PRINCALS scoresentered in the MANOVAs. It is evident thathigher differences were found for flashbulbmemory attributes and confidence.

Emotion, surprise, and appraisals of noveltyand importance. The ratings of negative emo-tional feeling state, surprise, and novelty did notchange over time. Contrasting with these results,importance assessed at the retest was significantlyhigher than when assessed at the first data col-lection (see also Figure 5). Thus, despite theimpairment of memory for flashbulb attributesand of feeling of confidence, the encoding factorstraditionally regarded as flashbulb memory pre-dictors kept their impact stable over time.

Rehearsal, prior knowledge, and attitudes.The ratings of amount of social sharing aboutMitterrand’s death, delay before the first sharing,frequency of rehearsal by media, and ruminationremained unchanged with time. In other words,

Figure 5. Mean differences between the two data collections for all PRINCALS scores entered in the MANOVAs. Points representthe differences between the means at the first data collection and the means at the retest; vertical lines depict confidence intervals ofthe mean differences. A positive mean difference indicates that scores from the first data collection were higher than scores at theretest; on the contrary, a negative mean difference indicates that scores at the retest were higher than scores at the first data collection.

94 CURCI ET AL.

TA

BL

E4

MA

NO

VA

onth

eso

cial

sha

ring

mea

sure

s

Pro

vena

nce

Tes

t–re

test

Inte

ract

ion

Shar

ing

mea

sure

sF (df)

M(F

r)(S

D)

M(B

e)(S

D)

effe

ctsi

ze(r

)F (df)

M(t

1)(S

D)

M(t

2)(S

D)

effe

ctsi

ze(r

)F (df)

M(F

r-t1

)(S

D)

M(B

e-t1

)(S

D)

M(F

r-t2

)(S

D)

M(B

e-t2

)(S

D)

Glo

bal

freq

uenc

y17

.62*

(1,2

06)

3.79

2(1

.106

)3.

163

(1.1

85)

.281

4.27

(1,2

06)

3.50

7(1

.209

)3.

441

(1.1

55)

.072

.41

(1,2

06)

3.84

6(1

.124

)3.

248

(1.2

31)

3.73

8(1

.089

)3.

071

(1.1

33)

Shar

ing

even

t11

.52

(1,2

06)

3.62

1(.

947)

3.24

4(.

983)

.230

.74

(1,2

06)

3.42

7(1

.029

)3.

475

(.93

0).0

30.7

4(1

,206

)3.

610

(.98

0)3.

206

(1.0

47)

3.63

3(.

916)

3.28

3(.

915)

Shar

ing

emot

iona

lrea

ctio

ns7.

84(1

,206

)2.

591

(1.1

35)

2.20

7(1

.120

).1

923.

46(1

,206

)2.

484

(1.1

54)

2.34

6(1

.130

).0

65.0

6(1

,206

)2.

656

(1.1

70)

2.28

2(1

.106

)2.

525

(1.1

00)

2.13

0(1

.134

)

Shar

ing

cont

ext

8.08

(1,2

06)

2.12

9(1

.061

)1.

746

(.89

5).1

944.

56(1

,206

)1.

874

(.96

0)2.

037

(1.0

47)

.074

.08

(1,2

06)

2.03

3(1

.024

)1.

686

(.84

4)2.

227

(1.0

93)

1.80

8(.

944)

*p

<.0

01.

Fr=

Fre

nch

grou

p,B

e=

Bel

gian

grou

p,t1

=1–

2m

onth

saft

erM

itte

rran

d’sd

eath

,t2

=1

year

afte

rfi

rstd

ata

colle

ctio

n,F

r-t1

=F

renc

hgr

oup

1–2

mon

thsa

fter

Mit

terr

and’

sde

ath,

Be-

t1=

Bel

gian

grou

p1–

2m

onth

saf

ter

Mit

terr

and’

sde

ath,

Fr-

t2=

Fre

nch

grou

p1

year

afte

rfi

rst

data

colle

ctio

n,B

e-t2

=B

elgi

angr

oup

1ye

araf

ter

firs

tda

taco

llect

ion.

