Auditorium
SYMPOSIA 9:30 - 10:45 a.m.

Focusing on the internal in memories of trauma
Chair: ROBYN FIVUSH
Contact: psyrf@emory.edu
Keywords: Autobiographical memory, Narrative production and comprehension.

Much of our autobiographical memory focuses on what Bruner has called "the landscape of consciousness," our thoughts and feelings about our life experiences. Because trauma often creates a problem for processing and comprehending experience, many theorists have proposed that traumatic memory may focus on the internal to a greater extent than non-traumatic memory. The papers in this symposium explore this proposal from developmental and longitudinal perspectives. The emerging pattern of results provides a framework for understanding how traumatic memories may be similar and different from non-traumatic memories, and how traumatic memories may change over time.

Memories of positive and negative events
Jennifer Bohanek

Whereas most memory research has focused on the accuracy of memories for particular events, the focus of this study was to compare the content and evaluation of women's memories of events differing by both emotional valence and intensity. Forty-four women (age 18-21) were instructed to write about an intensely negative, moderately negative, intensely positive, and moderately positive event including both the facts of what happened, as well as their thoughts and feelings surrounding the event, and they also subjectively rated their memories for characteristics such as vividness and significance. The narratives were coded for 5 categories of internal state language: positive and negative emotion, as well as words reflecting cognition, insight, and causation. The narratives were also coded for a global level of coherence. Overall, emotional valence was related to the content of the women's narratives, and emotional intensity was related to the subjective ratings of the memories. Negative narratives contained more negative emotion and cognitive processing words than positive narratives, and positive narratives contained more positive emotion words than negative narratives. Interestingly, there was little relation between the objective content of the narratives and the women's subjective ratings of their memory experiences. Women rated intensely emotional events as more frequently talked/thought about, significant, unique, emotional, and vivid than moderately emotional events, and negative narratives were rated as more emotional than positive narratives. Intensely negative narratives were the longest, and intensely positive narratives were the most coherent. These results have important implications for flashbulb memories, narrative research in general, methodological issues (such as the validity of subjective memory ratings), and the quality of traumatic memories.

Narratives of 9/11
Robyn Fivush, Valerie Edwards

Several theories of traumatic memory highlight the narrowing of attention, and specifically the focus on internal states, that may occur during the experience and recollection of trauma. It has been further suggested that this focus on the internal may dissipate over time, as the individual begins to process and cope with the trauma. In order to examine this possibility more systematically, we compared three groups of undergraduate students' narratives of the 9/11 terrorist attacks over time: students who knew someone who was killed, students who lived in the New York area but did not know someone who was killed, and students who did not live in New York or know a victim. Students filled out a battery of stress and anxiety measures, and then wrote narratives of 9/11 for 5 days in a row within 6 weeks of the event, and then completed the measures again and wrote one narrative 6 months later. We predicted that those directly involved would experience higher stress, and would include more words indicative of emotional and cognitive processing than those less directly involved and, further that this group would show a less steep decrease in stress and in the use of these kinds of internal state words over time. Inspection of a subset of the transcribed data support these predictions. Ongoing analyses will include the full sample, as well as relating the cognitive and emotional content of the narratives to individual ratings of stress and well-being both immediately and over time.

Communicating emotions and cognitions about traumatic and nontraumatic
Patricia J. Bauer

A question of major significance is whether memories of traumatic and nontraumatic events differ quantitatively and/or qualitatively. Because traumatic experiences engender such strong emotions, as well as cognitive reflection and evaluation, language about internal states is a prime domain for address of this question. Examination of use of internal states language in the context of mother-child conversations has the added benefit of informing as to whether the process of socialization of internal states language in autobiographical narrative is different for traumatic and nontraumatic events. We examined mother-child conversations about a devastating tornado and about two nontraumatic events to determine whether there are (a) differences in mother's and/or children's use of internal states language in narratives about the two types of events, and (b) similarities in mothers' and children's use of internal states language. Four months after the tornado, with conversational length controlled, there was no evidence of differential use of internal states language as a function of event, for mothers or children. Six months later (10 months after the tornado), older children's narratives about the tornado were more saturated with internal states language. There was evidence of internalization by children of their mothers' emotion language use: For both traumatic and nontraumatic events, there were correlations between maternal use of emotion language at Session 1 and children's use of emotion language at Session 2. The findings suggest differential socialization across internal states categories, but that the process of socialization of emotion language is similar for traumatic and nontraumatic events.

Stressing memory: Long-term relations among children's stress, recall, and psychological outcome following Hurricane Andrew
Jessica McDermott Sales, Robyn Fivush, Janat Parker, Lorraine Bahrick

Hundreds of children were affected by Hurricane Andrew, one of the most destructive natural disasters to occur in the United States. This study examined relations among individual level of stress, content of children's recall of Hurricane Andrew, and long term psychological functioning. Three to 4-year-old children were interviewed within a few months following the storm, and again six years later. The adult literature suggests that the emotional and cognitive content of one's memories has a significant impact on one's well-being. Thus interviews were coded to examine the emotional and cognitive content of children's recall. Additionally, children's stress levels were assessed by multiple measurements collected at either one or both interviews. Children remembered the hurricane in vivid detail, and they continued to be emotionally affected by the hurricane even years later. At the initial interviews, children who were more stressed included less positive emotion, fewer cognitive processing words, provided less information overall and less free recall in their narratives. In contrast, children who were less stressed included more cognitive processing words and recalled more information. When re-interviewed 6 years later, the children who had been more stressed initially included more negative emotion and cognitive processing words, but provided less information during free recall. Children who had initially used more positive emotion words and recalled more information displayed lower stress scores 6 years later. Implications for children's remembering and coping with traumatic events will be discussed.

Discussant: Lynne Baker-Ward

KCF7
SYMPOSIA 9:30 - 10:45 a.m.
Memory Conformity: How do people influence each other's memories?
Chair: FIONA GABBERT
Contact: f.gabbert@abdn.ac.uk
Keywords: General memory, Social cognition

Five papers investigate how one's knowledge of another person's memory can influence one's own memory. Each paper within the symposium approaches this topic from a different perspective, from exploring the effects of discussion between eyewitnesses, to questioning how and why memory conformity might occur. To meet the growing interest in this area of research, the symposium brings together a varied collection of studies each addressing a common theme.

Co-witness contamination: The effects of witness discussion on subsequent recall and recognition
Helen Paterson, Richard Kemp

This presentation focuses on the effects of co-witness information on the accuracy and comprehensiveness of eyewitness memory. Co-witness information is defined as information that one eyewitness conveys to another eyewitness about an event that they both observed. It is standard practice for police and other legal personnel to attempt to prevent eyewitnesses from discussing the event with each other, however it is clear that witnesses do often talk to each other about the event. Experiments are described which examine the effects of post event information encountered through discussion with other witnesses. In each study, participants were shown a crime video and then asked to discuss the video in groups (some of which received experimentally induced misinformation). Finally, participants were asked to give their individual accounts of what happened. It was found that there are both advantages and disadvantages of co-witness discussion, depending on the methodology employed. Furthermore, some evidence suggests that it may be possible to distinguish between 'real' memory and information obtained from a co-witness. These results are discussed in terms of methodological considerations, theory, and policy implications.

