Confidence-accuracy Relationship Research Articles

Eyewitness Confidence Does Not Necessarily Indicate Identification Accuracy

Eyewitness identifications are prone to error. Scholars and legal stakeholders are exploring whether and when eyewitness confidence predicts identification accuracy. Scholars agree on a strong but imperfect relationship between initial confidence and accuracy under “pristine” (ideal) conditions. However, pristine conditions are not unfailingly present in criminal cases. Even when pristine conditions are present, high confidence is not always associated with high accuracy because other factors influence the confidence–accuracy relationship. Researchers have not yet reached a consensus about the many variables that may impact the confidence–accuracy relationship. Ultimately, legal practitioners should be cautious in using confidence as an indicator of accuracy.

Sensitization instructions can reduce the misinformation effect and improve the eyewitness confidence–accuracy relationship.

Sensitization instructions can reduce the misinformation effect and improve the eyewitness confidence–accuracy relationship.

Supplemental Material for Sensitization Instructions Can Reduce the Misinformation Effect and Improve the Eyewitness Confidence–Accuracy Relationship

Supplemental Material for Sensitization Instructions Can Reduce the Misinformation Effect and Improve the Eyewitness Confidence–Accuracy Relationship

On the confidence-accuracy relationship in memory: inferential, direct access, or indirect access?

On the confidence-accuracy relationship in memory: inferential, direct access, or indirect access?

0058 The sleepy eyewitness: Self-reported sleep predicts eyewitness memory

Abstract Introduction Sleep is critical for memory, but research indicates that eyewitness identification may be an exception. We conducted three experiments to assess the effect of self-reported sleep duration and quality, and current sleepiness on eyewitness recall for a mock crime and suspect identification. Experiments 1 and 2 looked at pre-encoding sleep on lineups and showups, respectively. Experiment 3 investigated lineups across two days to assess retrieval effects. Methods Participants viewed a mock crime video then answered sleep questions as well as central and peripheral recall questions over the crime. Participants were then randomly assigned to view a fair 6-person target-present (TP) or target-absent (TA) lineup for experiments 1 and 3, or a TP or TA showup for experiment 2. In experiment 3, the recall questions and lineups occurred on day 2, but the sleep questions were asked both days. Decision confidence was provided after the identification procedure. Results Sleep quality predicted central recall for experiment 1 [F(1, 3834)=12.96, p<.001] whereas sleep quantity predicted central recall for experiment 2, F(1, 2712)=36.22, p< .001. Choosing rate for pre-encoding sleep was higher for low quality [X2(1)=29.10, p<.001] and low duration [X2(1)=20.15, p<.001], but false IDs were lower for high duration [X2(1)=29.76, p<.001] and high quality, X2(1)=35.47, p<.001. High quality self-reported sleep increased discriminability for lineups, G=2.14, p=.016. False IDs for showups were lower for high duration [X2(1)=5.27, p=.024] and all three sleep questions predicted discriminability for showups: more sleep (G=2.85, p=.002), better sleep (G=2.60, p=.005), and less sleepiness, G=2.44, p=.007. Showups had greater accuracy (p=.038) and a stronger confidence-accuracy relationship (p<.001) for the average of the “good sleep” conditions than the average of the three “bad sleep” conditions, p=.022. In experiment 3, involving post-encoding self-reported sleep, there were still effects on recall but minimal effects on eyewitness identification (and no effects on discriminability). Conclusion Self-reported sleep for the night prior to the crime could be an important individual difference variable for eyewitness memory, affecting encoding more than post-encoding processes. Support (if any) This project was supported by Grant 2018-R2-CX-0027 awarded by the National Institute of Justice, Office of Justice Programs, and U.S. Department of Justice.

