How can you study eyewitness testimony

The first concerns the way memory is studied and discussed. In the author's opinion, contradictions are essentially the result of excessive simplifications in the presentation of research results, and of too far-reaching conclusions drawn from examination of a particular memory process.

It seems that the influence of emotions on memory may not be broad and general, as it is sometimes portrayed. The impact seems to vary depending on memory function under investigation recollection and recognition , the paradigm used to test memory free recall, structured recall, etc. Considering those nuances in study reports should be a common practice rather than a brief element of research discussion.

The second issue concerns emotions. While analysing research papers on the relations between affective states and memory performance, one may conclude that the vast body of theoretical and empirical studies on emotions are taken into account insufficiently.

Thus, this paper will discuss three major issues concerning the way emotions are addressed and investigated in eyewitness testimony research, and which may be why we face difficulties when comparing study results. These concerns involve the way emotions are 1 defined, 2 induced and 3 measured in laboratory settings. It is a convenient truism that emotions are difficult to define. As interest in research on emotions intensified in the late s, almost every major theorist developed a distinctive concept of what emotions are and how they differ from other related affective states, such as moods or attitudes.

The wealth of terms and definitions should not, however, be an excuse to ignore the body of knowledge related to the theory of emotions.

Typically, most research investigating human emotions falls into one of two theoretical categories—discrete or dimensional. The first approach views emotions as specific affective states that can be labelled e. Even though it cannot be said that one or the other is more or less correct, the dimensional approach is more likely to focus on the fundamental organisation of emotions Scherer, When this approach is employed, at least two affective dimensions arousal and valence should also be considered.

However, studies are typically focused on one or the other Bradley, The lack of precision in defining the area of interest is noticeable in eyewitness testimony research. A wide range of terms is used to name affective states, such as negative emotions, arousal or stress. There is often no reflection on what these terms imply, or on the theoretical methodological consequences of their use. When we look more closely at research subjects and objectives, we can see that the terms chosen for independent variables are rarely set within the theoretical framework.

Furthermore, they only seemingly relate to the approaches described above. In consequence, as the concepts are used arbitrarily and thoughtlessly, some studies answer questions other than those which they ask. When using the term negative emotion, emphasis seems to be on the valence—the subjective meaning an individual gives to a stimulus. However, eyewitness testimony research commonly uses the term primarily to investigate the discrete emotions—fear, anger, anxiety—not the pleasant—unpleasant dimension.

The term also seems to describe the implicit negative tone of the criminal event, not the affective state of an eyewitness per se. However, a research problem formulated in this fashion entails the risk of investigating artifacts. In quantitative studies, the impact of the first may be neutralised by the very different effect caused by the latter.

Anger serves as preparedness for attack, as motivation to approach the stimulus. In consequence, while the subject experiencing anger may be strongly focused on the stimulus that caused it, the other may divert their attention away from the stimulus.

Thus, the inconclusive study results. It is also reasonable to ask which of these discrete emotions or rather, what kind of their configuration reflect the actual experiences of eyewitnesses and can, therefore, serve as the best analogy of those experiences. Seeing a verbal, non-physical attack may cause anger rather than fear.

Thus, when adopting the discrete approach, it seems crucial to analyse the impact of emotions on testimony in a more nuanced way. On the other hand, if the concept of negative emotion is used in terms of the dimensional approach, it may have different implications for research design.

While drawing attention to the subjective valence of emotional experience, the next logical step would be to examine how memory is affected depending on stimulus assessment on a scale from unpleasant to pleasant. Moreover, as the psychology of eyewitness testimony is interested in unpleasant events, it would be ideal to differentiate the level of unpleasantness of stimuli by manipulating it from the least unpleasant to the most unpleasant. This approach would facilitate accurate and appropriate investigation into valence.

However, to examine how negative emotions impact memory, it is crucial to create an event that truly evokes the required emotional reaction. Given ethical constraints, laboratory experiments may not include a stimulus that evokes negative emotions—only one that does not evoke positive ones.