The effects of discussion and prior acquaintance on the recall of contradictory details of a simulated crime.
Samantha Foster, James Ost

This study investigated the effects of discussion and self-reported level of acquaintance on the recall of contradictory elements of a videotaped, staged theft. The experiment employed a 2 (pre-discussion recall vs. post-discussion recall) x 4 (stranger/friend/partner/individual) mixed design. The dependant variables were the (a) proportion of correct responses, (b) mean confidence and (c) testimonial validity. Fifty-nine participants (37 females and 22 males) were recruited. Participants individually watched a videotape of a staged theft, completed a multiple-choice questionnaire concerning details of the crime and provided confidence ratings for those details. Participants then either thought about the video for 10 minutes (no discussion condition), or discussed the video for 10 minutes in a dyad (discussion condition). Critically, in the discussion condition, each member of the dyad saw a slightly different version of the videotaped theft in which five details had been deliberately altered. Further, to examine the possible effects of familiarity with the other person in the dyad, participants in the discussion condition were asked to self-rate their level of acquaintance. All participants were then asked individually to complete the multiple-choice questionnaire for a second time. Twelve between-within subjects ANOVAs were conducted. For details of the video that were seen the same by both members of the dyad, discussion had little positive or negative effect on individual recall and testimonial validity. For details of the video that had been seen differently by each member of the dyad, discussion negatively effected both individual recall and testimonial validity. Contrary to expectation, there were no differences as a function of participants' self-rated level of acquaintance. Implications will be discussed.

Asymmetric Effects in Memory Conformity
Daniel B. Wright, Sorcha Mathews

Two experiments show that people's recognition memory for words and pictures of automobiles can be affected by what another person says. Participants viewed either a set of words or automobiles with a confederate, and then they were tested using an old/new recognition test. The confederate responded first. Both experiments showed similar patterns. The memory conformity effects, where the participant responded with what the confederate said, were of similar magnitudes in the studies. For both studies, there the effect was not the same for items previously seen and unseen. The effect was larger for convincing someone that a previously unseen item was present, than to make them say that an item which they had seen was not present. We discuss this asymmetry in relation to social and cognitive explanations for the memory conformity effect. We also varied the frequency of the words (high and low frequency, and non-words) and how unique/memorable the automobiles were. These had different levels of accuracy, but did not vary in their susceptibility to conformity effects.

Investigating Eyewitness Conformity Effects for a Simulated Crime
Fiona Gabbert, Amina Memon, Kevin Allan, Daniel B. Wright

In everyday life it is natural for people to discuss their shared experiences. However, because memories are not infallible, sometimes one person will introduce inaccurate information into the conversation. Our past research has shown that this information can be incorporated into the other person's memory, and subsequently reported as their own recall of the event. This 'conformity' effect is stronger than the effect of more commonly used experimental methods of misleading eyewitnesses, such as post-event narratives.

Our current research investigates factors underlying memory conformity. We address forensically relevant questions, asking whether memory conformity is an inevitable result of discussion, or whether it only occurs in certain interactions. Participants were tested in pairs. Each member watched a different version of the same event. They later discussed the event together before individually providing a written summary of what they had seen. A battery of individual difference measures were used to explore whether memory conformity was related to particular personality characteristics. In addition the naturally occurring discussions were audio-taped, transcribed and analysed. Our most striking finding was that the witness initiating the discussion was most likely to influence the other witness's memory report. In other words, the witness initiating the discussion was the most resistant to influence even when their memory was disputed by their co-witness.

Memory distortions as a function of social influence and meta-cognitive knowledge
Eva Walther

It is asked under which conditions individuals fall prey to the influence of a group of others when they are uncertain as to whether an event has actually occurred. We suggested that judgments of others have an impact on memory reports only if other sources of information (e.g., metacognitive knowledge) are not sufficiently applicable to resolve uncertainty. Using a standard recognition paradigm, we examined in two experiments how individuals were influenced by the judgments of a bogus group when they were asked about stimuli that had never been presented. Following traditional research on social influence (cf. Asch, 1951; Festinger, 1954; Sherif, 1936), it is assumed that social influence on memory is similar to the influence that others exert on behaviors and attitudes, with three variables determining the impact of a group: uncertainty, group size, and the presence of dissenters. In two experimental studies we investigated how false memory reports provided by anonymous group members affect individuals' recognition. Participants were presented with a series of salient and non-salient objects and recognition memory was subsequently assessed for presented and non-presented stimuli. In this recognition test, participants were confronted with (in)correct responses from bogus group members. The findings indicate that the bogus answers influenced memory reports only in the absence of a clear recollection and if the stimulus was not judged to be particularly memorable. Moreover, we found no evidence that motivational influences affected participants' judgments. However, the influence of the bogus group decreased with the presence of dissenters who broke the unanimity of the group, and it increased with higher uncertainty.

KCF8
SYMPOSIA 9:30 - 10:45 a.m.

Detection of malingering
Chairs: MARKO JELICIC, HARALD MERCKELBACH
Contact: m.jelicic@psychology.unimaas.nl
Keywords: Malingering, Reasoning and decision making

People involved in legal matters sometimes simulate memory and other cognitive impairments. Such malingering may take place in order to obtain financial compensation for alleged brain damage or to avoid imprisonment by claiming insanity. This symposium will deal with different aspects of the detection of malingering. First, methodological issues of research on identifying malingerers will be discussed. The following three presentations will deal with the usefulness of different malingering tests and questionnaires. The last presentation will discuss the influence of coaching on the detection of malingering with special malingering tests.

From north to south and back - evaluating different tests of "lowered effort" with simulators and patients
Jonathan K. Foster

Malingering tests (or tests of `lowered effort`) can be evaluated in two main contexts: a) in simulators who have been instructed to `fake bad` and b) in patients who are suspected of exaggerating their cognitive symptoms. In the case of a), simulators are often psychology students who (it is often assumed) may possess a relatively sophisticated knowledge of memory and memory impairment. Performance of these simulators on tests of malingering is typically compared with the memory performance of a demographically comparable control group of individuals who are instructed to perform to the best of their ability. Tests of lowered effort are evaluated with respect to their sensitivity and specificity in determining (via a cut-off score) which participants belong to each of these two `simulator` and `best effort`groups. In the case of b) patients who are suspected of exaggerating their cognitive symptoms, it is of course more problematic to determine whether or not the individual is in fact `faking bad` without reference to cut-off scores which have typically been established through alternative means. The focus of our research to date has been to evaluate (in a comparative manner)novel and more established tests of lowered effort in simulators (and controls) rather than in patients, but we are now beginning to apply these tests in litigating versus non-litigating patients who have sustained an alleged brain injury. Relevant findings concerning sensitivity, specificity and other relevant test characteristics will be discussed.

Symptom validity testing: Clinical applications and implications for research
Ben Schmand

Malingering tests have been designed for application in forensic settings. Originally, their purpose was to test the validity of cognitive symptoms in the context of litigation, workmen's compensation claims, and the like. However, these tests may also be applied very fruitfully in clinical settings. In cases with cognitive complaints of unclear aetiology, such as those in chronic fatigue syndrome, late whiplash, other chronic pain syndromes, and chronic toxic encephalopathy ('painter's disease'), malingering tests may serve to separate 'bona fide' patients from somatisers and aggravators. After removing the noise caused by these latter groups, there remains very little evidence for cognitive disorders in these syndromes. When applied in psychiatric syndromes, such as depression and schizophrenia, malingering tests may serve as a methodological check of the effort invested by the patients during cognitive evaluation. Examples of these clinical applications from the literature and from own research will be reviewed. Some of the most surprising findings will be discussed, as well as their implications for research in cognitive neuropsychiatry.

