A memory like an elephant? The consistency of memory for
emotional events.
Giezen, A.E. van
Citation
Giezen, A. E. van. (2007, November 8). A memory like an elephant? The consistency of memory for emotional events. Retrieved from
https://hdl.handle.net/1887/12420
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/12420
Note: To cite this publication please use the final published version (if applicable).
5
Memory for Emotionally Arousing and Neutral Stimuli:
An Experimental Study on Accuracy and Consistency
of Memory in Victims of Violent Crime
This study examined accuracy and consistency of memory for emotional and neutral events in two experiments. Victims of assault, students who had experienced a traumatic event and general population controls viewed an emotional or neutral slide show and were tested for free recall and recognition memory 2 and 4 weeks later. Results showed that memory for both emotional and neutral stimuli was fairly consistent and became more accurate over time. Central aspects were more accurately and consistently remembered than peripheral aspects. An important finding was that accuracy and consistency were highly correlated. Consistently recalled information was significantly more often accurately recalled, with the central aspects being better recalled than the peripheral aspects.
Van Giezen, A.E., Arensman, E., & Spinhoven, Ph.
Manuscript submitted for publication
Introduction
In the recent debate about the nature of memory for traumatic events, some authors argue that memory for emotionally arousing events is indelible and accurate over the lifetime, while others state that it is malleable and subject to substantial distortion. Although the nature of memories for arousing events and possible underlying processes has been subject of investigation in many research disciplines (e.g., Brewin & Andrews, 1998; Brown, Scheflin, & Hammond, 1998;
McNally, 1998, 2003a) differences in opinion about the reliability of memories for trauma remain.
The notion of some authors that memory for emotional events is malleable and subject to distortion is mainly based on experimental research on the inaccuracies of normal memory and on the misinformation effect (Loftus, 1993;
Loftus & Ketcham, 1994). Deficits in recalling traumatic events are attributed to the same processes of decay and interference as involved in deficient memories for ordinary events. Moreover, proponents of this view state that by giving false post- event information, memory distortions can be induced and even that entire scenes of stressful events can be fabricated and inserted into autobiographical memory (Hyman & Pentland, 1996).
On the other hand, different sources of data support the notion that memories for emotionally arousing events are stable and consistent over the lifetime. In a high percentage of individuals, unexpected stress arousing events that were experienced by or known to a large group of people, such as assassinations or assassination attempts of public persons produce so-called
“flashbulb memories” that resemble a “photographic print” (Brown & Kulik, 1977).
Memory for traumatic experiences, such as earthquakes, sniper attacks, or kidnappings, is usually persistent and accurate, although some distortion occurs (Schacter, 1996).
Easterbrook‟s attentional narrowing hypothesis states that high levels of arousal narrow the focus of attention, which allows an individual to focus on the more salient cues (Easterbrook, 1959). Therefore, emotional arousal should lead to enhanced memory for central information, but diminished retention of peripheral detail. Laboratory research into emotion and memory indicates that memory for
emotionally arousing events is generally accurate, with distortion occurring at the level of specific details (Christianson, 1992), and that emotional arousal leads to enhanced long term memory (Cahill & McGaugh, 1995). On consecutive recall tests a net gain in memory for details of the traumatic scene (i.e. hypermnesia) occurs (Christianson, 1984; Erdelyi & Becker, 1974; Scrivner & Safer, 1988).
Although memory for emotional events is widely debated, few studies have focused on consistency of memory. Among these studies, different definitions are used to indicate memory consistency. Although not the same, consistency of memory is often used as an indicator of memory accuracy (McNally, 2003a). In legal practice, inconsistency is often seen as strongly predictive of inaccuracy (Brewer, Potter, Fisher, Bond, & Luszcz, 1999; Fisher & Cutler, 1995). However, they refer to different concepts. A report can be consistent, without necessarily being accurate. Consistency of memory refers to the same information being reported at different time points, while accuracy can be described as the agreement between recall of the traumatic event and an objective report of what has occurred. In order to assess accuracy, memory reports need verification, which is often not feasible.
