Reference Manual on Scientific Evidence: Fourth Edition (2025)
Chapter: Reference Guide on Eyewitness Identification
Reference Guide on Eyewitness Identification
THOMAS D. ALBRIGHT AND BRANDON L. GARRETT
Thomas D. Albright, Ph.D., is Professor and Director, Center for the Neurobiology of Vision, Conrad T. Prebys Chair in Vision Research at The Salk Institute for Biological Studies.
Brandon L. Garrett, J.D., is L. Neil Williams Professor of Law and Faculty Director, Wilson Center for Science and Justice at Duke University School of Law.
CONTENTS
The Promise and Peril of Eyewitness Testimony
Definition of Eyewitness Identification
A Word About Probabilities and Prediction
Organization of This Reference Guide
The Eyewitness as an Information-Processing “Instrument”
How Is the Instrument Used?: A Taxonomy of Identification Procedures
Nonstandard Identification Procedures
Predicting the Accuracy of Eyewitness Identification
Manson v. Brathwaite: Accuracy Prediction by the Court
Accuracy Prediction by the Scientific Community
Scientific Questions Regarding Eyewitness Identification
Basic Science of Vision, Memory, and Choice
Upper Bounds on Visual Performance
How memory goes missing: The “forgetting function”
How memory goes wrong: Source memory failure and false memories
The effect of emotion on memory storage
Applied Eyewitness Studies in the Laboratory
Viewing Distance, Lighting, and Exposure Duration
Timing of identification and conditions of confidence assessment
Communication Between Law Enforcement Personnel and Witness
Statements that convey prior probability of suspicion
Communication during lineups: The importance of blinded lineup administration
Video recording of lineup procedures
Communication Between Community and Witness: Social Influences on Eyewitness Performance
The social utility of misinformation
The long reach of modern social networks
Prior Exposure and Transference Effects
Corroborating Variables: Machine-Based Facial Recognition
Applied Eyewitness Studies in the Field
Effects of Eyewitness Testimony on Juries
Sources of Expertise on Matters of Eyewitness Science
The Legal Framework for Eyewitness Evidence
State Eyewitness Evidence Statutes
Police Practice Recommendations
FIGURES
1. Sensitivity to luminance contrast as a function of overall illumination
3. Visual “crowding” impairs object recognition
4. Viewing angle affects face recognition
5. Focus of attention affects recognition of more complex objects
8. Perceptual uncertainty in response to visual “noise”
9. Clear image will resolve uncertainty in Figure 8
10. Perceptual state falls on a continuum between pure stimulus and pure imagery
11. Ebbinghaus forgetting function
12. Natural variation in human face recognition ability
13. Summary of empirical support for simultaneous lineup advantage
Introduction
The Promise and Peril of Eyewitness Testimony
Witnessing the world and reporting what we have seen are among the most elemental of human abilities. We use these abilities to learn, to understand, and to assess what is true. We find things that we misplaced. We recognize our friends. Perhaps most importantly, we share our experiences with others. We testify to the beauty of the full moon rising, to the birth of our children, and to the terror of war.
Societies have long mined the power of this native human ability by recruiting eyewitnesses to help resolve disagreements about past events. In many cases, the outcome is of little consequence (did he swing the bat, or foul his opponent?), and we readily defer judgment to the witness. In other cases, we seek out eyewitness testimony to settle fraught conflicts over civil or criminal responsibility. A firsthand report of what happened and who was there can be extremely valuable, particularly when no physical evidence exists. But in court, a factfinder’s decision rests not simply on what an eyewitness reports, but also on inference about the probability that the testimony is correct. It may not matter much if you misperceive, misremember, and misreport who was at the birthday party, but in a courtroom, being wrong about who assaulted you and hijacked the car can render a uniquely tragic form of injustice.
Eyewitness reports of criminal acts rarely involve deceit;1 witnesses are generally well-intentioned people who offer a valued piece of information that they alone possess and believe to be true. Inaccurate testimony, should it occur, typically reflects an unwitting failure of human vision and memory, a failure that gets smoothed over, in the witness’s telling, by candor and confidence. The result imperils the innocent,2 leaves society at continued risk,3 and undermines public trust. However, scientific advances shed light on the accuracy of eyewitness testimony. These advances have increasingly informed legal actors, including law enforcement, lawmakers, and courts. In this guide, we summarize several decades of scientific research and legal approaches, including those that incorporate scientific insights. We also discuss more recent discoveries and the challenges and opportunities posed by new technology.
1. A noteworthy exception is “understandable” deceit, in which a witness intentionally fails to identify a suspect (e.g., a gang member) for fear of retribution.
2. Brandon L. Garrett, Convicting the Innocent: Where Criminal Prosecutions Go Wrong (2011), https://doi.org/10.4159/harvard.9780674060982.
3. Jee Park, Eyewitness Identification and Innocence, 64 Loy. L. Rev. 669, 670 (2018) (explaining that “[o]f the 158 cases [of exonerations] where the true perpetrators were identified by DNA, these actual perpetrators went on to commit 150 additional violent crimes”).
Definition of Eyewitness Identification
Eyewitness identification is defined in operational legal terms as “a naming or description by which one who has seen an event testifies from memory about the person or persons involved.”4 In the more analytical idioms of science, it is a type of visual object recognition task used for forensic purposes. Object recognition, in turn, is a ubiquitous function of human vision and memory, in which an observer must decide whether a currently viewed stimulus is the same as a previously viewed stimulus. People commonly perform such tasks in a variety of contexts and do so very successfully when the objects are familiar, such as one’s coffee mug or the office stapler. By contrast, the majority of eyewitness reports in criminal cases are based on events in which a witness saw an unfamiliar person or persons during a single encounter, often briefly and under highly constrained conditions. Because people are poorer at identification of unfamiliar faces, relative to familiar ones,5 the unfamiliar-face scenario presents a significant challenge to our criminal justice system. For purposes of this guide, eyewitness identification refers exclusively to the unfamiliar case.
A Word About Probabilities and Prediction
Probabilistic reasoning is a ubiquitous feature of courtroom decisions. Evidence is rarely certain; it varies continuously in the strength with which it is probative. In a criminal case, in order to convict, for example, a jury must infer whether eyewitness evidence is more likely under one hypothesis (the chef pulled the trigger) versus a different hypothesis (someone else pulled the trigger).
Analogously, a trial judge may apply probabilistic reasoning to make decisions about the accuracy and admissibility of evidence. The best approach to this problem is to assess conditions associated with the viewing and identification events that are known from prior empirical studies to increase or decrease accuracy. The result is a form of prediction in which the evidence is assigned (often implicitly) a probability of being true. From a continuous range of such probabilities, a judge would then apply a threshold, or decision criterion, to make a binary decision about whether to permit the evidence at trial.
Building on this framework of probabilistic inference, scientific approaches to the eyewitness accuracy problem fall into two broad categories: prospective and retrospective. The prospective approach uses novel techniques to elicit greater accuracy in proffered testimony. We review some of these techniques herein, but the problem faced by the courts is necessarily retrospective; the goal here is to
4. Black’s Law Dictionary 894 (11th ed. 2019).
5. Bruce Vicki et al., Matching Identities of Familiar and Unfamiliar Faces Caught on CCTV Images, 7 J. Exp. Psych.: Applied 207 (2001), https://doi.org/10.1037//1076–898x.7.3.207.
better predict the level of accuracy. This is a problem that the sciences of visual perception and memory are well equipped to address.
We stress these points at the outset because, as we will show, the assessment of the reliability of eyewitness evidence is not only the task of the jurors as factfinders, but it is also central to the judge’s task under the Supreme Court’s 1977 Manson v. Brathwaite ruling on eyewitness evidence. Modern scientific research provides an empirical basis for assessing such probabilities and making predictions about the accuracy of eyewitness identifications, which is the subject of this guide.
Organization of This Reference Guide
The topic of eyewitness identification has been an intense focus of scientific research and legal policy development for decades. In this guide we have synthesized a vast amount of information from a variety of sources. We offer here a brief precis to assist with navigation of the text.
We begin with a generic description of the eyewitness and the task they are confronted with, which serves as an introduction to the problem faced by the criminal justice system. Casting the eyewitness as a biological information-processing “instrument” helps to focus on general factors that limit instrument performance under the constraints of the task.
With this orientation to the problem, we turn to accuracy prediction, which lies at the heart of any judicial decision about use of eyewitness testimony. We summarize the Supreme Court ruling in Manson v. Brathwaite, which recognized the significance of accuracy prediction and promoted a rule-based method for doing so. We follow this with a brief account of modern scientific approaches to prediction, which have ready application to the eyewitness problem and may improve upon the existing legal approach.
The next section provides a brief taxonomy of scientific questions that have been raised in research aimed at understanding eyewitness performance and use of eyewitness testimony by the courts. This taxonomy is followed by a dive into the science itself and the implications of this knowledge for assessment of the accuracy of eyewitness testimony.
The relevant science is rich and broad. We subdivide it into (1) basic science of vision and memory, with a focus on how these systems work and how their operational characteristics limit the accuracy of eyewitness reports; and (2) applied eyewitness science, which has experimentally manipulated variables commonly associated with the eyewitness experience, such as viewing conditions or identification procedures employed by the police, and assessed effects of these variables on identification performance. These variables—together with an understanding of how vision and memory work—provide a foundation for predicting the accuracy of eyewitness reports in criminal cases.
The final section of this guide turns to legal policy developments over the past half-century, which are intended to inform and regulate the use of eyewitness testimony. While much of this legal framework will be familiar to members of the judiciary, we show that these developments—which include court rulings, legislative actions, and changes in police practices—are increasingly rooted in the growing body of basic and applied science summarized herein.
The Eyewitness as an Information-Processing “Instrument”
From a scientific or technical perspective, it is useful to consider an eyewitness as a biological instrument that records, recovers, and reports previously encountered events—a sequence known generically as “information processing.”6 As for any functioning instrument, there are three pieces of knowledge needed to understand its utility for application: (1) How is it going to be used? (2) How does it work? and (3) How well does it work? In the following section, we briefly summarize the ways in which eyewitnesses are commonly used in criminal investigation and prosecution, and we provide an example from an actual criminal case to illustrate some of the strategies and pitfalls. In later sections on the basic sciences of vision and memory, and the applied science of eyewitness identification, we describe how the instrument works and how well it can perform the task at hand. This knowledge is critical for understanding the predictive relationship between variables associated with an eyewitness event and the accuracy of an identification.
How Is the Instrument Used?: A Taxonomy of Identification Procedures
The procedures used by law enforcement for eyewitness identification have become standardized through adherence to disseminated guidelines.7 There are three basic procedures in use today: (1) showups, (2) photo arrays, and (3) live lineups. In addition, there are a variety of nonstandard identification procedures that are employed occasionally by police, or by eyewitnesses themselves.
6. “Information” here refers to content acquired through the senses, retrieved from memory, or derived from comparisons thereof, which improves the observer’s ability to make predictions about the state of the world.
7. See, e.g., U.S. Dep’t of Just., Eyewitness Identification: Procedures for Conducting Photo Arrays (2017), https://perma.cc/TP4F-NXCC.
Showups
In a showup, which usually occurs at or near the crime location, police officers present a single live suspect to a witness. Showup procedures are only useful shortly after the crime, as when a person matching the witness’s description is discovered in proximity to the crime scene and the witness’s memory of the events is believed to be fresh. Any benefit gained by this proximity is countered, however, by the inherent suggestiveness of a showup: The procedure presents a witness with a single choice, which may cause the witness to infer that the police believe that the suspect is the perpetrator.8 Because of this suggestiveness, showups have been, as the U.S. Supreme Court has put it, “widely condemned.”9
Photo Arrays
Photo arrays are the most commonly used eyewitness identification procedure. In this case, the eyewitness is presented with six facial photographs,10 each in a standard portrait view. One photo is that of a suspect; the other five photos are of “fillers”—people known to be innocent who are drawn from a preexisting database. The fillers serve to challenge recognition memory and reduce suggestiveness. Ideally, the filler faces should be chosen to match the witness’s description of the perpetrator (a common police practice),11 not the appearance of the suspect, because the witness’s description is a direct reflection of what they have stored in memory from the crime scene.
Photo arrays are of two types: simultaneous and sequential. In a simultaneous presentation, officers display all facial photos at the same time, commonly in a two by three arrangement known as a “six-pack” photo array. In a sequential presentation, by contrast, photos are displayed one at a time.12 Problems with suggestiveness in photo arrays may still arise if fillers are not carefully chosen. For example, a lineup that includes some fillers who do not match the witness’s description of the perpetrator reduces the number of sensible choices.
