Remembering September 11th: The role of retention interval and ...

16 downloads 0 Views 160KB Size Report
Aug 9, 2005 - Retention interval and rehearsal effects on flashbulb and event memory for 11th September 2001 (9/11) were examined. In Experiment 1 ...
MEMORY, 2006, 14 (2), 129±147

Remembering September 11th: The role of retention interval and rehearsal on flashbulb and event memory Lauren R. Shapiro Emporia State University, KS, USA Retention interval and rehearsal effects on flashbulb and event memory for 11th September 2001 (9/11) were examined. In Experiment 1, college students were assessed three times (Groups 1 and 2) or once (Group 3) over 11 weeks. In Experiment 2, three new groups assessed initially at 23 weeks (Group 4), 1 year (Group 5), or 2 years (Group 6) were compared at 1 year and at 2 years with subsamples of those assessed previously. No effects of retention interval length or rehearsal were found for flashbulb memory, which contained details at each assessment. Event memory, but not consistency, was detrimentally affected by long retention intervals, but improved with rehearsal. Recall was higher for the reception event than for the main events. Also, consistency from 1 day to 11 weeks, but not 1 year to 2 years, was higher for flashbulb memory than for event memory. Event recall was enhanced when respondents conceived of their memory as vivid, frozen, and encompassing a longer period of time. Positive correlations were found for event memory with confidence in accuracy and with rehearsal through discussion at 2 years.

On the morning of 11th September 2001, 19 terrorists hijacked four commercial jets to attack the United States. Two planes were flown into the north and south towers of the World Trade Center (WTC), a third flew into the Pentagon, and a fourth crashed in a field near Shanksville, Pennsylvania (close to Pittsburgh) before reaching its final destination. In addition to the 266 people who were killed on board the planes and 2948 killed in the three buildings, 403 rescue workers were also killed when the WTC collapsed unexpectedly. The attacks were allegedly masterminded by Osama Bin Laden, the leader of the alQaida terrorist group in Afghanistan. Are the circumstances surrounding the moments in which you were informed of these tragedies fixed in your mind, seemingly unaffected by the passage of time? Brown and Kulik (1977) noted that people were able to recall with clarity and detail, even after many years, various aspects of their personal circumstances (i.e., reception event) when learning about unexpected, novel

public events, such as the assassination of President Kennedy. In their study, people were asked to indicate whether they recalled the reception event, to provide accounts of their memory, to rate personal consequentiality on a 5-point scale, and to estimate overt rehearsal for each of nine public (and one private) events. Six types of information (i.e., canonical categories) were commonly found in people's accounts of their personal circumstances, specifically their location, the activity interrupted by the news, the source of the news, their own emotional reactions, others' emotional reactions, and aftermath. Brown and Kulik proposed that flashbulb memory formed through a special mechanism in the brain that is activated when an event is surprising (albeit not overly so) and has personal consequentiality or significance. Flashbulb memories differ from other types of memory because they are elaborate, yet incomplete accounts of events that contain perceptually clear, idiosyncratic details. Despite 25 years of research, the existence of flashbulb

Correspondence should be addressed to Lauren R. Shapiro, Emporia State University, Psychology and Special Education, 1200 Commercial Street, CB#4031, Emporia, KS 66801, USA. Email: [email protected]

http://www.psypress.com/memory

# 2006 Psychology Press Ltd

DOI:10.1080/09658210544000006

130

SHAPIRO

memory is still hotly debated by supporters (e.g., Conway, 1995) and refuters (e.g., Neisser & Harsch, 1992; Rubin & Kozin, 1984). Resolution of this dispute is prevented because of a discrepancy in the minimum requirements for determining the existence of flashbulb memory, variations in both coding systems and research designs that prevent generalisation across studies, and use of inappropriate or no comparison events. The gap between supporters and refuters is fuelled by their disagreement about which categories are essential rather than extra, how much detailed information must be reported in each category, and the degree to which information must be consistently reported from initial to delayed assessments in order for the recall to qualify as flashbulb memory. Brown and Kulik's requirements of acknowledging recall of the reception event and providing at least one canonical category out of six seems far too lenient. However, insisting that a person include detailed information for every canonical category and/or additional categories about the reception context at each delay interval, regardless of length (Curci, Luminet, Finkenauer, & Gisle, 2001; Tekcan & Peynircioglu, 2002; Wright, Gaskell, & O'Muircheartaigh, 1998), is not reflective of how flashbulb memory is supposed to operate (Conway, 1995; Pillemer, 1990). Brown and Kulik indicated that the number and type of details that people report about their reception context vary in each assessment. Certain details are included in some retellings but omitted in other retellings, and new details may also appear in various retellings. Therefore, inconsistency by itself does not indicate the absence of flashbulb memory. A sensible premise for determining the existence of flashbulb memory for personally relevant events is: (a) the inclusion of location, activity, and source at the substantive (general) level as essential categories, but other types of information as optional, and (b) consistency of the three essential categories across reports at the general level. Errors at the fundamental level that change the meaning of the information should qualify as absence of flashbulb memory (Christianson, 1992), whereas errors in or omission of peripheral details would be inconsequential, as individual difference in completeness is the norm (Pillemer, 1990). Several researchers have indicated that the presence and consistency of location, activity, and source serve as major attributes in ascertaining the existence of flashbulb memory (e.g., Hornstein, Brown, & Mulligan, 2003; McCloskey, 1992;

Neisser & Harsch, 1992). For example, Christianson (1992) reported that 97% of respondents gave location, 92% provided their interrupted activity, and 100% gave source information about their reception event in an initial survey given 6 weeks after the death of Swedish Prime Minister Palme. At the delayed retest at 52±54 weeks, 89% of respondents provided location, 72% gave interrupted activity, and 92% provided source information. Kvavilashvili, Mirani, Schlagman, and Kornbrot (2003, Experiment 1) found that a high proportion of British and Italian respondents recalled their location (100%), activity (98%), and source (100%) for the reception event of September 11th assessed initially 3 months after the tragedy. In contrast, emotional reactions of the responder (and others, when the respondent was alone) are not always assessed. Previous studies have shown that past negative emotions are often overestimated in reports (Thomas & Diener, 1990) and that self-ratings of emotional reactions are rarely consistent over time (Christianson & Engelberg, 1999). Additionally, aftermath is rarely assessed by researchers, probably because it is not always possible for respondents to change their activity (e.g., being on a train, in a class) immediately upon learning of a tragedy. Supporters and refuters also argue about the degree of consistency between reports needed to confirm the existence of flashbulb memory. Refuters use the presence of inaccurate, vague, or incomplete details in second reports as evidence that information was based on reconstruction rather than on recall (e.g., McCloskey, Wible, & Cohen, 1988; Neisser & Harsch, 1992). There are two problems with this conclusion. First, it is important to differentiate errors and omissions made in canonical categories such as location, source, and activity, which focus on the central aspects of the reception event, from those made in extraneous categories such as time, day, others present, and clothing worn, which focus on peripheral aspects of the reception event. In their examination of autobiographical and event memory for September 11th, Tekcan, Ece, Gulgoz, and Er (2003) indicated that location, activity, and source were consistent across three assessments from 3 days to 6 months to 1 year, whereas consistency in time and in others present declined over time due to a loss of information. Time and day is inferred from other information (Brewer, 1988). Second, researchers use different coding systems to calculate consistency, making it impossible to interpret findings across studies.

REMEMBERING SEPTEMBER 11th

Many researchers (e.g., Cohen, Conway, & Maylor, 1994; Tekcan et al., 2003) have followed Neisser and Harsch's (1992) 3-point coding system, but then applied either lenient or strict criteria. There are also researchers who have developed their own coding system. For example, McCloskey et al. (1988) indicated that 61% of the responses to the location, activity, source, and reaction questions in their study across a 9-month period were ``the same'', 6% were ``more specific'', 19% were ``more general'', 8% were ``inconsistent'', and 6% were ``don't remember''. Inconsistency (and omissions) in accounts of reception events for public tragedies does not constitute evidence that flashbulb memories do not exist if researchers failed to assess personal consequentiality and surprise. Conway (1995) argued that without such measures there is no reason to believe that people would have formed flashbulb memories. To understand the durability of flashbulb memory, researchers have used a variety of designs, such as cross-sectional (e.g., Bohannon, 1988; Wright, 1993), longitudinal (e.g., Christianson, 1989; Neisser & Harsch, 1992; Smith, Bibi, & Sheard, 2003), and sequential (Conway et al., 1994; McCloskey et al., 1988; Pillemer, 1984; Tekcan et al., 2003). The effect of different retention intervals on recall of the reception event is tested through between-group comparisons in cross-sectional studies (e.g., 2 weeks versus 8 months after the Challenger space shuttle disaster, in Bohannon, 1988) and through within-group comparisons in longitudinal studies (e.g., 1 week and 6 months after September 11th, in Smith et al., 2003). However, because it is impossible to ascertain accuracy in autobiographical memory, it is inferred from consistency in information reported by the same respondent across time. In longitudinal studies, respondents may provide initial reports at the same retention interval, but second reports at different retention intervals (e.g., 3 days initially and either 15 months or 32 months follow-up for the announcement of O.J. Simpson's acquittal, in Schmolck, Buffalo, & Squire, 2000) or initial reports at different retention intervals, but second reports at the same retention interval (e.g., same day versus 1 week and again 8 weeks after Simpson's acquittal announcement, in Winningham, Hyman, & Dinnel, 2000). Unfortunately, retention interval and rehearsal via repeated testing are confounded in longitudinal studies. In contrast, sequential designs

