Absolute-judgment models better predict eyewitness decision-making than do relative-judgment models

When presented with a lineup, the witness is tasked with identifying the culprit or indicating that the culprit is not present. The witness then qualifies the decision with a confidence judgment. But how do witnesses go about making these decisions and judgments? According to absolute-judgment models, witnesses determine which lineup member provides the strongest match to memory and base their identification decision and confidence judgment on the absolute strength of this MAX lineup member. Conversely, relative-judgment models propose that witnesses determine which lineup member provides the strongest match to memory and then base their identification decision and confidence judgment on the relative strength of the MAX lineup member compared to the remaining lineup members. We took a critical test approach to test the predictions of both models. As predicted by the absolute-judgment model, but contrary to the predictions of the relative-judgment model, witnesses were more likely to correctly reject low-similarity lineups than high-similarity lineups (Experiment 1), and more likely to reject biased lineups than fair lineups (Experiment 2). Likewise, witnesses rejected low-similarity lineups with greater confidence than high-similarity lineups (Experiment 1) and rejected biased lineups with greater confidence than fair lineups (Experiment 2). Only a single pattern was consistent with the relative model and inconsistent with the absolute model: suspect identifications from biased lineups were made with greater confidence than suspect identifications from fair lineups (Experiment 2). The results suggest that absolute-judgment models better predict witness decision-making than do relative-judgment models and that pure relative-judgment models are unviable.

Feeling the future of eyewitness research

For decades, eyewitness memory research has had the worthy goal of minimizing the chances that an innocent suspect is falsely identified. However, this is not the only goal. Partial receiver operating characteristic (ROC) curves provide a way to identify lineup procedures that keep the false alarm rate low while also maximizing the hit rate. Recently, there have been attempts to extend the ROC curve into high false alarm rate regions that fair lineups are intentionally designed to avoid. These new full ROCs could provide a way for the police to circumvent the protections offered by fillers in a fair lineup. Moreover, these attempts to extend the ROC curve are not based on a mathematically coherent model of latent diagnostic signals. In this article, we empirically demonstrate how this lack of a solid foundation can lead to dubious conclusions, such as eyewitnesses possessing precognition and being able to reliably identify the person they will see commit a crime in the future.

Improving the diagnostic value of lineup rejections

Erroneous eyewitness identification evidence is likely the leading cause of wrongful convictions. To minimize this error, scientists recommend collecting confidence. Research shows that eyewitness confidence and accuracy are strongly related when an eyewitness identifies someone from an initial and properly administered lineup. However, confidence is far less informative of accuracy when an eyewitness identifies no one and rejects the lineup instead. In this study, I aimed to improve the confidence-accuracy relationship for lineup rejections in two ways. First, I aimed to find the lineup that yields the strongest confidence-accuracy relationship for lineup rejections by comparing the standard, simultaneous procedure used by police worldwide to the novel "reveal" procedure designed by scientists to boost accuracy. Second, I aimed to find the best method for collecting confidence. To achieve this secondary aim, I made use of machine-learning techniques to compare confidence expressed in words to numeric confidence ratings. First, I find a significantly stronger confidence-accuracy relationship for lineup rejections in the reveal than in the standard procedure regardless of the method used to collect confidence. Second, I find that confidence expressed in words captures unique diagnostic information about the likely accuracy of a lineup rejection separate from the diagnostic information captured by numeric confidence ratings. These results inform models of recognition memory and may improve the criminal-legal system by increasing the diagnostic value of a lineup rejection.