8. See Nancy Steblay et al., Eyewitness Accuracy Rates in Police Showup and Lineup Presentations: A Meta–Analytic Comparison, 27 Law & Hum. Behav. 523, 523–24 (2003), https://doi.org/10.1023/a:1025438223608.
9. Stovall v. Denno, 388 U.S. 293, 302 (1967).
10. Six is the standard used in the United States today, where it is deemed sufficient to challenge recognition memory. But there is not universal agreement on this number: In the United Kingdom, for example, lineups (“identity parades”) are commonly composed of nine faces.
11. Police Exec. Research Forum, A National Survey of Eyewitness Identification Procedures in Law Enforcement Agencies (2013), https://perma.cc/REX2-Y29T.
12. Nat’l Research Council, Identifying the Culprit: Assessing Eyewitness Identification 24 (2014) [hereinafter Identifying the Culprit], https://perma.cc/4246-37D2.
Live Lineups
Before the ready availability of standardized facial photographs, live lineups were used. In this procedure, the suspect and fillers are presented in person, either as a group (simultaneous) or one at a time (sequential). While some agencies still use live lineups, they are less common today because police may need probable cause to place a nonconsenting person in a lineup, presence of counsel may be required, and it can be practically difficult to find live fillers who fairly resemble the witness’s description of the perpetrator.13
Nonstandard Identification Procedures
A range of additional procedures may be used in situations in which officers do not have a suspect, including showing mug books or sets of photographs or asking the eyewitness to help prepare a composite image or drawing of a culprit. Police will, on occasion, display a single photograph to a witness in an effort to confirm the identity of a person of interest. Police typically limit this “confirmatory photograph” method to situations in which the person is previously known to or acquainted with the witness. Police may sometimes take a witness to view people in the field, perhaps at a location the culprit is known to frequent. Finally, eyewitnesses may try to make identifications not arranged by police, in person or on social media, offender registries, or other online image collections. All of these situations are less controlled than a lineup procedure and potentially suggestive.
An Illustrative Case
To help illustrate the process by which eyewitness evidence emerges and its potential unfolds, and to illuminate the manifold variables that come into play when a trier of fact is evaluating the accuracy of the evidence, we begin with an example from the real world. We reference this example occasionally in the remainder of the guide in order to place scientific discussions in context.
As a juror in a lengthy criminal trial, you are listening to the prosecution recount how while walking to a friend’s house, a 16-year-old student was abruptly abducted, dragged into roadside bushes, and sexually assaulted by an unknown man.14 After repeatedly punching him in the face, the victim managed to break free and run away. A passerby, in his car, saw the face of the attacker. Later, during
13. Id. at 25.
14. Thomas D. Albright, Why Eyewitnesses Fail, 114 PNAS 7758, 7758–64 (2017) [hereinafter Albright, Why Eyewitnesses Fail], https://doi.org/10.1073/pnas.1706891114.
the trial, you listen to the victim’s and passerby’s testimony. When asked if the victim was confident that the defendant sitting in the courtroom was the culprit, the victim replies: “Yes, I’m sure . . . I will never forget what he looks like.” The passerby confirms the identification, and confirms that he has no doubts at all. As a juror, your instinct is to believe the two witnesses who observed the crime and confidently conveyed their testimony. How could you not? One saw—and courageously punched—the assailant, and the other saw the assailant just moments afterward. How could either be wrong about the attacker’s identity?
The situation just described is the case of Uriah Courtney, who was sentenced to life in prison. After Courtney served eight years, however, the court granted a motion to retest DNA from the victim’s clothing. The results excluded Courtney, and he was released in 2013. The DNA evidence was linked, through a hit in a DNA database, to a man who lived near the crime scene. The wrongful conviction of Uriah Courtney naturally raises questions about the accuracy of the victim’s testimony and points to three general factors that often undermine criminal investigations and prosecutions based on eyewitness testimony: uncertainty, bias, and overconfidence. In simple terms, appreciation of the extent to which these three factors are in play can help improve inferences about the accuracy of eyewitness testimony.
One cause of uncertainty is the fact that the second witness and Courtney were of different races and ethnicities. It is well known that people generally have greater difficulty discriminating faces of a race different from their own (relative to same-race face discrimination),15 which manifests as disproportionately large eyewitness errors in cross-racial identifications and has significant implications for racial justice.
There were also potential sources of bias built into the identification procedure used in Uriah Courtney’s case. For example, the police showed a traditional simultaneous photo array to the two eyewitnesses, who had described the culprit as white, in his mid-twenties, with brown hair and a goatee. Only two of the five fillers in this lineup had any facial hair at all—and Courtney’s goatee was the most conspicuous. This lineup configuration, in conjunction with the witnesses’ descriptions, naturally affected the statistics of the identification process, leading to a high prior likelihood—a bias—for picking Courtney, which both eyewitnesses did. Additional potential for bias was introduced by the fact that the officer administering the lineups knew which participant was the suspect, which allowed for inadvertent behavioral signals that could have influenced the witness’s decisions.
15. Rankin W. McGugin et al., Race-Specific Perceptual Discrimination Improvement Following Short Individuation Training with Faces, 35 Cognitive Sci. 330–47 (2011), https://doi.org/10.1111/j.1551-6709.2010.01148.x; Hoo Heat Wong, Ian D. Stephen & David R. T. Kebble, The Own-Race Bias for Face Recognition in a Multiracial Society, 11 Frontiers Psych. 208 (2020), https://doi.org/10.3389/fpsyg.2020.00208.
Finally, during the Courtney trial the witnesses expressed marked confidence in their identifications, which likely moved the jury to convict, despite the fact that upon the victim’s first viewing of a photo lineup, she was tentative, telling police that the defendant’s photo was “most similar,” but that she was “not sure,” and had a confidence level of 60%.16 By the time of the trial, her confidence had become inflated, while the accuracy of her testimony remained unchanged.
Uriah Courtney’s wrongful conviction is but one in a lengthy and growing list that involved eyewitness misidentifications, in which accuracy of eyewitness testimony was incorrectly inferred by the court. Of the over 375 individuals in the United States who have been exonerated as of this writing based on postconviction DNA testing, approximately 70% were convicted, in part, based on eyewitness misidentifications.17 We do not know how often eyewitness identifications are conducted, but according to one estimate, they may be used in tens of thousands of cases a year.18
Predicting the Accuracy of Eyewitness Identification
As emphasized above, it is impossible for a user—law enforcement or the courts—to rationally and responsibly employ eyewitness testimony without information about the probability that the testimony is correct. But where does that information come from? There is no oracular source, and witness statements about their own accuracy are naturally suspect. As with other indeterminate systems, such as medical prognoses or weather forecasting, we must draw from other pieces of information to predict eyewitness accuracy.
In this section we briefly review principles of accuracy prediction developed by law and science communities. Before doing so, we highlight a distinction between (1) estimates of the probability that witness testimony would be accurate under a particular set of known conditions, and (2) the probability that a specific witness is accurate.19 As we will show, knowledge gained from both
16. See Uriah Courtney, Innocence Center, https://theinnocencecenter.org/case/uriah-courtney/.
17. See Brandon L. Garrett, Convicting the Innocent: DNA Exoneration Resource Database, available at https://www.convictingtheinnocent.com (providing detailed information concerning DNA exoneration cases); Garrett, supra note 2 (summarizing DNA exoneration data); Brandon L. Garrett, Judging Innocence, 108 Colum. L. Rev. 55 (2008) (“The vast majority of the exonerees (79%) were convicted based on eyewitness testimony.”).
18. Identifying the Culprit, supra note 12 (citing Alvin G. Goldstein, June E. Chance & Gregory R. Schneller, Frequency of Eyewitness Identification in Criminal Cases: A Survey of Prosecutors, 27 Bull. Psychonomic Soc’y 71, 73 (1989) https://doi.org/10.3758/bf03329902); Garrett, supra note 2, at 50.
19. See David L. Faigman, John Monahan & Christopher Slobogin, Group to Individual (G2i), Inference in Scientific Expert Testimony, 81 U. Chi. L. Rev. 417, 417–80 (2014).
basic and applied sciences makes it possible, in principle, to offer quantitative estimates of accuracy under known conditions. By contrast, we would discourage attempts by experts to apply a numerical probability to a specific witness, recognizing that doing so wades into the province of the court, and there are numerous unknown variables that bear on individual cases.
Manson v. Brathwaite: Accuracy Prediction by the Court
In its landmark 1977 Manson v. Brathwaite ruling on the use of eyewitness evidence, the U.S. Supreme Court explicitly promoted a predictive strategy for assessing the accuracy of eyewitness reports, which was intended to provide orderly grounds for decisions about the admissibility of the evidence under the Due Process Clause. This strategy involves application of a five-factor test, in which specific conditions of witnessing and reporting are compared to five conditional standards—often termed “reliability criteria”—believed to be correlated with (and potentially causal of) the probability that a witness’s testimony is correct. We briefly review these criteria and the logic behind them here because they provide the constitutional judicial standard for admissibility of eyewitness evidence (although as discussed, federal and state rules of evidence also govern eyewitness testimony in court, as well as state statutes). For each, we suggest how growing scientific knowledge of human information processing (reviewed in greater detail below) supports a predictive strategy and may bring us closer to precisely estimating the probability that eyewitness testimony is correct.20
The first Manson reliability criterion—“the opportunity of the witness to view the criminal at the time”—is intended to address the question of visual fidelity, or accuracy of a witness’s perceptual experience. An “opportunity to view” (commonly termed “line-of-sight visibility” by the scientific community21)
20. The disconnect between the Manson “reliability criteria” and modern sciences of vision and memory has been noted by others: Gary L. Wells & Deah S. Quinlivan, Suggestive Eyewitness Identification Procedures and the Supreme Court’s Reliability Test in Light of Eyewitness Science: 30 Years Later, 33 L. & Hum. Behav. 1, 1–24 (2009); Randolph N. Jonakait, Reliable Identification: Could the Supreme Court Tell in Manson v. Brathwaite?, 52 U. Colo. L. Rev. 511 (1981); Ruth Yacona, Manson v. Brathwaite: The Supreme Court’s Misunderstanding of Eyewitness Identification, 39 J. Marshall L. Rev. 539 (2006); Laura Smalarz et al., Psychological Science on Eyewitness Identification and the U.S. Supreme Court: Reconsiderations in Light of DNA-Exonerations and the Science of Eyewitness Identification, in The Witness Stand and Lawrence S. Wrightsman, Jr. (Cynthia Willis-Esqueda & Brian Bornstein eds., 2016).
21. M.L. Benedikt, To Take Hold of Space: Isovists and Isovist Fields, 6 Environ. & Planning B: Urb. Analytics & City Sci. 1 (1979), https://doi.org/10.1068/b060047.
is essential, to be sure, but the current state of vision science, much of which is focused on the very question of visual sensitivity and fidelity, is such that estimates of accuracy can now be made beyond the limiting case of what is within an observer’s line of sight. In particular, observers do not simply view or not view things in their environment; there is a predictive relationship between variables that introduce uncertainty in visual experience and the accuracy of observer reports.
The second Manson criterion—“the witness’ degree of attention”—acknowledges that attention promotes greater visual fidelity and certainty of measurement, and is thus predictive of accuracy. Modern science informs our understanding of this variable as well. Specifically, attention is a limited resource; direction of attention to one part of a visual scene routinely precludes attention to other parts. Thus, the “degree of attention” is primarily meaningful in the context of what is being attended to. As detailed under discussions of scientific research (see sections titled “Visual attention” and “Weapons focus” below), a high degree of attention to a weapon, for example, will naturally draw attentional focus away from the face, introducing uncertainty in visual experience of the face and reduction of identification accuracy. Building upon this information, it is now possible to make refined predictions of accuracy based on both the degree and the likely target of attention.
The third Manson criterion—“the accuracy of [the witness’s] prior description of the criminal”—is based on the premise that a witness’s ability to recall what the perpetrator looked like (“accuracy of his prior description”) can be used to infer the probability that an identification made by the witness is correct. This is a valid premise for familiar objects, on which attention has been focused, perceptual processing is commonly “deep,” and the level of detail recalled is significant.22 By contrast, modern science reveals that the correlation between recall and recognition is not strong under conditions in which unfamiliar faces are witnessed under viewing conditions constrained by time, distance, lighting, stress, and poorly focused attention.23 All of which suggests that the predictive value of a good description may be substantially limited by the conditions of viewing.