131

include within-group comparisons in which at least two reports at different retention intervals are collected from the same respondents and between-group comparisons in which new (control) groups are added at later retention intervals. For example, McCloskey et al. (1988) assessed recall for the Challenger disaster twice in the repeated testing group (3 days and 9 months later) and once in the control group (9 months later). Everyone in the repeated testing group was able to provide substantive information about their location, activity, and source in the initial interview. Moreover, the repeated testing and control groups did not differ in their ability to provide reception event information in their 9-month reports (100% recalled source, 96% vs 94% recalled location, and 89% vs 87% recalled activity), demonstrating little benefit of rehearsal. Unfortunately, it is not clear if these findings are the norm or the exception because few sequential studies report statistical analysis for rehearsal (e.g., Tekcan et al., 2003). Dissention between supporters and refuters for the claim that flashbulb memory is different from other types of memory may be resolved by examining changes over time in the recall of the reception event and a suitable comparison event. Previous studies (e.g., Kvavilashvili et al., 2003) have used two or more shocking events that occurred at various time periods within the person's lifetime, which Conway (1995) argued was inappropriate because personal consequentiality of the events would vary for the recipient. Bohannon (1988) advocated comparing the reception event to the news event itself, as information would have the same emotional valence and was learned at approximately the same time (Finkenauer, Luminet, Gisle, El-Ahmadi, Van der Linden, & Philippot, 1998). Research has shown that facts in the news eventÐwhich have predominantly included difficult-to-remember numerical data (e.g., ``How many men were on the shuttle?'' for the Challenger explosion or ``How many planes were hijacked?'' for September 11th) rather than core features of what happenedÐwere remembered less well than attributes in the reception event (e.g., Bohannon, 1988; Curci et al., 2001; Er, 2003; McCloskey et al., 1988; Smith et al., 2003). For example, Tekcan et al. (2003) reported that 3 days after September 11th, 91% of the respondents reported details about major and minor categories of their reception event compared to 36% who had ``perfect'' memory for the date/time, number of planes, and crash locations

132

SHAPIRO

of the news event. Percentages at 6 months and 1 year remained high for recall of the reception event (94% and 92%, respectively), but declined for recall of the news event (13% and 10%, respectively). Investigators have also reported that consistency was higher for the reception event than for the news event (Smith et al., 2003; Tekcan et al., 2003). Unfortunately, it is impossible to determine if recall differs for the reception and news events when the generality of information is not the same for both. Nachson and Zelig (2003) reported no differences when the news event consisted of central and peripheral information and was classified using similar categories to the reception event (e.g., activity, source). In their study, factual memory decreased by 25% and flashbulb memory decreased by 36% from 2 weeks to 11 months after the assassination of Israel's Prime Minister Rabin. Unfortunately, they did not use control groups to determine the role of rehearsal on retention. Media exposure for September 11th was very distinct and was expected to impact on the encoding and retrieval of reception and news events (Neisser, 2003). For example, veteran newscasters were visibly shaken by information they were presenting (for example, Peter Jennings of ABC in New York wept), coverage was extended several weeks after the attacks, and vivid footage was repeatedly shown of the destruction of the World Trade Center (e.g., planes crashing and buildings collapsing), the Pentagon burning, and victims from these locations. Televised news reports containing multiple emotional and graphic images provide visual-verbal redundancy which results in better retrieval of the publicised event (Collins, Taylor, Wood, & Thompson, 1988; Gunter, 1980; Robinson & Levy, 1986; Son, Reese, & Davie, 1987). Media coverage of a traumatic, personally relevant public event also facilitates flashbulb memory formation by aiding viewers' awareness of the economical and political consequentiality of an event on their daily lives (Pillemer, 1998, 2003) and encouraging social sharing about the emotionally charged circumstances through discussion and rumination (Curci et al., 2001; Luminet, Bouts, Delie, Manstead, & Rime, 2000; Martin & Tesser, 1989, 1996). The goal of the current study was to improve psychological understanding of memory for stressful events by determining whether recall patterns for the reception and news events of September 11th varied over a 2-year period, whether flashbulb and factual memory were

differentially affected by retention interval, and whether rehearsal moderated retention interval effects on recall. In the original experiment, the role of retention interval was examined by varying the onset of the initial test (1 day vs 11 weeks) and the timing of the interim tests (2 vs 6 weeks). Also, the role of rehearsal on long-term recall was tested by varying the number of prior assessments (0 or 2). In the follow-up experiment, three additional retention intervals were examined (23 weeks, 1 year, and 2 years) and the effects of rehearsal on very long retention (i.e., 0, 1, or 3 at 1 year and 0, 1, or 4 at 2 years) were examined by comparing a subsample of original participants with new (control) participants. A diagram of the sequential design is shown in Table 1. The presence of flashbulb memory was based on respondents' ability to report location, activity, and source, with those in repeated conditions providing information in these three categories consistently over time at the general level. For the purposes of this research, the immediate assessment of the reception event served as a proxy for encoding against which consistency was assessed. Based on past findings, respondents who indicated personal relevance were expected to have flashbulb memory, regardless of the length of retention interval (Hypothesis 1a) or number of rehearsals (Hypothesis 1b). In contrast, recall of the news event was expected to be less elaborate and consistent at long rather than short retention intervals (Hypothesis 2a), but rehearsal was predicted to ameliorate long-term recall elaboration and consistency (Hypothesis 2b). Four other factorsÐ metacognitive self-assessments (e.g., memory conceptualised as vivid or vague), self-reported rehearsal (i.e., discussions with others), confidence in recall, and media exposure (i.e., number and type of media sources used to learn about 9/ 11)Ðwere examined in Experiment 2 to determine what effects, if any, they had on flashbulb and news event memory. TABLE 1 Number of post-event assessments of each retention interval in Experiments 1 and 2 Group 1 2 3 4 5 6

1st

2nd

3rd

1±2 days 1±2 days 11 wks 23 wks 1 year 2 years

2 wks 6 wks 1 year 1 year 2 years

11 wks 11 wks 2 years 2 years

4th 1 year 1 year

5th 2 years 2 years

REMEMBERING SEPTEMBER 11th

EXPERIMENT 1 Method Participants and design The participants were 61 students enrolled in three undergraduate psychology courses at a small Midwestern university. Students participated as part of their class project examining flashbulb and event memory, and subsequently signed permission letters allowing use of their data for research. Data were not included for one female student who declined research participation and from five female and two male students whose data were not available for all assessments. The age of the final sample on 11th September 2001 ranged from 17 years to 45 years (M = 21.21 years, SD = 4.25 years). Three groups were assessed one or three times over an 11-week period for their reception context and event recall of September 11th. Group 1 (11 women, 2 men) and Group 2 (14 women, 6 men) were initially assessed at 1 to 2 days post-event. The second assessment for Group 1 occurred at 2 weeks post-event and for Group 2 at 6 weeks postevent. The third assessment for Groups 1 and 2 occurred at 11 weeks post-event. Group 3 (12 women, 8 men) was initially assessed at 11 weeks post-event. Materials and procedure A survey called the Flashbulb and Event Memory Survey (FEMS) was constructed to elicit three of Brown and Kulik's (1977) six original canonical categories of reception context (Section 1) and a description of the news event itself (Section 2). In Section 1, respondents were instructed to indicate the following information concerning their learning of the tragic events of September 11th: (a) location (i.e., ``Where were you?''), (b) interrupted activity (i.e., ``What were you doing?), and (c) source (i.e., ``How did you find out?''). A fourth item assessed perceived personal relevance of this public tragedy to the respondent by indicating on a Likert-type scale from 1 to 5 the impact the events of September 11th had on their lives, with ``1'' representing a general relevance (e.g., as a USA citizen), ``3'' representing a specific relevance (e.g., family/ friends reside in New York City or Washington, DC), and ``5'' representing a very specific relevance (e.g., loss of family/friend). Although some

133

respondents provided information about their own and others' immediate emotional reaction in their reports, no specific request was made for this information. People were visibly upset (e.g., shaking, crying) in response to this tragedy and told classmates that they were feeling shock, surprise, and fear. In Section 2, respondents were instructed to write down everything they knew about the tragedy. Instructions for the 2-week through 11-week surveys included an additional prompt to label information as ``old'' if they knew about it on September 11th, and to label information as ``new'' if they had learned about it after September 11th. The respondents' ability to differentiate old from new information was used as a measure of their source-monitoring ability. Coding Completeness of response. RECEPTION EVENT. The content for the three canonical categories of location, interrupted activity, and the source was identified and organised by groups. Location responses were categorised as public (e.g., work, outside, lobby) or private (e.g., home, friend's home, personal vehicle). The interrupted activity was categorised as morning activities (e.g., sleeping, waking, dressing), working (e.g., in school, at a job), in transit (e.g., driving, walking), or other (e.g., watching television). Responses for source were identified as television, radio, or person. The fourth canonical category, relevance of the event to the respondent, was coded on a self-reported Likert-type scale ranging from 1 to 5. To obtain an index of total recall, responses to each category were identified as either substantive (i.e., an answer was provided, even if not highly specific) and assigned a score of ``1'', or absent (including ``I don't remember'' responses) and assigned a score of ``0''. For elaborated recall, substantive responses were reclassified as general information, which received a score of ``1'', or as detailed information, which received a score of ``2''. For example, in response to the location prompt, the statement ``in my Economics class'' received 2 points, whereas the statement ``in class'' received 1 point. The scores from each category were summed into a WAS composite score (ranging from 0 to 6 points). NEWS EVENT. A coding manual was developed for coding accurate and inaccurate responses in the FEMS. Also, a count was made of any source misattributions (e.g., stating information was new