On the advantages of using AI-generated images of filler faces for creating fair lineups

Recent advances in artificial intelligence (AI) enable the generation of realistic facial images that can be used in police lineups. The use of AI image generation offers pragmatic advantages in that it allows practitioners to generate filler images directly from the description of the culprit using text-to-image generation, avoids the violation of identity rights of natural persons who are not suspects and eliminates the constraints of being bound to a database with a limited set of photographs. However, the risk exists that using AI-generated filler images provokes more biased selection of the suspect if eyewitnesses are able to distinguish AI-generated filler images from the photograph of the suspect's face. Using a model-based analysis, we compared biased suspect selection directly between lineups with AI-generated filler images and lineups with database-derived filler photographs. The results show that the lineups with AI-generated filler images were perfectly fair and, in fact, led to less biased suspect selection than the lineups with database-derived filler photographs used in previous experiments. These results are encouraging with regard to the potential of AI image generation for constructing fair lineups which should inspire more systematic research on the feasibility of adopting AI technology in forensic settings.

New Insights on Expert Opinion About Eyewitness Memory Research

Experimental psychologists investigating eyewitness memory have periodically gathered their thoughts on a variety of eyewitness memory phenomena. Courts and other stakeholders of eyewitness research rely on the expert opinions reflected in these surveys to make informed decisions. However, the last survey of this sort was published more than 20 years ago, and the science of eyewitness memory has developed since that time. Stakeholders need a current database of expert opinions to make informed decisions. In this article, we provide that update. We surveyed 76 scientists for their opinions on eyewitness memory phenomena. We compared these current expert opinions to expert opinions from the past several decades. We found that experts today share many of the same opinions as experts in the past and have more nuanced thoughts about two issues. Experts in the past endorsed the idea that confidence is weakly related to accuracy, but experts today acknowledge the potential diagnostic value of initial confidence collected from a properly administered lineup. In addition, experts in the past may have favored sequential over simultaneous lineup presentation, but experts today are divided on this issue. We believe this new survey will prove useful to the court and to other stakeholders of eyewitness research.

Identifying unfamiliar voices: Examining the system variables of sample duration and parade size

Voice identification parades can be unreliable due to the error-prone nature of earwitness responses. UK government guidelines recommend that voice parades should have nine voices, each played for 60 s. This makes parades resource-consuming to construct. In this article, we conducted two experiments to see if voice parade procedures could be simplified. In Experiment 1 (N = 271, 135 female), we investigated if reducing the duration of the voice samples on a nine-voice parade would negatively affect identification performance using both conventional logistic and signal detection approaches. In Experiment 2 (N = 270, 136 female), we first explored if the same sample duration conditions used in Experiment 1 would lead to different outcomes if we reduced the parade size to include only six voices. Following this, we pooled the data from both experiments to investigate the influence of target-position effects. The results show that 15-s sample durations result in statistically equivalent voice identification performance to the longer 60-s sample durations, but that the 30-s sample duration suffers in terms of overall signal sensitivity. This pattern of results was replicated using both a nine- and a six-voice parade. Performance on target-absent parades were at chance levels in both parade sizes, and response criteria were mostly liberal. In addition, unwanted position effects were present. The results provide initial evidence that the sample duration used in a voice parade may be reduced, but we argue that the guidelines recommending a parade with nine voices should be maintained to provide additional protection for a potentially innocent suspect given the low target-absent accuracy.

Enabling witnesses to actively explore faces and reinstate study-test pose during a lineup increases discriminability

Accurate witness identification is a cornerstone of police inquiries and national security investigations. However, witnesses can make errors. We experimentally tested whether an interactive lineup, a recently introduced procedure that enables witnesses to dynamically view and explore faces from different angles, improves the rate at which witnesses identify guilty over innocent suspects compared to procedures traditionally used by law enforcement. Participants encoded 12 target faces, either from the front or in profile view, and then attempted to identify the targets from 12 lineups, half of which were target present and the other half target absent. Participants were randomly assigned to a lineup condition: simultaneous interactive, simultaneous photo, or sequential video. In the front-encoding and profile-encoding conditions, Receiver Operating Characteristics analysis indicated that discriminability was higher in interactive compared to both photo and video lineups, demonstrating the benefit of actively exploring the lineup members' faces. Signal-detection modeling suggested interactive lineups increase discriminability because they afford the witness the opportunity to view more diagnostic features such that the nondiagnostic features play a proportionally lesser role. These findings suggest that eyewitness errors can be reduced using interactive lineups because they create retrieval conditions that enable witnesses to actively explore faces and more effectively sample features.