Diagnostic accuracy of the SIMS in detecting malingering of psychopathological symptoms
Harald Merckelbach

The Structured Inventory of Malingered Symptomatology (SIMS) is a self-report measure that intends to screen for malingering tendencies. The 75 yes-no items of the SIMS tap commonly malingered conditions, like psychosis and amnestic disorder. The items describe symptoms in such way that they are plausible to naïve people, but highly suspicious to specialists (e.g., "I have difficulty remembering the day of the week").In this presentation, we address the psychometric properties of a Dutch translation of the SIMS when administered to diverse samples of undergraduate psychology students and psychiatric inpatients. In the normal samples, some participants were instructed to simulate symptoms, while others were asked to respond honestly. Findings show that our SIMS version possesses good test-retest reliability and internal consistency. Also, simulation findings indicate that undergraduate students instructed to simulate pathology display higher SIMS scores than either normal controls or psychiatric inpatients. Data pooled over several samples (N= 298) yielded sensitivity, specificity, and Positive Predictive Power (PPP) rates that were all relatively high (≥0.90). However, we also found robust correlations ( r's > 0.50) between SIMS scores, depression, and trait anxiety. The precise meaning of this overlap is unclear: does it mean that people with high SIMS scores also overendorse symptoms on scales that are thought to tap real psychopathology or does it mean that depressive and anxious people have a tendency to exaggerate their symptoms? All in all, our findings provide a basis for cautious optimism regarding the usefulness of the SIMS as a screening tool for malingering.

"Supernormality" in psychopathy: Development of a new instrument
Maaike Cima

It has often been suggested that psychopaths are successful malingerers. This assumes that psychopathic individuals have a tendency to display one particular type of deception (i.e., exaggeration). However, it may well be the case that the link between psychopathy and deceptive behaviour is strongly affected by the context. For instance, the psychopathic patient in a mental institution might try to deny psychopathology in an attempt to make a healthy impression on his evaluators. The term 'supernormality' refers to the tendency to systematically deny the presence of common symptoms (e.g., intrusive thoughts). A previous study described the psychometric qualities of a 37 item self-report measure of supernormality (i.e., Supernormality Scale; SS). Results showed that the accuracy parameters for the SS (i.e., sensitivity and specificity) were rather poor. Although, the findings indicated that the SS was a useful research tool for measuring feigning of healthy behaviour, it certainly needed improvement. In a follow-up study, the SS was revised (i.e., Supernormality Scale-Revised; SS-R). This 72 item questionnaire was administered to normal individuals, noncriminal psychiatric patients, nonpsychiatric delinquents, and a heterogeneous sample of forensic patients. In the forensic patient sample, SS-R scores were correlated with psychopathic personality characteristics as measured with the Psychopathic Personality Inventory (PPI). In addition, as a way to establish the validity of the SS-R, a thought suppression paradigm was conducted within two different samples of sexual offenders. We predicted that high scores of faking good on the SS-R would go hand in hand with a suspicious low frequency of target thoughts during thought suppression. Results of this study will be discussed.

Effects of coaching on instruments developed to detect malingering of memory impairments
Marko Jelicic

Faking of memory and other cognitive impairments occurs in people seeking financial compensation for alleged brain damage. Because clinicians have great difficulty identifying malingering of memory impairments, special instruments have been developed to detect malingering. Most of these instruments are based on the assumption that malingerers do not know that genuine patients exhibit relatively normal performance on some tests and questionnaires. On these instruments malingerers show lower performance or more symptoms than real patients. It has been suggested in the literature that coaching, i.e. providing information about brain damage, would render special malinger tests invalid.We studied the effects of coaching on two instruments used to identify malingering of memory impairments, the Amsterdam Short Term Memory test (ASTM) and the Structured Inventory of Malingered Symptomatology (SIMS). These two instruments were administered to three groups of undergraduate students. The control group consisted of 30 participants asked to perform honestly, a naïve malinger group consisting of 30 participants asked to simulate memory impairments brought about by brain damage, and a informed malinger group comprising 30 participants who were given some information about brain damage and then asked to simulate memory impairments. The ASTM and the SIMS were able to identify both naïve and informed malingers (sensitivity rates > .70). Control participants were also identified as non-malingerers (specificity rates > .90). Our findings suggest that the ASTM and the SIMS are relatively immune to the effects of coaching.

Auditorium
SYMPOSIA 11:05 a.m. - 12:20 p.m.

When we change autobiographical belief, do we also change memory?
Chair: GIULIANA MAZZONI
Contact: mazzongi@shu.edu
Keywords: Autobiographical memory, False memories, Suggestibility.

"We investigated the role of fluency and familiarity in the formation of false memories. Subjects completed a Life Events Inventory in which one-half of the items contained scrambled key words (e.g., ""broke a dnwoiw playing ball""). Subjects rated whether these life events had occurred in their childhood before the age of 10. In a series of experiments, we show how the act of unscrambling a key word in an event description increases subjects` confidence that the event occurred not only in their own childhood but also in the childhood of others. In another line of studies involving incidental and deliberate exposure to key words (subjects first count vowels in or vividly imagine key words, e.g., ""window,"" that later appear in an event description, e.g., ""broke a window playing ball""), we failed to increase subjects` confidence. These results indicate that.prior exposure to words and unscrambling words can, in some cases, produce fluency that leads to a feeling of familiarity. When the source of the fluency is obscure, people will search for potential reasons (sources) to explain the fluency. In the absence of an obvious or reasonable source, people misattribute the fluency to their own childhood. The act of correctly unscrambling an anagram can produce a surge of fluency, akin to an ""aha"" experience, that people fail to attribute to the act of unscrambling. Instead, they misattribute their feeling of familiarity to their own childhood. These findings suggest that false memories can arise, in part, through the misattribution of familiarity.

Unscrambling one's past
Daniel M. Bernstein, Ryan Godfrey, Elizabeth F. Loftus

Recently, Wade et al. (2002) showed that 50% of adult subjects developed false memories when presented with doctored photos of a fictitious childhood event. We adapted this doctored photo method to investigate the effects of plausibility and childhood amnesia on children's false memories. Six and 10-year old children were interviewed three times over the course of a week. At each interview children were presented with four photos, two real and two false. We varied two factors. First, we varied the plausibility of the false events (plausible, implausible). Second, we varied the age at which the event purportedly occurred (age 2 or 6). Children were asked to describe what they could remember about the event, and then rate "how much" they could remember and "how sure" they were that the event had happened. Six year olds reported high belief but did not provide a great deal of information about the events. For the 10-year old children, those who were recalling events from age 2 were more likely to produce false reports than those recalling events from age 6. These results suggest that childhood amnesia may play a role in the development of false memories.

The effect of false information and imagination on false beliefs and false memories
Stefanie Sharman, Giuliana Mazzoni, Irving Kirsch, Maryanne Garry

Subjects completed a Life Events Inventory in which one-half of the items contained scrambled key words (e.g., "broke a dnwoiw playing ball"). Subjects were asked to determine whether these life events had occurred in their childhood before the age of 10. In a series of experiments, we show how the act of unscrambling a key word in an event description increases subjects` confidence that the event personally occurred. In another line of studies involving incidental and deliberate exposure to key words (subjects count vowels in or vividly imagine key words), we failed to increase subjects` confidence. We discuss the role of familiarity misattribution in the formation of false memories. Specifically, we discuss how fluency can be misattributed to one`s past by altering the availability of potential sources that could explain the fluency.