There is also a possibility that the reaction of a subject exposed to a stimulus is not at all an emotion but a startlement or surprise, which are, for some scholars, instinctive reactions e. To sum up, it is suggested that, if the term negative emotion in fact applies the discrete approach, a study should examine the variations between the influence of different types of emotions on the reliability of eyewitness testimony. In addition, research into the valence dimension should focus on how experiences evaluated as unpleasant are remembered, compared to neutral and pleasant ones.

Besides negative emotions, a term frequently used in eyewitness testimony research to describe an affective state is arousal. Arousal, along with valence, is considered a core dimension of emotional experience.

The ability to evoke arousal distinguishes between emotions and other affective states, e. It is a psychophysiological reaction to a stimulus, representing the activation of the autonomic nervous system ANS.

Thus, a study involving arousal may examine an objectively measurable state, not only an individual's subjective assessment. It is even more important, as some studies show, that self-reporting of arousal may differ from an objective measurement of ANS activation e. In eyewitness testimony, research into arousal is often although not exclusively associated with the weapon focus effect.

Thus, even though arousal is a neutral state, it is in eyewitness testimony research sometimes defined as negative e. This implicit negative valence may explain why, in contrast to general research on the arousal—memory relationship that suggests increased accuracy of recollection of arousing stimuli e.

Bradley et al. However, it is important to underline alternative explanations that focus on the differences between central versus peripheral details e. Kensinger, , as well as the interview procedure e. When analysing studies on arousal and eyewitness testimony, we can also notice that arousal is often treated as a nominal variable. Yet, arousal as a physiological reaction can be accurately measured.

Thus, research does not need to limit its inference to a reduced problem. The issue of objective measures of emotional experience is further elaborated in the subsection on emotion measurements. To sum up, it is postulated, that research should, in terms of arousal, focus primarily on examining how different levels and patterns of ANS activity affect the processing of visual information and its coding in long-term memory.

It is also important to keep in mind that arousal is fundamentally neutral, but can be evoked by stimulus with negative, neutral and positive valence. The lack of precision in the use of terms in eyewitness testimony research is even more evident when we look at the term stress. One may feel that research on the eyewitness testimony mimics trends in society at large, in which the term is applied to describe very different mental states—anxiety, fear, frustration, fatigue, sense of overwhelming or even anger.

However, in eyewitness testimony research stress seems to describe properties of stimulus rather than eyewitness affective state. However, this stimulus-based conceptualisation of stress is questionable, as it ignores the core concept of psychological stress—individual differences in coping with stimuli.

Applying the concept of psychological stress scientifically has its theoretical connotations and methodological consequences. Therefore, the term stress should be used neither to describe a specific external stimulation as it is not synonymous with negative stimulus nor to define a specific pattern of emotional, physiological or behavioural reactions. Even though emotions are the focal point of the paper, the issue of chronic moods requires a brief comment, as it is also a subject of research on eyewitness testimony.

However, it is crucial to understand the difference between moods and emotions. While the latter are considered short, rapidly transient and intense affective states caused by a specific internal and external stimulus, moods are harder to define in terms of specific cause. As Frijda , p. Despite this clear distinction, many empirical studies use these concepts interchangeably. It is quite possible that arbitrary application of these terms may be a legacy of research on the mood-state dependent effect.

Since imagination guided by hypnotic suggestion was employed to induce them, mood is the term which seems more appropriate to describe the experience.

Thus, it is postulated to use distinctive terms for short affective states evoked in response to specific stimulus, and for chronic, general, diffuse affective state, which can be considered a disposition to experience certain emotions mood-trait or a long-lasting affective state mood-state. As some evidence suggests that chronic moods may also influence eyewitness testimony, thus control for a dispositional mood can be also considered. To sum up the deliberations on defining emotions, it is important to highlight the misuse of terms and concepts of emotions, with little regard to theory and achievements of research on emotions.