95

evaluations of the amount of rehearsal were con-sistent over time, as respondents were questionedabout their sharing and rumination about theevent in the weeks immediately following it, bothat the first data collection and the retest. Priorknowledge about Mitterrand and his politics wasimpaired, while there were no effects of the test–retest factor on the variables assessing personalinterest and attitudes towards Mitterrand. Thus,the only mean difference between the two datacollections found to be significant was that forprior knowledge scores (see Figure 5), whilepeople’s ratings about the other social determi-nants of flashbulb memories were not affected bytime. No interaction effect of provenance by test–retest factor was found for any of the compositemeasures used in the present study.

Analyses of consistency

Finally, two one-way ANOVAs with provenance(two levels) as the independent factor were run onthe PRINCALS optimal scores for the consistencymeasures. Table 5 summarises the resultsobtained. The French had higher scores than theBelgians for memory for the original event con-sistency, suggesting that this aspect of memoryremained more stable over time for Frenchpeople. For the flashbulb memory attributes,however, the comparison between the groups didnot reach significance level, indicating that thetwo groups did not differ on the consistency oftheir memories for the reception context.

DISCUSSION

A first aim of the present study was to examine inmore detail how flashbulb memories vary acrosssocial groups. Memory for President Mitterrand’s

death was investigated in two ‘‘national groups’’,of French and Belgian citizens. Flashbulb memoryattributes, confidence in flashbulb memory,memory for the original event, emotional deter-minants, rehearsal, and previous knowledge aboutMitterrand were compared across groups.

The present study focused on the impact ofvariables that were assessed as structural featuresof flashbulb memory by previous studies (Finke-nauer et al., 1998). The theoretical variablesinvolved in the formation and maintenance offlashbulb memory were measured by nominal andnumerical indicators, and the PRINCALSapproach was chosen for aggregation purposes.This approach allowed us to summarise all mea-sures in composite variables and to compare thesevariables across groups.

Flashbulb memories for expectedevents

The present study found high percentages ofpeople answering the questions about the cano-nical categories of the reception context. The factthat the death of Mitterrand was expected becauseof his serious illness did not seem to make a dif-ference to the frequencies of these answers. Inother words, recollection may score highly onflashbulb memory attributes, even if the elicitingevent does not strictly have all the characteristicsrequired by the flashbulb memory literature(Brown & Kulik, 1977).

The research on flashbulb memories has gen-erally dealt with unexpected events. Yet, as Pil-lemer suggested (1984), the findings on flashbulbmemories need to be generalised by examining avariety of everyday memories. Some authors havealready extended the research to expected andpredictable events, such as the death of the Spanishdictator Francisco Franco, the American attack onIraq during the Gulf War, or the resignation of theItalian judge Antonio Di Pietro (Bellelli, 1999;Morse, Woodward, & Zweigenhaft, 1993; Neisser,1982; Ruiz-Vargas, 1993; Weaver, 1993). Thepresent study represents a new attempt to gen-eralise the findings on flashbulb memory forunexpected events to memory for expected events.

Concerns, emotions, and flashbulbmemories

The present study shows that memories for theoriginal event, flashbulb memory attributes, andconfidence ratings were generally higher for

TABLE 5

One-way analysis of variance on the PRINCALS optimalscores for the consistency measures

ProvenanceConsistencyfeatures

F(df)

M (Fr)(SD)

M (Be)(SD)

Effectsize (r)

Memory for theevent consistency

25.11*(1, 227)

.290(.996)

7.343(.900)

.316

Flashbulb memoryconsistency

9.27(1, 227)

.182(.994)

7.215(.973)

.198

* p < .001Fr = French group, Be = Belgian group.

96 CURCI ET AL.

French people, whose recollections of the originalevent were also more consistent. French peopleshowed a stronger intensity of emotion, impor-tance, and social sharing, and had more priorknowledge about Mitterrand and his politics.Time decreased the strength of the flashbulbmemory attributes, confidence, and prior knowl-edge about Mitterrand, but increased the relatedimportance ratings. Finally, no interaction effectof provenance by time was found in our data set.