The fourth Manson criterion—“the level of certainty demonstrated at the confrontation”—is rooted in the simple premise that “level of certainty,” or confidence, demonstrated at the “confrontation” (at the time of the initial identification test) is a predictor of eyewitness identification accuracy. This premise matches human intuition but, as we discuss in some detail below, modern science has shown that the relationship between confidence and accuracy of recognition
22. Fergus I.M. Craik & Robert S. Lockhart, Levels of Processing: A Framework for Memory Research, 11(6) J. of Verbal Learning & Verbal Behav. 671–84 (1972).
23. Melissa Pigott & John C. Brigham, Relationship Between Accuracy of Prior Description and Facial Recognition, 70 J. Applied Psych. 547 (1985), https://doi.org/10.1037//0021-9010.70.3.547.
memory is nuanced. Under some well-defined conditions,24 the initial confidence report may correlate with the probability that the testimony is accurate.25 As we see from the Courtney case, the predictive value of this measure can grow worse with the passage of time, largely owing to outside influences (e.g., reinforcement from media and counsel) that boost confidence but naturally have no effect on accuracy, reaching a nadir at the time of the in-court identification.
The fifth Manson criterion—“the length of time between the crime and the confrontation”—acknowledges that time-dependent memory loss increases uncertainty of recalled information, which means that memory retention time is inversely correlated with accuracy. Because accuracy is a continuous function of time and admissibility decisions are binary, a judge must apply a time threshold for evaluation. But where does that threshold come from? Scientific studies of the time–accuracy relationship show that memory loss can be modeled as a power law function, by means of which “not only can the percentage of remaining memory strength be determined for any retention interval, but this strength estimate can be translated into an estimated probability of being correct on a fair lineup of a specified size.”26 In other words, modern science can provide a quantitative, empirical foundation for the time threshold that a judge employs when evaluating evidence by the fifth Manson criterion.
The day-to-day judicial approach to eyewitness evidence has been largely driven over the past half-century by the predictive framework of Manson v. Brathwaite, at least in the federal courts. State courts, as we will discuss, have introduced more recent changes, through legislation and judicial rulings, seeking to more directly incorporate scientific insights into approaches toward eyewitness evidence. In federal courts, as in state courts, scientific insights have influenced the interpretation of the Manson factors and other related approaches, such as the use of expert witnesses to explain eyewitness evidence to the jury, and the use of jury instructions regarding eyewitness evidence. In the meantime, the scientific community has continued to weigh in on the important matter of inferring the probability that eyewitness testimony is correct, based primarily on new scientific discoveries in the area of human information processing and the use of new statistical tools.
24. “Pristine” lineup conditions include choosing lineup fillers such that they similarly match the witness’s description of the perpetrator, making the lineup “fair,” and imposing limits on biasing factors such as nonblinded lineup administration.
25. John T. Wixted & Gary L. Wells, The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis, 18 Psych. Sci. Pub. Int. 10 (2017), https://doi.org/10.1177/1529100616686966.
26. Kenneth A. Deffenbacher et al., Forgetting the Once-Seen Face: Estimating the Strength of an Eyewitness’s Memory Representation, 14 J. Experimental Psych.: Applied 139, 142 (2008), https://doi.org/10.1037/1076-898x.14.2.139.
Accuracy Prediction by the Scientific Community
Motivated by societal concerns over wrongful convictions, which began to mount significantly with the development of efficient and effective procedures for forensic genotyping (starting approximately a decade after the Supreme Court’s 1977 Manson ruling), a number of science–law collaborations emerged to explore how new scientific knowledge and tools might improve collection and use of eyewitness testimony.27 The scientific foundations continue to evolve and sharpen, as do various recommendations for practice and reform.
In 2013, the National Research Council (NRC) convened a panel of experts to consider the use and validity of eyewitness identification for forensic purposes. The panel was composed of individuals representing a variety of relevant fields, including the scientific study of visual perception and memory, sociology, statistics, law, and policing. After reviewing evidence from many sources, the panel released a comprehensive report in the fall of 2014.28
Other organizations have subsequently performed reviews. Particularly noteworthy and commendable is the 2020 scientific review paper from the American Psychology-Law Society,29 which covers much of the same scientific ground as the NRC report, but concentrates almost exclusively on recommendations to law enforcement for planning, designing, and conducting eyewitness identification procedures. In the remainder of this section, we focus primarily on the NRC report and its recommendations for law enforcement, for the courts, and for the advancement of eyewitness science.
The NRC report recognized the great value that eyewitness testimony can bring to criminal proceedings and the concerted efforts of law enforcement and
27. Gary L. Wells et al., Eyewitness Identification Procedures: Recommendations for Lineups and Photospreads, 22 Law & Hum. Behav. 603 (1998), https://doi.org/10.1023/a:1025750605807; U.S. Dep’t of Just., Off. of Just. Programs & Nat’l Inst. of Just., Eyewitness Evidence: A Guide for Law Enforcement (1999), https://perma.cc/Y9EA-B4W3; John W. Turtle, R.C. Lindsay & Gary L. Wells, Best Practice Recommendations for Eyewitness Evidence Procedures: New Ideas for the Oldest Way to Solve a Case, 1(1) Canadian J. of Police & Sec. Servs. 5 (2003); Report of the United States Court of Appeals for the Third Circuit Task Force on Eyewitness Identifications, 92 Temp. L. Rev. 1, 6 (2019), https://perma.cc/Q4CH-5NLU; Gary L. Wells et al., Policy and Procedure Recommendations for the Collection and Preservation of Eyewitness Identification Evidence, 44 Law & Hum. Behav. 3 (2020), https://doi.org/10.1037/lhb0000359; Margaret Bull Kovera et al., Science-Based Recommendations for the Collection of Eyewitness Identification Evidence, 58 Ct. Rev. 130 (2022).
28. Identifying the Culprit, supra note 12 (one author of the present guide served as co-chair of the National Resarch Council consensus report committee, and the other author served as a committee member).
29. Wells et al., Policy and Procedure Recommendations (2020), supra note 27.
legal professionals. At the same time, however, the NRC report identified factors that may lead to wrongful conviction and made recommendations for improvement in three areas: law enforcement procedure, use of eyewitness evidence in the courts, and scientific research. With respect to eyewitness performance, the NRC recommendations are both prospective, with an eye toward eliciting greater eyewitness accuracy through procedural reforms, and retrospective, with a focus on identifying and quantifying variables that offer predictive insight into the probability that eyewitness testimony is correct.
These recommendations have been embraced by both scientific and legal communities and have had, in a few short years, a significant positive impact on the field. The NRC report itself, which is freely available from the National Academies Press, has been downloaded more than 17,000 times and has elicited rich discussion, new scientific research, and improved legal practice. In 2017, U.S. Deputy Attorney General Sally Yates issued new Department of Justice (DOJ) guidelines for the conduct of eyewitness investigations,30 which hew tightly to the recommendations and language of the NRC report. Though the charter of the independent NRC defines and limits its work to be strictly advisory—unlike the DOJ, the NRC has no enforcement powers—we briefly highlight the NRC recommendations on eyewitness identification in this section, as they offer a scholarly complement to the predictive framework established by the Supreme Court in 1977.
Law Enforcement
The five NRC recommendations for law enforcement are intended to foster “adherence to guidelines that are consistent with scientific standards for data and reporting.”31 These include:
- Train all law-enforcement officers in eyewitness identification
- Implement double-blind lineup and photo array procedures
- Develop and use standardized witness instructions
- Document witness confidence judgments
- Videotape the witness identification process
These law enforcement recommendations are aimed prospectively at improving the quality of evidence elicited from eyewitnesses and presented to
30. See U.S. Dep’t of Just., supra note 7.
31. See Identifying the Culprit, supra note 12, at 105.
the court. Adherence to these recommendations can reduce the likelihood that uncertainty, bias, and overconfidence come into play, and make it easier to retrospectively predict the probability that an identification hypothesis is true.
The Courts
The four NRC recommendations for the courts encourage adoption of procedures intended to help courts distinguish evidence that does or does not have high predictive validity. These include:
- Conduct pretrial judicial inquiry
- Make juries aware of prior identifications
- Use scientific framework expert testimony
- Use jury instructions as an alternative means to convey information
Pretrial inquiry offers an opportunity for the judge to indulge in a careful consideration of variables (those identified by Manson, or others) that might serve retrospectively as good predictors of the accuracy of an identification. Use of scientific framework testimony (testimony about relevant scientific facts and principles, without explicit application to the case in question) and use of “clear and concise” jury instructions both provide similar opportunities for jurors to consider variables that are good predictors of the accuracy of an identification.32 Finally, the recommendation that juries be made aware of prior identifications is intended to establish the shortest retention interval, reveal inconsistencies, and combat the influence of confidence inflation on jury deliberations. All of which should improve the quality of evidence used retrospectively to predict the accuracy of an identification.
Scientific Research
Finally, the NRC report made several recommendations for additional scientific research, recommendations aimed at developing strategies to both prospectively elicit more accurate testimony from eyewitnesses, and—for retrospective purposes—to gain greater understanding of the causal relationships between variables associated with witnessed events and accuracy of the resulting
32. Jury instruction was an explicit recommendation of the NRC eyewitness committee. Nonetheless, as summarized below in the context of scientific studies conducted in the laboratory, some research suggests that jury instructions are not always effective at preventing wrongful convictions.
testimony. As summarized in detail in the next sections, this scientific research has progressed at a rapid pace, and particularly so since the publication of the NRC report.
Scientific Questions Regarding Eyewitness Identification
The past half-century has seen the accumulation of a vast trove of scientific knowledge bearing on the validity of eyewitness testimony. This knowledge has greatly enhanced understanding of the capabilities and limitations of this form of evidence, and has advised new policies and practices. In the foregoing sections we distilled some of this science to provide ready insights for common judicial decisions. In the following sections we take a deep dive into the scientific discoveries that informed our summary, with the conviction that this knowledge will serve as a comprehensive reference that allows for greater judicial flexibility and understanding of when, why, and with what probability eyewitnesses provide accurate testimony. First, we briefly outline the types of scientific questions that are commonly addressed. This is followed by a summary of findings from: (1) basic laboratory studies that ask how brain systems for vision and memory work; (2) applied laboratory studies that specifically focus on human performance in an eyewitness task; (3) field studies that test the effectiveness of new eyewitness tasks in real criminal justice settings; and (4) studies that explore the efficacy of communication between the eyewitness and the trier of fact.
Most scientific studies of eyewitness identification conducted over the past half-century have focused on three central questions:
- Can we improve prediction of eyewitness identification accuracy?
When an eyewitness makes an identification, the question naturally of foremost importance for the criminal justice system is: What is the probability that the eyewitness is correct? This is the retrospective question, which focuses on variables associated with things past (e.g., conditions of witnessing and collection of evidence by law enforcement) that can be used to infer the accuracy of eyewitness reports. Nearly all of the science relevant to eyewitness evidence has addressed this question, either directly or indirectly.
- Are there procedural improvements that can increase accuracy?
The second, prospective, question seeks strategies that the criminal justice system might use to increase the probability of accurate testimony. The scientific literature is thick with empirical comparisons of performance (e.g., accuracy) elicited by different identification procedures, such as showups versus lineups, or
simultaneous versus sequential lineup presentations. The answer may favor the use by police of one procedure over the other. Most importantly for the present context, awareness by the trier of fact of these prospective improvements may serve retrospectively to help infer the probability that eyewitness testimony is correct.
- Can we educate jurors to be better evaluators of accuracy?
The third scientific question addresses the outsized influence of eyewitness testimony on decisions by juries. The importance of this issue can be appreciated from a scientific perspective in which the use of eyewitness evidence by the courts is viewed as a process of communication: The eyewitness is the “transmitter” of information. The jury, conversely, is the “receiver” of information. If that receiver is flawed by an inability to interpret the signals that were transmitted—either by uncritical acceptance or by lack of technical knowledge—then the fidelity of information transmitted (the accuracy of the testimony) makes little difference to the outcome. It follows that the quality of judicial decisions might benefit from plain language explanations of (1) the underlying processes of vision and memory, (2) the ways that those processes can fail, contrary to intuition, and (3) why eyewitness candor and confidence are not necessarily predictive of accuracy. Carefully worded jury instructions and scientific expert testimony have been implemented in a number of jurisdictions. While some evidence challenges the effectiveness of this approach,33 recent experiments suggest that instructions may be more effective if they explain the reasons why caution is critical when evaluating eyewitness accuracy, rather than simply directing caution.34 We summarize the relevant science below.