In the present series of two studies, accuracy and consistency of memory were investigated in an experimental setting. An advantage of laboratory research is that it allows verification of the accuracy of memory. Therefore, is it possible to determine whether inconsistency in memory reflects diminished non-recall of events that did occur or the addition of pseudo-memories of events that did not occur. Altogether, we tested the hypothesis that a) memory for central information is more accurate and consistent than memory for peripheral information; b) memory for emotional stimuli is more accurate and consistent than memory for neutral stimuli; c) repeated testing will lead to amplification of memory. Furthermore we explored the relationship between accuracy and consistency of memory. In study 1, we examined accuracy and consistency of memory in victims of violent assault and matched control subjects. The purpose of the second experiment was to extend the generalizability of our findings from our first experiment by including a student population who had experienced a traumatic event in the last year.
Study 1
Materials and Methods Subjects.
Twenty-two victims of violent assault were tested. The victims were part of a prospective field study on consistency of memory for violent assault (Van Giezen, Arensman, & Spinhoven, submitted). Subjects were recruited through Victim Support Services and the emergency department of a general hospital. The study was also advertised through information on the internet, posters in public places, advertisements in local newspapers and interviews broadcasted on local and national television. Assault victims (N=76) were invited to participate in the experimental study after the second assessment of the prospective study (3-5 months after the assault). The study was approved by the Medical Ethical Committee of Medical Centre The Hague. T-tests and chi-square analyses did not show significant differences between the 22 victims who chose to participate in the experiment and those who chose not to participate, with respect to demographic and psychopathology variables, suggesting that no selection bias took place. The 22 victims of assault were compared with 22 healthy controls. The groups were matched by gender and age (+ 5 years). The control group was recruited by advertisements in local newspapers. Both victims and normal control subjects received 20 Euros for participating in the study. A total of 44 subjects (20 male, 24 female), with a mean age of 40 years (SD= 12.8, range 18-63) participated in the study.
Procedure and materials.
The experiment took place at the Leiden University laboratory. After complete description of the study, subjects gave their written informed consent. They were told that they were participating in an experiment on attention and concentration.
The subjects were then presented with a series of 11 identical slides in a neutral or emotionally arousing version (T1). Apart from minor adjustments, the stimulus materials were the same as those used by Cahill and McCaugh (1995) and Van Stegeren, Everaerd, Cahill, McGaugh, and Gooren (1998). The verbal commentary accompanying the slides was identical for the first and third phase of the slides, but the neutral or emotionally arousing content of the commentary differed for the
second phase of the slides. In the arousal version of the story, the subjects heard and saw a story about a boy and his mother going to visit his father at work. The father works at the laboratory of a hospital. On their way over there, the son is hit by a car and is transported to the hospital. At the hospital, surgeons operate the boy‟s disfigured legs in order to safe his life. Afterwards, the mother leaves the hospital to call the school of her other child. In the neutral version, the mother and son again visit the father at work, but this time on the way over there they witness a small car-accident. They go over to the hospital where the father works and together they watch a training procedure. The mother leaves to call the school of her other child. Up front, the participants were told that they would find some of the slides pleasant, some unpleasant and some neutral. They were instructed to pay attention to the slides and watch them like they would watch a movie.
After watching the slides the subjects were asked to rate on a 10-point Visual Analogue Scale (VAS) the emotional impact of the story. Subsequently, subjects were tested for detail memory in two consecutive recall tests two weeks after the experiment (T2) and again two weeks later (T3). The subjects had no expectation of a memory test. When they returned to the laboratory, free recall and recognition memory for central and peripheral elements of the presented slides and accompanying commentary were tested (Heuer & Reisberg, 1990; Van Stegeren et al., 1998).
For the free recall test, subjects were asked to give as much information as possible of the slides and accompanying narrative. The recognition test consisted of 76 four-choice multiple-choice questions. There were 5 to 9 questions for each slide and accompanying narrative. All questions and possible answers were read out to the subjects. Subjects had to choose one of the alternatives, even if they had to guess. VAS scores of the emotionality of the story were also assessed both two and four weeks after the initial experiment.
Scoring.
With regard to the recognition memory test, accuracy of memory was operationalized as the percentage of correct answers. Consistency was operationalized as the percentage of the same answers given on the recognition
memory test at T2 and T3. In accordance with Cahill et al. (1994) (see also Van Stegeren et al., 1998; Cahill & Van Stegeren, 2003), 18 questions of the multiple- choice recognition memory test were labelled as central to the story line, all others referred to peripheral information. For both central and peripheral questions, the percentage of correct and consistent answers was computed.