Police trainees versus laypeople: Identification performance and confidence–accuracy relationship for facial and body lineups

AbstractCourt instructions and public perception endorse that eyewitness evidence provided by police should weight more heavily than laypeople's in court. Evidence is inconsistent. The current experiment provides a nuanced analysis of identification performance of police and laypeople at different levels of confidence. Laypeople and advanced police trainees (N = 192) viewed portrait, profile, and body‐only lineups for central and peripheral targets. Police trainees displayed higher hit and correct rejection rates than laypeople for portrait lineups, and higher correct rejection rates in profile lineups for central targets. Calibration was similar for both groups, although police trainees had an advantage at low target presence base rates. Calibration was best for central targets' portrait and profile lineups. Participants displayed poor calibration and strong overconfidence for body‐only lineups and peripheral target lineups. We conclude that experience and specialization of police might be important when investigating a possible superiority of police who serve as eyewitness.

The confidence-accuracy relationship for lineup decisions holds for the Dutch identification procedure.

Post-decision confidence from witnesses who make a positive identification decision can serve as a valuable indicator of identification accuracy under certain conditions. International best-practice guidelines therefore recommend asking witnesses to indicate their confidence following a selection from a lineup. Three experiments that used Dutch identification protocols, however, reported no significant post-decision confidence-accuracy association. To examine this conflict between the international and the Dutch literature, we tested the strength of the post-decision confidence-accuracy relationship for lineups that followed Dutch protocol in two ways: we conducted an experiment and re-analyzed two experiments that implemented Dutch lineup protocols. As expected, the post-decision confidence-accuracy relationship was strong for positive identifications and weak for negative identification decisions in our experiment. The re-analysis of the pre-existing data showed a strong effect for positive identification decisions of participants up to the age of 40 years. For exploratory purposes, we also tested the confidence-accuracy relationship between lineup administrators' perception of witnesses' confidence and eyewitness identification accuracy. In our experiment, the relationship was strong for choosers and weak for nonchoosers. The re-analysis of pre-existing data showed no correlation between confidence and accuracy, unless we excluded adults over 40 of age. We recommend adapting the Dutch identification guidelines to reflect the current and previous findings on the post-decision confidence-accuracy relationship.

Warning weakens retrieval-enhanced suggestibility only when it is given shortly after misinformation: The critical importance of timing.

Recalling details from an experienced event can sometimes exacerbate eyewitnesses' susceptibility to subsequent misinformation. This finding, known as retrieval-enhanced suggestibility (RES), can be eliminated when participants are warned about possible inaccuracies in the misinformation source (Thomas et al., 2010). In three experiments, we investigated whether this warning benefit persists across delays. When the warning was issued shortly after participants were exposed to misinformation, it inoculated participant witnesses against RES, regardless of whether the final memory test occurred immediately or 48 hr after the warning. However, the warning lost its effectiveness when it was delivered 48 hr after participants were exposed to misinformation. These results applied to both recognition memory and the confidence-accuracy relationship. We considered these data from the perspective of temporal distinctiveness, and we argue that a warning serves a similar function to a forget cue in the directed forgetting paradigm. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Long retention intervals impair the confidence–accuracy relationship for eyewitness recall.

Long retention intervals impair the confidence–accuracy relationship for eyewitness recall.

Assessing the effect of eyewitness identification confidence assessment method on the confidence-accuracy relationship.

Assessing the effect of eyewitness identification confidence assessment method on the confidence-accuracy relationship.

Simulated viewing distance impairs the confidence–accuracy relationship for long, but not moderate distances: support for a model incorporating the role of feature ambiguity

There is an increasing need in eyewitness identification research to identify factors that not only influence identification accuracy but may also impact the confidence–accuracy (CA) relationship. One such variable that has a notable impact on memory for faces is viewing distance, with faces encoded from a shorter distance remembered better than faces encoded at longer differences. In four pre-registered experiments, using both laboratory and online samples, we compared faces viewed at a simulated viewing distance at two different levels (medium and far) to faces that were viewed at a very near-simulated distance. Distance was simulated using a Gaussian blur function with higher levels of blur corresponding to greater simulated distance. We found that both medium and far simulated distances impaired memory performance overall relative to no simulated distance, with increased distance resulting in poorer memory. However, only far simulated distances impaired the CA relationship. In a fourth experiment, we found that a pre-test warning did not ameliorate this impairment of the CA relationship for faces viewed at a far simulated distance. These findings suggest that even high-confidence identifications made for faces viewed from long distances should be disregarded, and that estimator variables that impact memory may degrade the CA relationship when memory is reduced to critical levels.