33. Amanda N. Bergold et al., Eyewitnesses in the Courtroom: A Jury-Level Experimental Examination of the Impact of the Henderson Instructions, 17 J. Experimental Criminology 433, 433–55 (2021), https://doi.org/10.1007/s11292-020-09412-3; Marlee Kind Berman, Eyewitness Identification Jury Instructions: Do They Enhance Evidence Evaluation?, C.U.N.Y. Acad. Works (2015); Angela M. Jones & Amanda N. Bergold, Examining the Effectiveness of the Henderson Eyewitness Instructions, 13 J. Forensic Psych. Prac. 1, 1–24 (2017), https://doi.org/10.1007/s11292-016-9279-6; Brandon L. Garrett et al., Factoring the Role of Eyewitness Evidence in the Courtroom, 17 J. Empirical Legal Stud. 556, 556–79 (2020), https://doi.org/10.1111/jels.12259; Marlee Kind Dillon et al., Henderson Instructions: Do They Enhance Evidence Evaluation?, 17 J. Forensic Psych. Rsch. & Practice 1 (2017), https://doi.org/10.1080/15228932.2017.1235964; Angela M. Jones et al., Comparing the Effectiveness of Henderson Instructions and Expert Testimony: Which Safeguard Improves Jurors’ Evaluations of Eyewitness Evidence?, 13 J. Exp. Crim. 29, 29–52 (2017), https://doi.org/10.1007/s11292-016-9279-6; Cindy E. Laub, Christopher D. Kimbrough & Brian H. Bornstein, Mock Jurors’ Perceptions of Eyewitnesses Versus Earwitnesses: Do Safeguards Help?, 34(2) Am. J. Forensic Psych. 33 (2016); Athan P. Papailiou, David V. Yokum & Christopher T. Robertson, The Novel New Jersey Eyewitness Instruction Induces Skepticism but Not Sensitivity, 10(12) PLoS One e0142695 (2015), https://doi.org/10.1371/journal.pone.0142695.
34. Brandon L. Garrett et al., Sensitizing Jurors to Eyewitness Confidence Using “Reason-Based” Judicial Instructions, 12(1) J. Applied Rsch. Memory & Cognition 141 (2022), https://doi.org/10.1037/mac0000035.
Basic Science of Vision, Memory, and Choice
Vision, memory, and decision-making have been subjects of rigorous scientific investigation since the middle of the nineteenth century, with particularly rapid advances over the past 50 years. In simple terms, the goal has been to understand how and how well the machine works. If we’re considering the purchase of a new car, violin, or patio heater, the first thing we want to know is what are the use conditions under which it performs well or fails. Unlike engineered products, however, human vision and memory do not come with performance specs and a user manual. The purpose of basic research in this area is to obtain information that can enable prediction of human performance under specific conditions, such as those associated with witnessing and reporting a crime.35 In the following sections, we summarize what has been learned from this basic science.
How Vision Works
Vision is usefully understood as the process of detecting informative signals about the external world and using those signals to recognize objects, make decisions, and guide behavior.36 Our focus here is on aspects of visual function that place constraints on what can be sensed and perceived. In two sections, we address (1) characteristics of vision that place upper bounds on performance, and (2) vision as a constructive process that seeks to overcome input noise and uncertainty.
Upper Bounds on Visual Performance
In this section we review some specific aspects of visual processing that place upper bounds on the quantity and quality of visual information available under defined viewing conditions. This knowledge offers a starting point for evaluating the probability that an eyewitness could have seen what they said they saw.37
35. See Identifying the Culprit, supra note 12; Albright, Why Eyewitnesses Fail, supra note 14, at 7758–64; Marc Green et al., Forensic Vision with Application to Highway Safety (3rd ed. 2008).
36. Charles D. Gilbert, The Constructive Nature of Visual Processing, in Principles of Neural Science 496 (Eric R. Kandel et al. eds., 6th ed. 2021); Charles D. Gilbert, Intermediate-Level Visual Processing and Visual Primitives, id. at 545; Thomas D. Albright & Winrich A. Freiwald, High-Level Visual Processing: From Vision to Cognition, id. at 564; Thomas D. Albright, On the Perception of Probable Things: Neural Substrates of Associative Memory, Imagery, and Perception, 74 Neuron 227 (2012), https://doi.org/10.1016/j.neuron.2012.04.001.
37. See Identifying the Culprit, supra note 12; Albright, Why Eyewitnesses Fail, supra note 14; Green et al., supra note 35.
Visual sensitivity
At the earliest stages of visual processing there are several factors that limit the visual information available for accurate perception and object recognition. Some of these factors are inherent to the visual system and largely uncontrollable, such as scattering of light by the fluid and tissues of the eye, and can be exacerbated by common observer-specific visual deficits, such as myopia, poor contrast sensitivity, or color blindness. Other factors are associated with viewing conditions, such as viewing duration and ambient illumination, which predictably influence the quantity of information that a viewer gains from a visual scene, and thus the fidelity of experience.
Anyone with a functioning visual system will report that their ability to resolve fine details of a viewed scene declines steeply in dim light; this is why we use flashlights, and why we cannot read a restaurant menu under candlelight. The basic science of vision puts numbers behind this experience, which can be used to predict the accuracy of eyewitness testimony.38 Figure 1 shows how this effect can be quantified. The y-axis plots the amount of luminance contrast between an image feature and the background—for example, the brightness difference between a small black letter on a white page—that is required to resolve the image feature.39 The x-axis plots ambient illumination. The plot shows that under very dim light (near total darkness; left side of x-axis), the image feature can be resolved only if it is extremely bright relative to background. By contrast, at high levels of illumination (daylight; right side of x-axis), feature resolution is easy. But there are intermediate levels of illumination (moonlight or candlelight; middle of x-axis), which are frequently encountered by human observers and make resolution very difficult.
The plot in Figure 1 represents one of the operating characteristics of human vision. In conjunction with measured values of ambient illumination present at a witnessed crime scene—knowable, in principle, from measurements of light sources, such as daylight, moonlight, street lights, illuminated windows, or signage—these operating characteristics can be used to predict the fidelity of experience and, in the spirit of Manson, assess with respectable precision whether a witness had an opportunity “to view the criminal at the time.” It is precisely for this reason that human operating characteristics—hard-won products of basic research—are such a valuable resource for accuracy predictions in the eyewitness context.
38. Percy W. Cobb, The Influence of Illumination of The Eye on Visual Acuity: I. Introductory and Historical, 29 Am. J. Physiology 76 (1911), https://doi.org/10.1152/ajplegacy.1911.29.1.76; Simon Shlaer, The Relation Between Visual Acuity and Illumination, 21 J. Gen. Physiology 165 (1937), https://doi.org/10.1085/jgp.21.2.165.
39. J.L. Barbur & A. Stockman, Photopic, Mesopic and Scotopic Vision and Changes in Visual Performance, in Encyclopedia of the Eye, vol. 3, 323 (Darlene A. Dartt et al. eds., 2010), https://doi.org/10.1016/b978-0-12-374203-2.00233-5.
Source: Used with permission of Elsevier Science & Technology Journals, from J. L. Barbur & A. Stockman, supra note 39, at 323 (permission conveyed through Copyright Clearance Ctr., Inc.).
Another compelling example comes from quantitative information about visual acuity.40 In Figure 2A, the y-axis is a measure of visual acuity. The x-axis is a measure of angular distance from the center of gaze. Normal maximum acuity is 20/20, which is achieved at the center of gaze (0 deg “retinal eccentricity” in this plot). At a viewing angle of 2 degrees from the center of gaze (about the width of a thumb at arm’s length), acuity declines by half. The perceptual consequences of this decline in acuity are illustrated in Figure 2B, where letters are scaled to be equally resolvable when directing eye gaze to the center of the image. Under time-limited, distracting, and stressful viewing conditions associated with viewing a typical crime, off-axis gaze direction—looking directly at other compelling features (such as weapon, a bleeding victim, or a second perpetrator)—can be expected to place severe constraints on the ability to detect, resolve, and memorize the face of the primary culprit, because that face image will be registered only by low-acuity regions of the sensory field.41 It follows that if one has facts about the scene, such as viewing position, distances, and angles, in conjunction
40. S.M. Anstis, A Chart Demonstrating Variations in Acuity with Retinal Position, 14(7) Vision Rsch., 589 (1974).
41. Haus Strasburger, Ingo Rentschler & Martin Jüttner, Peripheral Vision and Pattern Recognition: A Review, 11 J. Vision 13 (2011), https://doi.org/10.1167/jov.10.1167/11.5.13.
Source (Panel A): Stanley Lambertus et al., Highly Sensitive Measurements of Disease Progression in Rare Disorders: Developing and Validating a Multimodal Model of Retinal Degeneration in Stargardt Disease, 12 PLoS One (2017), https://doi.org/10.1371/journal.pone.0174020. Copyright: © 2017 Lambertus et al. This is an open access article distributed under the terms of the Creative Commons Attribution License. (Panel B): Used with permission of Elsevier Science & Technology Journals, from Anstis, supra note 40, at 589–92 (permission conveyed through Copyright Clearance Ctr., Inc.).
with the acuity function shown above in Figure 2A, it should be possible to draw probabilistic inferences about the degree to which the witness had an opportunity “to view the criminal,” bearing (per Manson) on the likelihood that the witness testimony is accurate.42
The ability to detect, resolve, and interpret image content is further affected in known ways by a myriad of other environmental factors, such as atmospheric conditions (rain, fog), glare, and object surface reflectance.43 “Crowding,” in which objects in a visual scene are closely spaced, significantly reduces the ability
42. There is a large literature on this same topic applied to the ability of field athletes (e.g., soccer, football, tennis) to perceive critical events as a function of gaze angle, e.g., Karl Marius Aksum et al., What Do Football Players Look At? An Eye-Tracking Analysis of the Visual Fixations of Players in 11 v. 11 Elite Football Match Play, 11 Frontiers Psych. (2020), https://doi.org/10.3389/fpsyg.2020.562995; Francisco Belda Maruenda, Can the Human Eye Detect an Offside Position During a Football Match?, 329 BMJ 1470–72 (2004).
43. Green et al., supra note 35.
of observers to quickly detect and identify specific objects.44 Figure 3 illustrates the perceptual consequence of visual crowding. In this image, letters are scaled, as in Figure 2B, to compensate for loss of visual acuity. In this case, however, the letters are “crowded” together in the central rings. As a consequence, the ability to detect and recognize specific letters is impaired. This effect suggests that a crime committed in a visually complex scene, such as a sporting event or a crowded street, will place severe and predictable limits on the ability of a witness to accurately perceive the facial features of a perpetrator.
Vision is also limited by image distortion (foreshortening) associated with angle of view. The retinal pattern generated by a face viewed directly from the front differs considerably—with changes in aspect ratio and relative placement of facial features—from that generated by a face viewed from an oblique side angle. Viewing a face from an angle above or below center (if the criminal were standing over you, or below you on the stairs) also yields retinal distortions
Source: Used with permission of Elsevier Science & Technology Journals, from Anstis, supra note 40, at 589-592 (permission conveyed through Copyright Clearance Ctr., Inc.).
44. Dennis M. Levi, Crowding—An Essential Bottleneck For Object Recognition: A Mini-Review, 48 Vision Rsch. 635 (2008), https://doi.org/10.1016/j.visres.2007.12.009; David Whitney & Dennis M. Levi, Visual Crowding: A Fundamental Limit on Conscious Perception and Object Recognition, 15 Trends Cognitive Sci. 160 (2011), https://doi.org/10.1016/j.tics.2011.02.005; Ryan V. Ringer et al., Investigating Visual Crowding of Objects in Complex Real-World Scenes, 12 i-Perception (2021), https://doi.org/10.1177/2041669521994150.
(foreshortening) of facial features, as shown in Panel A of Figure 4.45 Panel B of Figure 4 plots the ability of human observers to discriminate (d’) a frontal view (0 degrees) from the same or different face tilted up or down. Discriminability declined markedly with tilt angle. (The seven data points in Panel B correspond to the seven view angles in Panel A.) These effects are largely overcome when viewing familiar people because we have extensive prior knowledge of appearance that is independent of view angle. When viewing unfamiliar faces under time-limited conditions, however, this image distortion can severely impair subsequent recognition.46
Note: Panel A was created by the authors using parts extracted from Figure 1 on page 3 of source. Panel B was created by the authors by modifying Figure 2, Panel A, on page 5 of source.Source: Adapted from Simone K. Favelle & Stephen Palmisano, The Face Inversion Effect Following Pitch and Yaw Rotations: Investigating the Boundaries of Holistic Processing, 3 Frontiers Psych. 563 (2012), https://doi.org/10.3389/fpsyg.2012.00563. Copyright © 2012 Favelle and Palmisano. This is an open-access article distributed under the terms of the Creative Commons Attribution License.
45. Simone K. Favelle & Stephen Palmisano, The Face Inversion Effect Following Pitch and Yaw Rotations: Investigating the Boundaries of Holistic Processing, 3 Frontiers Psych. 563 (2012), https://doi.org/10.3389/fpsyg.2012.00563.