In order to understand these findings, the the-oretical construct of concerns can be taken intoaccount. This construct allows us to link theresearch on flashbulb memory to research on thecognitive determinants of emotion (Frijda, 1987;Frijda et al., 1989; Scherer, 1984, 1997; Smith &Ellsworth, 1985). Concerns are hypothetical con-structs which are defined as motives or reasons forstriving to reach or maintain a given statefavourable to the individual (Frijda, 1994). In thissense, they can be viewed as the most basic ante-cedents of emotional experience: They direct thecognitive appraisal of the original event, and thenelicit differentiated subjective feeling states. Inother words, an event is appraised as emotionallyrelevant and has an emotional impact on theindividual only if it favours or harms the indivi-dual’s concerns (Frijda, 1994; Frijda et al., 1989).

The effect of concerns is not limited to personalemotional events. The impact of public events isdetermined by the degree to which these eventsaffect people’s concerns. Appraisals of publicevents stem from the attempts people make toprotect or promote their own concerns. Differentsocial groups might be differently concerned bythe same public event (Ciompi, 1997; Kenwyn &Crandell, 1984). For instance, the death of a well-known politician in one country does not concernpeople from other countries to the same extent.People whose political leader dies are likely tohave been more affected by his or her politics, andare therefore likely to experience the effects of theleader’s death much more than people of foreigncountries. Although foreign people couldappreciate the relevance of the leader’s politics asregards international matters, their ratings willalways reflect the outsider’s perspective. Peoplefrom different social groups form and maintainmemories for important public events whoseintensity varies as a function of their own groupconcerns.

In the present study, we based our predictionson the assumption that the two national groups ofFrench and Belgians differ with respect to their

concern in the original event. French citizens wereassumed to have been more concerned by Mit-terrand’s death because of the direct impact Mit-terrand’s politics had on them. The differences inconcerns between the two national groups werehypothesised to account for the different extent towhich the two groups recalled the news and thedifferent impact of the encoding and rehearsaldeterminants of flashbulb memories.

The memory for Mitterrand’s death and theflashbulb attributes were stronger for the Frenchparticipants. As expected, flashbulb memories forthis event were more prominent for the groupmore concerned by Mitterrand’s politics. TheFrench respondents were also more consistent intheir memory for the original event, but, contraryto expectations, they did not differ from the Bel-gian respondents with respect to their consistencyfor the flashbulb memory attributes. The feelingof confidence, which is among the main features offlashbulb memories (Brown & Kulik, 1977;Neisser & Harsch, 1992), was higher for theFrench respondents but decreased over time forboth groups. Time seemed to have weakened thememory for the flashbulb attributes and peoplewere aware of their forgetting. Despite the highlevel of confidence people exhibited shortly afterthe event, flashbulb memories then seemed proneto decay, thus sharing the same destiny as ordinarymemories.

Flashbulb memory and its emotionaldeterminants

The fact that the French respondents had astronger memory for the original event and theflashbulb memory attributes at least shortly afterthe event enabled us to examine more closely theimpact of the supposed encoding and rehearsaldeterminants of flashbulb memory. More specifi-cally, the present study also aimed to examinewhether there were any differences between thetwo groups on the encoding and rehearsal factorsusually related to flashbulb memory.

As expected, and consistent with the assump-tion that the French respondents were more con-cerned than the Belgian respondents, theemotional impact of the news was stronger for theFrench group. With respect to the type of emotionfelt after the news, sadness was the most intenselyexperienced emotion. French respondents, whowere assumed to be the most concerned, alsoexperienced the highest level of sadness. Having

FLASHBULB MEMORIES IN SOCIAL GROUPS 97

adopted the theoretical perspective about con-cerns, it follows that concerns should primarilyaffect appraisals, which, in turn, determine emo-tional responses (Frijda, 1987; Frijda et al., 1989;Scherer, 1984, 1997; Smith & Ellsworth, 1985).The present study focused on the novelty andimportance appraisal of the news. Although nodifferences were found between the two groupsfor novelty, importance was higher for the Frenchrespondents. The absence of difference on noveltybetween the two groups can be explained by thefact that people appraise the novelty of a newsitem independently from their own concernsabout it. The appraisal of novelty theoreticallyprecedes all the other appraisal dimensions(Leventhal & Scherer, 1987). It can be consideredas a very primary process that merely serves as anorientation reaction. In this sense, the personaland social concerns with respect to the news mayonly have affected the cognitive appraisal of thenews at a secondary level. For importance, ratingsincreased for both groups as time passed. It seemsthat the long-term consequences of the death weremore seriously appraised by people. Anotherexplanation of why importance increases could bethat only after some time are people able toappreciate the differences that arose after Mit-terrand died. Some changes in politics could havebeen implemented that seem very different, butpeople can only evaluate these differences afterone year or more.