Thus, it is postulated that a study should define the spectrum of affective experience precisely and in accordance with the chosen theoretical approach. It must go hand in hand with awareness of the consequences of that choice, imposed by the research subject and methodology.

Moreover, although both approaches to study emotions have their justification in the theory of emotions, it seems that in the case of eyewitness testimony research, the most appropriate would be to combine both. On the one hand, it would allow us to indicate what feelings accompany witnesses depending on the type of crime observed discrete approach. On the other, dimensional approach may help to determine the characteristics of the emotional stimulus which seems to be particularly threatening the reliability of the testimony.

Another area of doubt concerning research on eyewitness testimony is related to experimental manipulation, particularly the methods of inducing affective states. In experiments conducted in psychological labs, short films or movie clips presenting a simulated crime are the most frequently employed emotive stimuli. In fewer instances, a crime is staged, making the subjects real eyewitnesses to a false event. Researchers also use slide presentations with short narratives or photographs.

Other methods of inducing emotion include exposure to aversive stimuli, such as threat of an injection or mild electric shocks see Deffenbacher et al. In research on chronic mood and its impact on eyewitness testimony, a different approach is often employed. Moods may be induced by techniques such as hypnotic suggestion, instruction to recall past events, reading statements related to certain moods e.

In other cases, a different methodological approach with quasi-experimental research and subjects recruited on the basis of their dispositional mood or mood disorders e. While focusing on emotions, the fundamental question when evaluating research design is whether such experimental manipulations are sufficient to evoke real emotions, similar to those experienced by eyewitnesses and consequently, the extent to which experimental studies can be generalised in the forensic context.

As Yuille and Tollestrup argue, a typical laboratory eyewitness is a passive observer, not experiencing a sense of danger or involvement in the event. Their physical and psychological well-being is not under threat. Thus, the behaviour of witnesses in this setting is not representative of that of witnesses to actual crimes. With this in mind, one should consider what type of stimulus manipulation increases the chances of evoking real emotions, not just a declaration of emotions.

Thus, in order to induce states as similar as possible to those experienced by real eyewitnesses, the distance both physical and psychological between subject and stimulus should be minimised. However, it may be difficult to achieve through short films, presentations or photos presented on a TV or computer screen.

Although there is evidence that films can induce emotions, and they are widely regarded as ecologically valid stimuli, they must meet several conditions. After examination of movie databases serving as archives of stimuli useful in laboratory studies e. The duration of the video clip may also be meaningful. Footnote 2. The hypothetical mechanism explaining how movies can influence emotions, and in consequence beliefs, attitudes, or behaviour, is the transportation into narrative effect.

This describes psychological immersion into a narrative, and its effect is particularly pronounced when the emotional state of individual pre-reading or pre-watching content is consistent with the emotional tone of the narrative e. It seems that videos capable of inducing this effect have a greater chance of evoking emotional experiences comparable to those of eyewitnesses. However, it is difficult to meet these criteria with videos lasting 30—40 s, devoid of context and lacking vivid, identifiable characters with whom a viewer can form an empathic bond, which seems to be an important component of an emotional experience of a film Tan, It would also be good practice to assess the film before the experiment in a pilot study or after the presentation manipulation check , to investigate its impact on at least three-dimensional space, which represents the valence of the stimulus, the intensity of subjectively perceived excitement, and the dominance the degree to which the film was involving and allowed subjects to detach from external stimuli.

Another way to simulate eyewitness experience in the safe and controllable setting of a laboratory is to use modern technologies. Virtual reality VR with high-quality equipment provides a great opportunity for research into memory, as it allows the creation of a complex and rich environment that is fully under the control of the researcher. As goggles cut off external stimuli, VR minimises the distance between the observer and the scene, which increases immersion and allows transportation into a fictional event.