Evaluating the impact of first-yes-counts instructions on eyewitness performance using the two-high threshold eyewitness identification model

In eyewitness research, multiple identification decisions in sequential lineups are typically prevented by telling participants that only their first identification decision counts. These first-yes-counts instructions are incompatible with standard police protocols prescribing that witnesses shall see the entire lineup. Horry et al. were the first to experimentally test how this discrepancy between eyewitness research and standard police protocols affects eyewitness identification decisions. Here, the two-high threshold eyewitness identification model was used to disentangle the effect of the first-yes-counts instructions on the detection and guessing processes underlying eyewitness identification decisions. We report both a reanalysis of Horry et al.'s data and a conceptual replication. Both the reanalysis and the results of the conceptual replication confirm that first-yes-counts instructions do not affect the detection of the culprit but decrease the probability of guessing-based selections. To improve the ecological validity, research on sequential lineups should avoid first-yes-counts instructions.

Measuring lineup fairness from eyewitness identification data using a multinomial processing tree model

The mock-witness task is typically used to evaluate the fairness of lineups. However, the validity of this task has been questioned because there are substantial differences between the tasks for mock witnesses and eyewitnesses. Unlike eyewitnesses, mock witnesses must select a person from the lineup and are alerted to the fact that one lineup member might stand out from the others. It therefore seems desirable to base conclusions about lineup fairness directly on eyewitness data rather than on mock-witness data. To test the importance of direct measurements of biased suspect selection in eyewitness identification decisions, we assessed the fairness of lineups containing either morphed or non-morphed fillers using both mock witnesses and eyewitnesses. We used Tredoux's E and the proportion of suspect selections to measure lineup fairness from mock-witness choices and the two-high threshold eyewitness identification model to measure the biased selection of the suspects directly from eyewitness identification decisions. Results obtained in the mock-witness task and the model-based analysis of data obtained in the eyewitness task converged in showing that simultaneous lineups with morphed fillers were significantly more unfair than simultaneous lineups with non-morphed fillers. However, mock-witness and eyewitness data converged only when the eyewitness task mimicked the mock-witness task by including pre-lineup instructions that (1) discouraged eyewitnesses to reject the lineups and (2) alerted eyewitnesses that a photograph might stand out from the other photographs in the lineup. When a typical eyewitness task was created by removing these two features from the pre-lineup instructions, the morphed fillers no longer lead to unfair lineups. These findings highlight the differences in the cognitive processes of mock witnesses and eyewitnesses and they demonstrate the importance of measuring lineup fairness directly from eyewitness identification decisions rather than indirectly using the mock-witness task.

fullROC: An R package for generating and analyzing eyewitness-lineup ROC curves

A police lineup is a procedure in which a suspect is surrounded by known-innocent persons (fillers) and presented to the witness for an identification attempt. The purpose of a lineup is to test the investigator's hypothesis that the suspect is the culprit, and the investigator uses the witness' identification decision and the associated confidence level to inform this hypothesis. Whereas suspect identifications provide evidence of guilt, filler identifications and rejections provide evidence of innocence. Despite the capacity of lineups to provide exculpatory information, past research has focused, almost exclusively, on inculpatory behaviors. We recently developed a method for incorporating all lineup outcomes in a single receiver operator characteristic (ROC) curve. The area under the full lineup ROC curve reflects the total capacity of a lineup procedure to discriminate guilty suspects from innocent suspects. Here, we introduce a Comprehensive R Archive Network (CRAN) package, fullROC, to support eyewitness researchers in using the full ROC approach to analyze lineup data. The fullROC package provides functions for adjusting identification rates, generating full ROC curves for lineup data, computing the area under the ROC curves (AUC), and statistically comparing the AUCs of different lineups. Using both simulated and empirical data, we illustrate the functionality of the fullROC CRAN package. In brief, the fullROC package provides a useful tool for eyewitness researchers to analyze lineup data using the full ROC method, which incorporates both the inculpatory and exculpatory information of eyewitness behaviors.