Flashbulb memory and its socialdeterminants

With respect to the social determinants of flash-bulb memory, the present study showed only amain effect of provenance on social sharing.French respondents shared their experience to alarger extent than Belgian respondents, withoutdiscriminating between factual aspects, emotionalfeatures, or contextual information. Sharing was awidespread phenomenon that encompassed dif-ferent aspects. French respondents showed thehighest scores for all four investigated features ofsharing, indicating that their discussions focusedon a large variety of topics. On the contrary,rumination and rehearsal through the media didnot differ across the provenance of the subgroupsand did not change over time. It could be arguedthat French respondents shared the news morefrequently for two reasons. First, they may havehad a greater urge to share the news with others.

Second, their social environment may have invitedmore sharing and discussion of the news. There-fore, the social availability of commentaries abouta public event can be hypothesised to affect thefrequency of social sharing of the news (Bellelli,1999).

This explanation is also supported by the factthat French respondents showed better priorknowledge about Mitterrand and his politics.Mitterrand was a well-known politician whosepolitical activities affected French societal life formany years. French respondents had betterknowledge of Mitterrand than Belgian respon-dents independent of their attitudes towards himand their personal interest in French politics.Information about Mitterrand and his politicalactivity as more available in the French context.As a consequence, opportunities for conversationsabout him and the consequences of his death mayhave increased shortly after the event.

Destiny of flashbulb memory

Finally, the present study also supports the ideathat flashbulb memories are not immune to for-getting (Christianson, 1989; Christianson &Engelberg, 1999; Weaver, 1993). They share thesame long-term destiny as ordinary memories. Inother words, these memories are affected bytemporal modifications, as are any ordinarymemories, and as a consequence they are not asspecial as some authors have suggested (Brown &Kulik, 1977; Conway, 1995).

Consistency of the flashbulb attributes for theFrench group was not significantly higher thantheir consistency for the Belgian group. Flashbulbmemory attributes were impaired over time andthe lack of interaction effects of time by proven-ance on the flashbulb attributes showed that decayof memory did not differ across the two groups.On the contrary, the memory for the originalevent remained stable for both groups, but theFrench were more consistent than the Belgiansabout their remembering. As Larsen suggested(1992), memory for the event could be enhancedto some extent by the public importance of thenews. This explanation fits well with the presentstudy where the most concerned participantsshowed the highest consistency on memory for theevent, but time had the same effect on flashbulbmemory attributes as on the ordinary memories.

According to the literature, flashbulb mem-ories are characterised by the great confidence

98 CURCI ET AL.

people exhibit in their recollections (Brown &Kulik, 1977). In the present study, however, con-fidence decreased over time, independent of theprovenance of the respondents. As expected,negative emotional feeling state, rehearsal of theexperience (i.e., social sharing, following media,and rumination), and interest and attitudestowards Mitterrand were not affected by time.Only importance was higher at the retest. Theseresults suggest that people are very capable ofremembering their emotions and the emotionalaftermath of an event. Although they may forgetdetails, they seem to continue to rememberaccurately how they felt and how they reacted forsome time (Christianson & Engelberg, 1999).Some authors have argued that studies based onrecall are highly prone to retrospective bias (Ross,1989). In the present study, retrospective bias doesnot seem to play an important role in distortingmemories for emotional experiences, at leastwhen the recall interval does not exceed one year.Thus, the findings of the present study are note-worthy in that they support to some extent the useof recall procedures.

CONCLUSIONS

The present study attempted to discriminate theimpact of flashbulb memory attributes and theirencoding and rehearsal determinants in two dif-ferent national groups. The results partly con-firmed the hypotheses: French respondentsexperienced a stronger memory for the originalevent and the Flashbulb memory attributes. Theywere more confident about their recollections andtheir memories for the original event were alsohighly consistent. Their ratings were higher forboth the emotional (i.e., emotional feelings andimportance appraisal) and social determinants(i.e., social sharing and prior knowledge). Timedid not seem to have any impact on these deter-minants—except on importance and prior know-ledge—but it impaired memory for the flashbulbattributes and confidence ratings across the twogroups.