Providing eyewitness confidence judgments during versus after eyewitness interviews does not affect the confidence–accuracy relationship.

Providing eyewitness confidence judgments during versus after eyewitness interviews does not affect the confidence–accuracy relationship.

Supplemental Material for Long Retention Intervals Impair the Confidence–Accuracy Relationship for Eyewitness Recall

Supplemental Material for Long Retention Intervals Impair the Confidence–Accuracy Relationship for Eyewitness Recall

Child witness expressions of certainty are informative.

Children are frequently witnesses of crime. In the witness literature and legal systems, children are often deemed to have unreliable memories. Yet, in the basic developmental literature, young children can monitor their memory. To address these contradictory conclusions, we reanalyzed the confidence–accuracy relationship in basic and applied research. Confidence provided considerable information about memory accuracy, from at least age 8, but possibly younger. We also conducted an experiment where children in young (4–6 years), middle (7–9 years), and late (10–17 years) childhood (N = 2,205) watched a person in a video and then identified that person from a police lineup. Children provided a confidence rating (an explicit judgment) and used an interactive lineup—in which the lineup faces can be rotated—and we analyzed children’s viewing behavior (an implicit measure of metacognition). A strong confidence–accuracy relationship was observed from age 10 and an emerging relationship from age 7. A constant likelihood ratio signal-detection model can be used to understand these findings. Moreover, in all ages, interactive viewing behavior differed in children who made correct versus incorrect suspect identifications. Our research reconciles the apparent divide between applied and basic research findings and suggests that the fundamental architecture of metacognition that has previously been evidenced in basic list-learning paradigms also underlies performance on complex applied tasks. Contrary to what is believed by legal practitioners, but similar to what has been found in the basic literature, identifications made by children can be reliable when appropriate metacognitive measures are used to estimate accuracy.

Supplemental Material for Providing Eyewitness Confidence Judgments During Versus After Eyewitness Interviews Does Not Affect the Confidence–Accuracy Relationship

Supplemental Material for Providing Eyewitness Confidence Judgments During Versus After Eyewitness Interviews Does Not Affect the Confidence–Accuracy Relationship

Providing Eyewitness Confidence Judgments During Versus After Eyewitness Interviews Does Not Affect the Confidence-Accuracy Relationship

Recent studies suggest that highly confident eyewitnesses are likely to provide highly accurate identification evidence, at least in some conditions. Yet few studies have investigated the confidence-accuracy relationship in witness interviews or exactly when confidence judgments should be taken. Across three experiments, 831 adults answered questions about a mock crime and rated their confidence in each response. Participants gave their confidence immediately after each response or at the end of the memory test. The timing of the confidence judgment did not affect the confidence-accuracy relationship, and the confidence-accuracy relationship remained strong even when participants encoded the event under poor visibility conditions. When participants were unknowingly exposed to misinformation, however, the confidence-accuracy relationship was substantially weakened—participants became highly over-confident in the accuracy of their memories. These findings help to refine the parameters in which witness confidence serves as a useful indicator of memory accuracy.

Eyewitness identification: The complex issue of suspect-filler similarity.

Current guidelines for selecting lineup fillers are imprecise. Consequently, filler characteristics likely vary widely across lineups, potentially affecting identification decisions in important but unexplained ways. We report 2 experiments investigating the impact of 1 source of variation, the number of relatively high versus low similarity match-description fillers, on identification outcomes. Identifications following a retention interval of a few minutes (Experiments 1 and 2) and several weeks (Experiment 2) were examined. Increasing the number of high similarity lineup members within match-description lineups increased choosing (characterized by more filler but fewer suspect identifications), decreased accuracy and confidence and caused a poorer confidence-accuracy relationship. The changes to identification outcomes as the number of high similarity fillers increased were attributable to decreased discriminability and a relaxed criterion, with positive identifications requiring a lesser evidential discrepancy between target and fillers. There was limited evidence of guilty and innocent suspect identifications being differently affected by variations in the number of high similarity fillers and no evidence of suspect discriminability being influenced. For decisions after a short retention interval, high confidence suspect identifications were good predictors of accuracy; however, high similarity match-description fillers undermined the predictive value of high confidence suspect identifications at the long retention interval. Our results suggest that future research must explore methods for curbing variation in identification patterns resulting from uncontrolled filler characteristics. Until then it is critical that identification evidence be interpreted acknowledging that, even in lineups constructed following best practice guidelines, filler characteristics could have profoundly influenced the outcome.