Measuring autobiographical memories and beliefs
Alan Scoboria, Giuliana Mazzoni, Irving Kirsch, Mark Relya

Research has shown that giving people false suggestions can make them recall events that never happened (eg, Loftus & Pickrell, 1995). Research has also shown that asking people to imagine fictitious events makes them more confident that those events really happened (Garry, Manning, Loftus, & Sherman, 1996). We examined the combined and separate effects of false suggestions and imagination on the creation of false beliefs and false memories, using a three-stage procedure. First, participants provided belief and memory ratings for a list of childhood events. Second, participants (a) were told that the false haircut event happened to them; (b) imagined the haircut; (c) were both told it had happened and imagined it, or (d) were in the control condition. Third, all participants completed the belief and memory ratings again.

We found that only false information produced false beliefs: participants who were told the event happened became more confident that it really had. Imagination enhanced memories of the false haircut event, but only for participants who were not told that the event had happened. Our results have both theoretical and practical implications. They confirm predictions from a theoretical model (Mazzoni & Kirsch, in press); they suggest that imagination alone is powerful enough to create false memories, and they reveal the ease with which memories for events that never happened can be implanted.

Inferring probable constituents: Belief versus memory in the 'memory wars.'
James Ost

In the false memory literature, measures of one construct are sometimes used as stand-ins for related constructs. Measures of autobiographical belief are often used as a measure of autobiographical memory, and script knowledge has been used to infer the plausibility of an event. The present study tested these assumptions by investigating the relationship between plausibility, belief, memory, and script knowledge for events. 685 undergraduate students completed measures of plausibility, belief and memory, and script descriptions for a series of 10 hypothetical childhood events. Results indicate that the constructs of plausibility, belief and memory are distinct, and correlate with one another in a predictable manner. Script knowledge did not correlate with any of the constructs.

A right royal road to false childhood memory
Deryn Strange, Rachel Sutherland, Maryanne Garry, Harlene Hayne

A fundamental disagreement exists over the findings of laboratory studies of false memory. Some researchers claim these provide impressive demonstrations of false memory, whilst others use exactly the same findings to support the argument that such memories are extremely difficult to 'implant'. There is nevertheless agreement that 'memory' should be the central issue. But is this really the case? This paper will review the disagreements concerning false memory in the published literature, as well as presenting the findings from our own research with retractors (individuals who have repudiated their earlier claims of abuse). Analysis, both of the literature, and of retractors' accounts, indicates that 'memory' should not be the sole focus of the false/recovered memory debate. Examination of retractors' accounts suggests that, in some cases, a belief is, in itself a sufficient condition for making an accusation of abuse. There are also similarities between retractors' accounts of coming to make accusations of abuse, and those of individuals who have come to confess falsely in police interrogations. It would seem that what we believe about the past may, in some cases, be more important than what we remember about it.

KCF7
SYMPOSIA 11:05 a.m. - 12:20 p.m.

True and false memories for events. The effects of emotional arousal
Chairs: LYNN HULSE, AMINA MEMON
Contact: l.hulse@abdn.ac.uk
Keywords: Emotion, General memory.

The relationship between emotional arousal and memory is complex. Accordingly, the papers in this symposium investigate this relationship using various interesting approaches. The first paper measures the effects of physiological arousal on event recall and identification accuracy for a sequential lineup. The second paper investigates the association between arousal (state and trait anxiety), source identification and ability to detect misleading details, while the third focuses on the role of emotion in the creation of false memories using the Deese-Roediger-McDermott paradigm. The final paper examines the effect of visually- versus thematically-induced emotion on memory for central and peripheral details. A discussion, led by Paula Hertel, follows these four papers.

Affecting memories: Emotional arousal and eyewitness testimony
Lynn Hulse, Amina Memon, Kevin Allan

At present, there are three somewhat conflicting main theories about the effects of emotional arousal on memory for events. One theory suggests that emotional arousal impairs memory, another suggests that it enhances memory, and the third proposes that it has selective effects on memory for different types of details. The current study aimed to explore the relationship between emotional arousal and memory by applying both physiological and behavioural measures of arousal, recall and recognition to an eyewitness paradigm. Emotional arousal is somewhat inherent in crime events due to their often violent or upsetting nature. Moreover, the reliability and accuracy of eyewitness memory is crucial to help the police and courts uphold justice. Thus, we presented sixty participants (university undergraduates) with either an emotionally arousing video event depicting a crime, or a relatively neutral event. Memory for event details was then assessed by performance on a written cued-recall measure and a sequential lineup task, which included a picture of the target from the witnessed event. Physiological activity (skin conductance, heart rate) was recorded while participants viewed the event stimuli and lineup. In addition, participants were asked to complete self-report questionnaires about emotions they experienced and, finally, Spielberger's (1983) Trait Anxiety Scale. The results of this study are discussed in the light of the aforementioned theories of emotion and memory and provide insight into whether self-report and physiological measures of arousal predict memory for event details and accuracy of identification.

The effect of state and trait anxiety on suggestibility in a source identification paradigm
Anne M. Ridley, Brian R. Clifford

Source identification (ID) tests as used by Zaragoza and her colleagues (e.g. Mitchell & Zaragoza, 1996) provide a more stringent method for testing suggestibility effects than the standard suggestibility paradigm because they reduce response bias and experimental demand characteristics. They also provide a way of differentiating between reductions in suggestibility due to two very different processes: correct source identification, and the non-recognition or forgetting of misleading information. Being able to discriminate between these processes is particularly important when investigating individual differences and suggestibility. In this study using 56 undergraduates, we investigated how state and trait anxiety (as measured by the STAI; Spielberger, Gorsuch, Luschene, Vagg & Jacobs, 1983) affect the ability to identify correctly the source of misleading post-event information. Multiple regression was used to analyse the data. The results showed that state anxiety was negatively related to suggestibility (p <.01). This relationship was moderated by trait anxiety. The strongest relationship between state anxiety and suggestibility was found when trait anxiety was high. Non-recognition or forgetting of post-event information was positively related to trait anxiety. A source identification task cannot take place if post-event information is not recognised. It was found that the exclusion of unrecognised misleading items from the analyses increased the size of the effect of state anxiety on suggestibility. The key finding that state anxiety is associated with reduced suggestibility will be discussed in relation to the literature on individual differences and suggestibility, source monitoring, and anxiety and cognition.

Bad experiences, bad memory?
Joanne Sangster, Kevin Allan

The effect of emotion on memory is a complex one. To try and establish how and why emotion can exert a disruptive affect on memory it is necessary to investigate paradigms in which emotional stimuli results in false memory. Therefore, the underlying processes can be uncovered and an understanding of what influences accurate memory recall can be achieved. There has been a vast amount of research into false memory for words using the DRM paradigm (Roediger & McDermott, 1995). However, there has been a serious lack of research into how emotional content of stimuli in the DRM paradigm can lead to differences in false memory. This research focuses on the use of categorised emotional pictures to establish how the effect of categorization and, more importantly, the effect of emotion disrupts accurate recognition of previously experienced items as well as false alarms to novel items. To date, this paradigm has discovered that there are no differences in the false alarm rates for different emotions (positive, negative and neutral), whereas there are significant differences between baseline false memory and false memory for categorised items. These results seem to suggest that the influence of emotion may only occur when certain types of memory processes are involved (e.g. source memory) and not when responses are highly influenced by perceptual and conceptual similarity. False alarming in this modified version of the DRM paradigm may be solely related to the converging characteristics of the pictures.