In order to score the results of the free recall test a list with all possible elements was constructed by two independent judges. The list was first piloted and subsequently modified. For the arousal version, the list consisted of 146 items, for the neutral version this list contained 149 elements. In order to assess accuracy, three judges scored the free recall by counting all reported correct elements.
Elements that were incorrect were also counted. Furthermore, an accuracy index was calculated by dividing the number of correctly recalled items by the sum of the number of correctly recalled items and the number of incorrectly recalled items (Brewer et al, 1999; Smeets, Candel, & Merckelbach, 2004). Consistency of free recall was calculated by counting the elements that were reported at both T2 and T3. Omissions were operationalized as the number of elements reported at T2 but not at T3, while commissions referred to the number of elements only reported at T3. Also, an index of consistency was calculated by dividing the number of consistently recalled items by the sum of the number of consistently recalled items and the number omissions and the number of commissions (Brewer et al, 1999;
Smeets et al, 2004). For the free recall test, no distinction in central and peripheral information was made.
Finally, for all answers on the recognition test, scores on consistency and accuracy were combined. Answers could be either - consistent and accurate, consistent and inaccurate, inconsistent and accurate or inconsistent and inaccurate -. Proportions of all four combinations were calculated.
Statistical analyses.
Memory accuracy and consistency measurements were analysed by means of (repeated measures) ANOVA using general linear models. Within subjects factors were type of information (central vs. peripheral) and time of assessment (T2 vs.
T3). Between subjects factors were group (victims of assault vs. matched controls)
and condition (arousal vs. neutral). Paired sample t-tests were performed to analyse proportions of combined consistency and accuracy scores. An α level of .05 was used for all statistical tests.
Results
Manipulation check for emotional impact
With regard to the emotionality of the slides, the arousal and neutral condition differed significantly in their mean rating of emotionality on all 3 assessments.
Subjects in the arousal condition rated the emotional impact of the slides significantly higher than the subjects in the neutral condition. The mean VAS-score for the arousal condition was 4.7 (SD=2.6) versus 1.8 (SD=2.0) for the neutral condition at T1 [t(42)=-4.06, p<.001]. At T2 VAS scores were 4.0 vs. 1.4 [t(42)=- 3.48, p<.002] and at T3 VAS scores were 4.1 vs. 1.7 [t(42)=-2.91, p<.007]. Thus, the stories produced different levels of emotional arousal in the two conditions.
Although assault victims rated the emotional impact of the slides somewhat higher than the control group (T1: 3.6 vs. 2.9; T2: 3.1 vs. 2.4; T3: 3.2 vs. 2.6), no significant differences for group were found.
Accuracy.
A 2 (group) x 2 (condition) x 2 (type of information) x 2 (time) repeated measures ANOVA on recognition memory accuracy revealed a significant main effect of time [F(1,40)=19.72, p<.001] and type of information [F(1,40)=252.41, p<.001].
Subjects in all conditions were significantly more accurate at T3 (52.6%) compared to T2 (48.7%) and recognized central information (71.4%) more accurately than peripheral information (44.3%). No main effects of group or condition were found, nor any interaction effects involving group, condition, type of information and time (see also Table 1).
With regard to accuracy of the free recall, a 2 (group) x 2 (condition) x 2 (time) repeated measures ANOVA on memory accuracy also revealed a significant main effect of time for both accuracy measures (Number of correctly reported items [F(1,40)=38.04, p<.001]; accuracy index: [F(1,40)=8.80, p<.01]). No main effects of group or condition were found, nor any interaction effects involving group, condition or time. Results show that overall, at T2 8.9 (SD=5.6) story elements were recalled
compared to 12.5 (SD=5.1) elements at T3. Subjects also recalled elements which were not part of the story, and therefore not accurate. At T2, 1.45 (1.2) elements were confabulated, at T3 1.7 (1.7) incorrect elements were reported (see also table 2).
Consistency.
Overall, 67.1% of the questions of the recognition memory test were answered consistently. A 2 (group) x 2 (condition) x 2 (type of information) repeated measures ANOVA on consistency of memory revealed a significant main effect of type of information [F(1,40)=31.12, p<.001].