46. In the extreme, unconventional views of a common object, such as a bucket viewed from above, make recognition nearly impossible in the absence of context.
Visual attention
William James, the great nineteenth-century experimental psychologist, famously articulated a subjective phenomenon familiar to us all:
Everyone knows what attention is. It is the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought. Focalization, concentration, of consciousness are of its essence. It implies withdrawal from some things in order to deal effectively with others.47
By this rule, attention, which is required for awareness, is a limited resource. Thus, only a small fraction of the information falling on the retina reaches awareness or is used by an observer for recognition, action, or storage in memory. Allocation of selective attention is an active process that can be directed by external factors—visual attributes with high salience, such as a bright light or an unfamiliar object—or internal control.48 For example, if you are searching for your phone, you may explicitly direct your attention to the countertop where it was last seen.
Attended image content is transiently enhanced—equivalent to turning up the volume of specific sensory signals—to increase the fidelity of perceptual experience. Image content not falling within the focus of attention is processed with less fidelity.49 In some cases, unattended content is effectively invisible: It does not reach awareness, it is not perceived, and it is not available for use in guiding decisions or actions, or for storage in memory.50
Different pieces of visual information compete dynamically for attentional selection based on their changing physical attributes, locations in space, and relevance to the observer’s needs and behavioral goals.51 An eyewitness must select what to attend to, often within a short window of time, without advance warning, in the presence of many novel objects and events, and under the confounding influences of anxiety and fear. Reductions in efficiency are common under such conditions, causing sensitivity to unattended items to be markedly reduced when there are many objects simultaneously competing for attention.52
47. William James, The Principles of Psychology (Henry Holt ed., 1890).
48. Michael I. Posner, Orienting of Attention, 32 Q. J. Experimental Psych. 3 (1980), https://doi.org/10.1080/00335558008248231.
49. Id.
50. Arien Mack & Irvin Rock, Inattentional Blindness (1998).
51. Harold Pashler, James C. Johnston & Eric Ruthruff, Attention and Performance, 52 Ann. Rev. Psych. 629 (2001), https://doi.org/10.1146/annurev.psych.52.1.629; Robert Desimone & John Duncan, Neural Mechanisms of Selective Visual Attention, 18 Ann. Rev. Neuroscience 193 (1995), https://doi.org/10.1146/annurev.ne.18.030195.001205.
52. James, supra note 47.
A simple demonstration of the complexity of attentive effects on awareness comes from the well-studied phenomenon of “visual search.”53 Consider the two images in Figure 5. In panel A, the green “target” stimulus immediately “pops out” perceptually from the field of red objects and commands attention, simply because it is distinguished by a difference in color; shape varies but is irrelevant to the task. By contrast, the target stimulus in panel B is distinguished by a unique conjunction of color and shape (green and circular); detection of this target requires an active attentional search that proceeds serially through the objects in the display.54 What this means, of course, is that during time-limited viewing of an unfolding crime, a witness’s attention will be automatically drawn to objects that are readily distinguished by a unique visual attribute—the woman in the red dress. Objects that are defined by unique conjunctions of features—the man on the bus in the gray sweater, surrounded by a field of men in gray coats and blue sweaters—will be harder to attend and, perhaps most importantly, harder to remember because of combinatorial complexity.
The effect of feature conjunctions on the perceptual experiences (also called percepts) that we commit to memory is illustrated by the image in Figure 6.55 Look briefly at this image and then look away. Ask yourself whether the image contained a green letter A, a B surrounded by a diamond, or a red D in a circle. The red D in a circle is indeed present, but the A is blue and the B is surrounded
53. Anne M. Treisman & Garry Gelade, A Feature-Integration Theory of Attention, 12 Cognitive Psych. 97 (1980), https://doi.org/10.1016/0010-0285(80)90005-5.
54. This search process has been described metaphorically as scanning the visual display using a “spotlight” of attention, which brings features in and out of awareness. Id.
55. This demonstration appears in Jeremy M. Wolfe, Visual Search, in Stevens’ Handbook of Experimental Psychology and Cognitive Neuroscience (John T. Wixted ed., 2022), https://doi.org/10.1002/9781119170174.epcn213.
by a triangle. Your inability to accurately report what you have seen in this demonstration reflects the visual system’s propensity to render “illusory conjunctions” of features.56 The individual feature parameters—color, surrounding shape, and letter—are physically bound together in the image, but they are phenomenally unbound by vision, such that they easily recombine perceptually in forms that do not exist in the image. Once again, the implications for accurate eyewitness identification are significant: The person on the train platform who pulled the trigger had a mustache, brown hair, and light skin, but you may easily confuse these features with those of other people present in the scene, such that you describe him as possessing a goatee, blond hair, and light skin.
Another adverse consequence of attentional selection is that competing elements in a scene can hijack the attentional focus. The technique of misdirection—one of the essential tools of performance magic—directs attention to uninformative image content and exploits the invisibility of unattended features.57 The well-studied phenomenon of inattentional blindness is an example of misdirection,
Source: Used with permission of John Wiley & Sons, Inc., from Jeremy M. Wolfe, Visual Search, in Stevens’ Handbook of Experimental Psychology and Cognitive Neuroscience (John T. Wixted ed., 2022), https://doi.org/10.1002/9781119170174.epcn213.
56. Anne Treisman & Hilary Schmidt, Illusory Conjunctions in the Perception of Objects, 14 Cognitive Psych. 107 (1982), https://doi.org/10.1016/0010-0285(82)90006-8.
57. Gustav Kuhn & Luis M. Martinez, Misdirection–Past, Present, and the Future, 5 Frontiers Hum. Neuroscience 172 (2012), https://doi.org/10.3389/fnhum.2011.00172; Stephen Macknik & Susana Martinez-Conde with Sandra Blakeslee, Sleights of Mind: What the Neuroscience of Magic Reveals About Our Everyday Deceptions (1st ed. 2010).
which occurs readily in real-world situations.58 In this case, attention directed to one behaviorally significant property of a visual scene precludes awareness of other potentially important features.59 As a result, an individual can be surprisingly unaware of surreptitious changes to the physical appearance of an unfamiliar person while the two are engaged in conversation.60 This finding suggests that events that briefly divert attention have the potential to markedly impair the accuracy of eyewitness identifications.61
Attentional hijacking is particularly characteristic of stimuli that elicit strong emotional responses. Visual stimuli that trigger fear responses act as powerful external cues that command attention. While this potentiates sensitivity to those stimuli, at the expense of lost sensitivity to others, it is often the case that the attended emotional stimuli are not the ones with relevant informational content. The so-called weapon focus effect (reviewed below) is a real-world case in point for eyewitness identification, in which attention is compellingly drawn to emotionally laden stimuli, such as a gun or a knife, at the expense of acquiring greater visual information about the face of the perpetrator.62 Similarly, viewing a crime with multiple perpetrators (see section titled “Multiple Perpetrators” below) divides attentional resources such that the fidelity of perceptual information about each face is limited.63
58. Mack & Rock, supra note 50; Ulric Neisser & Robert Becklen, Selective Looking: Attending to Visually Specified Events, 7 Cognitive Psych. 480 (1975), https://doi.org/10.1016/0010-0285 (75)90019-5; Daniel J. Simons, Attentional Capture and Inattentional Blindness, 4 Trends Cognitive Scis. 147 (2000), https://doi.org/10.1016/s1364-6613(00)01455-8.
59. For a dramatic demonstration of this effect, produced by Daniel Simons and Christopher Chabris, see Daniel J. Simons, Selective Attention Test, http://tinyurl.com/inattentional-blindness; Daniel J. Simons & Christopher F. Chabris, Gorillas in our Midst: Sustained Inattentional Blindness for Dynamic Events, 28 Perception 1059 (1999), https://doi.org/10.1068/p281059.
60. Daniel J. Simons & Daniel T. Levin, Failure to Detect Changes to People During a Real-World Interaction, 5 Psychonomic Bull. & Rev. 644 (1998), https://doi.org/10.3758/bf03208840.
61. Graham Davies & Sarah Hine, Change Blindness and Eyewitness Testimony, 141 J. Psych. 423 (2007), https://doi.org/10.3200/jrlp.141.4.423-434.
62. Thomas H. Kramer, Robert Buckhout & Paul Eugenio, Weapon Focus, Arousal, and Eyewitness Memory: Attention Must be Paid, 14 Law & Hum. Behav. 167 (1990), https://doi.org/10.1007/bf01062971; Kerri L. Pickel, Stephen J. Ross & Ronald S. Truelove, Do Weapons Automatically Capture Attention?, 20 Applied Cognitive Psych. 871 (2006), https://doi.org/10.1002/acp.1235; Elizabeth F. Loftus, Geoffrey R. Loftus & Jane Messo, Some Facts About “Weapon Focus,” 11 Law & Hum. Behav. 55 (1987), https://doi.org/10.1007/bf01044839.
63. Brian R. Clifford & Clive R. Hollin, Effects of the Type of Incident and the Number of Perpetrators on Eyewitness Memory, 66 J. Applied Psych. 364 (1981), https://doi.org/10.1037//0021-9010.66.3.364; Zoe J. Hobson & Rachel Wilcock, Eyewitness Identification of Multiple Perpetrators, 13 Int’l J. Police Sci. & Mgmt. 286 (2011), https://doi.org/10.1350/ijps.2011.13.4.253; Robert F. Lockamyeir et al., One Perpetrator, Two Perpetrators: The Effect of Multiple Perpetrators on Eyewitness Identification, 35 Applied Cognitive Psych. 1206 (2021), https://doi.org/10.1002/acp.3853; Alicia Nortje, Colin G Tredoux & Annelies Vredeveldt, Eyewitness Identification of Multiple Perpetrators, 33 S. Afr. J. Crim. Just. 348 (2020), https://perma.cc/CN2K-JHSY.
Categorical perception
The precision of object recognition can easily be derailed because visual perception is categorical.64 Although the objects of our experience vary extensively and uniquely along multiple sensory dimensions, we lump them into perceptual categories based on prior associations, many of which stem from common functions, physical properties, meanings, or emotional valence.65 Many behavioral and cognitive processes are greatly simplified by treating all members of a category as the same, despite their physical differences. It rarely matters, for example, whether the pear we choose to eat is dappled on one side or irregular in shape, nor does the text font bear greatly on our ability to read. Evidence indicates that the structure of object memory is also categorical, suggesting that perceived objects are encoded in memory as a category type, often without specific detail, such that fine details needed for subsequent individuation are lost.66
One of the functional corollaries of categorical perception is that observers are far better at discriminating between objects from different categories than objects from the same category. It is easier, for example, to discriminate between yellow and orange than it is to discriminate different shades of orange, even when the wavelength differences are the same.67 Another functional corollary is that it is possible to acquire better discriminative ability—perceptual expertise—that is sharply delimited by category boundaries.68 Through exposure and practice, we can become experts at discriminating objects within a category, such as types of sports cars, x-radiographs, and tennis shoes, that each give us little benefit when judging objects outside of that category.
Not surprisingly, a well-known phenomenon of categorical expertise is manifested as better performance on a task that involves discriminating between faces of one’s own race relative to faces of a different race. This cross-race effect
64. Robert L. Goldstone, Influences of Categorization on Perceptual Discrimination, 123 J. Exper. Psych.: Gen. 178 (1994), https://doi.org/10.1037//0096-3445.123.2.178.
65. Robert R. L. Goldstone & Andrew T. Hendrickson, Categorical Perception, 1 Wiley Inter-disc. Revs.: Cognitive Sci. 69 (2010), https://doi.org/10.1002/wcs.26.
66. Robert L. Goldstone, Yvonne Lippa & Richard M. Shiffrin, Altering Object Representations Through Category Learning, 78 Cognition 27 (2001), https://doi.org/10.1016/s0010-0277(00)00099-8.
67. W. D. Wright & F. H. G. Pitt, Hue-Discrimination in Normal Colour-Vision, 46 Proceedings Physical Soc’y 459 (1934), https://doi.org/10.1088/0959-5309/46/3/317.