Moreover, as many studies show, emotional, behavioural, and social reactions of people are the same in virtual reality as in everyday life e. Gamberini et al. With today's technological capabilities and relatively inexpensive equipment, VR may serve as a substitute not only for insufficiently involving films or video clips, but also field studies, hazardous because of possible uncontrolled variables affecting the result. This method of experimental manipulation is nothing new.

It has been successfully employed in many fields, eyewitness testimony included e. Kloft et al, A particularly interesting approach is to use tools that are already utilised in law enforcement training, based on real-life simulators or large-screen video projections, not a programmed, thus artificial environment e. Provided they use equipment at the highest technological level, they can be highly realistic and involving, and thus, simulating a real experience.

It seems that the most reliable way to simulate the experience of real eyewitnesses is to expose subjects to staged crime. Even though the crime is faked, an individual may for a moment experience real emotions, thus it meets the criteria of high ecological validity. This kind of experimental manipulation is particularly useful in group conditions, where all participants watch the same event. Considering the high demands on the sample numbers, it may be harder to reproduce an identical event for a single participant.

When deciding on this type of experimental manipulation, it is crucial to consider all ethical concerns as well. To sum up the subsection on inducing emotion in a laboratory setting, it is crucial to underline the need to simulate experiences resembling at least to some extent those of witnesses to the crime.

This can be achieved when experimental manipulation meets several criteria: 1 the stimulus is subjectively meaningful for the subjects; 2 psychological and physical even imagined distance between stimulus and observer is short; 3 regardless of the chosen way of presenting the crime event, subjects find it absorbing, 4 emotions and involvement are meticulously examined.

This approach avoids the risk of drawing incorrect conclusions about the relationship between emotions and memory based on ineffective stimulus manipulation, or on one that induces states other than those intended. The choice of theoretical background in research on the impact of emotions on eyewitness testimony should have consequences related not only to how research questions are formulated and hypotheses tested, but also to the way in which the emotions are assessed and measured.

Measures of the effects of an emotion can be generally categorised as objective behavioural, physiological or subjective self-reports. The latter, in the form of quantitative questionnaires, are the most common in eyewitness testimony research. They are used primarily due to convenience, arising from the ease and speed of collecting data and the lack of additional costs. However, as is often pointed out e. Subjects may not only be unaware of their emotions, but could also intentionally or unintentionally, mislead the researcher as they try to fulfil the internalised expectations of the research.

Another concern about self-reports is related to their structure. The choice between the dimensional versus discrete approach should be reflected in a properly chosen method of assessment. When it comes to the discrete approach, some doubts may arise when the list of labels is short, forcing an individual to choose one of them, even when they are uncertain about their own emotions. For instance, Yuille and Cutshall asked eyewitnesses of real crime to assess their level of stress on a seven-point scale, and to indicate any negative effects engendered by the incident nightmares, sleeplessness.

While the question of negative consequences may be a useful counterbalance for self-reports, limiting the questionnaire to one scale may have resulted in discovering artifacts. When subjects have only one option, they feel obliged to report anything, even if they perceive the event differently than assumed by the researcher. Thus, it is postulated, any research in eyewitness testimony which assumes a priori that the results of stimulation are negative should also give subjects sufficient options to assess a wider spectrum of emotions, positive included.

The experimental situation itself can be seen as novel and interesting, and some subjects in contact with unpleasant stimuli may react with a mixture of excitement and curiosity. Still on the issue of scale, concerns about the psychometric properties of Likert-type scales depending on rating format are also worth noting. As there is no room here for a detailed discussion, readers seeking a more advanced overview are directed to other papers focused on the issue e. However, it should be mentioned that there is substantial evidence showing that, as the internal structure of the scale does not lose its properties regardless of the number of points, the more points a scale has, the more skewness and kurtosis is reduced.