With regard to consistency of free recall, 7.4 (SD=4.5) story elements were consistently recalled. A 2 (group) x 2 (condition) ANOVA on consistency of free recall revealed no significant main or interaction effects for the number of consistently reported items or for the consistency index. Furthermore, results show a mean number of commissions of 5.2 (SD=3.4) and 1.6 (SD=1.8) omissions, which is a significant difference [t(43)=6.34, p<.001]. Victims of assault showed significantly more commissions (M=2.3, SD=1.9) than normal controls (M=0.9, SD=
1.3) [t(42)=2.67, p<.02].
Relationship between consistency and accuracy.
Consistency of recognition memory and T2 (r=.61, p<.001) and T3 accuracy (r=
.61, p<.001) were significantly correlated. Consistency of central aspects was also significantly correlated with accuracy of central information at T2 (r=.59, p<.01) and T3 (r= .47, p<.01), consistency of peripheral aspects was significantly correlated with accuracy of peripheral information at T2(r=.43, p<.01) and T3 (r= .48, p<.01).
Results from the free recall test show a significant correlation between T2 (r=.96, p<.001) and T3 number of correctly reported items and number of consistently reported items (r=.75, p<.001). When looking at the computed accuracy and consistency index, results from the free recall test also show a significant correlation between T2 and T3 accuracy and consistency (r=.51, p<.01;
r= .47, p<.05 ). Correlations between accuracy and consistency did not differ between victims of assault and normal controls with regard to recognition memory.
Descriptives of the combined scores on accuracy and consistency of recognition memory show that of the consistently recalled items, almost 80% of the
central items are also accurate at T2 and T3, compared to 51.9% of the peripheral items (see also table 5). Looking at the proportion of inconsistent but accurate responses, over 25% of the central (30.4%) and peripheral (26%) responses are accurately recalled at T2, compared to 54% (central items) and 37% (peripheral items) at T3. Paired sample t-tests comparing the proportion of consistent and accurate responses with consistent but inaccurate responses, showed significant differences (p<.001), except for the responses on the peripheral aspects.
Comparing the proportions of inconsistent but accurate responses with the proportion of inconsistent and inaccurate responses, results of all paired sample t- tests appeared to be significant (p<.001), except for those on T3 central aspects.
STUDY 2
Material and Methods
Subjects and Procedure.
In addition to study 1, undergraduate students who had experienced a traumatic event within the last year were tested. The same procedure and stimulus materials as in study 1 were applied. The study was approved by the Medical Ethical Committee. A total of 28 subjects participated in the study, 5 men and 23 women. Their mean age was 22.8 (SD=6.5). Thirteen participants were randomly assigned to the neutral condition, 15 were assigned to the arousal condition. The undergraduate students were recruited by posters and internet. Students completed the General Health Questionnaire (GHQ-12; Goldberg, 1972), to rule out psychopathology and completed the Impact of Event Scale (IES; Horowitz, Wilner, & Alvarez, 1979) to assess the level of posttraumatic stress symptoms in reaction to the experienced traumatic event. Subjects with a score of 30 and higher on the IES were included. Subjects who received professional treatment at the time of the experiment were excluded in order to prevent interference. Mean scores on the IES were 40.2 (SD=9.9), indicating a rather high level of posttraumatic stress symptoms (Neal et al., 1994). All participants gave informed consent. They received course credits for their participation in this experiment.