134

SHAPIRO

when it was known on 9/11). However, there were few inaccurate responses (M = 0.13 for immediate, M = 0.27 at short delay, and M = 0.13 at long delay) or source misattributions (Ms = 1.24 at short and long delays) so these measures are not discussed further. Eight features, identified a priori, were grouped into one of three categories: main events, which included the attack on the World Trade Center (WTC), the attack on the Pentagon, and the plane crash in Pennsylvania; attackers/weapons, which comprised Osama Bin Laden as the mastermind of the plot, the terrorists who enacted the plot, and the planes that served as the weapons; and aftermath, which consisted of the death toll from WTC, Pentagon, and all four planes, as well as the rescue efforts at the tragedy sites (see Table 2). An additional feature, the President as an official target, was identified, but this item was dropped due to extremely low responses (M = .03). Shortly after it occurred, essential facts about the main events of the 9/11 attack were presented by the media. In contrast, various details about the attackers/weapons and aftermath were updated by the media several times as information became available from official investigation reports. To obtain total recall in each category, responses to each item were identified as substantive (1 point) or absent (0 points) and then summed. Substantive answers were reclassified as incomplete information (e.g., location and action or hint of tragedy), complete information (i.e., must indicate the method of the attack and up to one detail), or detailed information (i.e., complete information and at least one additional detail) and assigned 1, 2, and 3 points, respectively. An

TABLE 2 Questions corresponding to the eight news event features organised into three components Main events 1. What happened in New York City? [World Trade Centre] 2. What happened in Washington, DC? [Pentagon] 3. What happened in Pennsylvania? Major players and weapons 4. Who was the prime suspect (mastermind) behind the terrorist attacks? 5. How many terrorists were on the planes? 6. What was the number of planes used to achieve the event? Aftermath 7. What was the death toll for the attacks at all locations? 8. What rescue efforts were made for the three locations and how long did they last?

example of an incomplete response would be stating ``The WTC was attacked.'' An example of a complete response would be stating ``Two hijacked planes crashed into both towers of the WTC.'' An example of a detailed response would be stating ``WTC was attacked by 2 hijacked planes and collapsed, causing damage to #7 that led to it collapsing as well.'' Details for the WTC included stating that there were 110 stories per tower, the North Tower was attacked first or the South Tower was attacked second, which floors the planes entered, information about the destruction or structural damage to surrounding buildings, etc. Elaborated recall in each category was computed by tallying the number of elaboration points assigned across items. Consistency. RECEPTION EVENT. To assess the consistency of initial responses in location, interrupted activity, and source with those given at subsequent retention intervals, Neisser and Harsch's (1992) coding scheme was used. A score of ``2'' was assigned when information in a category was essentially (i.e., conveyed the same meaning) or exactly the same across assessments, whereas a score of ``1'' was assigned when information was basically the same across assessments (e.g., some of the same elements were present). For example, because the statements ``at home'' and ``in my house'' convey the same meaning, location would be scored as ``2''. In contrast, stating ``I was in Morse Hall'' initially and then stating ``I was in the dorm lounge'' subsequently shows only some convergence and would be scored as ``1''. A score of ``0'' was assigned when information in a category was completely inconsistent across assessments, including when it was absent in at least one of the assessments. For example, in the initial assessment reporting the source as the radio, such as ``I heard it on the radio'', but in a subsequent assessment reporting it as a person, such as ``My mom came in my room to tell me'', would be wholly inconsistent. A WAS composite score ranging from 0 (completely inconsistent) to 6 (completely consistent) was constructed by summing the consistency scores in the three categories. NEWS EVENT. The FEMS was also coded for content consistency only in main events, as information in this category was given immediately and facts did not vary over time. Consistency was assessed in the initial and second surveys, initial and third surveys, and second and third surveys using a modified version of Neisser and Harsch's

REMEMBERING SEPTEMBER 11th

(1992) scoring system. A score of ``2'' was assigned when the main idea or action and most of the details in an item were essentially or exactly the same across assessments, whereas a score of ``1'' was assigned when the main idea/action was the same across assessments, but some of the details were missing or new. For example, because the statements ``Two planes were hijacked and flown in the WTC. The two buildings fell to the ground'' and ``Two hijacked planes flew into each tower causing them to collapse'' convey the same actions and details, WTC would be scored as ``2''. In contrast, stating ``hijackers crashed two planes into each of the towers, the towers went up in flames and collapsed an hour later, people jumped from the 81st floor'' in one assessment and stating ``2 planes flew into the towers several minutes apart and the towers collapsed'' in another assessment shows convergence only for the action, as two details are missing and a new detail was given, so WTC would be scored as ``1''. A score of ``0'' was assigned when information in a category was completely inconsistent across assessments, including when it was absent in one or both assessments. For example, reporting in the initial assessment ``a bomb exploded in the twin towers'' but reporting ``hijacked planes crashed into the twin towers'' in a subsequent assessment would be totally inconsistent. Reliability. Two judges independently coded 25% of the surveys, obtaining high reliability for recall and consistency. Percent agreement for the reception event averaged 100% in recall and 92% in consistency, and for the news event averaged 96% in recall and 94% in consistency. Any discrepancies between coders were subsequently discussed until they reached full agreement on coding. One judge proceeded to code the remaining surveys.

Results Content description Reception event. All three attributes were included in prompted-recall reports by 97% of the respondents at 1 day and by 100% of the respondents at 11 weeks. Respondents indicated in their initial assessment that they were more likely to be in a private setting (62%, e.g., in their own or a friend's home) than in a public setting (38%, e.g., at work or school) when they learned of the September 11th attack. Many people (53%) reported

135

that their morning routine was interrupted (e.g., getting up, dressing), although some people were already working or in school (19%), some were in transit (24%, such as walking or driving to class or work), and some were doing other activities (2%, e.g., watching television, on the internet). Finally, most respondents were informed of the event by a person (55%) rather than by news reports broadcast on television (24%) or radio (21%). In terms of relevance, 36% of the respondents indicated ``1'', 34% indicated ``2'', 25% indicated ``3'', 4% indicated ``4'', and 1% indicated ``5'' out of 5 on the scale. News event. An examination of free-recall reports at 1 day indicated that 45% of respondents provided all three items in main events, 27% provided two items, and 19% provided only one item, whereas reports at 11 weeks showed that 41% of the respondents provided all three items, 30% provided two items, and 19% provided only one item. In contrast, only 1% of respondents provided all three items for attackers/weapons in their 1-day reports, 13% provided two items, and 49% provided only one item, versus 1% of respondents who provided all three items in their 11-week reports, 15% provided two items, and 47% provided only one item. No one could provide both items for aftermath in 1-day or 11-week reports, but 13% provided one item at 1 day and 11% provided one item at 11 weeks. Recall over time Groups 1 and 2 were compared to determine whether differences in the short-term retention interval produced variations in the pattern of recall over the 11-week retention interval and in long-term retention. Separate 2 6 3 (Condition 6 Retention interval) repeated-measures analyses of variance (ANOVAs) were performed using two measures for the reception event (WAS elaboration recall and relevance scores), and six measures of the event (total recall and elaboration recall of main events, attackers/weapons, and aftermath). WAS total recall was not analysed because all respondents, except for one person who omitted activity for the first assessment, provided substantive responses for all three categories on every assessment. Condition (2-week vs 6-week shortterm delay) was the between-subjects variable and retention interval (immediate vs short-term vs long-term) was the within-subjects variable.

136

SHAPIRO

Reception event. A main effect of retention interval, F(2, 62) = 3.63, p < .04, Z2 = .11, indicated that WAS elaborated recall was higher when assessed immediately (M = 4.56, SE = .18) than after a short delay (M = 4.17, SE = .20), p < .02. However, recall after a long delay (M = 4.30, SE = .18) did not differ significantly in elaboration from either immediate (p > .07) or short delay recall (p > .39). In addition, respondents' ratings of relevance did not differ from the initial (M = 1.88, SE = .15) to the third assessment (M = 1.91, SE = .18), p > .80. The absence of condition effects indicated that varying the short-term assessment did not differentially affect recall of the reception event over time, nor did it impact negatively on longterm recall. News event. Table 3 displays the means and standard errors for total and elaborated recall of main events, attackers/weapons, and aftermath at the immediate, short-term, and long-term assessments. An effect of retention interval for total recall of attackers/weapons, F(2, 62) = 3.56, p < .04, Z2 = .10, was interpreted within the Condition 6 Retention interval interaction, F(2, 62) = 4.20, p < .03, Z2 = .12. Planned comparisons of the interaction revealed different patterns of recall over time for Groups 1 and 2 for attackers/weapons. Total recall for attackers/weapons was better

at the second and third assessments than at the initial assessment for Group 1 (p < .05), whereas recall was only better at the second than the third assessment for Group 2 (p < .01). The two groups did not differ in total recall of attackers/weapons in their first (p > .57), second (p > .41), or third assessments (p = .067), although the trend in the latter was for Group 1 to have higher recall of attackers/weapons than Group 2. Main effects of retention interval were also found for elaborated recall of attackers/weapons, F(2, 62) = 3.20, p < .05, Z2 = .09, and of aftermath, F(2, 62) = 3.18, p < .05, Z2 = .09. Elaborated recall of attackers/weapons was lower initially than at the second assessment (p < .01), whereas elaborated recall of aftermath was higher in both the initial and second assessments than at the third assessment (p < .05). In conclusion, varying the timing for the short-term assessment had minimal effect on the pattern of recall and long-term recall of attackers/weapons, but no effect on recall of main events and aftermath. In summary, respondents demonstrated no decline in memory for general facts in the reception event or in the main events of 9/11 over 11 weeks, and long-term retention for reception context and main event details did not decline significantly from the immediate assessment. In contrast, memory for specific details about the