Therefore, it is often suggested not only to abandon five-point scales, but even to use points scales that increase sensitivity and follow the normal distribution. In light of this, methodological considerations may be formulated in relation to studies that rely on scales with a small point range. Hypothetically, when a four-point scale is used e.

Thus, abnormal data distribution is highly probable. Consequently, when our objective is to examine if there are differences in the assessment of emotions between conditions, we should not use parametric estimation tests unless our sample size is big enough to be resistant to non-normality. The T test, for instance, is invalid with small samples from non-normal distributions, and by using it we risk falsely rejecting the null hypothesis.

On the other hand, non-parametric tests are underpowered to detect an effect, so they increase the risk of accepting a false null hypothesis Conover, Moreover, it is postulated to include the measurement of chronical affective states in research plan.

As mentioned, moods can influence the formulation of testimony and, due to mood-congruency effect, enhance the impact of experimental manipulation. Thus, controlling for that variable will help us estimate the interactions between emotions and moods, as well as eliminate the disruptive impact of opposite mood on stimulus manipulation. Questionnaires that allow diagnosis of dispositional mood, affective style or mood disorder are also recommended, as they can be help controlling for individual differences.

Another way to assess emotions uses objective measurements of physiological correlates of affective states. Emotions are accompanied by physiological arousal, which reflects the activation of the autonomic nervous system. Each has multiple indexes, which are evidence of ANS activation and allow the conclusion that a subject is experiencing not only general arousal, but even emotions.

Thus, when adopting more than just one method to investigate autonomic response patterns, it may be useful to discuss not only the dimension of core properties of emotional experience, but also the type of discrete emotion. However, it is crucial to note that consideration of only one index is insufficient to indicate what kind of emotion is experienced.

Moreover, individual differences in ANS activation should be always taken into account. Taking into consideration the shortcomings of self-reporting, as well as the nature of emotions and their dimensions that allow us to objectively measure some aspects of emotional experience, it is postulated that, in the case of research on how emotional events are remembered, a heterophenomenological approach should be adopted.

This indicates both subjective self-reports and objective psychophysiological assessment of investigated variables, which may allow us to address discrepancies in previous research. When psychophysiological measurements are included, this may provide us with information on the nature of emotional engagement, demonstrating the effectiveness of experimental manipulation. Thus, it may also serve as an additional manipulation check.

On the other hand, using self-reports gives us insight into subjective experiences, personal appraisals and assessments. The purpose of this paper is to present the theoretical and methodological shortcomings of research into the relationship between emotions and eyewitness testimony.

In light of contradicting results, it is crucial to indicate those areas of research that may be responsible for misrepresentation of memory performance. The summary of the discussion on emotions and their impact on testimony constitutes a proposition of a framework that would allow better comparisons between studies and explanations of conflicting results, while structuring knowledge that will eventually provide a complete theory of the influence of emotions on eyewitness testimony.

The framework consists of the following steps:. Even though there are no reasons to consider the discrete approach as theoretically unsound, the advantage of the dimensional model of emotions is primarily due to the methodological consequences it imposes.

When both valence and arousal are considered, the researcher is advised to apply two different methods of measurement—self-reports with a multi-point Likert-like scale to assess valence, and psychophysiological measurement to study arousal. Psychological scientist Elizabeth Loftus studies memories. A two-part 60 Minutes news story focusing on the case of Ronald Cotton and Jennifer Thompson, one of the best documented cases of false conviction. Extensive interviews with the people involved in the case as well as Elizabeth Loftus and Gary Wells.

Scientific American Mind. Website of Dr. Gary Wells , who has done extensive research on the validity of police line-ups.

His website is a wealth of information, links, and videos. Eyewitness Testimony and Memory Biases. Benton, T. Eyewitness memory is still not common sense: comparing jurors, judges and law enforcement to eyewitness experts.

Applied Cognitive Psychology, 20, — Hastorf, A. They saw a game; a case study. The Journal of Abnormal and Social Psychology, 49 1 , Loftus, E. Zaragoza, M. Misinformation Effects and the Suggestibility of Eyewitness Memory. Hayne Eds. Students may bring responses to class or post them online.