68. Eleanor J. Gibson, Principles of Perceptual Learning and Development (1969); Geoffrey R. Norman et al., The Correlation of Feature Identification and Category Judgments in Diagnostic Radiology, 20 Memory & Cognition 344 (1992), https://doi.org/10.3758/bf03210919; Rosalynn M. Peron & Gary L. Allen, Attempts To Train Novices For Beer Flavor Discrimination: A Matter of Taste, 115 J. Gen. Psych. 403 (1988), https://doi.org/10.1080/00221309.1988.9710577; Irving Biederman & Margaret M. Shiffrar, Sexing Day-Old Chicks: A Case Study and Expert Systems Analysis of a Difficult Perceptual-Learning Task, 13 J. Exper. Psych.: Learning, Memory & Cognition 640 (1987), https://doi.org/10.1037//0278-7393.13.4.640.
is pronounced and present in a variety of contexts.69 In Figure 7, the y-axis plots a measure of face discriminability (d’) for faces of four different races. The x-axis identifies the race of the human observers performing the face discrimination task. Bar texture identifies the race of the faces being discriminated. Regardless of their own race, all observers are best at discriminating between faces of their own race. (Error bars in Figure 7 represent standard errors of the mean.) The differential perceptual expertise that underlies the cross-race effect is a natural consequence of perceptual learning from exposure biases (e.g., most children grow up surrounded by family and friends of the same race),70 though there may be some consequence of in-group versus out-group social biasing effects.71
Category-specific perceptual expertise is an operating characteristic of biological sensory systems that affords the ability to gain proficiency in specialized domains, such as medical diagnosis of skin lesions, or the ability to quickly
Source: Adapted from Wong et al., supra note 69, at 208. Copyright © 2020 Wong, Stephen and Keeble. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY).
69. Rankin W. McGugin et al., Race-Specific Perceptual Discrimination Improvement Following Short Individuation Training with Faces, 35 Cognitive Sci. 330–47 (2011), https://doi.org/10.1111/j.1551-6709.2010.01148.x; Hoo Heat Wong et al., The Own-Race Bias for Face Recognition in a Multiracial Society, 11 Frontiers Psych. 208 (2020), https://doi.org/10.3389/fpsyg.2020.00208.
70. McGugin et al., supra note 69.
71. Eric Hehman, Eric W. Mania & Samuel L. Gaertner, Where the Division Lies: Common Ingroup Identity Moderates the Cross-Race Facial-Recognition Effect, 46 J. Exper. Soc. Psych. 445 (2010), https://doi.org/10.1016/j.jesp.2009.11.008.
identify the gender of a baby chick.72 But this same category specificity has obvious potential to yield inequity of recognition decisions under conditions of uncertainty in the real world.
The Constructive Nature of Visual Experience
Any effort to assess the validity of what one says they saw must come to grips with the fact that visual perception is, by its very nature and for very good reasons, not always an accurate reproduction of the world. Vision does not operate like a camera. Rather, visual perception is a construct of what we predict the world to be. To understand this constructive mandate, consider the optical and biological processes that underlie visual perception: Light reflected off of objects in the world around us is refracted by the lens in the anterior portion of the eye and projected as a two-dimensional (2D) patterned image onto the back surface of the eye, known as the retina. The goal of vision is to infer, from these 2D retinal patterns, the structure and meaning of objects and events in the three-dimensional (3D) world around us. This inferential problem is constrained, however, by the fact that the projected image does not contain—indeed cannot contain—all of the information needed to accurately determine the real-world cause of the image.73 To put it bluntly, there is an infinite set of real-world environments that can give rise to any specific retinal image.
Visual perception overcomes much uncertainty about the cause of sensation through reference to prior knowledge or experience. From these priors, or “biases,” the human brain fills in perceptual gaps with what is most likely to be out there. To illustrate this effect, consider the abstract visual pattern that appears in Figure 8. This pattern elicits a high degree of perceptual uncertainty; most people struggle to identify what it is. Figure 9 provides an additional bit of information that instills a bias. When returning to Figure 8, that bias enables the observer to improvise perceptual structure and meaning where formerly there was none.
Biases born from experience thus play a ubiquitous and essential role in perception. This concept is illustrated schematically in Figure 10. Information about the world gathered from sensory input, and information previously stored in memory, normally collaborate to yield coherent percepts. Thus, under most conditions, what we actually see falls somewhere along a continuum of mixtures of sensation and memory—a condition that facilitates behavioral interaction with the world under conditions of sensory uncertainty. The two ends of this continuum, however, are problematic states. At one end, pure stimulus is the
72. Goldstone, supra note 64; Biederman & Shiffrar, supra note 68.
73. This is a classic “inverse problem” in that there is not sufficient information in the retinal image to uniquely infer the values of the parameters that gave rise to it.
Source: Paul B. Porter, Another Puzzle-Picture, 67 Am. J. Psych. 550–51 (1954), https://doi.org/10.2307/1417952. Copyright 1954 by the Board of Trustees of the University of Illinois. Used with permission of the University of Illinois Press.
Source: Paul B. Porter, Another Puzzle-Picture, 67 Am. J. Psych. 550–51 (1954), https://doi.org/10.2307/1417952. Copyright 1954 by the Board of Trustees of the University of Illinois. Used with permission of the University of Illinois Press.
condition in which a meaningful interpretation is impossible. An observer’s initial viewing of Figure 8 leads to that state. This type of perceptual failure is rare in adult visual experience, but it is debilitating and potentially dangerous when it occurs, because the observer is confronted with a stimulus for which rational decisions and actions are impossible. At the other end of the continuum, pure imagery is the hallucinatory condition in which perceptual experience is detached from all sensory input, which can be entertaining or terrifying, but not at all adaptive. There are thus selective pressures for perceptual improvisation, through integration of sensation and memory.
Though evolution has thus afforded us the ability to employ prior knowledge to perceive an unambiguous world under conditions of uncertainty, there is a well-known hazard associated with this process: Our priors are sometimes wrong. In one of the most provocative demonstrations of this phenomenon, Jerome Bruner and Leo Postman used “trick” playing cards to demonstrate a persistent biasing influence of prior knowledge on perception.74 The trick cards were created by altering the color of a given suit—a red eight of spades, for example. Observers were shown a series of cards with brief presentations; some cards were trick and the others normal. With surprising frequency, observers failed to identify the trick cards and instead reported them as normal, thus demonstrating that strongly learned associations are capable of sharply, persistently, and unconsciously biasing visual perception. The heightened uncertainty associated with rapidly unfolding events of a crime is similarly ripe for unconscious introduction of biases held by the observer.
To make matters worse, perceptual inaccuracy born from uncertainty and bias is often associated with overconfidence—a special form of bias in which the observer implicitly rates the certainty of the experience as greater than it warrants. This phenomenon was pronounced in the “trick card” experiment cited above: Upon questioning about what they had perceived, the observers staunchly
74. Jerome S. Bruner & Leo Postman, On the Perception of Incongruity: A Paradigm, 18 J. Personality 206 (1949), https://doi.org/10.1111/j.1467-6494.1949.tb01241.x.
defended their mistaken identifications. Both eyewitness experiments conducted in the laboratory75 and studies of actual criminal cases76 demonstrate that while confidence may be low at the time of the initial lineup identification, confidence grows with reinforcement and with receipt of what appears as independent corroborating information—someone else saw something similar—which may drive the observer’s certainty in line with a larger story (see section below titled “Confidence”).77
In summary, the sense of vision is a biological information-processing device. As with any such device, the information gained is plagued by uncertainty, bias, and overconfidence. Uncertainty is a natural consequence of the incompleteness and ambiguity of sensory information. Biases drawn from memory fill in the gaps to ensure a seamless perceptual experience. Confidence in what was perceived grows as those experiences are reinforced by other sources of information. The significance of this perspective for eyewitness testimony is profound, as it helps us to realize that the accuracy of information about the environment—the face of a criminal perpetrator, for example—gained through vision is necessarily, and often sharply, limited by ambiguity and noise. It naturally follows that having an opportunity “to view the criminal at the time” (per Manson) may not be sufficient for assessment of the accuracy of eyewitness testimony, because what was viewed is not necessarily what was perceived and remembered.
75. Gary L. Wells, Tamara J. Ferguson & R. C. Lindsay, The Tractability of Eyewitness Confidence and Its Implications for Triers of Fact, 66 J. Applied Psych. 688 (1981), https://doi.org/10.1037//0021-9010.66.6.688; Melissa Paiva et al., Influence of Confidence Inflation and Explanations for Changes in Confidence on Evaluations of Eyewitness Identification Accuracy, 16 Legal & Criminological Psych. 266 (2011), https://doi.org/10.1348/135532510x503340; Siegfried Ludwig Sporer et al., Choosing, Confidence, and Accuracy: A Meta-Analysis of the Confidence-Accuracy Relation in Eyewitness Identification Studies, 118 Psych. Bull. 315 (1995), https://doi.org/10.1037//0033-2909.118.3.315; Kenneth A. Deffenbacher, Eyewitness Accuracy and Confidence: Can We Infer Anything About Their Relationship?, 4 Law & Hum. Behav. 243 (1980), https://doi.org/10.1007/bf01040617; Kenneth A. Deffenbacher & William Sturgill, Some Predictors of Eyewitness Memory Accuracy, Practical Aspects of Memory, 4 Law & Hum. Behav. 219 (1980), https://psycnet.apa.org/doi/10.1007/BF01040617; Mark Tippens Reinitz et al., Confidence–Accuracy Relations for Faces and Scenes: Roles of Features and Familiarity, 19 Psychonomic Bull. & Rev. 1085 (2012), https://doi.org/10.3758/s13423-012-0308-9; Robert K. Bothwell, Kenneth A. Deffenbacher & John C. Brigham, Correlation of Eyewitness Accuracy and Confidence: Optimality Hypothesis Revisited, 72 J. Applied Psych. 691 (1987), https://doi.org/10.1037//0021-9010.72.4.691; Amy Bradfield Douglass & Eric E. Jones, Confidence Inflation in Eyewitnesses: Seeing Is Not Believing, 18 Legal & Criminological Psych. 152 (2020), https://doi.org/10.1111/j.2044-8333.2011.02031.x; Laura Smalarz & Gary L. Wells, Do Multiple Doses of Feedback Have Cumulative Effects on Eyewitness Confidence?, 9 J. Applied Rsch. Memory & Cognition 508 (2020), https://doi.org/10.1037/h0101857; Nancy K. Steblay, Gary L. Wells & Amy Bradfield Douglass, The Eyewitness Post Identification Feedback Effect 15 Years Later: Theoretical and Policy Implications, 20 Psych., Pub. Pol’y. & L. 1 (2014), https://doi.org/10.1037/law0000001; Amy Bradfield Douglass & Nancy Steblay, Memory Distortion in Eyewitnesses: A Meta-Analysis of the Post-Identification Feedback Effect, 20 Applied Cognitive Psych. 859 (2006), https://doi.org/10.1002/acp.1237.
76. Garrett, supra note 2.
77. Daniel Kahneman & Amos Tversky, On the Psychology of Prediction, 80 Psych. Rev. 237 (1973), https://doi.org/10.1037/h0034747.
How Memory Works
Functional Processes of Memory
Visual perceptual experiences are commonly stored as declarative memories, which are accessed and expressed as knowledge about the world. Declarative memories are of two types: Semantic memories are of facts and concepts, whereas episodic memories are of events, such as those witnessed during a crime or a walk in the park. Declarative memories are processed in three stages—encoding, storage, and retrieval—which refer to placement of items in memory, maintenance therein, and subsequent access to the stored information.
In the following sections, we briefly review memory encoding, storage, and retrieval, with emphasis on the limits of these processes as they pertain to eyewitness identification. We also discuss the phenomenon of “recognition memory,” which refers to the specific type of memory retrieval in which a stimulus (e.g., a face) is used to make an identification.
Memory Encoding
Encoding is the process through which perceived objects and events are placed into storage. This process occurs in two temporal stages, distinguished by the quantity of information stored, the duration of storage, and the susceptibility to interference.78 Short-term or working memory is the conscious content of recent perceptual experiences, or information recently recalled from long-term storage. Short-term memory is of limited duration and capacity79 and is labile, decaying quickly and easily disrupted by other perceptual or cognitive processes.80
Through cellular and molecular events that play out over time, the contents of short-term memories may be encoded and consolidated into long-term memory,81 which is more enduring (albeit evolving with ongoing experience), and of greater capacity. Memories are particularly labile during this encoding process. The contents can be disrupted by interference and biased by prior
78. R. C. Atkinson & R. M. Shiffrin, Human Memory: A Proposed System and Its Control Processes, in 2 The Psychology of Learning and Motivation 89 (1968), https://doi.org/10.1016/s0079-7421(08)60422-3; James, supra note 47; Alan Baddeley, Working Memory: Looking Back and Looking Forward, 4 Nature Revs. Neurosci. 829 (2003), https://doi.org/10.1038/nrn1201; Alan Baddeley, Working Memory (1986).
79. George A. Miller, The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information, 63 Psych. Rev. 81 (1956), https://doi.org/10.1037/h0043158.