TABLE 3 Recall: Experiment 1 Total recall Assessment

Main events

Elaborated recall

Attackers

Aftermath

Main events

Attackers

Aftermath

Immediate (1 day) Group 1 2.38 (.20) Group 2 2.60 (.16) Total 2.49 (.18)

1.15 (.20) 1.30 (.16) 1.23 (.18)

.08 (.12) .40 (.10) .24 (.11)

6.00 (.52) 5.75 (.42) 5.88 (.47)

3.61 (.44) 3.40 (.35) 3.51 (.40)

1.00 (.27) 1.90 (.22) 1.45 (.25)

Short-term Group 1* Group 2** Total

2.38 (.22) 2.45 (.18) 2.42 (.20)

1.62 (.20) 1.40 (.16) 1.51 (.18)

.15 (.16) .40 (.13) .28 (.15)

5.54 (.38) 5.30 (.31) 5.42 (.35)

4.46 (.37) 4.00 (.30) 4.23 (.34)

1.54 (.35) 1.40 (.28) 1.47 (.32)

Long-term (11 weeks) Group 1 2.46 (.24) Group 2 2.40 (.20) Total 2.43 (.22)

1.54 (.18) 1.05 (.15) 1.30 (.17)

.15 (.18) .25 (.07) .20 (.13)

5.31 (.49) 5.25 (.39) 5.28 (.44)

4.31 (.38) 3.20 (.31) 3.76 (.35)

.85 (.23) 1.00 (.19) .93 (.21)

Control (11 weeks) Group 3 1.70 (.20)

.40 (.15)

.00 (.07)

3.60 (.39)

1.75 (.31)

.15 (.19)

Mean total recall and elaborated recall (and Standard Errors) for event categories at immediate, short-term, and long-term assessments by condition in Experiment 1. * = 2 weeks, ** = 6 weeks.

REMEMBERING SEPTEMBER 11th

attackers/weapons improved over the 11 weeks, which may be attributed to media reports during this time period. Role of retention interval The effect of varying retention interval on initial recall was examined using separate MANOVAs for reception context variables and for event variables. For these analyses, Groups 1 and 2 were combined because they had the same initial assessment. Reception event. The condition effect was marginally significant, F(51) = 3.68, p = .06, Z2 = .07, indicating that respondents tended to provide more details when their initial assessment for the reception context was 1 day (M = 4.52, SE = .18 for Group 1/2) rather than 11 weeks post-event (M = 3.95, SE = .24 for Group 3). Not surprisingly, Group 1/2 (M = 1.88, SE = .15) and Group 3 (M = 2.25, SE = .25) did not differ on initial relevance (p > .18). News event. Condition effects were significant for total recall of main events, F(51) = 10.98, p < .01, Z2 = .18, attackers/weapons, F(51) = 21.61, p < .01, Z2 = .30, and aftermath, F(51) = 7.22, p < .02, Z2 = .12, as well as for elaborated recall of main events, F(51) = 16.30, p < .01, Z2 = .24, attackers/weapons, F(51) = 18.74, p < .01, Z2 = .27, and aftermath, F(51) = 31.93, p < .02, Z2 = .39. As shown in Table 3, both indices were significantly higher when initial assessment was at 1 day (see immediate total for Group 1/2) than at 11 weeks post-event (see Group 3 on the bottom row). Role of rehearsal Separate MANOVAs were used to examine the effect of rehearsal (i.e., two previous rehearsals for Groups 1 and 2, albeit at different shortterm retention intervals, versus no rehearsal for Group 3) on long-term recall. Reception event. Group 1 (M = 4.31, SE = .29 and M = 2.15, SE = .25), Group 2 (M = 4.31, SE = .23 and M = 1.75, SE = .24), and Group 3 (M = 3.95, SE = .24 and M = 2.25, SE = .25) did not differ on their 11-week assessment of WAS elaborated recall and relevance, respectively (ps > .49).

137

News event. Significant condition differences for total recall of main events, F(2, 50) = 4.29, p < .02, Z2 = .15, of attackers/weapons, F(2, 50) = 12.77, p < .001, Z2 = .34, and of aftermath, F(2, 50) = 2.91, p = .06, Z2 = .10, as well as for elaborated recall of main events, F(2, 50) = 5.67, p < .03, Z2 = .19, of attackers/weapons, F(2, 50) = 14.40, p < .001, Z2 = .37, and of aftermath, F(2, 50) = 5.80, p < .01, Z2 = .19, were found (see Table 3). Planned comparisons revealed different patterns. First, although Groups 1 and 2 did not differ from each other (ps > .61), both provided higher levels of total and elaborated recall of main events and elaborated recall of aftermath than did Group 3 (ps < .02). Second, Group 2 tended to provide a higher level of total recall of aftermath than Group 3 (p < .02). Third, Group 1 provided higher levels of total and elaborated recall of attackers/ weapons than did Group 2 (ps < .04), who in turn provided higher levels of both indices than did Group 3 (ps < .01). The fact that there were two prior rehearsals seemed to play a more important role in long-term recall than the timing of the second rehearsal per se. Consistency Separate MANOVAs were conducted for WAS consistency and for main event consistency. Reception event. Only one significant condition effect was found for WAS consistency scores, F(1, 31) = 6.79, p < .02, Z2 = .18: Higher consistency was found from the first to second assessment when the latter was done at 2 weeks (Group 1: M = 6.00, SE = .00) rather than at 6 weeks (Group 2: M = 5.50, SE = .15). However, the two groups did not differ on WAS consistency scores from the first to third assessments (Group 1, M = 5.92, SE = .07 versus Group 2, M = 5.60, SE = .18; p > .18) or from the second to third assessments (Group 1, M = 5.85, SE = .10 versus Group 2, M = 5.50, SE = .21; p > .22). Overall, consistency across assessments was very high, ranging from 92% to 100% for the three categories combined. News event. There were no significant differences found for main event consistency from the first to second (Group 1, M = 4.46, SE = .51 versus Group 2, M = 3.80, SE = .41; p > .32), first to third (Group 1, M = 4.46, SE = .53 versus Group 2, M = 4.10, SE = .18; p > .42), or second to third assessments (Group 1, M = 4.69, SE = .37 versus Group 2, M = 4.20, SE = .30; p > .31). Unlike consistency

138

SHAPIRO

in the reception context, consistency in main events ranged from 0% to 100% for the three items combined. Direct comparison of recall for reception event and main events of 9/11 Total recall was higher for the reception event than for the main events of 9/11 at 1 day, t(32) = ± 3.92, p < .001 (M = 2.97, SE = .17 vs M = 2.51, SE = .12), and at 11 weeks, t(52) = 76.66, p < .001 (M = 3.00, SE = .00 vs M = 2.15, SE = .13). In addition, consistency from 1 day to 11 weeks was higher for the reception event than for the news event, t(32) = 4.51, p < .001 (M = 5.73, SE = .12 vs M = 4.24, SE = .33). However, these differences may be due to variations in elicitation, given that free recall was used for factual memory and probed recall for flashbulb memory.

Discussion All respondents demonstrated flashbulb memory for September 11th despite indications of low personal relevance. This sample exceeded the minimal requirements by providing some details of their reception event at each retention interval and consistently reporting these details beyond the general level (i.e., WAS elaboration and consistency scores were higher than 3). Consistent with Hypothesis 1a, retention interval length did not affect total recall or consistency of information about the reception event over the 11-week period and only marginally affected the degree of recall elaboration provided in reports. Also, as predicted in Hypothesis 1b, long-term recall elaboration and consistency for the reception event was not improved by having two rehearsals or by having the second rehearsal at a shorter rather than a longer time lag (2 vs 6 weeks). Unlike recall of the reception event, recall of various aspects of the news event was detrimentally affected by a long retention interval. Partial support for Hypothesis 2a was found because respondents provided less elaborate information about the main events, attackers/ weapons, and aftermath when reports were initially assessed at 11 weeks rather than 1 day. Contrary to expectations, the news event was recalled consistently over the 11-week period.

Some support for Hypothesis 2b was also indicated, as rehearsal enhanced long-term retention of the September 11th attacks, but not recall consistency. Interestingly, both short-term rehearsal intervals inoculated respondents against forgetting details for main events and for aftermath at 11 weeks, but the 2-week rehearsal was more effective than the 6-week rehearsal in aiding long-term elaborated recall of attackers/weapons. Overall, long-term retention of the news event was more dependent on the number of rehearsals per se rather than the timing of the rehearsals.