The results of the present study are also con-sistent with some of the findings of the model offlashbulb memory formation developed by Fin-kenauer et al. (1998). In both studies, the emo-tional and social determinants of flashbulbmemories had a stronger effect for peopleassumed to be the most concerned by the originalnews.

Recently, research about this topic hasfocused on structural models of formation andmaintenance without any reference to the meanstructure of the interacting factors. The presentfindings suggest that flashbulb memory differsnot only qualitatively but also quantitatively as afunction of variables associated with respon-dents’ provenance. Although further researchstill needs to investigate the relationships amongvariables in order to explain the complex pheno-menon of flashbulb memories, studies shouldalso take the influence of potential moderatorsinto account, such as personal and social con-cerns.

Finally, further research should investigate thedirect effects of concerns on memory. Flashbulbmemories decay in a fashion comparable toordinary memories. Moreover, they seem to decayindependent of people’s concerns for the originalevent. Indeed, forgetting of flashbulb memoriesoccurred in both national groups. Therefore,flashbulb memories seem not to be as special asshort-term investigation suggests (Christianson,1989). Research should directly focus on theinvestigation of concerns by including some directmeasures of them (such as those assessed by theConcern Strength and Concern Relevance ques-tionnaires; see Sonnemans & Frijda, 1995). Inaddition, research should better explore the short-and long-term relationship between concerns andmemory, with reference to different kind of ori-ginal events.

Manuscript received 10 February 2000Manuscript accepted 4 December 2000

REFERENCES

Bellelli, G. (Ed.) (1999), Ricordo di un giudice. Unostudio sulle flashbulb memories [Remembering ajudge: A study on flashbulb memories]. Napoli:Liguori.

Bellelli, G., Leone, G., & Curci, A. (1998). Auto-biographical memory and memorability of emotionalpublic events. Paper presented at the 10th Con-ference of the International Society for Research onEmotions, Wurzburg, Germany.

Bohannon, J.N. (1988). Flashbulb memories for theSpace Shuttle disaster: A tale of two theories. Cog-nition, 29, 179–196.

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). Memoryand emotional consistency: The MS Estonia ferrydisaster. Memory, 7, 471–482.

FLASHBULB MEMORIES IN SOCIAL GROUPS 99

Ciompi, L. (1997). On the emotional bases of thinking.Fractal affect-logic and communication. Systemfamilie, 10, 128–134.

Conway, M.A. (1995). Flashbulb memories. Hove, UK:Lawrence Erlbaum Associates Ltd.

Conway, M.A., Anderson, S.J., Larsen, S.F., Donelly,C.M., McDaniel, M.A., McClelland, A.G.R.,Rawles, R.E., & Logie, R.H. (1994). The formationof flashbulb memories. Memory & Cognition, 22,326–343.

Finkenauer, C., Gisle, L., & Luminet, O. (1997). Whenindividuals memories are socially shaped: Flashbulbmemories of socio-political events. In J.W. Penne-baker, D. Paez, & B. Rime (Eds.), Collective memoryof political events: Social psychological perspectives(pp. 191–207). Mahwah, NJ: Lawrence ErlbaumAssociates Inc.

Finkenauer, C., Luminet, O., Gisle, L., El-Ahmadi, A.,Van Der Linden, M., & Philippot, P. (1998). Flash-bulb memories and the underlying mechanism oftheir formation: Toward an emotional-integrativemodel. Memory & Cognition, 26, 516–531.

Frijda, N.H. (1987). Emotion, cognitive structure, andaction tendency. Cognition & Emotion, 1, 115–143.

Frijda, N.H. (1994). Universal antecedents exist, andare interesting. In P. Ekman & R.J. Davidson (Eds.),The nature of emotion: Fundamental questions (pp.155–162). New York: Oxford University Press.

Frijda, N.H., Kuipers, P., & ter Schure, E. (1989).Relations among emotion, appraisal, and emotionalaction readiness. Journal of Personality and SocialPsychology, 57, 212–228.

Gaskell, G.D., & Wright, D.B. (1997). Group differ-ences in memory for a political event. In J.W. Pen-nebaker, D. Paez, & B. Rime (Eds.), Collectivememory of political events: Social psychologicalperspectives (pp. 175–189). Mahwah, NJ: LawrenceErlbaum Associates Inc.