The contrast between thematically- and visually-induced emotion: A broader view of memory narrowing
Cara Laney, Friderike Heuer, Daniel Reisberg

Previous research suggests a multiplicity of effects of emotional arousal on memory. First, memory is better for emotionally arousing events than for matched emotionally neutral events. Second, attention seems to be narrowed in emotionally arousing circumstances, so that emotion specifically enhances memory for the "centre" of the to-be-remembered event, but at the expense of that event`s "periphery." We argue, however, that these previous studies may not be generalizable to all emotional memories. In particular, the memory narrowing shown in these studies may be an artifact of how emotion was induced in these earlier procedures through the use of an attention-grabbing emotional stimulus (e.g., the sight of a bloody wound). We report data showing that this "visually-induced arousal" is atypical for emotional events outside of the laboratory; the vast majority of naturally-occurring emotional memories instead involve "thematically-induced arousal." We then report two laboratory studies, asking how thematically-arousing events are remembered. In both studies, arousal was produced through participants` empathy with the themes of the to-be-remembered materials, and not through the introduction of shocking or arousing visual stimuli. Participants` physiological and subjective levels of arousal were assessed and recognition memory was tested for both gist and details of the two stimulus sets. In both studies, participants in the arousal groups had better memories overall, suggesting that the enhanced memory is a feature of all emotional events. However, participants showed no indication of narrowed memories, relative to those in the neutral groups, suggesting that the narrowing phenomenon is not a product of arousal.

Discussant: Paula Hertel

KCF 8
SYMPOSIA 11:05 a.m. - 12:20 p.m.

Eyewitness identification: Witness confidence, identification
accuracy and investigator beliefs
Chair: NEIL BREWER
Contact: neil.brewer@flinders.edu.au
Keywords: Confidence-accuracy relationship, Eyewitness testimony, Face perception and identification.

Given the important role played by eyewitness identifications in criminal investigations, researchers have focused on the diagnosticity of independent markers of eyewitness identification accuracy. One influential marker within the justice system has been eyewitness confidence. With strong empirical support for the confidence-accuracy relationship not forthcoming psychologists remain pessimistic about its diagnostic value. In this symposium we present new research on the confidence-accuracy relationship. This research (a) highlights new aspects of the relationship in eyewitness identification and face recognition, (b) explores the source of confidence inflation produced by post-identification feedback, and (c) shows how identifications shape investigators? confidence and actions.

Confidence-accuracy calibration in eyewitness identification
Neil Brewer, Gary L. Wells

There is a now a substantial body of empirical research on the confidence-accuracy relation in eyewitness identification which demonstrates weak or, at best, modest relations between the two variables, thus suggesting that eyewitness confidence should be of little diagnostic value. Recent research in our laboratory and elsewhere with large samples, while confirming the typically modest confidence-accuracy correlation, has also used a procedure known as confidence-accuracy calibration to demonstrate impressive confidence-accuracy relations, at least under certain experimental conditions. Here we report a new study in which participants (N = 1,200) viewed a simulated crime and attempted identifications of two targets from separate 8-person photoarrays under one of several experimental conditions: (1) unbiased vs biased lineup instructions (i.e., witnesses warned or not warned about the possible absence of the offender/target from the photoarray); (2) high vs low similarity of photoarray foils to the target; and (3) target-present vs target-absent array. The typical modest confidence-accuracy correlations were detected across all experimental conditions. There were sizable variations in identification response patterns across the different identification stimuli and experimental conditions. Variations in the closeness of confidence-accuracy calibration were examined across these conditions and, importantly, in response to different base rates of target-absent lineups.

Confidence-accuracy calibration in face recognition: Effects of judgment type vs. task difficulty
Nathan Weber, Neil Brewer

Differences between sequential and simultaneous lineups are largely attributed to the processing strategies encouraged by the two different lineups. Specifically, simultaneous lineups are posited to encourage relative judgment strategies, whereas sequential lineups predispose witnesses to the use of absolute judgment strategies. Following Weber and Brewer (in press), who demonstrated differences in confidence-accuracy calibration between the two judgment types, we report two experiments that investigate the processes underlying relative and absolute face recognition judgments. Specifically, we examine the difference in confidence-accuracy calibration between absolute and relative judgments across different levels of task difficulty. Both experiments employed a face recognition paradigm where participants viewed a series of faces in a study phase then, after completing a distracter task, made recognition judgments about a series of individual faces (absolute judgments) or pairs of faces (relative judgments). In experiment 1 the difficulty of the recognition task was varied by manipulating the exposure duration of the faces in the study phase. In experiment 2 the difficulty of the recognition task was varied by manipulating the exposure duration of faces in the test phase. Results from both experiments indicated that participants tended to become more overconfident (or less underconfident) with increasing task difficulty. Further, equivalent calibration was observed for both judgment types in situations of equivalent task difficulty suggesting that the calibration differences observed previously were likely due to differences in difficulty between absolute and relative judgments not to differences in decision making processes. Implications for face recognition memory models and eyewitness identification are discussed.

The effects of intervening confidence assessments on the eyewitness post-identification feedback effect
Carolyn Semmler, Neil Brewer

Several studies have established that feedback to eyewitnesses indicating the accuracy of their identification decision is likely to influence their judgments of the subjective probability that they made the correct decision. Despite the robust nature of the post-identification feedback effect, we have little knowledge of the mechanisms underlying the effect. This study explored the possibility that the extent of the post-identification feedback effect is related to the ease or extent to which confidence cues are accessed. Specifically, we examined the impact on the post-identification feedback effect of requiring participants first to make an identification, then to make a series of confidence judgments in relation to subsequent identifications, and finally to make a confidence judgment about the original identification. Thus, the experimental conditions varied the number and type of judgments made between the first identification and the eventual delivery of feedback, with participants given confirming or no feedback just prior to making their confidence estimate for the initial identification. We examined whether participants who made multiple identifications and confidence judgments either (a) might have had more difficulty accessing relevant confidence cues for the initial identification and hence were more susceptible to the effects of post-identification feedback than participants who only accessed confidence cues associated with a single identification, or (b) might have been more likely to access confidence cues for the initial identification and to be less dependent on externally provided cues. The findings are discussed in terms of their impact on theories of post-identification confidence inflation.

Investigating witnesses' influence on investigators: A new method and some preliminary findings
Leora C. Dahl, D. Stephen Lindsay, C.A. Elizabeth Brimacombe

Eyewitness identifications often play a key role in police investigations and prosecutions. Inspired by practical concerns of wrongful imprisonment, psychologists have studied eyewitness identifications for nearly a century. Eyewitnesses? lineup performance is an important issue, but this area of research neglects another important player in the forensic drama: The investigator who collects the evidence. There is little research on how investigators conduct lineup tests, and even less is known about how investigators weigh identification evidence in the context of other evidence pertaining to the crime. To address such issues, we have developed a new method in which subject/investigators interview witnesses, search a computer database of potential culprits to find a suspect, and rate the probability that the suspect they chose is the culprit. Thereafter, investigators administer a lineup (that includes their chosen suspect) to the witnesses. Finally, investigators re-rate their confidence in the suspect?s guilt. We speculate that investigators tend to be greatly swayed by eyewitness identification testimony (even when it is inaccurate, and even when it conflicts with more solid grounds for decisions). In the studies we report here, the ?investigators? were university undergraduates and the witness was a confederate, who either identified the investigator?s suspect, identified a similar-looking lineup member, identified a dissimilar-looking lineup member, or indicated that the perpetrator was not present in the lineup. The studies revealed that an eyewitness? identification decision greatly influences investigators? confidence in their choice of suspect, confidence in the witness, and decision to arrest the suspect.