EXPERIMENT 2 In Experiment 2, new retention intervals of 23 weeks, 1 year, and 2 years were added to determine whether the pattern of forgetting for the reception and news event memory differed at longer delays. Some researchers have found that flashbulb memory was consistent up to 1 year after the event (e.g., Christianson, 1989; Pillemer, 1984), with relatively little distortion of information (i.e., inconsistency rather than accuracy, as the latter cannot be determined without an objective record), yet few investigators have compared consistency in long-term recall of the reception and news events. For example, Bohannon and Symons (1992) indicated that reception and news event memory for the Challenger disaster declined from 2 weeks to 8 months by 25%, whereas an additional 20% decline in memory was shown from 15 to 36 months only for the news event. Tekcan et al. (2003) reported almost no loss of details about the reception event of September 11th by Turkish participants over 1 year, but recall of the news event declined 23% from 3 days to 6 months before stabilising until 1 year. However, the discrepancy in recall is likely due to differences in the level of specificity in information obtained for the two types of events. The follow-up study, rather than obtaining only numerical facts (i.e., number of planes) as had been done in past studies, used prompted recall to elicit main events and specific details about the attackers/weapons, and aftermath of the news event. A subsample of respondents from Experiment 1 agreed to participate in the 1-year and 2year assessments, and their recall was compared with that of new respondents whose flashbulb and factual memory had not been assessed previously.

REMEMBERING SEPTEMBER 11th

Method New participants The new participants were 85 undergraduate students (M = 20.2 years, SD = 2.45 years, range = 18 to 37 years) enrolled in three psychology courses at a small Midwestern university. They participated as part of their class project examining flashbulb and event memory, and signed informed consent forms allowing their data to be used for research. Design and Procedure To examine how delays in the initial assessment affected flashbulb and event recall, three new retention intervals were included for comparison with the two retention intervals used in the original study (i.e., 1 day and 11 weeks). Initial assessment for Group 4 (16 women, 6 men) occurred at 23 weeks post-event, for Group 5 (20 women, 7 men) at 1 year post-event, and for Group 6 (26 women, 10 men) at 2 years postevent. To examine rehearsal effects, the students from the original study, and subsequently from this follow-up study, were invited through letters and emails to participate in additional assessments and were compared against students who had only one assessment (Group 5 at 1 year and Group 6 at 2 years). The number of people who agreed to participate in subsequent assessments varied. At 1 year post-event (‹ 8 weeks), 35 participants agreed to a second (Group 3/4, n = 15) or a fourth assessment (Group 1/2, n = 20). At 2 years postevent (±2 weeks/+6 weeks), 35 participants agreed to a second (Group 5, n = 13), third (Groups 3/4, n = 5), or fifth assessment (Group 1/2, n = 17). Materials The original FEMS was used for the 23-week assessment by Group 4. However, changes and additions were made to the original survey for all of the groups to use during the 1-year (FEMS1) and 2-year assessments (FEMS2). Section 1. For the reception context assessment, the questions in Section 1 corresponding to location, source, activity, and relevance remained the same in all versions. There were additional questions corresponding to own and others' emotional reactions (FEMS1) and to aftermath (FEMS2) that were not scored. At the end of this

139

section in FEMS1, respondents were asked to report what consequences 9/11 had on their lives. There were two additions and one modification in FEMS2. Respondents were asked to indicate for each of the six questions how confident they were that reported information was accurate using a Likert-type scale from 1 (very confident) to 5 (not very confident). To assess rehearsal, a question was added in which respondents provided estimations of the number of times that they had spoken to people about their personal circumstances for learning of September 11th (a) in the week that it happened, (b) from September 2001 to August 2002, and (c) from September 2002 to August 2003, using a Likert-type scale ranging from 1 (constantly) to 7 (seldom). For the consequence question, respondents were asked to indicate why they felt the 9/11 event was important to them/their lives even though they were not at one of the three eastern locations at the time of the crisis and had few to no personal ties to the affected areas. Section 2. For the event memory assessment, nine cued-recall prompts were used in FEMS1 and FEMS2 instead of the free recall prompt to ensure that a complete report would be obtained. The prompts corresponded to eight items grouped into three categoriesÐthe main events, attackers and weapons, and aftermathÐplus an unused item of official target. Table 2 contains the questions corresponding to the eight news event features used in the study. In FEMS2, respondents were also asked to indicate their confidence level that information for each of the nine items was accurate using a Likert-type scale from 1 (very confident) to 5 (not very confident). Additionally, respondents were instructed to provide estimations of the number of times that they had spoken to people about various aspects of September 11th itself: (a) in the week that it happened, (b) from September 2001 to August 2002, and (c) from September 2002 to August 2003, using a Likerttype scale ranging from 1 (constantly) to 7 (seldom). Section 3. A new section containing a set of three metacognitive self-assessment questions was created for the FEMS1 and also used for the FEMS2 to elicit respondents' conceptualisation of their memory for September 11th as: (a) vivid versus vague, (b) encompassing initial seconds versus a longer period, and (c) frozen versus fading over time.

140

SHAPIRO

Two additional questions were constructed for the FEMS1, but not used in the FEMS2. The items assessed the effect of media exposure on recall, specifically the number of sources participants used to learn about the event (e.g., radio, internet) and whether they were exposed to a 6-month televised special on 9/11. Coding Coding for the completeness of response and for consistency in the reception and news events was the same as in Experiment 1 and data were coded by the same judge. Metacognitive self-assessment of memory. Each response was artificially coded as ``1'', ``2'', or ``3'', depending on the number of choices for that item. For example in the item ``My memory for this event is __'', respondents who selected the first option of ``vivid'' would receive a ``1'', whereas those who indicated the second option of ``vague'' would receive a ``2''. Media exposure. For the media sources question, responses were coded as ``1'' if the source was selected and ``0'' if it was not. The total number of sources was then tallied. In addition, exposure to televised 6-month specials was artificially coded as ``1'' for yes and ``0'' for no.

Results Content description Reception event. All three attributes were provided by 98% of the respondents at the 1-year assessment and by 97% of the respondents at the 2-year assessment; the balance of the respondents produced two attributes. According to their initial assessment at 1 year or at 2 years, respondents were more likely to be in a private setting (56%, such as in their own or a friend's home) than in a public setting (44%, such as at work or school). Many people indicated that their job or class (34%) was interrupted, although some people were doing their morning routine (25%), some were in transit (18%, such as walking or driving to class or work), and some were doing other activities (23%, e.g., watching television, on the internet). Finally, respondents were informed of the event by a person (54%) rather than by news

reports broadcast on television (29%) or radio (16%). This sample was similar to the previous one for location and for source, but not for activity interrupted. In terms of their initial self-reported personal relevance, 41% of respondents indicated ``1'', 39% indicated ``2'', 12% indicated ``3'', 5% indicated ``4'', and 2% indicated ``5'' out of 5 on the scale. News event. An examination of main events showed that 66% of the respondents at 1 year provided all three items, 18% provided two items, and 13% provided one item, whereas 79% of respondents at 2 years provided all three items, 10% provided two items, and 9% provided one item. In contrast, 13% of respondents at 1 year were able to provide all three items for attackers/ weapons, 34% provided two items, and 51% provided only one item versus 19% of the respondents at 2 years who provided all three items, 52% provided two items, and 25% provided only one item. Poor recall was also indicated for aftermath, as 21% of respondents at 1 year provided both items and 53% provided one item versus 13% of the respondents at 2 years who provided both items and 43% provided one item. Recall over time Three sets of 2 6 2 (Condition 6 Retention interval) repeated-measures ANOVAs were conducted. In the first analysis, Group 1/2 (immediate) was compared to Group 3/4 (long-delay) across initial and 1-year assessments for reception context variables (i.e., WAS elaborated recall, relevance). In the second analysis, Group 1/2 (immediate) was compared to Group 5 (very long delay) across initial and 2-year assessments for reception context variables. In the third analysis, Group 1/2 was compared to Group 5 across 1-year and 2-year assessments for reception context and for event variables. WAS total recall was not included in any analyses because all respondents, except for one person who omitted activity and one person who omitted source, provided substantive responses for all three categories on their initial assessment. Group 3/4 was not included in the second and third set of analyses because this group had too few respondents at 2 years. Also, event variables were not included in the first two sets of analyses because different recall prompts were used in the 1-day to 23-week assessments than for the 1- to 2-year assessments.

REMEMBERING SEPTEMBER 11th

Reception event. No retention effects were found in the first analysis, but condition effects, F(1, 32) = 6.52, p < .02, Z2 = .17, indicated that WAS elaborated recall was higher for Group 1/2 (M = 4.74, SE = .20) than for Group 3/4 (M = 3.97, SE = .23). There were no significant effects found for WAS elaborated recall (M = 4.65, SE = .23 and M = 4.74, SE = .25) or relevance (M = 2.08, SE = .19 and M = 2.17, SE = .27) in the second (ps > .41) and third set (ps > .10) of analyses, respectively. News event. Retention interval effects were found for total recall of main events, F(1, 23) = 4.99, p < .04, Z2 = .18, and attackers/weapons, F(1, 23) = 6.67, p < .001, Z2 = .42, and for elaborated recall of attackers/weapons, F(1, 23) = 6.16, p < .03, Z2 = .21. Condition effects were found for total recall of attackers/weapons, F(1, 23) = 16.74, p < .001, Z2 = .42, and of aftermath, F(1, 23) = 4.78, p < .04, Z2 = .17, and for elaborated recall of attackers/weapons, F(1, 23) = 16.57, p < .001, Z2 = .42. The pattern was the same in each case. Surprisingly, recall was better at 2 years than at 1 year (M = 2.89, SE = .09 vs M = 2.61, SE = .12 for total recall of main events; M = 2.09, SE = .15 vs M = 1.46, SE = .09 and M = 4.66, SE = .28 vs M = 3.84, SE = .19 for total and elaborated recall of attackers/weapons, respectively). Also, contrary to prediction, Group 5 had higher recall than Group 1/2 (M = 2.17, SE = .14 vs M = 1.39, SE = .13 and M = 4.96, SE = .25 vs M = 3.54, SE = .24 for total and elaborated recall of attackers/ weapons, respectively; M = 1.21, SE = .17 vs M = .69, SE = .16 for total recall of aftermath). Role of retention interval Three sets of MANOVAs were conducted. In the first set, WAS elaborated recall and relevance were compared across five initial retention intervals (1 day, 11 weeks, 23 weeks, 1 year, 2 years). In the second and third sets, event variables were compared across three retention intervals (1 day, 11 weeks, 23 weeks) and two retention intervals (1 year, 2 years), respectively. Reception event. There were no differences in WAS elaborated recall (M = 4.38, SE = .22, ranging from M = 4.00 to M = 4.66) due to increasingly longer retention intervals between the event and the assessment (p > .32). However, retention interval effects were found for relevance, F(4, 131) = 3.37, p < .02, Z2 = .09: 11-week (M = 2.26, SE = .22) and 1-year (M = 2.33, SE = .18) assessments