The effect of pre-event instructions on eyewitness identification

Research on eyewitness identification often involves exposing participants to a simulated crime and later testing memory using a lineup. We conducted a systematic review showing that pre-event instructions, instructions given before event exposure, are rarely reported and those that are reported vary in the extent to which they warn participants about the nature of the event or tasks. At odds with the experience of actual witnesses, some studies use pre-event instructions explicitly warning participants of the upcoming crime and lineup task. Both the basic and applied literature provide reason to believe that pre-event instructions may affect eyewitness identification performance. In the current experiment, we tested the impact of pre-event instructions on lineup identification decisions and confidence. Participants received non-specific pre-event instructions (i.e., "watch this video") or eyewitness pre-event instructions (i.e., "watch this crime video, you'll complete a lineup later") and completed a culprit-absent or -present lineup. We found no support for the hypothesis that participants who receive eyewitness pre-event instructions have higher discriminability than participants who receive non-specific pre-event instructions. Additionally, confidence-accuracy calibration was not significantly different between conditions. However, participants in the eyewitness condition were more likely to see the event as a crime and to make an identification than participants in the non-specific condition. Implications for conducting and interpreting eyewitness identification research and the basic research on instructions and attention are discussed.

Evaluating earwitness identification procedures: adapting pre-parade instructions and parade procedure

Voice identification parades can be unreliable, as earwitness responses are error-prone. In this paper we tested performance across serial and sequential procedures, and varied pre-parade instructions, with the aim of reducing errors. The participants heard a target voice and later attempted to identify it from a parade. In Experiment 1 they were either warned that the target may or may not be present (standard warning) or encouraged to consider responding "not present" because of the associated risk of a wrongful conviction (strong warning). Strong warnings prompted a conservative criterion shift, with participants less likely to make a positive identification regardless of whether the target was present. In contrast to previous findings, we found no statistically reliable difference in accuracy between serial and sequential parades. Experiment 2 ruled out a potential confound in Experiment 1. Taken together, our results suggest that adapting pre-parade instructions provides a simple way of reducing the risk of false identifications.

A conjunctive analysis of false accusations, official misconduct, and race in violent and sexual exonerations cases

The National Registry of Exonerations tracks cases of individuals who have been convicted and exonerated in the United States since 1989. The objective of the current study was to explore the misconduct patterns in violent and sexual offense cases using a novel method with a focus on cases where suspects were falsely accused, either deliberately or by mistaken witness identification. A conjunctive analysis of case configurations (CACC) was conducted using 1690 males convicted of murder/manslaughter, rape/sexual assault, and child sexual abuse. The results showed that compared to Whites, Blacks had a disproportionate amount of official misconduct in both sexual and non-sexual homicide cases. Blacks also had a disproportionate amount of misconduct by multiple actors and multiple types. Blacks had more cases of mistaken witness identification and a disproportionate amount of official misconduct in sexual assault cases. Implications and recommendations are discussed.

Experimental validation of a multinomial processing tree model for analyzing eyewitness identification decisions

To improve police protocols for lineup procedures, it is helpful to understand the processes underlying eyewitness identification performance. The two-high threshold (2-HT) eyewitness identification model is a multinomial processing tree model that measures four latent cognitive processes on which eyewitness identification decisions are based: two detection-based processes (the detection of culprit presence and absence) and two non-detection-based processes (biased and guessing-based selection). The model takes into account the full 2 × 3 data structure of lineup procedures, that is, suspect identifications, filler identifications and rejections in both culprit-present and culprit-absent lineups. Here the model is introduced and the results of four large validation experiments are reported, one for each of the processes specified by the model. The validation experiments served to test whether the model’s parameters sensitively reflect manipulations of the processes they were designed to measure. The results show that manipulations of exposure duration of the culprit’s face at encoding, lineup fairness, pre-lineup instructions and ease of rejection of culprit-absent lineups were sensitively reflected in the parameters representing culprit-presence detection, biased suspect selection, guessing-based selection and culprit-absence detection, respectively. The results of the experiments thus validate the interpretations of the parameters of the 2-HT eyewitness identification model.