Begin by posing the myth, shown on Slide 1 of the accompanying PowerPoint slides. Eyewitness Testimony slides. Next discuss the assignment. Have students discuss their responses.

Different people can see the same event and come away with very different memories. This is a good entry point to discuss the nature of long-term memory and how it is both constructive at encoding and reconstructive at retrieval. The instructor can discuss work on the misinformation effect and eyewitness testimony here. Schema theory can also be discussed.

Slides 5 and 6 review the myth and the reality of memory. The goal of this day is to give an overview of how memory works. There are various ways of organizing the memory unit: historically from Ebbinghaus to current models, or focusing on a particular model such as Information processing almost 50 years out of date, but still a powerful way of organizing concepts.

Select an organization that makes it easiest for you to discuss key memory concepts. Use Activity 3 see below to show how quickly and easily memory can be distorted. The balance of the class time can be spent discussing applications of memory research.

Possible topics are listed in the concept map on Slide 7. All of these topics have important implications. Students often appreciate learning how to leverage memory research to improve their study habits because it directly relates to their current experience. Instructors may want to refer students to my video series on how to study effectively www.

Concept Checks The following formative assessment questions allow both you and the students to gauge their level of understanding of the unit. Be sure to have students respond individually so both you and they can see how well the class understands the concepts. Then they can discuss as pairs or as a class. There are a variety of high tech clickers and low-tech fingers methods that can be used to administer these concept check questions.

How Bias and Expectations Shape Perception and Memory For this assignment, you will listen to an excerpt from a podcast that tells the story of the most famous football game in psychology. What makes it famous? You will have to listen to find out, but it is directly relevant to our next unit. Below is the link to the podcast. Start and end at the designated times. Of course, you are welcome to listen to the second half as well if you are interested.

Reflection questions Write a paragraph in reflection to each question. Bring your reflections with you to class. We will be discussing them. Eyewitness Memory Activity This activity introduces students to the challenge of accurate eyewitness testimony and the misinformation effect. The students will watch a video of a bicyclist assisting police in chasing a thief. They will then be asked questions about the video. Some of the questions contain misleading post-event information MPI.

Then, students are asked about the presence or absence of certain details in the video, some of which are present and some of which are absent. Students can see their results, and the class results can also be compiled. This activity follows a typical misinformation effect paradigm: subjects witness an event, are introduced to a mix of accurate and inaccurate post-event information, then tested for the accuracy of their memories of the event.

The MPI is introduced through leading questions. In the post-event information, half the items mentioned were present in the video and half were not. Then during the detailed recall portion, half of the items listed were also listed in the post-event information and half were not.

Although it has been a matter of some debate over the years, the now prevailing view is that there is a consistent positive, albeit not perfect, relationship between confidence and recognition accuracy Wixted et al. Confidence has also been a prime interest in studies on verbal eyewitness recall, such as eyewitness testimony. While the strength of the relationship between confidence and accuracy in witness recall has varied somewhat throughout studies, the overall trend is consistent with, and mirrors the results of recognition studies; people are more confident in recalled memories that are correct, compared to incorrect Robinson and Johnson, ; Robinson et al.

Moreover, while confidence may be based on the indirect accuracy of cues, it seems plausible that the cues people rely on are not always those that are the most accurate predictors. One cue that has been found to predict both accuracy and confidence is response latency, that is, the speed with which a memory is produced.

As shown by Kelley and Lindsay , people are more confident in quickly produced as compared to more slowly produced verbal responses. The same results were obtained in a study by Robinson et al. Higher confidence and shorter response latency for correct answers was found both for verbal recall as well as for recognition judgments.