141

contained higher relevance than 23-week (M = 1.50, SE = .20) and 2-year (M = 1.74, SE = .16) assessments. No other differences were significant. News event. In the second set of analyses, condition effects were found for total recall of main events, F(2, 72) = 6.28, p < .01, Z2 = .15, of attackers/weapons, F(2, 72) = 15.90, p < .001, Z2 = .31, and aftermath, F(2, 72) = 5.60, p < .01, Z2 = .14, as well as for elaborated recall of main events, F(2, 72) = 9.27, p < .001, Z2 = .21, of attackers/ weapons, F(2, 72) = 13.16, p < .001, Z2 = .27, and aftermath, F(2, 72) = 18.43, p < .001, Z2 = .34. Planned comparisons showed the same pattern for total and elaborated recall of all three categories: initial assessments at 1 day (Group 1/2) contained more information than those at 11 weeks (Group 3) and 23 weeks (Group 4), ps < .02 (see initial assessment in Table 4). In the third set of analyses, condition effects were found for total, F(2, 72) = 10.19, p < .01, Z2 = .15, and elaborated recall, F(2, 72) = 12.93, p < .01, Z2 = .18, of aftermath: initial assessment at 1 year was higher than that at 2 years (see Group 5 at 1 year and Group 6 at 2 years in Table 4). No other differences were significant. Role of rehearsal Separate MANOVAs were conducted using the reception context variables and the event variables at 1 year and at 2 years. Group 1/2 (three rehearsals), Group 3/4 (one rehearsal), and Group 5 (no rehearsal) were compared at 1 year, whereas Group 1/2 (four rehearsals), Group 5 (one rehearsal), and Group 6 (no rehearsal) were compared at 2 years. Group 3/4 was not included in the 2-year analysis because there were too few respondents. Reception event. There were no significant differences due to the number of rehearsals for WAS elaborated recall (M = 4.48, SE = .25 and M = 4.55, SE = .29, ps > .14), or relevance (M = 2.32, SE = .36 and M = 1.96, SE = .24, ps > .34) at 1 year or 2 years, respectively. News event. At 1 year, total recall, F(2, 58) = 12.88, p < .01, Z2 = .31, and elaborated recall, F(2, 58) = 10.17, p < .01, Z2 = .26, for attackers/ weapons were significant. As shown in Table 4, Group 5 provided more detailed information than Group 3/4 (p = .05) and both groups provided

142

SHAPIRO TABLE 4 Recall: Experiment 2 Total recall

Elaborated recall

Assessment

Main events

Attackers

Aftermath

Main events

Attackers

Aftermath

Initial Group 1/2 Group 3 Group 4

2.52 (.17) 1.70 (.21) 1.77 (.204)

1.24 (.11) .40 (.14) .50 (.13)

.27 (.06) .00 (.07) .05 (.07)

5.85 (.36) 3.60 (.46) 4.00 (.44)

3.49 (.23) 1.75 (.30) 2.09 (.28)

1.55 (.15) .15 (.19) .55 (.19)

1 year Group 1/2 Group 3/4 Group 5

2.53 (.19) 2.13 (.22) 2.59 (.16)

1.05 (.14) 1.53 (.16) 2.00 (.12)

.74 (.15) .80 (.17) 1.19 (.13)

6.37 (.40) 5.27 (.45) 6.07 (.33)

3.32 (.26) 4.13 (.29) 4.85 (.22)

2.53 (.35) 2.80 (.39) 2.89 (.29)

2 years Group 1/2 Group 5 Group 6

2.73 (.19) 3.00 (.21) 2.49 (.11)

1.67 (.18) 2.33 (.21) 1.77 (.12)

.80 (.18) 1.08 (.20) .57 (.12)

6.33 (.39) 6.58 (.45) 5.60 (.25)

3.93 (.32) 5.17 (.38) 4.34 (.21)

2.33 (.41) 2.92 (.46) 1.60 (.27)

Mean total recall and elaborated recall (and Standard Errors) for news event categories at initial, 1-year, and 2-year assessments by condition in Experiment 2.

more detailed information than Group 1/2 (p < .001 and p < .05, respectively). Total recall for aftermath at 1 year showed borderline significance, F(2, 58) = 2.99, p = .058, Z2 = .09, suggesting a tendency for Group 5 to provide more information than Group 1/2 (p < .03). At 2 years, total recall of attackers/weapons was lower for Groups 1/2 and 6 than for Group 5, F(2, 62) = 3.23, p < .05, Z2 = .10, and elaborated recall of attackers/weapons was lower for Group 1/2 than for Group 5, F(2, 62) = 3.06, p = .05, Z2 = .09. Elaborated recall of aftermath, F(2, 62) = 3.39, p < .05, Z2 = .10, was lower in Group 6 than in Group 5. No other condition effects were significant. Consistency of recall over time In the first independent-samples t-test, Group 1/2 was compared to Group 3/4 using initial to 1year WAS consistency. The next independentsamples t-tests compared Group 1/2 to Group 5 on initial to 2-year WAS consistency. Finally, two independent-samples t-tests compared Group 1/2 to Group 5 on 1-year to 2-year WAS consistency and main event consistency. Reception event. There were no significant differences in initial to 1-year WAS consistency (M = 5.61, SD = .15; p > .44), initial to 2-year WAS consistency (M = 5.17, SE = .31; p > .14), or 1-year to 2-year WAS consistency (M = 5.10, SE = .31; p > .25). Unlike Experiment 1, however, consistency

across assessments for the combined categories varied greatly from 33% to 100% for Group 5, but from 66% to 100% for Groups 1/2 and 3/4. News event. There were no significant differences found in consistency for 1-year to 2-year main events (p > .75). Consistency for the main events ranged from 33% to 100% for Group 1/2 (M = 4.69, SE = .38) and Group 5 (M = 4.85, SE = .32). Direct comparison of recall for reception event and main events in 9/11 Total recall was higher for the reception event than for the main events of 9/11 at 1 year, t(51) = 74.50, p < .001 (M = 2.98, SE = .02 vs M = 2.42, SE = .12), and at 2 years, t(63) = ±3.57, p < .01 (M = 2.98, SE = .02 vs M = 2.66, SE = .09). However, no differences (p > .62) in consistency from 1 year to 2 years were found for the reception (M = 4.96, SE = .30) and main events of 9/11 (M = 4.77, SE = .24). Other factors impacting on recall Metacognitive self-assessment of memory. For the 1-year and 2-year assessments of WAS and of event variables, separate MANOVAs were conducted using each of the self-reported classifications (vivid vs vague, long period vs seconds, frozen vs fading) as the grouping variable. A higher number of participants believed that their memory of 9/11 was vivid with lots of details (N =

REMEMBERING SEPTEMBER 11th

37 at 1 year, N = 41 at 2 years) rather than vague with little detail (n = 23 at 1 year, n = 21 at 2 years). No significant effects were found for WAS at either 1 year or 2 years. However, significance was found for total recall of attackers/weapons at 1 year, F(1, 58) = 7.84, p < .01, Z2 = .12, and of main events at 2 years, F(1, 60) = 8.43, p < .01, Z2 = .12, and for elaborated recall of main events at 2 years, F(1, 60) = 7.93, p < .01, Z2 = .12. Self-classified ``vivid'' recallers included more information about attackers/weapons (M = 1.78, SE = .12 vs M = 1.26, SE = .15) and main events (M = 2.85, SE = .10 vs M = 2.38, SE = .13) and provided more elaboration about main events (M = 6.44, SE = .21 vs M = 5.43, SE = .29) than did ``vague'' recallers. There were more participants who indicated that recall of 9/11 involved a longer period of time encompassing the news of the attacks and related information (n = 41 at 1 year, n = 45 at 2 years) than only the seconds surrounding learning of the attacks (n = 19 at 1 year, n = 18 at 2 years). Although no significant differences were found at 1 year for WAS or event variables, participants who believed recall involved a long period rather than a few seconds provided higher total recall (M = 2.78, SE = .10 vs M = 2.33, SE = .17), F(1, 61) = 5.20, p < .03, Z2 = .08, and elaborated recall (M = 6.29, SE = .23 vs M = 5.28, SE = .36), F(1, 61) = 5.54, p < .03, Z2 = .08, for main events at 2 years. Participants either described their memory of 9/11 as ``frozen in time'' (n = 34 at 1 year, n = 36 at 2 years) and resistant to forgetting, or as ``similar to other memories which fade over time'' (n = 26 at 1 year and 2 years). This grouping, like the other two, failed to differentiate WAS at 1 year and 2 years. However, significant effects were found for attackers/weapons in total recall at 2 years, F(1, 60) = 4.70, p < .04, and in elaborated recall at 1 year, F(1, 58) = 4.33, p < .05, and at 2 years, albeit marginal, F(1, 60) = 3.62, p = .06. In comparison to people who described their memory as ``fading'', those who described their memory as ``frozen'' had higher recall at 2 years (M = 1.69, SE = .13 vs M = 2.06, SE = .11) and higher elaborated recall at 1 year (M = 3.81, SE = .25 vs M = 4.50, SE = .22) and 2 years (M = 4.23, SE = .22 vs M = 4.78, SE = .19). Rehearsal through discussion. Self-reported rehearsal scores for discussion of personal circumstances during the week of 9/11 (M = 3.62), the first year after 9/11(M = 5.23), and the second year after 9/11 (M = 6.06) were not correlated with 2-year WAS scores (ps > .38). Similarly, self-