Auditorium
SYMPOSIA 1:20 - 2:35 p.m

Face composite production by eyewitnesses:
New technologies and developments
Chairs: DAWN MCQUISTON, COLIN TREDOUX
Contact: dmcquist@utep.edu
Keywords: Eyewitness testimony, Face perception and identification.

Crime witnesses are often asked by police to create visual likenesses of faces. This is most frequently achieved with the assistance of technical devices or specialized software (e.g., IdentiKit). The use of these composite technologies is explored in the current symposium for its theoretical and applied implications. Topics will include (a) the development of alternative technologies that use `eigenfaces` and genetic algorithms to improve composite production; (b) alternative construction techniques and their influence on cross-race composite production; (c) decision criterion and its impact on composite production, and (d) the use of composite faces to improve subsequent identification performance. Finally, Professor Malpass will act as discussant for the symposium, raising important theoretical and applied issues in this domain.

Parallel approaches to composite production: recognition rather than recall
Peter Hancock

Traditional approaches to the production of a facial likeness of a crimina suspect are very feature-based: "What did the nose look like?", "Are these the right eyes?", etc. Much psychological research indicates that people do not recognise faces in that way; we perceive the whole face rather than individual features. It is rather hard to describe someone`s face, and much easier (though still error-prone) to say whether or not you saw a given face. The question is how best to extract from someone`s memory what the face looked like. This research describes two new approaches to producing a facial likeness that draw on this recognition ability. EvoFIT is a completely new approach to the problem, using a principal components analysis of faces as the underlying representation. Eighteen randomly generated faces, matching the general description of the target, are generated and the witness selects several that look like the target. The system combines these in an evolutionary algorithm to create a new set for selection. The process continues, gradually getting closer to the target. The second approach is a parallel interface to PROfit, a standard feature-based composite system. However, rather than working serially through the selected set of noses, six at a time are presented for comparison. Counter to our intuitions, neither system works particularly well in practice, with EvoFIT only equal to current systems and the parallel PROfit slightly worse. We are currently investigating reasons for this outcome, one possibility being that users are simply confused by the number of faces that they are shown. These two systems will be discussed, along with possible reasons for the performance.

Investigating the own-race bias in composite production
Dawn McQuiston, Roy S. Malpass

Empirical research has demonstrated that people are experts at recognizing own-race faces versus faces of another, less familiar race (Meissner & Brigham, 2001). This expertise has been shown to be associated with between-feature (configural) facial processing, while novices rely on more featural information (e.g., Diamond & Carey, 1986). How these theories of face processing extend to the recall of facial information is largely unknown. Interestingly, the configural-featural hypothesis would suggest a difference in the accuracy/quality of recalled information by novices versus experts, but in the opposite direction of what is observed with the own-race bias in face recognition. The purpose of this research was to investigate the role of perceptual expertise in face recall and improve our understanding of the configural-featural theory of face processing underlying the own-race bias. Participants were shown a target face for 60 sec. and then completed a distractor task for 5 min. Following instructional manipulations designed to encourage configural or featural processing, participants were asked to generate a facial composite of the target face using Faces 3.0. It was predicted that (1) composite quality would be better for those based on cross-race faces than own-race faces, and (2) instructional manipulations would affect composite quality. Results indicate a floor effect in terms of composite quality. The theoretical basis of these results will be discussed in terms of constrasting processes between recall and recognition. The limitations of composite production technologies and the implications of these results for law enforcement practices will also be discussed.

The influence of instructional bias on the generation of facial composites
Christina Sendina, Christian A. Meissner

Previous research has demonstrated a verbal overshadowing effect in face identification -- the phenomenon that verbally describing a face later impairs identification of the face from a photo lineup (Schooler & Engstler-Schooler, 1990). Recent empirical research (MacLin, Tapscott, & Malpass, in press; Meissner, in press; Meissner, Brigham, & Kelley, 2001) and a meta-analytic review of previous studies (Meissner & Brigham, 2001) has indicated that "instructional bias" on the description task may be partly responsible for the overshadowing effects observed in eyewitness memory. Interestingly, the verbal overshadowing phenomenon has not been directly extended to the creation of facial composites, although an analysis across previous studies provided some evidence for verbal facilitation (Meissner & Brigham, 2001). The purpose of the present study was to assess the influence of verbally describing a face on the subsequent generation of a facial composite. Participants were shown a target face for 5 sec and subsequently completed a distractor task for 5 min. Participants were then asked either to describe the facial characteristics of the target, or to engage in a "no description" control condition by continuing with the distractor task. Instructional bias was manipulated in the description conditions as in previous studies by either heightening or lowering participants` criterion of recall on the description task. Following the description task, all participants were asked to generate a facial composite of the target face using Faces 3.0. It was predicted that the instructional bias manipulation would influence the quality of composites generated such that heightening participants` criterion on the description task would improve quality of the reproduction when compared with those participants whose output criterion was lowered, and results confirm this prediction. The theoretical basis of this effect will be discussed, and procedures for the collection of verbal descriptions from witnesses will be suggested.

Optimizing facial composite hit-rates
Graham Pike, Nicola Brace, Richard Kemp, Sally Kynan

The current paper examines how best to present facial composites in order to maximize the chances of obtaining a positive identification. It has been standard practice in the UK to obtain just a single facial composite per investigation and to only release the image if the witness rated the likeness above a pre-set threshold. Several field studies, employing live staged crimes and trained police artists, were conducted to determine whether these procedures could be improved. Initial results revealed that witness ratings are not an accurate indicator of the quality of the composite. In addition, the presentation of multiple composites of the suspect led to a higher hit rate than presenting any one image, even the best likeness. This suggests that the people viewing the composites were able to make use of the variation between the different images in order to build a more accurate description of the suspect's face. Studies were also conducted which looked at what textual information should be included along with the facial image. These showed that the inclusion of information to do with the time and place of the crime improved hit-rate, but that the inclusion of a description of the suspect resulted in a lower hit-rate. One explanation of this is that verbal descriptions tend to be more inaccurate than the visual description contained in the facial composite image itself. The results of a final study revealed that constructing a facial composite does not seem to interfere with the witness' performance at an identification procedure

Testing featural and holistic composite technologies under different methods of recall
Bhavesh Prag & Colin Tredoux

In previous presentations, we reported the development and testing of a holistic face composite system. The system (ID) uses 'eigentexture' and 'eigenshape' models to interactively construct likenesses of faces, and results thus far have been promising. In the present research we tested ID against a software program widely used by US and South African (SA) police (FACES), under three recall conditions: in view, a standard SA police interview, and a Cognitive Interview. One group of participants was exposed to one of four targets during a tutorial session, and required to reconstruct the faces from memory between one and two weeks later. A second group reconstructed the same faces with their photographs in full view. A third group of participants evaluated the composites in a range of tasks. Results suggest that the FACES program gives better reconstructions when the target is in full view, but that the ID program gives better reconstructions from memory.

KCF7
SYMPOSIA 1:20 - 2:35 p.m.

Questioning children
Chairs: HELEN WESTCOTT, ROBYN HOLLIDAY
Contact: H.L.Westcott@open.ac.uk
Keywords: Eyewitness testimony, Interviewing children, Suggestibility.

This symposium comprises four papers and a discussant exploring how best to elicit and understand information from children. Two experimental studies re-examine the misinformation effect. The first paper reviews children's difficulties in distinguishing between correct and incorrect responses to misleading and unbiased questions. The second paper queries to what extent compliance, rather than deception or secrecy, contributes to suggestibility effects. A third paper explores the content of children's verbal descriptions of unfamiliar faces in relation to constructing facial composites (e.g. E-fit). The final paper examines insights and issues associated with the application of different coding protocols to a sample of transcripts from suspected child sexual abuse interviews.