The relations between confidence, response latency and accuracy demonstrated in these studies in recall of episodic memories, are consistent with findings from a body of research on recognition of verbal information Koriat and Ackerman, ; Ackerman and Koriat, , semantic memory recall Smith and Clark, as well as in eyewitness identification studies e.

Given the evidence that memory accuracy is related to retrieval ease as measured by response latency, other cues of the ease with which a memory is retrieved should also predict accuracy. Lindholm et al. In two studies, participants were interviewed about their memory of a simulated crime event.

In transcripts of these interviews, measures of effort were obtained by identifying a number of cues indicating retrieval difficulty. These effort cues included delays pauses between or within statements , hedges , that is, commitment avoidance e.

While witness confidence was found to be positively related to accuracy, confidence did not contribute with any unique variance in predicting accuracy when the effort cues were included. Moreover, the effort cues fully mediated the relationship between confidence and accuracy, supporting the notion in cue-utilization theory that confidence is based on cues during memory retrieval, rather than a direct monitoring of memory strength Koriat, , The finding of new, objectively verifiable cues that may be linked to eyewitness accuracy constitutes an important first step for developing methods to improve evaluations of eyewitness memory.

However, before initiating attempts at methodological development, it is essential to further test the replicability of these initial findings. It seems possible that the exact latency a continuous measure before initiation of a response is a more fine-tuned and better predictor of memory accuracy than a courser delay discrete measure, and that such a latency measure may even make other effort cues redundant.

On the other hand, while response latency gives the exact timing before response initiation, pauses and hesitations during the response are not included in this measure.

As memory retrieval is rarely instantaneous, but often unfolds as the memory is reported Clark and Tree, ; Warren, , delays during a response could also be critical cues to retrieval effort, and carry information about memories correctness.

Thus, the role of response latency vs. The aim of the current study is to test the robustness of the Lindholm et al. Based on their results, it is hypothesized that retrieval effort cues i. We further expect that confidence will not provide unique variance in predicting accuracy once the effort cues are accounted for. Extending the previous findings, the current study also measures the effort cue response latency and explores the contribution of this factor relative to the other effort cues in predicting accuracy.

As the theoretical assumption from cue-utilization theory is that confidence is based on cues rather than derived from memory accuracy directly, we examined whether effort cues mediated the relationship between confidence and accuracy. Participants were informed that they were to see a simulated crime event on video, and that they would later be videotaped while being asked questions about the event.

They all gave informed consent to participate. The materials and procedures were identical to those carried out by Lindholm et al. Participants were tested individually in the lab, where they watched a 1-min film sequence involving a staged crime on a computer monitor.

The film initially shows a man waiting at a bus stop. Shortly thereafter, a second man approaches the first man, attacks and stabs him in the gut, before leaving.

After seeing the film, participants were interviewed about their memory of the event. The interviews included a free recall phase, immediately followed by a cued recall task with open questions e. Since the details reported by the witness were noted during an ongoing interview, it was not possible for the interviewer to catch every detail. We asked for confidence after the interview had finished to allow witnesses to make a focused memory search without being interrupted repeatedly.

This also allowed us to better mimic a free-recall situation similar to that typical of eyewitness testimony. As we were interested specifically in cues to accuracy in memories of individual details, rather than in overall accuracy, witnesses did not provide overall confidence estimates, neither in free nor cued recall. Based on the information in the crime video, we first cataloged all scorable and objectively verifiable details. Statements including partly correct and partly incorrect information e.

Given that questions in the cued recall phase sometimes asked for a detail the participant had mentioned during free recall, we focused on responses during cued recall to avoid associating the same confidence score to two different reports of the same information. This yielded a total of correct answers and incorrect statements.

Of these, confidence was obtained for correct and incorrect statements. To make our results section less convoluted, we focus our analyses only on statements for which confidence ratings were made.