80. John Jonides et al., The Mind and Brain of Short-Term Memory, 59 Ann. Rev. Psych. 193 (2008), https://doi.org/10.1146/annurev.psych.59.103006.093615.
81. Eric Kandel & Larry Squire, Memory: From Mind to Molecules (1999).
knowledge, expectations, or beliefs, resulting in a distorted representation of experience. Short-term memories of events that happened early in a witnessed proceeding may simply be forgotten with the passage of time or compromised by attention directed to subsequent emotional events or cognitive and behavioral demands (e.g., anxiety, fear, the need to escape). In such cases, the compromised information may never be consolidated fully into long-term storage, or that storage may contain distorted content.82 At the same time, the quality of encoding of stimuli that are attended is commonly enhanced by highly emotional content.83
Memory Storage
Once memories have been encoded, they may be retained indefinitely in long-term storage. That stored information is not immutable, however, like information locked in a vault. On the contrary, stored memories fade with the passage of time, and they are commonly modified—often without awareness—as new knowledge and experiences are acquired. The quantitative features of these effects can provide insight into the validity of eyewitness reports.
How memory goes missing: The “forgetting function”
Much of the plasticity that plagues memory storage is the simple loss of information, which we call forgetting. The famous Ebbinghaus “forgetting function,” reproduced in Figure 11, illustrates that under the conditions used in Ebbinghaus’s nineteenth-century experiment (recall of nonsense syllables), two-thirds of the experienced content was forgotten after 24 hours.84 The rate of forgetting
82. James L. McGaugh, Memory—a Century of Consolidation, 287 Sci. 248 (2000), https://doi.org/10.1126/science.287.5451.248; James L. McGaugh & Benno Roozendaal, Role of Adrenal Stress Hormones in Forming Lasting Memories in the Brain, 12 Current Op. Neurobiology 205 (2002), https://doi.org/10.1016/s0959-4388(02)00306-9.
83. Kevin N. Ochsner, Are Affective Events Richly Recollected or Simply Familiar? The Experience and Process of Recognizing Feelings Past, 129 J. Exper. Psych.: Gen. 242 (2000), https://doi.org/10.1037//0096-3445.129.2.242; Deborah Talmi et al., Immediate Memory Consequences of the Effect of Emotion on Attention to Pictures, 15 Learning & Memory 172 (2008), https://doi.org/10.1101/lm.722908; Elizabeth A. Kensinger & Daniel L. Schacter, Neural Processes Supporting Young and Older Adults’ Emotional Memories, 7 J. Cognitive Neurosci. 1 (2008), https://doi.org/10.1162/jocn.2008.20080; Elizabeth A. Phelps, Emotion and Cognition: Insights from Studies of the Human Amygdala, 57 Ann. Rev. Psych. 27 (2006), https://doi.org/10.1146/annurev.psych.56.091103.070234.
84. Hermann Ebbinghaus, Memory: A Contribution to Experimental Psychology (Henry A. Ruger & Clara E. Bussenius trans., 1913) (1885).
has been studied in greater detail over the past 150 years. Effects depend upon the type of memory content and the circumstances under which it was encoded, but the time-dependent decline follows a power law: m = Lt−De−It, where m is a measure of memory retained, L is the strength of memory at time 0, t is retention time in seconds, D is decay rate, and I is interference rate.85 The modeled value of decay captures the rapid initial decline in memory retention, which is a consequence of failure to consolidate new information into long-term memory. As consolidation proceeds with the passage of time, there is a corresponding reduction of the rate of forgetting.
How memory goes wrong: Source memory failure and false memories
Memory loss is not simply due to decay. Memory mechanisms are inherently plastic throughout life, which means that content stored for the long term is labile in the face of new information. Without awareness, we regularly reconstruct,
85. Wayne A. Wickelgren, Single-Trace Fragility Theory of Memory Dynamics, 2 Memory & Cognition 775 (1974), https://doi.org/10.3758/bf03198154; John T. Wixted & Ebbe B. Ebbesen, On the Form of Forgetting, 2 Psych. Sci. 409 (1991), https://doi.org/10.1111/j.1467-9280.1991.tb00175.x.
update, and distort the things we believe to be true.86 Because new content can be added and the source of that content forgotten (source memory failure, in which we effectively forget how we know things),87 an observer may attribute updated memories to the originally witnessed events—in some cases, substantially changing what they believe they have seen. An eyewitness might learn from the police that a potential suspect has a goatee and then attribute that knowledge to the witnessed events, which could have disastrous consequences for the ability of the eyewitness to accurately report what was seen. Opportunities for this interference naturally increase with the passage of time, countering ongoing consolidation of new memories. Newly incorporated information also need not be true to fact. Research on false memories shows that it is possible to plant fabricated content in memory, which leads to recall of things that were never experienced.88
The effect of emotion on memory storage
Memories of highly arousing emotional stimuli tend to be more enduring than memories of nonarousing stimuli.89 Highly salient, unexpected, or arousing events—such as the Challenger space shuttle disaster, September 11, or episodes associated with a witnessed crime—are commonly stored strongly in memory, and their later retrieval is often associated with the subjective experience of high
86. Elizabeth F. Loftus & Geoffrey R. Loftus, On the Permanence of Stored Information in the Human Brain, 35 Am. Psych. 409 (1980), https://doi.org/10.1037//0003-066x.35.5.409.
87. D. Stephen Lindsay & Marcia K. Johnson, Recognition Memory and Source Monitoring, 29 Bull. Psychonomic Soc’y 203 (1991), https://doi.org/10.3758/bf03335235.
88. Elizabeth F. Loftus, Planting Misinformation in the Human Mind: A 30-Year Investigation of the Malleability of Memory, 12 Learning & Memory 361 (2005), https://doi.org/10.1101/lm.94705; Elizabeth F. Loftus & Jacqueline E. Pickrell, The Formation of False Memories, 25 Psych. Annals 720 (1995), https://doi.org/10.3928/0048-5713-19951201-07; Marcia K Johnson & Carol L Raye, False Memories and Confabulation, 2 Trends Cognitive Scis. 137 (1998), https://doi.org/10.1016/s1364-6613(98)01152-8.
89. Lewis J. Kleinsmith & Stephen Kaplan, Paired-Associate Learning as a Function of Arousal and Interpolated Interval, 65 J. Exper. Psych. 190 (1963), https://doi.org/10.1037/h0040288; Michael W. Eysenck, Arousal, Learning, and Memory, 83 Psych. Bull. 389 (1976), https://doi.org/10.1037//0033-2909.83.3.389; Friderike Heuer & Daniel Reisberg, Vivid Memories of Emotional Events: The Accuracy of Remembered Minutiae, 18 Memory & Cognition 496 (1990), https://doi.org/10.3758/bf03198482; Tali Sharot & Elizabeth A. Phelps, How Arousal Modulates Memory: Disentangling the Effects of Attention and Retention, 4 Cognitive, Affective & Behav. Neurosci. 294 (2004), https://doi.org/10.3758/cabn.4.3.294; Elizabeth A. Kensinger, Rachel J. Garoff-Eaton & Daniel L. Schacter, Memory for Specific Visual Details Can Be Enhanced by Negative Arousing Content, 54 J. Memory & Language 99 (2006), https://doi.org/10.1016/j.jml.2005.05.005; Elizabeth A. Kensinger, Remembering Emotional Experiences: The Contribution of Valence and Arousal, 15 Revs. Neuroscis. 241 (2004), https://doi.org/10.1515/revneuro.2004.15.4.241.
vividness and a sense of reliving.90 The stronger encoding and storage of emotional memories results from the engagement of a specialized system of stress hormones, which is triggered by arousing content and has potentiating effects on memory consolidation and storage.91 Despite the vividness that characterizes retrieval of emotional memories, there are many indications that such memories are prone to errors.92 Although emotional memories are often inaccurate in detail, one important corollary of their vividness is that they are frequently held with high confidence.93 This breakdown of the relationship between accuracy and confidence can undermine eyewitness accounts.94
90. Gezinus Wolters & Joop J. Goudsmit, Flashbulb and Event Memory of September 11, 2001: Consistency, Confidence and Age Effects, 96 Psych. Reps. 605 (2005), https://doi.org/10.2466/pr0.96.3.605-619; Elizabeth A. Kensinger, Anne C. Krendl & Suzanne Corkin, Memories of an Emotional and a Nonemotional Event: Effects of Aging and Delay Interval, 32 Exper. Aging Rsch. 23 (2006), https://doi.org/10.1080/01902140500325031; Ulric Neisser & Nicole Harsch, Phantom Flashbulbs: False Recollections of Hearing the News about Challenger, in Affect and Accuracy in Recall: Studies of “Flashbulb” Memories 9 (Emory Symposia in Cognition, 1992), https://doi.org/10.1017/CBO9780511664069.003; K. S. LaBar & E. A. Phelps, Arousal-Mediated Memory Consolidation: Role of the Medial Temporal Lobe in Humans, 9 Psych. Sci. 490 (1998), https://doi.org/10.1111/1467-9280.00090.
91. James L. McGaugh, Memory—A Century of Consolidation, 287 Sci. 248 (2000), https://doi.org/10.1126/science.287.5451.248; James L McGaugh & Benno Roozendaal, Role of Adrenal Stress Hormones in Forming Lasting Memories in the Brain, 12 Current Op. Neurobiology 205 (2002), https://doi.org/10.1016/s0959-4388(02)00306-9.
92. Elizabeth A. Kensinger, Remembering the Details: Effects of Emotion, 1 Emotion Rev. 99 (2009), https://doi.org/10.1177/1754073908100432; Tali Sharot, Mauricio R Delgado & Elizabeth A Phelps, How Emotion Enhances the Feeling of Remembering, 7 Nature Neurosci. 1376 (2004), https://doi.org/10.1038/nn1353; H. Schmolck, E.A. Buffalo & L.R. Squire, Memory Distortions Develop Over Time: Recollections of the O.J. Simpson Trial Verdict After 15 and 32 Months, 11 Psych. Sci. 39 (2000), https://doi.org/10.1111/1467-9280.00212; Stephen R. Schmidt, Autobiographical Memories for the September 11th Attacks: Reconstructive Errors and Emotional Impairment of Memory, 32 Memory & Cognition 443 (2004), https://doi.org/10.3758/bf03195837; Tony W. Buchanan & Ralph Adolphs, The Role of the Human Amygdala in Emotional Modulation of Long-Term Declarative Memory, in Emotional Cognition: From Brain to Behavior, in Advances in Consciousness Research 9 (2002), https://doi.org/10.1075/aicr.44.02buc.
93. Ulrike Rimmele et al., Emotion Enhances the Subjective Feeling of Remembering, Despite Lower Accuracy for Contextual Details, 11 Emotion 553 (2011), https://doi.org/10.1037/a0024246; Kensinger, supra note 92; Neisser & Harsch, supra note 90; Elizabeth A. Phelps & Tali Sharot, How (and Why) Emotion Enhances the Subjective Sense of Recollection, 17 Current Directions Psych. Sci. 147 (2008) https://doi.org/10.1111/j.1467-8721.2008.00565.x.
94. Kate A. Houston et al., The Emotional Eyewitness: The Effects of Emotion on Specific Aspects of Eyewitness Recall and Recognition Performance, 13 Emotion 118 (2013), https://doi.org/10.1037/a0029220; Robin S. Edelstein et al., Emotion and Eyewitness Memory, in Memory and Emotion 308 (D. Reisberg & P. Hertel eds., 2004), https://doi.org/10.1093/acprof:oso/9780195158564.003.0010; Sven-Åke Christianson, Emotional Stress and Eyewitness Memory: A Critical Review, 112 Psych. Bull. 284 (1992), https://doi.org/10.1037//0033-2909.112.2.284.
Memory Retrieval
Memory retrieval refers to the process by which stored information is accessed and brought into consciousness, where it can be used to make decisions and guide actions. Retrieval of long-term declarative memories is commonly triggered through association with an external stimulus—the sound of the coffee grinder is a powerful retrieval cue for visual and olfactory memories of the morning espresso. A corollary of this association-based phenomenon is that memory retrieval is often context dependent; a memory may be more readily retrieved if the observer is in physical surroundings that are the same as or similar to those in which the original experiences took place,95 because the surroundings provide additional cues to trigger memory retrieval.