143

reported rehearsal scores for discussion of the 9/11 event during the week of 9/11 and the first year after 9/11 were not correlated with 2-year event variables. However, discussion the second year after 9/11 was related to 2-year total recall of attackers/weapons, r(64) = 7.25, p < .05: participants who had many discussions had higher recall. Confidence in recall. There were no significant correlations between recall confidence and actual recall at 2 years of location, activity, and source (ps > .11). In contrast, recall confidence was negatively correlated with total recall and elaborated recall of main events, rs(66) = 7.46, ps < .001, of attacker/weapons, rs(66) > 7.35, p < .01, and aftermath at 2 years, rs(66) > 7.43, p < .001. Thus, high confidence was associated with a high level of detailed recall of the news event. Media effects. The number of media sources that participants used to learn about 9/11 (0±3 sources, n = 19; 4 sources, n = 23; 5±7 sources, n = 19) and their exposure to a 6-month televised special on 9/11 (n = 33 for yes; n = 28 for no) were used as grouping variables in separate MANOVAs. Surprisingly, neither grouping variable affected the amount of total or elaborated recall reported at 1 year for main events, attackers/ weapon, and aftermath (ps > .24).

Discussion Everyone in Experiment 2 had a flashbulb memory for September 11th as demonstrated by WAS elaboration and consistency scores that exceeded the general level of reporting. Moreover, the absence of retention interval and rehearsal effects on (total and) elaborated recall of the reception event confirmed Hypotheses 1a and 1b. One caveat was that having an immediate rather than a delayed initial assessment aided detailed retention of the reception event at 1 year. There also was no impact of differences in relevance on formation and maintenance of flashbulb memory. The findings partially supported Hypothesis 2a, but not Hypothesis 2b. Respondents provided more complete and detailed information about all news event categories in their initial reports at shorter rather than longer retention intervals. Recall for main events and attackers/weapons improved over time from the 1-year to 2-year assessments, but consistency was unaffected by

144

SHAPIRO

rehearsal. Moreover, reports at 1 year and 2 years were more complete, detailed, and accurate about attackers/weapons and aftermath when produced by the control group who had no prior assessment at 1 year and one prior assessment at 2 years than by groups who had one or three assessments at 1 year or groups who had no prior or four assessments at 2 years. These findings can best be explained in two ways. First, rehearsal through repeated testing can prevent forgetting of both accurate and inaccurate information. Respondents who had earlier assessments may have consolidated their knowledge about attackers/ weapons and aftermath through rehearsal, but failed to update it when new information became available between 24 weeks and 1 year. Second, media coverage of attackers/weapons and aftermath at 6 months and 1 year may have produced an advantage for the 1-year control group over the 2-year control group. Consistent with past research, a direct comparison of elaborated recall at 1 year and at 2 years revealed that memory for the reception and news events differed. However, unlike newly published studies on 9/11 which found that flashbulb memory was more consistent than factual memory during the first year after the tragedy (i.e., from 3 days to 1 year in Tekcan et al., 2003; from 1 week to 6 months in Smith et al., 2003), Experiment 2 showed no differences in consistency from 1 to 2 years. Of course, the recent work differed from Experiment 2 not only in employment of earlier retention intervals, but also in their use of non-American participants and assessment of specific facts rather than information about the main events. The exploratory analyses yielded some interesting findings. Respondents who had positive metacognitive self-assessments (i.e., vivid recall, extended in time, and resistant to fading) had corresponding high levels of very long retention for main events and attackers/weapons of September 11th. Some support was found for the notion that overt rehearsal may serve to aid elaboration (Finkenauer et al., 1998) as discussion of September 11th during the second year was associated with high levels of recall for attackers/ weapons. This element of the news event may have been discussed more than other elements because President Bush justified the bombing of Afghanistan and war on Iraq as necessary in his campaign against terrorism. Military intelligence indicated that the Taliban government in Afghanistan was protecting the al-Qaida terrorists and

their leader Bin Laden, and that Iraqi leader Saddam Hussein was providing aid to al-Qaida and had ``weapons of mass destruction'' which may be used to attack America. Although there was no relationship between confidence and elaborate recall of the reception event, respondents who were highly confident in their recall of the news event were associated with higher levels of elaborate memory. Migueles and Garcia-Bajos (1999) reported higher confidence for correct than for incorrect information about a simulated kidnapping. Nachson and Zelig (2003) reported that 69% of their participants were confident in the accuracy of their factual memory of Rabin's assassination, however 65% were also confident of the accuracy of their flashbulb memory. Finally, 1-year retention of the reception or news events was not differentially affected by either the number of media sources accessed or exposure to televised specials about 9/11.

GENERAL DISCUSSION September 11th is the classic example of a national tragedy that has the potential to become a flashbulb memory. The main events and aftermath of the 9/11 terrorist attacks are among the most unforgettable and emotionally negative incidents in history (Pillemer, 2003). Everyone in the world who had access to television watched the videotape of the burning North Tower as the second hijacked plane flew into the South Tower; it was replayed over and over again in the days following the attacks between glimpses of both WTC towers and the Pentagon burning, rescue and recovery efforts, and the collapse of the North and South Towers and Tower #7. The government closed airports, the stock market, and all national landmarks for several days and the fear of future terrorism led to political (e.g., bombing Afghanistan, Iraq war), economic (e.g., low stock prices), and social repercussions (e.g., hate and anger directed at Muslims) in the USA that affected the rest of the world. Consequently, September 11th was the ideal to-be-remembered emotional stimulus for examining retention interval effects, rehearsal effects, and recall patterns for the reception and news events over a 2-year period. Recommended guidelines for determining the presence of flashbulb memory were provided in the introduction as a way of reducing the gap between supporters and refuters. Specifically, data analyses of the reception context should

REMEMBERING SEPTEMBER 11th

focus only on major canonical categories of location, source, and activity, and consistency in these categories should be at the general level. In the current research, the durability of flashbulb memory was clearly demonstrated by the participants who remembered detailed information about their location, activity, and source for their reception context of September 11th, even after 2 years. Respondents provided essentially or exactly the same detailed information for the three major attributes at each retention interval between the event and assessment, despite differences in the amount of overt rehearsal (i.e., surveys) they had had between assessments. Specifically, general recall of all three attributes was consistently provided by 100% of the respondents at all assessments from 1 day to 11 weeks and from 1 day, 11 weeks, and 23 weeks to 2 years, but by 92% from 1 year to 2 years. Detailed recall for two or three attributes was provided consistently by 100% of the respondents from 1 day to 2 weeks or 6 weeks, from 2 weeks to 11 weeks, and from 1 day to 2 years; by 95% from 1 day to 11 weeks and from 6 weeks to 11 weeks; by 87% from 11 weeks or 23 weeks to 1 year; and by 61% from 1 year to 2 years. The findings matched or exceeded those of previous studies in which flashbulb memory was examined over the course of 1 year (e.g., Christianson, 1992; Tekcan et al., 2003) or longer (e.g., Bohannon & Symons, 1992; Schmolck et al., 2000). Whether or not flashbulb memory should form for surprising/traumatic public events is dependent on personal consequentiality ascribed to the event by the rememberer (Brown & Kulik, 1977; Conway, 1995). The personal relevance scale in this investigation, which used relational anchors (e.g., ``general'' or national loss of lives, ``specific'' or personal loss of friends and family), failed to capture the importance of this event to the participants (initial M = 1.88 out of 5). Why have other researchers (e.g., Conway et al., 1994) been able to differentiate people who do and do not have flashbulb memory using a personal consequentiality scale? One reason may be that their scales used emotional anchors in which people determined if the event being studied was ``the most upsetting event ever experienced''. However, this type of anchor requires people to compare past events whose emotionality has waned with current, emotionally salient events. It is also possible that the present group of respondents were more homogeneous than groups used by past researchers. For example, Conway et al.