Greenacre, M.J. (1984). Theory and applications ofcorrespondence analysis. London: Academic Press.

Greenacre, M.J. (1993). Correspondence analysis inpractice. London: Academic Press.

Greenacre, M.J. (1994). Correspondence analysis andits interpretation. In M.J. Greenacre & J. Blasius(Eds.), Correspondence analysis in the social sciences(pp. 3–22). London: Academic Press.

Guy, S.C., & Cahill, L. (1999). The role of overtrehearsal in enhanced conscious memory for emo-tional events. Consciousness and Cognition, 8, 114–122.

Jodelet, D. (1998). El lado moral y afectivo de la historia.Un ejemplo de memoria de masas: el proceso a K.Barbie, ‘‘El Carnicero de Lyon’’ [The moral andaffective side of the history. An example of collectivememory: The trial of K. Barbie, ‘‘The Executioner ofLyon’’]. In D. Paez, J.F. Valencia, J.W. Pennebaker,B. Rime, & D. Jodelet (Eds.), Memorias colectivas deprocesos culturales y politicos (pp. 341–360). Bilbao:Servicio Editorial de la Universidad del Paõ´s Vasco.

Kenwyn, S.K., & Crandell, S.D. (1984). Exploring col-lective emotion. American Behavioral Scientist, 27,813–828.

Larsen, S.F. (1992). Potential flashbulb: Memories ofordinary news as the baseline. In E. Winograd & U.

Neisser (Eds.), Affect and accuracy in recall: Studiesof ‘‘flashbulb memories’’ (pp. 32–64). New York:Cambridge University Press.

Leventhal, H., & Scherer, K. (1987). The relationship ofemotion to cognition: A functional approach to asemantic controversy. Cognition & Emotion, 1, 3–28.

Livingston, R.B. (1967). Brain circuitry relating tocomplex behavior. In G.C. Quarton, T. Melnechuck,& F.O. Schmitt (Eds.), The neurosciences: A studyprogram (pp. 499–514). New York: Rockfeller Uni-versity Press.

Luminet, O., Bouts, P., Delie, F., Manstead, A.S.R. &Rime, B. (2000). Social sharing of emotion followingexposure to a negatively valenced situation. Cogni-tion & Emotion, 14, 661–688.

Luminet, O., Zech, E., Rime, B., & Wagner, H.L.(2000). Predicting cognitive and social consequencesof emotional episodes: The contribution of emo-tional intensity, the Five Factor Model and alex-ithymia. Journal of Research in Personality, 34, 471–497.

McCloskey, M., Wible, C.G., & Cohen, N.J. (1988). Isthere a special flashbulb-memory mechanism?Journal of Experimental Psychology, 117, 171–181.

Morse, C.K., Woodward, E.M., & Zweigenhaft, R.(1993). Gender differences in flashbulb memorieselicited by the Clarence Thomas hearings. TheJournal of Social Psychology, 133, 453–458.

Neisser, U. (1982). Snapshots or benchmarks? In U.Neisser (Ed.), Memory observed (pp. 43–48). SanFrancisco: Freeman.

Neisser, U., & Harsch, N. (1992). Phantom flashbulbs:False recollections of hearing the news about Chal-lenger. In E. Winograd & U. Neisser (Eds.), Affectand accuracy in recall: Studies of ‘‘flashbulb memo-ries’’ (pp. 9–31). New York: Cambridge UniversityPress.

Pennebaker, J.W., Paez, D., & Rime, B. (Eds.). (1997).Collective memory of political events. Social psy-chological perspectives. Mahwah, NJ: LawrenceErlbaum Associates Inc.

Pillemer, D.B. (1984). Flashbulb memories of theassassination attempt on President Reagan. Cogni-tion, 16, 63–80.

Pillemer, D.B. (1998). Momentous events, vivid memo-ries. Cambridge, MA: Harvard University Press.

Pillemer, D.B., Goldsmith, L.R., Panter, A.T., & White,S.H. (1988). Very long-term memories of the firstyear in college. Journal of Experimental Psychology,14, 709–715.