Table 1 Accurately and consistently recalled items (in percentages) by assault victims and matched controls on recognition test
Accuracy T2
Accuracy T3
Accuracy Central information T2
Accuracy Peripheral information T2
Accuracy Central information T3
Accuracy Peripheral information T3
Consistency T2-T3
Consistency central information T2-T3
Consistency peripheral information T2-T3
Arousal condition
49.4 53.6 67.2 43.9 73.5 47.4 68.4 75.3 66.0
Neutral condition
48.0 51.7 69.4 41.3 75.5 44.3 65.9 75.0 63.1
Assault victims
48.1 50.9 68.2 41.9 74.0 43.7 66.5 63.9 63.9
Controls 49.3 54.4 68.4 43.3 75.0 48.0 67.8 65.2 65.2
Total 48.7 52.6 68.3 42.6 74.5 45.9 67.1 75.1 64.5
Table 2 Accurately and consistently recalled items by assault victims and matched controls on free recall
Accuracy T2
Accuracy index T2
%
Accuracy T3
Accuracy index T3
%
Consistency T2-T3
Consistency Index T2-T3
%
Omissions T3
Commissions T3
Confabulation T2
Confabulation T3
Arousal condition
9.36 85.30 12.45 91.98 7.50 52.94 1.86 5.00 1.18 1.41
Neutral condition
8.50 77.43 12.55 85.64 7.23 47.39 1.32 5.32 1.73 2.00
Assault victims
9.41 86.46 13.09 90.81 7.18 45.29 2.27 5.95 1.27 1.41
Controls 8.45 76.28 11.91 86.81 7.55 55.04 .91 4.36 1.64 2.00
Total 8.93 81.37 12.50 88.81 7.36 50.17 1.59 5.16 1.45 1.70
92
Statistical analyses.
Memory accuracy and consistency measurements were analysed by means of (repeated measures) ANOVA using general linear models. Within subjects factors were detail (central vs. peripheral) and time of assessment (T2 vs. T3); the between subjects factor was condition (arousal vs. neutral). Paired sample t-tests were performed to analyse proportions of combined consistency and accuracy scores. An α level of.05 was used for all statistical tests.
Results
Manipulation check for emotional impact
With regard to the emotional impact of the slides, the arousal and neutral condition did not differ significantly in their mean rating of the emotional impact of the slides on the 3 assessments, although subjects in the arousal condition rated the emotionality of the slides higher than the subjects in the neutral condition. The mean VAS-score for the arousal condition was 5.2 (SD=2.5) versus 4.0 (SD=1.5) for the neutral condition at T1. At T2 VAS scores were 4.6 vs. 3.5 and at T3 VAS scores were 4.5 vs. 3.5.
Accuracy.
A 2 (condition) x 2 (type of information) x 2 (time) repeated measures ANOVA on memory accuracy on the recognition task revealed a significant main effect of time [F(1, 26)=22.69, p<.001] and type of information [F(1,26)=135.09, p<.001]. No main effect of condition was found, nor any interaction effects involving condition, type of information or time (see also table 3). Results show that overall, 43.5 % of the questions were answered correctly at T2, and 49.6 % at T3. Central information (66.6%) was more accurately recognized than peripheral information (40.3%).
With regard to accuracy on the free recall task, a 2 (condition) x 2 (time) repeated measures ANOVA on memory accuracy revealed a significant main effect of time [F(1, 26)=34.43, p<.001] for the number of correctly reported items and a borderline significant main effect of time for the accuracy index [F(1, 26)=4.01, p=.056]. No main effects of condition were found, nor any interaction effects involving condition or time. Results showed that 9.5 (SD=4.8) story elements were
recalled at T2, while 14.6 (SD=5.3) elements were recalled at T3. At T2, 2.4 (SD=1.2) incorrect story elements were recalled, as opposed to 2.6 (SD= 2.2) at T3 (see also table 4).
Consistency
Overall, 67.3% of the questions on the recognition task were answered consistently. A 2 (condition) x 2 (type of information) repeated measures ANOVA on consistency of memory revealed no significant main or interaction effects. On the free recall test, more than 7 story elements (M=7.7, SD=4.6) were consistently recalled. Almost 7 story elements (M=6.9, SD=3.7) were recalled at T3, but not at T2, while 1.9 (SD=1.6) elements were only recalled at T2. This difference between the number of reported omissions and commissions was statistically significant [t(27)=5.9, p<.001]. No effect of condition on consistency was found for both the consistency index and the number of consistently recalled items.
Relationship between consistency and accuracy
Consistency and T2 and T3 accuracy of recognition memory were not significantly correlated (r=-.15, r=-.03). Consistency of peripheral aspects was also not significantly related to accuracy of peripheral aspects at T2 and T3 (r=-.33, p<.09;
r=-.25, n.s.), while consistency of central aspects was significantly correlated with memory accuracy of central aspects at both T2 and T3 (r=.79, r=.53, p<.001).
Furthermore, results from the free recall test showed a significant correlation between T2 and T3 number of accurately reported items and number of consistent items (r=.95, r=.73, p<.001). Correlations of the free recall accuracy and consistency index were significant for T2 accuracy and consistency (r=.64, p<.001), but not for T3 accuracy and consistency (r=-.04).