Applying confidence accuracy characteristic plots to old/new recognition memory experiments

ABSTRACT Confidence-accuracy characteristic (CAC) plots were developed for use in eyewitness identification experiments, and previous findings show that high confidence indicates high accuracy in all studies of adults with an unbiased lineup. We apply CAC plots to standard old/new recognition memory data by calculating response-based and item-based accuracy, one using false alarms and the other using misses. We use both methods to examine the confidence-accuracy relationship for both correct old responses (hits) and new responses (correct rejections). We reanalysed three sets of published data using these methods and show that the method chosen, as well as the relation of lures to targets, determines the confidence-accuracy relation. Using response-based accuracy for hits, high confidence yields quite high accuracy, and this is generally true with the other methods, especially when lures are unrelated to targets. However, when analyzing correct rejections, the relationship between confidence and accuracy is less pronounced. When lures are semantically related to targets, the various CAC plots show different confidence-accuracy relations. The different methods of calculating CAC plots provide a useful tool in analyzing standard old/new recognition experiments. The results generally accord with unequal-variance signal detection models of recognition memory.

Elevated stress impairs the accuracy of eyewitness memory but not the confidence-accuracy relationship.

Although numerous studies have identified factors that affect eyewitness identification accuracy, recent studies report that many of these factors do not affect the accuracy of high-confidence identifications. This is critical because legal cases are more likely to be prosecuted if they involve high-confidence eyewitnesses. Using a confidence-accuracy characteristic (CAC) analysis, we explored whether stress affects the accuracy of high-confidence identifications. In two experiments, people viewed faces followed by an old/new recognition-memory test and provided confidence ratings. Stress was manipulated by pairing a low- or high-valence image with each studied face. Identification accuracy was higher in the low- than high-stress condition, yet the proportion correct for high-confidence positive identifications was similar in the two stress conditions. Elevated stress impairs eyewitness identification accuracy overall. However, the results of this study suggest that confidence is a better predictor of recognition-memory accuracy than is stress even though confidence alone is still an imperfect predictor. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

A decision processes account of the differences in the eyewitness confidence-accuracy relationship between strong and weak face recognizers under suboptimal exposure and delay conditions.

When pristine testing conditions are used, an eyewitness's high-confidence identification from a lineup can be a reliable predictor of their identification accuracy (Wixted & Wells, 2017). Further, Grabman, Dobolyi, Berelovich, and Dodson (2019) found that high-confidence identifications are more predictive of accuracy for individuals with stronger than weaker face recognition ability. We extend this research by investigating why strong face recognizers make more informative confidence judgments and fewer high-confidence errors through the framework of two different accounts: the optimality account (Deffenbacher, 1980) and the decision processes account (e.g., Kruger & Dunning, 1999). The optimality account holds that differences in the predictive value of confidence ratings made by strong versus weak face recognizers are a result of differences in the quality of their memory representations for faces, indicating that confidence-accuracy calibration would be equated between these two groups when overall accuracy is equated, whereas the decision processes account attributes differences in calibration to strong face recognizers' superior metacognitive skills, which allow them to better evaluate their performance in the domain of face recognition. Therefore, to distinguish between these accounts, we manipulated exposure and retention interval to create conditions that produced comparable levels of identification accuracy between stronger and weaker face recognizers, and then examined their confidence-accuracy calibration. The decision processes account was supported, as differences in calibration between stronger and weaker face recognizers persisted even when overall identification accuracy was equated. Stronger face recognizers are better able to regulate their use of the confidence scale points with changes in identification accuracy. (PsycInfo Database Record (c) 2021 APA, all rights reserved).