Rime, B., Finkenauer, C., Luminet, O., Zech, E., &Philippot, P. (1998). Social sharing of emotion: Newevidence and new questions. In W. Stroebe & M.Hewstone (Eds.), European Review of Social Psy-chology (Vol. 8, pp. 145–189). Chicester, UK: Wiley& Sons Ltd.

Rime, B., Philippot, P., Boca, S., & Mesquita, B. (1992).Long-lasting cognitive and social consequences ofemotion: Social sharing and rumination. In W.Stroebe & M. Hewstone (Eds.), European Review ofSocial Psychology (Vol. 3, pp. 225–258). Chicester,UK: Wiley & Sons Ltd.

Robinson, W.P. (1996). Deceit, delusion, and detection.Thousand Oaks, CA: Sage Publications.

100 CURCI ET AL.

Ross, M. (1989). Relation of implicit theories to theconstruction of personal histories. PsychologicalReview, 96, 813–819.

Rubin, D.C., & Kozin, M. (1984). Vivid memories.Cognition, 16, 81–95.

Ruiz-Vargas, J.M. (1993). ¿Como recuerda usted lanoticia del 23-F? Naturaleza y mecanismos de los‘‘recuerdos-destello’’ [How do you remember the 23-F news? Nature and mechanisms of ‘‘flashbulbmemories’’]. Revista de Psicolog õ a Social, 8, 17–32.

Scherer, K.R. (1984). On the nature and function ofemotion: A component process approach. In K.R.Scherer & P.K. Ekman (Eds.), Approaches to emo-tion (pp. 293–317). Hillsdale, NJ: Lawrence ErlbaumAssociates Inc.

Scherer, K.R. (1997). Profiles of emotion-antecedentappraisal: Testing theoretical predictions across cul-tures. Cognition & Emotion, 11, 113–150.

Schmolck, H., Buffalo, E.A., & Squire, L.R. (2000).Memory distortions develop over time: Recollec-tions of the O.J. Simpson trial verdict after 15 and 32months. Psychological Science, 11, 39–45.

Smith, C.A., & Ellsworth, P.C. (1985). Pattern of cog-nitive appraisal in emotion. Journal of Personalityand Social Psychology, 48, 813–838.

Sonnemans, J., & Frijda, N.H. (1995). The determinantsof subjective emotional intensity. Cognition &Emotion, 9, 483–506.

Tajfel, H. (1981). Human groups and social categories.Cambridge: Cambridge University Press.

Turner, J.C. (1987). Rediscovering the social group.Oxford: Blackwell.

Van de Geer, J.P. (1993). Multivariate analysis of cate-gorical data: Applications. Newbury Park, CA: Sage.

Weaver, C.A. III (1993). Do you need a ‘‘Flash’’ to forma flashbulb memory? Journal of Experimental Psy-chology: General, 122, 39–46.

Weller, S.C., & Romney, A.K. (1990). Metric scaling:Correspondence analysis. London: Sage.

Winograd, E., & Killinger, W.A. (1983). Relating age atencoding in early childhood to adult recall: Devel-opment of flashbulb memories. Journal of Experi-mental Psychology, 112, 413–422.

Winograd, E., & Neisser, U. (Eds.). (1992). Affect andaccuracy in recall: Studies of ‘‘flashbulb memories’’.New York: Cambridge University Press.

Wright, D.B. (1993). Recall of the Hillsborough disasterover time: Systematic biases of ‘flashbulb’ memories.Applied Cognitive Psychology, 7, 129–138.

Wright, D.B., & Gaskell, G.D. (1992). The constructionand function of vivid memories. In M.A. Conway,D.C. Rubin, H. Spinnler, & W.A. Wagenaar (Eds.),Theoretical perspectives on autobiographical mem-ory (pp. 275–293). The Netherlands: Kluver Aca-demic Publishers.

Wright, D.B., & Gaskell, G.D. (1995). Flashbulbmemories: Conceptual and methodological issues.Memory, 3, 67–80.

Young, F.W. (1981). Quantitative analysis of qualitativedata. Psychometrika, 46, 357–388.

Young, F.W., Takane, Y., & De Leeuw, J. (1978). Theprincipal components of mixed measurement levelmultivariate data: An alternating least squaresmethod with optimal scaling features. Psychome-trika, 43, 279–281.

FLASHBULB MEMORIES IN SOCIAL GROUPS 101