Social and metacognitive processes in children's suggestibility: Is there evidence for a general overconfidence effect?
Claudia Roebers, Pauline Howie

Two studies, conducted in Australia and Germany, investigated metacognitive monitoring in children's event memory, as reflected in their ability to give higher confidence judgments after correct than after incorrect responses. Eight- and 10- year olds answered either unbiased or strongly misleading questions about a video they had viewed two weeks earlier, and rated their confidence that each answer was correct, on a 3-point scale. In Study 1, the format of all key questions was strongly misleading for half the children and unbiased for the remaining children, while in Study 2, each child received a mixture of misleading and unbiased questions. Both studies revealed developmental progression in the ability to appropriately differentiate confidence judgments. In Study 1, 8-year-olds who received predominantly unbiased questions performed relatively well when rating confidence, but predominantly misleading questions produced serious monitoring impairments, reflected in higher confidence after incorrect than correct responses. This pattern was present in both countries. Study 2 explored whether the overconfidence in Study 1 was attributable to social factors associated with "bombardment" with misleading questions. With misleading questions now interspersed among unbiased questions, the younger children no longer showed an overconfidence effect. However, the absence of a difference between their confidence in correct and incorrect responses to misleading questions, combined with an appropriate difference in unbiased questions, indicates that they were still experiencing difficulty differentiating adequately between correct and incorrect responses to misleading questions. The results are discussed in terms of social as well as metacognitive factors influencing children's eyewitness testimony.

Re-examining the misinformation effect: Differentiating between secrecy, deception, suggestion and compliance in children
Clare Wilson, Martine Powell

The current study hypothesized that compliance and suggestibility would be positively correlated and both would be negatively correlated with deception and secrecy (that is, as the cognitive sophistication to lie and keep secrets develops, the willingness to agree with others decreases). It was further hypothesized a secret may be more resistant to suggestibility (as the child must mentally rehearse that the information is a secret and therefore have a stronger memory for that information). Two hundred children (aged 5 - 6 and 9-10) took part, in pairs, in a dance lesson. At the end of the lesson the instructor took the children`s photo and then asked the children not to tell "her boss" that she had used the camera and taken their photo (this was the secret). Immediately following the lesson, each child was asked to perform a deceptive task that was later rated. One week later the children were interviewed, either using highly suggestive questions followed by a multiple choice version of the same questions (either as the original test or modified tests outlined by McCloskey & Zaragoza, 1985) or a control test which contained multiple choice questions but no suggestive questions. The children were all finally given the opportunity to re-read each question and change any answer they thought was incorrect. Strong correlations were found as predicted. However, younger children did not show a stronger memory for the secret. There is a strong compliancy element to suggestibility and as children age they are more able to correct this compliancy tendency.

Children's verbal descriptions of unfamiliar composite faces
Carina Paine, Nicky Brace, Graham Pike, Helen Westcott

When the identity of an offender is unknown in a criminal investigation, witnesses are often asked to produce a facial description or a composite image with a police operator. However, spontaneous facial descriptions tend to be limited. Therefore, operators may use some form of cued recall, or list of adjectives, for eliciting a description. The majority of these lists are based on research which has collated adults' terms for describing faces. Although research has shown that children have difficulty understanding the meaning of adults` words and may use their own terms, to date there is no equivalent list of children's terms. Thus operators must interpret and "translate" children's descriptions. The current study therefore aimed to: Investigate the content of children`s verbal descriptions of an unfamiliar face and; establish the language and terms used by children. The study also investigated the quantity of children`s verbal descriptions of an unfamiliar face and determined: the effect of prompting children in a featural and a configural way. Thirty 6-, 8-, and 10-year olds were each shown two facial composite images. Children were asked to provide a free narrative, followed by a prompted description for each composite image. Children were finally asked to compare the two composite images. The results will include information about which facial features children provided descriptions of; examples of the terms children used the effect of describing two faces; and the effect of prompting on children descriptions. Implications for future research and practice developments in composite construction with children will be highlighted.

Questioning coding: An analysis of different approaches to coding suspected child sexual abuse interviews
Clare Wilson, Helen Westcott, Sharon Casey, Sally Kynan

Interview analysis is often conducted to find ways of improving the quality of investigation interviews. Three main types of interview analysis can be identified. First, Question Type Analysis focuses on the questions that result in the most accurate information. Second, Content Analysis focuses on the quality of the information asked for and obtained. It can further be split in two: semantic content analysis and lexical content analysis. Third, Story Narrative Analysis focuses on the overall quality of the statement obtained during the interview. How an interview is coded predicts what findings can be made, for example, in Question Type analysis it is commonly reported that open-ended questions are the most reliable. However, in Content Analysis, a passive voice open-ended question (eg "Tell me how you were touched by him") is less successful than the active voice (eg "Tell me how he touched you") in obtaining a complete account. The purpose of this study was to examine the different coding schemes and to integrate them, where possible, into the most helpful combination of factors that will allow the most complete and useful account of the complex dynamics of an interview. The present study coded the police transcripts of 20 interviews for suspected child sexual abuse. Each transcript was coded four ways using popular coding protocols for each of the four analysis types: Question type, Semantic Content, Lexical Content and Story Narrative. The results will discuss the usefulness of multiple analysis and associated issues and insights. The paper will review implications for both the coding of interviews and for the improvement of interviewing techniques.

Discussant: Robyn Holliday

Consistency of memory for traumatic events
A. E. van Giezen, Ella Arensman, Philip Spinhoven

Consistency of memory for traumatic events: A prospective and experimental study among victims of recent (i.e., within the last 2 months) physical or sexual assault, occurring outside the context of ongoing domestic violence.

"I know exactly what I did on September 11": Retrieval-based and retrieval-independent criteria in judgments of memories for shocking events
Gerald Echterhoff

Experiments conducted in Germany and in New York City investigated the criteria people use to judge how accurately they can remember their encounter with the September-11th news. An ease-of-retrieval heuristic was used only by participants in Germany six months later, when the public relevance of September 11th had declined.

Altering traumatic memories
Daniel M. Bernstein, Veronika Nourkova, Elizabeth Loftus

We falsely suggested to Russian participants that they had seen a wounded animal during one of two terrorist attacks (1999 Moscow or 2001 World Trade Center). While 12.5% of the Moscow group believed the suggestion, none of the WTC group did. These findings demonstrate the experimental malleability of traumatic memories.

Korsakoff patients' memories of September 11, 2001
Ingrid Candel, Marco Jelicic, Harold Merckelbach, Arie Wester

This study investigated whether Korsakoff patients are capable of remembering a highly emotional event and to what extent they form flashbulb memories of such an event. On two test occasions, Korsakoff patietns and healthy control participants were interviewed about their memories for September 11, 2001.

Evidence of a specific impairment of inhibition for trauma-related material in patients with PTSD
Tim Brennen, Ragnhild Dybdahl, Jasminka Jukic, Almasa Kapidzic

Patients with war-induced PTSD and a control group performed a directed forgetting task with positive, neutral and war-related words. The PTSD group recalled more to-be-forgotten war-related words, but fewer in every other condition, including to-be-remembered war-related words. This task provides a laboratory model of flashbacks experienced by PTSD patients.

Auditorium
SYMPOSIA 2:55 - 4:10 p.m.

Forensic factors in facial composites
Chair: CHARLIE FROWD
Contact: cdf1@stir.ac.uk
Keywords: Face perception and identification, Face composites.