A validation of the two-high threshold eyewitness identification model by reanalyzing published data

The two-high threshold (2-HT) eyewitness identification model serves as a new measurement tool to measure the latent cognitive processes underlying eyewitness identification performance. By simultaneously taking into account correct culprit identifications, false innocent-suspect identifications, false filler identifications in culprit-present and culprit-absent lineups as well as correct and false lineup rejections, the model capitalizes on the full range of data categories that are observed when measuring eyewitness identification performance. Thereby, the model is able to shed light on detection-based and non-detection-based processes underlying eyewitness identification performance. Specifically, the model incorporates parameters for the detection of culprit presence and absence, biased selection of the suspect and guessing-based selection among the lineup members. Here, we provide evidence of the validity of each of the four model parameters by applying the model to eight published data sets. The data sets come from studies with experimental manipulations that target one of the underlying processes specified by the model. Manipulations of encoding difficulty, lineup fairness and pre-lineup instructions were sensitively reflected in the parameters reflecting culprit-presence detection, biased selection and guessing-based selection, respectively. Manipulations designed to facilitate the rejection of culprit-absent lineups affected the parameter for culprit-absence detection. The reanalyses of published results thus suggest that the parameters sensitively reflect the manipulations of the processes they were designed to measure, providing support of the validity of the 2-HT eyewitness identification model.

Sources of Risk of AI Systems

Artificial intelligence can be used to realise new types of protective devices and assistance systems, so their importance for occupational safety and health is continuously increasing. However, established risk mitigation measures in software development are only partially suitable for applications in AI systems, which only create new sources of risk. Risk management for systems that for systems using AI must therefore be adapted to the new problems. This work objects to contribute hereto by identifying relevant sources of risk for AI systems. For this purpose, the differences between AI systems, especially those based on modern machine learning methods, and classical software were analysed, and the current research fields of trustworthy AI were evaluated. On this basis, a taxonomy could be created that provides an overview of various AI-specific sources of risk. These new sources of risk should be taken into account in the overall risk assessment of a system based on AI technologies, examined for their criticality and managed accordingly at an early stage to prevent a later system failure.

Eyewitness Identification Is a Visual Search Task

The simultaneous six-pack photo lineup is a standard eyewitness identification procedure, consisting of one police suspect plus five physically similar fillers. The photo lineup is either a target-present array (the suspect is guilty) or a target-absent array (the suspect is innocent). The eyewitness is asked to search the six photos in the array with respect to a target template stored in memory (namely, the memory of the perpetrator's face). If the witness determines that the perpetrator is in fact in the lineup (detection), then the next step is to specify the position of the perpetrator's face in the lineup (localization). The witness may also determine that the perpetrator is not present and reject the lineup. In other words, a police lineup is a detection-plus-localization visual search task. Signal detection concepts that have long guided thinking about visual search have recently had a significant impact on our understanding of police lineups. Expected final online publication date for the Annual Review of Vision Science, Volume 7 is September 2021. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Testing encoding specificity and the diagnostic feature-detection theory of eyewitness identification, with implications for showups, lineups, and partially disguised perpetrators