Descriptives of the combined scores on accuracy and consistency of recognition memory show that of the consistently recalled items 78.6% of the central items were also accurate at T2 and T3, compared to 46.1 of the peripheral items (see also table 5). With regard to the inconsistently but accurately recalled items, 20.4 % of the central items and 22% of the peripheral items were accurately recalled at T2, compared to 49% and 37.5% at T3.
Table 3 Accurately and consistently recalled items (in percentages) by students on recognition test
Accuracy T2
Accuracy T3
Accuracy Central information T2
Accuracy Peripheral information T2
Accuracy Central information T3
Accuracy Peripheral information T3
Consistency T2-T3
Consistency central information T2-T3
Consistency peripheral information T2-T3 Arousal
condition
46.1 52.6 64.4 40.5 72.2 46.6 67.3 73.0 65.5
Neutral condition
40.4 46.2* 61.1 34.0** 68.0 39.4** 67.2 73.5 65.3
Total 43.5 49.6 62.9 37.4 70.2 43.2 67.3 73.2 65.4
**p < .05, * p< .10
Table 4 Accurately and consistently recalled items by students on free recall
Accuracy T2
Accuracy index T2
%
Accuracy T3
Accuracy index T3
%
Consistency T2-T3
Consistency Index T2-T3
%
Omissions T3
Commissions T3
Confabulation T2
Confabulation T3
Arousal condition
9.93 74.24 15.13 85.49 8.40 49.05 1.53 6.73 2.40 2.53
Neutral condition
9.15 77.72 13.92 84.46 6.85 40.38 2.31 7.08 2.38 2.69
Total 9.57 75.85 14.57 85.01 7.68 45.03 1.89 6.89 2.39 2.61
95
With regard to the comparison of the combined scores for accuracy and consistency, all paired sample t-tests comparing the proportion of consistent and accurate responses with consistent but inaccurate responses showed significant differences, although less profound for the total questions and the peripheral questions (p<.05) than for the central questions (p<.001). When comparing the proportions of inconsistent but accurate responses with the proportion of inconsistent and inaccurate responses, results of all paired sample t-tests were significant (p<.001), except for the responses on T3 central aspects.
Table 5 Relationship between consistency and accuracy
Victims and normal controls Study 1
Central T2 Peripheral T2 Total T2 Central T3 Peripheral T3 Total T3
Mean number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
%
Consistent and accurate
10.7 79.9 19.3 51.9 30.0 59.3 10.7 79.9 19.3 51.9 29.9 59.3
Consistent and inaccurate
2.7 20.1** 17.9 48.1 20.6 40.7** 2.7 20.1** 17.9 48.1 20.7 40.7**
Inconsistent and accurate
1.4 30.4 5.4 26.0 6.8 26.9 2.5 54.3 7.8 37.5 10.3 40.6
Inconsistent and inaccurate
3.2 69.6** 15.4 74.0** 18.5 73.1** 2.1 45.7 13.0 62.5** 15.1 59.4**
Students study 2
Central T2 Peripheral T2 Total T2 Central T3 Peripheral T3 Total T3
Mean number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
% Mean
number of items
%
Consistent and accurate
10.3 78.6 17.5 46.1 27.8 54.4 10.3 78.6 17.5 46.1 27.8 54.4
Consistent and inaccurate
2.8 21.4** 20.5 53.9* 23.3 45.6* 2.8 21.4** 20.5 53.9* 23.3 45.6*
Inconsistent and accurate
1 20.4 4.4 22.0 5.4 21.7 2.4 49.0 7.5 37.5 9.9 40.0
Inconsistent and inaccurate
3.9 79.6** 15.6 78.0** 19.5 78.3** 2.5 51.0 12.5 62.5** 14.8 60.0**
Note. Central max=18, peripheral max=58 and total max=76 ** p< .001, * p<.005
97
Discussion
The aims of this study were to investigate the accuracy and consistency of memory for central and peripheral aspects of emotionally arousing versus neutral stimuli in subjects who had experienced a traumatic event versus control subjects, and furthermore to explore the relationship between accuracy and consistency of memory.