This symposium connects research projects relevant to the construction of facial composites. The presentations focus on a range of factors likely to affect or predict composite quality. These broadly cluster into witness factors (cross race effect, verbal overshadowing and verbal description) and target factors (target delay, target salience and construction view).

Investing the cross-race effect in facial composite construction
Richard I Kemp, Jennifer S Baldwin

It is easier to recognise same-race than other-race faces. This reliable finding is known as the cross race effect (CRE) and is usually demonstrated using a recognition memory paradigm. This paper reports the results of experiments designed to investigate the CRE using a facial composite construction task.

The cross-race effect (CRE) describes the finding that it is easier to recognise same-race than other-race faces. Recent meta-analytic reviews have shown this to be a reliable effect, and it is likely that the effect is a significant factor in a number of the false convictions recently uncovered by DNA analysis. Despite this high degree of theoretical and practical importance, the underlying causes of the CRE are not well understood. To date most studies of the CRE have employed a recognition-memory paradigm to measure recognition of same-race and other-race faces. However, some real forensic tasks require witnesses to recall a face. One such task is facial composite construction, where a witnesses works with an operator using a facial composite system to construct a facial likeness of a target. It is not clear whether tasks such as this will also show a CRE. This paper reports the results of experiments designed to determine if facial composite construction is prone to the CRE.

Participants were "Asian" and "Caucasian" male and female students who watched a short video of an unfamiliar Asian or Caucasian face. These participant witnesses were then interviewed by a trained interviewer who worked with them using a computerised composite construction system (E-Fit) to construct a likeness of the target seen on video. These likenesses were then rated by a different group of Asian and Caucasian judges. Results show little evidence for a CRE in composite construction. Possible reasons for this finding and its practical and theoretical implications are discussed.

Composites and the Verbal Overshadowing Effect
Derek Carson

A number of recent studies have demonstrated that when participants are asked to provide a verbal description of a previously seen face their subsequent recognition of that face can be hindered (Schooler & Engstler-Schooler, 1990; Finger & Pezdek, 1999). This phenomenon has been termed the verbal overshadowing effect. Not all attempts to replicate these studies have been successful and Meissner, Brigham and Kelley (2001) argue that the type of instruction given to participants determines the likelihood of finding the effect. When participants were instructed to lower their criterion levels and report everything, the effect was a strong one. Investigators when conducting cognitive interviews with witnesses currently utilise a similar type of instruction.

Experiment One adopts Finger and Pezdek's experimental design to investigate if verbal overshadowing can occur when witnesses are asked to construct facial composites from memory. Twenty participants were asked to construct E-fits from memory. Ten followed current police procedures: they described a previously seen face and immediately constructed the composite. Ten had a delay of 24 hours between the description phase and the construction phase. Subsequent tests have shown that composites constructed in the delay condition were more likely to be named than their non-delay counterparts. They were also more likely to be chosen in a 2AFC test when the task was to decide which composite was most visually similar to a recent photograph of the target face. A second experiment investigates whether the benefit shown for the composites in the delay condition is likely to be evidence of a release from verbal overshadowing.

Multiple Technique Composites: the problem of facial salience
Charlie Frowd

Recent research has found that target distinctiveness exerts a strong influence on the identification of facial composites and appears to interact with the production technique employed. In this work, E-FIT and PROfit composites were better than artist-composites when salience was high, the reverse was found when salience was low. There is a significant body of research indicating the importance of facial salience in face perception. Recent work has suggested that composites of target faces previously-rated as distinctive are much better identified than composites of more average-looking faces. This notion appears to apply not only to a number of current UK composite systems (E-FIT, PROfit and Artist Sketches) but also outdated (Photofit) and developmental (EvoFIT) systems.

This research involved facial stimuli with a large difference in rated distinctiveness. A replication of the work has now been carried out with a less distinctive target set (and a more realistic delay to interview). Whereas E-FIT and PROfit composites were named about 20% on average, composites of the new target set were named at floor level. In contrast, the sketch artist's composites were consistently named for both sets (about 10%). Findings also highlight an advantage for the current systems over Photofit, and the new EvoFIT system over the other electronic variants. Overall, the work implies that knowledge of an assailant's distinctiveness is valuable for police work, enabling the selection of appropriate techniques and an estimate of composite quality.

Composites at different views
Hayley Ness

Many modern composite systems only allow construction of a two-dimensional full-face image. However, a witness will have viewed a three-dimensional moving face. As research (e.g. Thomson and Tulving, 1970) has indicated that successful retrieval is dependent on how well retrieval cues correlate with those in the original encoded experience, it is possible that this two-dimensional image will not contain enough information to ensure successful recall. Similarly, research has shown an advantage for three-quarter views in unfamiliar face recognition (e.g. Krouse, 1981; Logie, Baddeley & Woodhead, 1987; Bruce, Valentine & Baddeley, 1987). These studies found that when three-quarter views were shown at presentation, recognition memory was superior at test, compared to full-face and profile views.

We are currently investigating the effectiveness of PROfit's new female database, developed by ABM UK, which allows participants to construct a composite at a three-quarter view. In our initial investigation we found that three-quarter view composites were identified as well as full-face images, but not better. However, participants were exposed to all views of the target face for equal amounts of time. The current experiment has attempted to examine the encoding specificity principle in more detail. Participants were asked to view a 30-second video clip, however this time the exposure phase was split into 3 conditions (three-quarter-view, full-face view or all views). Participants were asked to view one clip of one target. They were then asked to construct composites in both a three-quarter and full-face view. It is expected that identification performance will be superior when the view in the exposure and construction phase match.

The effect of delay on facial composites
Stephanie Plews

UK working practices state that a witness description of an offender should be given within 24-36 hours. This study investigated whether there would be a significant difference in recognition between facial composites made 2 days or one week after viewing a target. Only images that were high scoring on a familiarity scale were used as targets and participants were screened so only those unfamiliar with the targets constructed composites. Naming rates suggest that there is no detriment for recognition of a composite between these two delay intervals.

This study was undertaken to assess the effects of witness' ability to successfully recall the facial appearance of, and produce a recognisable facial composite of, a target image of a white male. Six 'target' famous male images were selected and rated for familiarity and distinctiveness. Only images that were high scoring on a familiarity scale were used in the experiment. Participants were screened so that those who were unfamiliar with the 'target' images constructed composites using the standard cognitive interview technique employed by police officers. Half of the composites were constructed following a two-day delay and half after a week delay from viewing of the target. A further group of participants were asked to attempt to name the composites and likeness ratings were also collected as to the similarity between the target image and the resulting composite. The number of composites that were correctly recognised was low overall and preliminary results show no significant difference between a delay of two days and one week.

KCF7
SYMPOSIA 2:55 - 4:10 p.m.

The role of social and emotional factors
on children's eyewitness testimony
Chair: LAUREN R. SHAPIRO
Contact: shapirol@emporia.edu
Keywords: Cognitive development, Eyewitness testimony.

Researchers have shown that testimony by child witnesses is affected by social pressure, interrogation techniques, and emotionality of the event. False information presented before or after the witnessed event is incorporated into reports. Consequently, even when children are the only witnesses, they are rarely asked to testify in criminal proceedings. The first presentation examines the extent to which peer conversations can affect children`s memories for earlier experienced events. The second presentation explores how the combination of pre-event character stereotyping and postevent suggestions influenced recall accuracy over time. The third presentation focuses on whether emotionally negative events are recalled better than neutral events over time.

Making memories: How peer conversations can lead to false reports