145

compared UK and non-UK groups on a British national event. Respondents may also have had difficulty determining the direct and indirect consequences that this recent event would have on their lives (college students are generally not very introspective). According to the narratives collected by respondents, initial reactions to 9/11 included shock, confusion, fear for their lives and the lives of loved ones, expectations that bombs would be dropped on them, and thoughts that they/loved ones would be drafted for World War III. One and two years later, some respondents mentioned that they felt vulnerable, anticipated additional terrorism on US soil, were afraid to fly/ travel, believed they would not be able to find a job after college, etc. Future investigators may want to use a variety of measures, such as how people internalise problems, to judge the impact of a traumatic event on the formation of flashbulb memory. The present research also addressed the issue of whether recall for reception and news events had the same durability and consistency. Unlike flashbulb memory, the three components of factual memory of 9/11 were differentially affected by retention interval and rehearsal. The main events of the tragedy were recalled well over the 2-year period, with rehearsal aiding retention at 11 weeks, but not at 1 and 2 years. Recall of the reception context was higher at these retention intervals than recall of main events and was more consistent from 1 day to 11 weeks, but not from 1 to 2 years. It was interesting that the effects of rehearsal on retention interval were strongest for recall of attackers/weapons, because this component consisted of specific facts similar (sometimes identical) to those obtained in previous studies in their assessment of the news event (e.g., Smith et al., 2003; Tekcan et al., 2003). Detailed recall of attackers/weapons at 11 weeks was better when assessed initially at 1 day than at 11 or 23 weeks, and interim rehearsal at 2 weeks was better than 6 weeks for 11-week retention. In contrast, rehearsal detrimentally affected detailed recall at 1 year and 2 years. Rehearsal effects for aftermath were puzzling, as retention at 11 weeks and 2 years improved, whereas retention at 1 year declined. Given these results, future researchers are advised to assess both general and specific facts in the news event and to include control groups to determine the role of repeated testing.

146

SHAPIRO

Conclusion Despite small sample size, the findings were robust. Changes in the assessment of news event precluded an examination of forgetting from 23 weeks to 1 year. However, the major strengths of this investigation were that the design allowed for both cross-sectional and longitudinal comparisons for recall of both the reception and news events, and the inclusion of control groups helped to determine the role of rehearsal at three retention intervalsÐ11 weeks, 1 year, and 2 years. Studying recall of September 11th over 2 years aided in scientific understanding of how rehearsal may affect long-term retention for traumatic experiences, such as violent crimes. Unlike flashbulb memory, factual memory differed due to the respondents' conceptualisation of memory, and was related to confidence in reported information and self-reported rehearsal. The relationship between emotion and memory is complex (Christianson, 1992) and lab studies cannot construct situations that cause the same degree of threat to eyewitnesses as do real crimes (Yuille & Cutshall, 1986). Although it is not possible to know everything about this tragedy, experimental control was maintained because the facts of this disaster were relayed by a variety of media sources. Thus, like simulated events, rememberers in this investigation were provided with a specific view of the event, so that information they reported could be checked against facts represented in the media. Manuscript received 23 January 2004 Manuscript accepted 8 February 2005 PrEview proof published online 9 August 2005

REFERENCES Brewer, W. F. (1988). Memory for randomly sampled autobiographical events. In U. Neisser & E. Winograd (Eds.), Remembering reconsidered: Ecological and traditional approaches to the study of memory (pp. 21±90). New York: Cambridge University Press. Bohannon, J. N. III (1988). Flashbulb memories for the space shuttle disaster: A tale of two theories. Cognition, 29, 179±196. Bohannon, J. N. III, & Symons, V. N. (1992). Flashbulb memories: Confidence, consistency, and quantity. In E. Winograd & U. Neisser (Eds.), Affect and accuracy in recall: Studies of flashbulb memories (pp. 65±91). New York: Cambridge University Press. 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. (1992). Emotional stress and eyewitness memory: A critical review. Psychological Bulletin, 112, 284±309. Christianson, S-A., & Engelberg, E. (1999). Memory and emotional consistency: The MS Estonia ferry disaster. Memory, 7, 471±482. Collins, R. L., Taylor, S. E., Wood, J. V., & Thompson, S. C. (1988). The vividness effect: Elusive or illusory? Journal of Experimental Social Psychology, 24, 1±18. Cohen, G., Conway, M. A., & Maylor, E. A. (1994). Flashbulb memories in older adults. Psychology and Aging, 9, 454±463. Conway, M. A. (1995). Flashbulb memories. Hillsdale, NJ: Lawrence Erlbaum Associates Inc. Conway, M. A., Anderson, S. J., Larsen, S. F., Donnelly, C. M., McDaniel, M. A., McClelland, A. et al. (1994). The formation of flashbulb memories. Memory & Cognition, 22, 326±343. Curci, A., Luminet, O., Finkenauer, C., & Gisle, L. (2001). Flashbulb memories in social groups: A comparative test±retest study of the memory of French President Mitterrand's death in a French and a Belgian group. Memory, 9, 81±101. Er, N. (2003). A new flashbulb memory model applied to the Marmara earthquake. Applied Cognitive Psychology, 17, 503±517. 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. Gunter B. (1980). Remembering television news: Effects of picture content. The Journal of Genetic Psychology, 102, 127±133. Hornstein, S. L., Brown, A. S., & Mulligan, N. W. (2003). Long-term flashbulb memory for learning of Princess Diana's death. Memory, 11, 293±306. Kvavilashvili, L., Mirani, J., Schlagman, S., & Kornbrot, D. E. (2003). Comparing flashbulb memories of September 11 and the death of Princess Diana: Effects of time delays and nationality. Applied Cognitive Psychology, 17, 1017±1031. Luminet, O., Bouts, P., Delie, F., Manstead, A. S. R., & Rime, B. (2000). Social sharing of emotion following exposure to a negatively valenced situation. Cognition and Emotion, 14, 661±688. Martin, L. L., & Tesser, A. (1989). Toward a motivational and structural theory of ruminative thought. In J. S. Ulman & J. A. Bargh (Eds.), Unintended thoughts (pp. 306±326). New York: Guilford Press. Martin, L. L., & Tesser, A. (1996). Some ruminative thoughts. In R. S. Wyer Jr. (Ed.), Advances in social cognition: Vol. 9. Ruminative thoughts (pp. 1±48). Mahwah, NJ: Lawrence Erlbaum Associates Inc. McCloskey, M. (1992). Special versus ordinary memory mechanisms in the genesis of flashbulb memories. In E. Winograd & U. Neisser (Eds.), Affect and accuracy in recall: Studies of flashbulb memories (pp. 227±235). New York: Cambridge University Press. McCloskey, M., Wible, C. G., & Cohen, N. J. (1988). Is there a special flashbulb-memory mechanism? Journal of Experimental Psychology: General, 117, 171±181.

REMEMBERING SEPTEMBER 11th

Migueles, M., & Garcia-Bajos, E. (1999). Recall, recognition, and confidence patterns in eyewitness testimony. Applied Cognitive Psychology, 13, 257±268. Nachson, I., & Zelig, A. (2003). Flashbulb and factual memories: The case of Rabin's assassination. Applied Cognitive Psychology, 17, 519±532. Neisser, U. (2003). New directions for flashbulb memories: Comments on the ACP Special Issue. Applied Cognitive Psychology, 17, 1149±1155. Neisser, U., & Harsch, N. (1992). Phantom flashbulbs: False recollections of hearing the news about Challenger. In E. Winograd & U. Neisser (Eds.), Affect and accuracy in recall: Studies of flashbulb memories (pp. 9±31). New York: Cambridge University Press. Pillemer, D. B. (1984). Flashbulb memories of the assassination attempt on President Reagan. Cognition, 16, 63±80. Pillemer, D. B. (1990). Clarifying the flashbulb memory concept: Comment on McCloskey, Wible, and Cohen (1988). Journal of Experimental Psychology: General, 119, 92±96. Pillemer, D. B. (1998). Momentous events, vivid memories. Cambridge, MA: Harvard University Press. Pillemer, D. B. (2003). Directive functions of autobiographical memory: The guiding power of the specific episode. Memory, 11, 193±202. Robinson, J. P., & Levy, M. R. (1986). The main source: Learning from television news. Beverly Hills, CA: Sage. Rubin, D. C., & Kozin, M. (1984). Vivid memories. Cognition, 16, 81±95. Schmolck, H., Buffalo, E. A., & Squire, L. R. (2000). Memory distortions develop over time: Recollections of the O. J. Simpson trial verdict after 15 and 32 months. Psychological Science, 11, 39±45.

147

Smith, M. C., Bibi, U., & Sheard, D. E. (2003). Evidence for the differential impact of time and emotion on personal and event memories for September 11, 2001. Applied Cognitive Psychology, 17, 1047±1055. Son, J., Reese, S. D., & Davie, W. R. (1987). Effect of visual-verbal redundancy and recaps on television news learning. Journal of Broadcasting & Electronic Media, 31, 207±216. Tekcan, A. I., Ece, B., Gulgoz, S., & Er, N. (2003). Autobiographical and event memory for 9/11: Changes across one year. Applied Cognitive Psychology, 17, 1057±1066. Tekcan, A. I., & Peynircioglu, Z. F. (2002). Effects of age on flashbulb memories. Psychology and Aging, 17, 416±422. Thomas, D.L., & Diener, E. (1990). Memory accuracy in the recall of emotions. Journal of Personality and Social Psychology, 59, 291±297. Winningham, R. G., Hyman, I. E., & Dinnel, D. L. (2000). Flashbulb memories? The effects of when the initial memory report was obtained. Memory, 8, 209±216. Wright, D. B. (1993). Recall of the Hillsborough disaster over time: Systematic biases of `flashbulb' memories. Applied Cognitive Psychology, 7, 129±139. Wright, D. B., Gaskell, G. D., & O'Muircheartaigh, C. A. (1998). Flashbulb memory assumptions: Using national surveys to explore cognitive phenomena. British Journal of Psychology, 89, 103±121. Yuille, J. C., & Cutshall, J. L. (1986). A case study of eyewitness memory of a crime. Journal of Applied Psychology, 71, 291±301.