Next, two new blind coders coded the frequency of verbal and paraverbal expressions of effort in in each statement. Both coders coded the entire set of statements, and inconsistencies were resolved by a third coder. Using the operationalizations by Lindholm et al. We also measured Delays — a pause longer than 2 s before or during a response. Finally, we measured a fifth effort cue, Response latency see Table 1.

Both response latency and delays were measured using the video editing software iMovie version The interviews of the participants were loaded into the program, and elapsed time was obtained by computing the temporal distance of silences between utterances as indicated by sound wave intensity. Hence for these cues, interrater reliability was not measured.

Table 1. Mean amounts of effort cues and confidence z- transformed in accurate and inaccurate statements for each variable are presented in Figure 1. Figure 1. Mean amount of retrieval effort cues and confidence z- transformed in correct and incorrect memories. As the design used repeated measures all participants provided both correct and incorrect responses , in combination with a varying number of responses produced by different participants, data were therefore organized as a multilevel data set with individual responses nested within participants Wright and London, Our analyses largely followed the procedure outlined in Field and Mansour et al.

Hence, we first ran a set of regressions to examine which individual variables predicted accuracy. Thus, a baseline, intercept-only model predicting accuracy Model 1 was compared with models including each effort cue and confidence separately Models 2—7. Table 2 illustrates the model parameter estimates and fit indices. In this table, effect sizes are given as Akaike Weights. The Akaike Weights varies between 0 and 1 and estimate the probability that the chosen model is the best-fitting model, relative to the other model s Burnham and Anderson, ; Wagenmakers and Farrell, Hence, larger values indicate better fit.

Table 2. Parameter estimates for predictors in models of accuracy observations. We next examined whether a model including all the significant variables from the first set of regressions improved fit relative to each of the separate models with significant predictors. Because delays and response latency were both significant, but partly based on the same data a 2-s pause before the beginning of a statement would be coded both as latency and as a delay , we first needed to determine which of the two would be optimal in a model including all significant variables we also checked for multicollinearity between all cues, and only response latency and delays were at risk, see Supplementary Table 1.

To assess which model had the best fit, we compared Akaike Weights for each model. In the subsequent analysis, therefore, we used the model with Hedges, Delays, Word Fillers, and Confidence and compared it to the models with each significant predictor. Table 3. Multilevel logistic regression analysis predicting response accuracy from effort cues and confidence z -transformed. In the final analysis, we examined the role of effort cues as mediators of the relationship between accuracy and confidence.

For this analysis, we created an effort index by summarizing hedges and delays, the two effort cues that uniquely predicted accuracy. The mediational analysis was run using the mediation Tingley et al. Results showed that the effort cues partially mediated Figure 2. Effort index as a mediator of the relationship between accuracy and confidence.

Values represent unstandardized parameter estimates for each path. Along the path from accuracy to confidence the numbers in parentheses represent the coefficients when the effort index was entered into the analyses. Dashed line indicates that the direct path is significantly mediated by the indirect path. The datasets analyzed for this study, and the code for the analyses, have been deposited in the Open Science Framework. The aim of this study was to further explore previously demonstrated relations between eyewitness accuracy and cues to retrieval effort Lindholm et al.

Our results largely replicate previous results, providing additional support for the use of effort cues in estimating eyewitness accuracy. Looking at the relationship between accuracy, effort cues and confidence, we found that effort cues partially mediated the relationship between confidence and accuracy Figure 2.

This study also measured the effort cue response latency, and found, in line with previous studies Brewer et al. However, a coarser, but more inclusive temporal measure of delays pauses before and during a response was a better predictor of accuracy than response latency. Out of the five effort cues examined in this study, four hedges, delays, word fillers, and response latency were significantly related to memory accuracy, but non-word fillers was not.

Thus, our results largely mirror our hypotheses, as well as the results obtained by Lindholm et al. These results pointed in the same direction for all the cues, as correct statements contained fewer cues to retrieval effort compared to incorrect statements see Figure 1.

Furthermore, in the current study, hedges and delays proved to be unique predictors of accuracy.