With regard to accuracy of memory, only half of the questions on the multiple-choice test were answered accurately. The levels of accuracy were partly in line with other experimental findings. They are comparable with those of Cahill and McGaugh (1995) and Van Stegeren et al. (1998). On the other hand, the level of accuracy determined in the present study is lower than the findings of Smeets et al. (2004). However, this difference can be partly explained by the different way of questioning.
Smeets et al. (2004) used a detailed written account to assess accuracy compared to multiple-choice questions in the present study. According to Ibabe and Sporer (2004), open ended and true or false questions lead to more correct answers than multiple-choice questions.
Furthermore, results on accuracy of memory were in line with our hypothesis. As expected, repeated memory testing led to hypermnesia. Two weeks following the intervention, free recall and recognition memory for both emotionally arousing and neutral stimuli became significantly more accurate. These findings concur with previous research which showed that repeated testing improves recall for both arousing and neutral stimuli (Dunning & Stern, 1992; Erdelyi & Becker, 1974; Payne, 1986, 1987; Scriver & Safer, 1988; Turtle & Yuille, 1994, though see Wheeler & Roediger, 1992).
With regard to the increased accuracy rates on the recognition memory task, the “testing effect" may be a likely explanation, which refers to the effect that an initial test increases performance of later recall (Glover, 1989; McDermott, 2006; Roediger & Marsch, 2005; Spitzer, 1939). According to Roediger and Marsch (2005), taking a multiple-choice test improves performance on a later cued recall test.The first memory test has a positive effect when the same information is tested later on. It is also possible that by the second memory test, participants choose the alternative that seems like a reasonable guess. The testing effect might also account for the enhanced recall on the free recall test.
Information obtained in the multiple-choice recognition memory task at first follow-up can be learned or may serve as a cue for the reported information at T3 in the free recall task. However, repeated testing is unavoidable when studying memory consistency. In order to examine consistency of memory properly, recall has to be assessed on different occasions, using the same materials.
With regard to consistency of memory, results showed that overall consistency rates were higher than accuracy rates. Not in line with the study results of Smeets et al. (2004) inconsistencies were mainly due to commissions instead of omissions. In the Smeets et al. study memory consistency was assessed by comparing detailed written accounts based on free recall over a period of 3-4 weeks and no multiple-choice questions were asked. Apparently, this procedure mimics the process of normal forgetting, i.e. reporting of correct information during the first test which is omitted during the second test. The relatively high incidence of commissions in the present study may be partly due to both accurate and inaccurate facts learned from the first multiple-choice tests being brought in as commissions at T3 free recall. Information that is still implicit at the first assessment may have become explicit at the second assessment because of the cues in the multiple-choice questions (e.g., multiple-
choice question at T2: „What colour was the police car?‟; free recall at T3: „What did you see on the slides?‟; Answer (among others): „A white police car.‟). Also, the multiple-choice questions can affect the learning of false facts. Errors made on a previous multiple-choice test often persist on a cued recall test conducted at a later stage. Subjects often provide one of the incorrect alternatives of a multiple- choice questionnaire as an incorrect answer on a later test (Roediger & Marsch, 2005).
In contrast with our hypothesis that the emotional stimuli would be more accurately and consistently remembered, no overall differences between the neutral and arousal condition in memory performance were found in both study 1 and 2.
In line with our hypothesis with respect to memory accuracy and consistency of type of information, our results showed that central aspects were not only more accurately remembered than peripheral aspects (Cahill & McGaugh, 1995; Ibabe & Sporer, 2004), but also more consistently remembered. Of note is that no differences in accuracy and consistency between traumatized and control participants nor between emotional and neutral conditions were found. These results suggest a relatively high accuracy and consistency of central aspects compared to more peripheral aspects irrespective of trauma history or emotionality of the event. This makes central information of events a key issue in the accurate and consistent reports of both emotional arousing and neutral stimuli.
The exploration of the relationship between accuracy and consistency resulted in the most remarkable finding of our study. Results showed a high correlation between accuracy and consistency.