The diagnostic feature-detection theory (DFT) of eyewitness identification is based on facial information that is diagnostic versus non-diagnostic of suspect guilt. It primarily has been tested by discounting non-diagnostic information at retrieval, typically by surrounding a single suspect showup with good fillers to create a lineup. We tested additional DFT predictions by manipulating the presence of facial information (i.e., the exterior region of the face) at both encoding and retrieval with a large between-subjects factorial design (N = 19,414). In support of DFT and in replication of the literature, lineups yielded higher discriminability than showups. In support of encoding specificity, conditions that matched information between encoding and retrieval were generally superior to mismatch conditions. More importantly, we supported several DFT and encoding specificity predictions not previously tested, including that (a) adding non-diagnostic information will reduce discriminability for showups more so than lineups, and (b) removing diagnostic information will lower discriminability for both showups and lineups. These results have implications for police deciding whether to conduct a showup or a lineup, and when dealing with partially disguised perpetrators (e.g., wearing a hoodie).

Optimizing the selection of fillers in police lineups

A typical police lineup contains a photo of one suspect (who is innocent in a target-absent lineup and guilty in a target-present lineup) plus photos of five or more fillers who are known to be innocent. To create a fair lineup in which the suspect does not stand out, two filler selection methods are commonly used. In the first, fillers are selected if they are similar in appearance to the suspect. In the second, fillers are selected if they possess facial features included in the witness's description of the culprit (e.g., "20-y-old white male"). The police sometimes use a combination of the two methods by selecting description-matched fillers whose appearance is also similar to that of the suspect in the lineup. Decades of research on which approach is better remains unsettled. Here, we tested a counterintuitive prediction made by a formal model based on signal detection theory: From a pool of acceptable description-matched photos, selecting fillers whose appearance is otherwise dissimilar to the suspect should increase the hit rate without affecting the false-alarm rate (increasing discriminability). In Experiment 1, we confirmed this prediction using a standard mock-crime paradigm. In Experiment 2, the effect on discriminability was reversed (as also predicted by the model) when fillers were matched on similarity to the perpetrator in both target-present and target-absent lineups. These findings suggest that signal-detection theory offers a useful theoretical framework for understanding eyewitness identification decisions made from a police lineup.

Distinguishing Between Investigator Discriminability and Eyewitness Discriminability: A Method for Creating Full Receiver Operating Characteristic Curves of Lineup Identification Performance

The conceptual frameworks provided by both the lineups-as-experiments analogy and signal detection theory have proven important to understanding how eyewitness lineups work. The lineups-as-experiments analogy proposes that when investigators use a lineup procedure, they are acting as experimenters and should therefore follow the same tried-and-true procedures that experimenters follow when executing an experiment. Signal detection theory offers a framework for distinguishing between factors that improve the trade-off between culprit and innocent-suspect identifications and factors that affect the frequency of suspect identifications. We integrate these two conceptual frameworks. We argue that an eyewitness lineup procedure is characterized by two simultaneous signal detection tasks. On one hand, the witness is tasked with determining whether the culprit is present in the lineup and identifying that person. On the other hand, the investigator knows which lineup member is the suspect and which lineup members are known-innocent fillers and is therefore tasked only with determining whether the suspect is the culprit. The investigator uses the witness's identification decision and associated level of confidence to decide whether the suspect is the culprit. We leverage this realization to demonstrate a method for creating full receiver operating characteristic curves for eyewitness lineup procedures.

The impact of sleep on eyewitness identifications

Sleep aids the consolidation of recently acquired memories. Evidence strongly indicates that sleep yields substantial improvements on recognition memory tasks relative to an equivalent period of wake. Despite the known benefits that sleep has on memory, researchers have not yet investigated the impact of sleep on eyewitness identifications. Eyewitnesses to crimes are often presented with a line-up (which is a type of recognition memory test) that contains the suspect (who is innocent or guilty) and fillers (who are known to be innocent). Sleep may enhance the ability to identify the guilty suspect and not identify the innocent suspect (i.e. discriminability). Sleep may also impact reliability (i.e. the likelihood that the identified suspect is guilty). In the current study, we manipulated the presence or the absence of sleep in a forensically relevant memory task. Participants witnessed a video of a mock crime, made an identification or rejected the line-up, and rated their confidence. Critically, some participants slept between witnessing the crime and making a line-up decision, while others remained awake. The prediction that participants in the sleep condition would have greater discriminability compared to participants in the wake condition was not supported. There were also no differences in reliability.