The level of consistency between the two memory accounts was significantly related to the accuracy ratings at both assessments. Results were found for both central and peripheral aspects, both in a forced and free format. The correlation was found for the number of accurately and consistently reported items as well as for the accuracy and consistency index. Results were found in all groups and conditions, except for a significant correlation between total consistency and T2 and T3 total accuracy of recognition memory in the student population of study 2. The correlation of consistency of peripheral aspects with accuracy of peripheral aspects in study 2 as well as the correlation of consistency of free recall with T3 accuracy was also not significant. The absence of a significant correlation between accuracy and consistency of peripheral aspects in study 2 may be due to limited power problem because also in study I the association of accuracy with consistency was far more pronounced for central than for peripheral aspects. Although our way of questioning and scoring is comparable to theirs, our findings are only partly in line with results by Smeets et al. (2004) and Brewer et al. (1999).
Smeets et al. (2004) did not find a significant correlation between accuracy and consistency. Brewer et al. (1999) did find a moderate correlation of 0.30 for overall accuracy and consistency. When interpreting our results on the relationship between consistency and accuracy, zero-order correlations do not seem to be the best way to understand this relationship. When the correlation between accuracy and consistency is significant, we still do not know if consistency also implies accuracy and vice versa. When looking at the combined accuracy and consistency scores, results show that when information is consistently recalled, the central aspects are more often also accurately recalled, in contrast to the peripheral aspects. Furthermore, results show that when information is inconsistently recalled it is more often inaccurately recalled, but still can be accurate, and may even become more accurate with repeated testing.
Some strengths and limitations of the present study have to be acknowledged. An important strength of the study is the use of two different operationalizations of consistency and accuracy, making a comparison of the results easier. Another strength is the inclusion of both assault victims and students who personally experienced a traumatic event, allowing the assessment of memory performance in a clinically more relevant population than most previous experimental studies.
However, it appeared to be difficult to find victims of assault and students who had experienced a traumatic event and who were willing to participate in the experiments. Only 29% of the assault victims who were approached agreed to participate, which could indicate a selection bias. However, no evidence for such a bias was found. Although there are no indications for selection bias, one should be cautious in generalizing the present study findings. Another aspect concerning the external validity of our findings relates to our stimulus material. Although ratings of emotionality showed that subjects from study 1 showed a stronger emotional reaction to the stimuli in the arousal condition than in the neutral condition, no overall differences between the neutral and arousal condition in memory performance were found in both study 1 and 2. This is striking because it is widely assumed that events are more accurately and vividly remembered when they are emotionally arousing (Bremner, Krystal, Charney, & Southwick, 1996; Rolls, 1990). The absence of an effect for condition in the present study might be due to the emotionality of the stimulus material used. Reported VAS scores were in line with those of Cahill and McGaugh (1995) and Van Stegeren et al. (1998). However, the emotionally arousing slides probably did not elicit an emotional reaction as strong as emotions elicited by a real-life personally experienced event. The outcomes of the present study may have implications for legal practice. Accuracy rates in the present study were generally low. Only half of the information was accurately remembered, which indicates that one cannot completely rely on the accuracy of eyewitnesses. No significant differences were found in accuracy or consistency rates between participants who personally experienced a traumatic event compared to those who had not. Also no differences were found in accuracy and consistency of emotional stimuli versus neutral stimuli. These results imply that a person‟s history and the emotionality of an event may not critically affect the accuracy and consistency of memory reports. In all groups and conditions, evidence was found supporting the notion that central aspects were more accurately and consistently remembered than peripheral aspects. When consistently recalled, central aspects are more often accurate compared to the more peripheral aspects of an event. This might imply that one can rely more on the central aspects of an eyewitness testimony than on the more peripheral aspects. When using consistency as a proxy measure for accuracy, this applies primarily to the central aspects of a testimony. Because accuracy of recall increased over time, and consistent responses with respect to central details were more often accurate than inaccurate, it is advisable to conduct multiple interviews with witnesses in order to increase the accuracy of the information provided (Peterson, Moores, & White, 2001).
With regard to clinical practice, results show once more that therapist should stay far from statements on the reliability of their client‟s reports (see also Brown et al., 1998; Simon & Gutheil, 1997). Although the present study shows that consistent responses are often accurate with respect to central aspects, consistent responses still can be inaccurate. On the other hand, inconsistent responses do not always have to be inaccurate. The focus of a therapist should be on the impact of
the event and the meaning of the memories, and not on finding the truth (Everaerd & Gersons- Wolfensberger, 2004).