Lineup fairness: propitious heterogeneity and the diagnostic feature-detection hypothesis

Researchers have argued that simultaneous lineups should follow the principle of propitious heterogeneity, based on the idea that if the fillers are too similar to the perpetrator even an eyewitness with a good memory could fail to correctly identify him. A similar prediction can be derived from the diagnostic feature-detection (DFD) hypothesis, such that discriminability will decrease if too few features are present that can distinguish between innocent and guilty suspects. Our first experiment tested these predictions by controlling similarity with artificial faces, and our second experiment utilized a more ecologically valid eyewitness identification paradigm. Our results support propitious heterogeneity and the DFD hypothesis by showing that: 1) as the facial features in lineups become increasingly homogenous, empirical discriminability decreases; and 2) lineups with description-matched fillers generally yield higher empirical discriminability than those with suspect-matched fillers.

Models of lineup memory

Face recognition memory is often tested by the police using a photo lineup, which consists of one suspect, who is either innocent or guilty, and five or more physically similar fillers, all of whom are known to be innocent. For many years, lineups were investigated in lab studies without guidance from standard models of recognition memory. More recently, signal detection theory has been used to conceptualize lineup memory and to motivate receiver operating characteristic (ROC) analysis of lineup performance. Here, we describe three competing signal-detection models of lineup memory, derive their likelihood functions, and fit them to empirical ROC data. We also introduce the notion that memory signals generated by the faces in a lineup are likely to be correlated because, by design, those faces share features. The models we investigate differ in their predictions about the effect that correlated memory signals should have on the ability to discriminate innocent from guilty suspects. A popular compound signal detection model known as the Integration model predicts that correlated memory signals should impair discriminability. Empirically, this model performed so poorly that, going forward, it should probably be abandoned. The best-fitting model incorporates a principle known as "ensemble coding," which predicts that correlated memory signals should enhance discriminability. The ensemble model aligns with a previously proposed theory of eyewitness identification according to which the simultaneous presentation of faces in a lineup enhances discriminability compared to when faces are presented in isolation because it permits eyewitnesses to detect and discount non-diagnostic facial features.

The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis

Summary

The U.S. legal system increasingly accepts the idea that the confidence expressed by an eyewitness who identified a suspect from a lineup provides little information as to the accuracy of that identification. There was a time when this pessimistic assessment was entirely reasonable because of the questionable eyewitness-identification procedures that police commonly employed. However, after more than 30 years of eyewitness-identification research, our understanding of how to properly conduct a lineup has evolved considerably, and the time seems ripe to ask how eyewitness confidence informs accuracy under more pristine testing conditions (e.g., initial, uncontaminated memory tests using fair lineups, with no lineup administrator influence, and with an immediate confidence statement). Under those conditions, mock-crime studies and police department field studies have consistently shown that, for adults, (a) confidence and accuracy are strongly related and (b) high-confidence suspect identifications are remarkably accurate. However, when certain non-pristine testing conditions prevail (e.g., when unfair lineups are used), the accuracy of even a high-confidence suspect ID is seriously compromised. Unfortunately, some jurisdictions have not yet made reforms that would create pristine testing conditions and, hence, our conclusions about the reliability of high-confidence identifications cannot yet be applied to those jurisdictions. However, understanding the information value of eyewitness confidence under pristine testing conditions can help the criminal justice system to simultaneously achieve both of its main objectives: to exonerate the innocent (by better appreciating that initial, low-confidence suspect identifications are error prone) and to convict the guilty (by better appreciating that initial, high-confidence suspect identifications are surprisingly accurate under proper testing conditions).