Memory retrieval is affected by various sources of noise. Similarities of meaning or appearance between retrieval cues and items in memory can easily lead to retrieval of the wrong item, producing disruptive cognitive experiences—the sound of thunder may terrorize a veteran with retrieved memories of the battlefield—or false reports of what had been stored.96 This is particularly a problem given the categorical nature of memory.97 A related form of noise consists of intrusion errors, in which information known to be commonly associated with events of a general type becomes incorporated into the retrieved content of a specific memory. For example, because guns are often associated with robbery, an observer may readily and unwittingly
95. D. R. Godden & A. D. Baddeley, Context Dependent Memory in Two Natural Environments: On Land and Underwater, 66 Brit. J. Psych. 325 (1975), https://doi.org/10.1111/j.2044-8295.1975.tb01468.x; Stephen M. Smith & Edward Vela, Environmental Context-Dependent Eyewitness Recognition, 6 Applied Cognitive Psych. 125 (1992), https://doi.org/10.1002/acp.2350060204; Stephen M. Smith & Edward Vela, Environmental Context-Dependent Memory: A Review and Meta-Analysis, 8 Psychonomic Bull. Rev. 203 (2001), https://doi.org/10.3758/bf03196157; Endel Tulving & Donald M. Thomson, Encoding Specificity and Retrieval Processes in Episodic Memory, 80 Psych. Rev. 352 (1973), https://doi.org/10.1037/h0020071.
96. John R. Anderson, A Spreading Activation Theory of Memory, 22 J. Verbal Learning & Verbal Behav. 261 (1983), https://doi.org/10.1016/s0022-5371(83)90201-3; Allan M. Collins & Elizabeth F. Loftus, A Spreading-Activation Theory of Semantic Processing, 82 Psych. Rev. 407 (1975), https://doi.org/10.1037//0033-295x.82.6.407; Henry L. Roediger III, David A. Balota & Jason M. Watson, Spreading Activation and Arousal of False Memories, in The Nature of Remembering: Essays in Honor of Robert G. Crowder 95 (2001), https://doi.org/10.1037/10394-006; C. J. Brainerd & V. F. Reyna, The Science of False Memory (2005), https://doi.org/10.1093/acprof:oso/9780195154054.003.0002.
97. Endel Tulving, Episodic and Semantic Memory: Where Should We Go From Here?, 9 Behav. & Brain Scis. 573 (1986), https://doi.org/10.1017/s0140525x00047257; Martha J. Farah & James L. McClelland, A Computational Model of Semantic Memory Impairment: Modality Specificity and Emergent Category Specificity, 120 J. Exper. Psych.: Gen. 339 (1991), https://doi.org/10.1037//0096-3445.120.4.339; Glyn W. Humphreys & Emer M. E. Forde, Hierarchies, Similarity, and Interactivity in Object Recognition: “Category-Specific” Neuropsychological Deficits, 24 Behav. & Brain Scis. 453 (2001), https://doi.org/10.1017/s0140525x01004150.
incorporate a gun into the retrieved version of his or her memory of a witnessed robbery.
As noted above, memory retrieval is sometimes facilitated by being in the physical surroundings where the memorized events were experienced. A related phenomenon is state-dependent memory, in which retrieval accuracy is best if the individual’s cognitive state at the time of retrieval matches the cognitive state at the time of encoding.98 When memories have an emotional component, retrieval may be best when the individual is induced to a corresponding emotional state,99 which is accomplished by verbally or physically placing the person in the same context.100
Recognition Memory
Recognition memory is the specific type of declarative memory retrieval in which an immediately present sensory stimulus elicits retrieval of the same stimulus stored following a prior encounter. The subjective determination of “sameness” is the key here, for recognition memory differs from other forms of retrieval (such as recalling a phone number or the recipe for lasagna) in that a comparison must be made between the retrieved memory and the sensory stimulus (the “cue”) that elicited retrieval. This comparison process ultimately yields a categorical decision about whether the cue and the retrieved memory of a sensory stimulus are sufficiently similar (same source) or not (different source).
Recognition memory for faces differs greatly between familiar and unfamiliar faces. Because we often identify familiar faces with ease, most people tend to think they are generally very good at face recognition. However, the ability to recognize unfamiliar faces differs widely across individuals. At one extreme are those people, referred to as “super recognizers,” who hardly forget a face. At the other end of the spectrum are “face-blind people,” who have great difficulty
98. Donald W. Goodwin et al., Alcohol and Recall: State-Dependent Effects in Man, 163 Sci. 1358 (1969), https://doi.org/10.1126/science.163.3873.1358; Endel Tulving & Donald M. Thomson, Encoding Specificity and Retrieval Processes in Episodic Memory, 80 Psych. Rev. 352 (1973), https://doi.org/10.1037/h0020071; Edward Girden & Elmer Culler, Conditioned Responses in Curarized Striate Muscle in Dogs, 23 J. Compar. Psych. 261 (1937), https://doi.org/10.1037/h0058634; Donald A. Overton, State Dependent or “Dissociated” Learning Produced with Pentobarbital, 57 J. Compar. & Physiological Psych. 3 (1964), https://doi.org/10.1037/h0048023.
99. Pamela M. Kenealy, Mood State-Dependent Retrieval: The Effects of Induced Mood on Memory Reconsidered, 50 Q. J. Exper. Psych. 290 (1997), https://doi.org/10.1080/713755711; Penelope A. Lewis & Hugo D. Critchley, Mood-Dependent Memory, 7 Trends Cognitive Scis. 431 (2003), https://doi.org/10.1016/j.tics.2003.08.005; G. H. Bower, Mood and Memory, 36 Am. Psych. 129 (1981), https://psycnet.apa.org/doi/10.1037/0003-066X.36.2.129; Craik & Lockhart, supra note 22; Kensinger, supra note 92; Kenneth A. Leight & Henry C. Ellis, Emotional Mood States, Strategies, and State-Dependency in Memory, 20 J. Verbal Learning & Verbal Behav. 251 (1981), https://doi.org/10.1016/s0022-5371(81)90406-0.
100. Smith & Vela, A Review and Meta-Analysis, supra note 95.
recognizing even highly familiar faces. Evidence indicates that the ability to recognize an unfamiliar face varies continuously between these extremes across the human population.101 Figure 12 shows the results of an experiment designed to demonstrate this variation. Observers performed a recognition memory task that involved matching faces of three difficulty levels, identified on the x-axis as “same” (easy), “novel” (difficult), and “novel with noise” (very difficult). The y-axis plots recognition performance (“cumulative score”), where perfect recognition corresponds to the case in which y = x. Each curve contains data from one of 25 subjects. For the easy category, performance was nearly equivalent across subjects and nearly perfect. For difficult faces, performance was more variable, with some subjects performing nearly perfectly and other subjects performing approximately 65% correct. For very difficult faces, performance was highly variable across subjects, with the best performers at nearly 100% correct and the
Source: Used with permission of Elsevier Science & Technology Journals, adapted from Brad Duchaine & Ken Nakayama, The Cambridge Face Memory Test: Results for Neurologically Intact Individuals and an Investigation of Its Validity Using Inverted Face Stimuli and Prosopagnosic Participants, 44 Neuropsychologia, 576–85 (2006), https://doi.org/10.1016/j.neuropsychologia.2005.07.001 (permission conveyed through Copyright Clearance Ctr, Inc.).
101. Brad Duchaine & Ken Nakayama, The Cambridge Face Memory Test: Results for Neurologically Intact Individuals and an Investigation of Its Validity Using Inverted Face Stimuli and Prosopagnosic.
worst at less that 60% correct.102 New efforts to assess where a given witness lies along this spectrum, summarized in the section below titled “Applied Eyewitness Studies in the Field,” may prove diagnostic for interpreting the validity of identification testimony.103
How People Make Decisions
An eyewitness observer is manifestly a biological instrument for information measurement, storage, and classification. The foregoing discussions of vision and memory highlight functional capacities and operational limits of brain systems that serve the measurement and storage functions. The information measured and stored by an eyewitness is ultimately used to make a decision about identity, which is typically rendered as a binary choice (“He’s the one,” or “He’s not the one”) applied to each face in a lineup. The accuracy of that classification is an index of eyewitness performance. An understanding of the processes that underlie human decisions sheds light on where information processing works accurately or may go wrong.
Many human decisions involve comparisons of two items that have been measured by the senses. These items are often experienced as percepts derived from immediately present sensory stimuli: Which chair is bigger? Which light is brighter? Which suitcase is heavier? In some cases, the items being compared come from different informational sources, one from an immediate sensory stimulus and the other from memory of a stimulus previously experienced: Does it look like an oak tree? Does it taste like chicken? In both sensory-sensory (one sensory stimulus versus another simultaneously present sensory stimulus) and sensory-memory (a present sensory stimulus versus a remembered stimulus) comparisons, the results fall on a continuous cognitive scale of similarity. In all of these examples, however, the expected decision is not continuous; it is categorical, for example, “This one is brighter,” or “No, it does not taste like chicken.”
Human decisions thus depend upon two quantifiable elements. The first of these is the “decision variable,” which, for the examples given, is the subjective measure of similarity. An observer’s report of recognition is also influenced by the level of similarity that the observer finds acceptable. This level, known as the “decision criterion,” is thus the second element that underlies human categorical decisions. Under some conditions, the sensory and memory measures being compared are so different that the threshold level for similarity is easy to assign.
102. Id.
103. See Jesse H. Grabman et al., Predicting High Confidence Errors in Eyewitness Memory: The Role of Face Recognition Ability, Decision-Time, and Justifications, 8 J. Applied Rsch. Memory & Cognition 233, 241 (2019), https://doi.org/10.1037/h0101835.supp.
There are, however, many similarity outcomes that are plausible under two different ground truth conditions (same versus different source), which make an observer’s placement of the criterion for a categorical decision particularly problematic.
The decision criterion we actually apply under difficult conditions depends, in part, upon how we prioritize the outcomes. The simple desire to be informative to law enforcement, for example, may lead an eyewitness to lower their criterion for an identification decision. Estimating (or controlling) the observer’s decision criterion is thus a critical step in efforts to judge the validity of an identification.
Applied Eyewitness Studies in the Laboratory
Overview and Significance
Coincident with Supreme Court rulings on eyewitness identification,104 the 1970s saw a rapid growth of research on the validity of eyewitness testimony. In a field-defining review published in 1978,105 Gary Wells developed a taxonomy of variables that would be expected to influence eyewitness performance, which unleashed a tide of scientific research on the effects of those variables in controlled laboratory studies designed to simulate real-world conditions.
The variables of interest, which reflect environmental and human factors, are particularly important because they are frequently at issue in criminal cases, and their known states may provide insight into the validity of eyewitness testimony. These variables fall into two classes, which Wells termed “estimator” and “system” variables.106 Estimator variables characterize viewing conditions and the perceptual and cognitive states of the witness at the time of the crime. System variables, by contrast, influence identification accuracy after the crime has occurred and are variables that can be controlled by police.
The most valuable feature of applied laboratory studies of the impact of estimator and system variables is that “ground truth” is known. We thus know whether a given eyewitness made a correct or incorrect identification decision, which provides information about the predictive value of conditions associated with witnessing events (estimator variables) and identification events (system variables). The standard experimental paradigm involves exposing human
104. Neil v. Biggers, 409 U.S. 188 (1972); Manson v. Brathwaite, 432 U.S. 98, 114 (1977).
105. G. L. Wells, Applied Eyewitness-Testimony Research: System Variables and Estimator Variables, 36 J. Personality & Soc. Psych. 1546 (1978) https://doi.org/10.1037/0022-3514.36.12.1546.
106. Id.
subjects to mock crimes, in the form of staged events or videos. Subjects are then presented with a lineup and asked to report whether it contains a person they recognize from the witnessed crime. One can measure their accuracy, or whether they made a correct or incorrect choice, as well as other useful gauges of performance, such as confidence in their decision, and how quickly they arrive at that decision.
Although findings from many applied eyewitness studies have reinforced what basic research long ago revealed about the operating characteristics of vision and memory, particularly studies of the effects of viewing conditions and the cognitive state of the observer (estimator variables), these studies do so using a real-world context that highlights relevance to the courts. Applied research on other eyewitness variables, such as those associated with the type of lineup used for identification (system variables), has led to novel insights that can be used to improve eyewitness performance.
Performance Measures
The eyewitness identification problem studied under known-source laboratory conditions is a two-choice classification task, in which the witness subject must report whether a suspect is, or is not, the person they viewed at the crime scene. A few words about assessment of performance on this type of task are useful here.
Discrimination
The ability of an eyewitness to discriminate the perpetrator from innocent suspects is a core question in many laboratory studies, as it can be used to evaluate the impact of estimator and system variables on eyewitness performance. The standard measure of discriminability used for this purpose today is assessed as the frequency with which a witness identifies the perpetrator, relative to the frequency with which the witness identifies an innocent suspect, integrated across all decision criteria, ranging from circumspect to reckless.107
The knowledge gained from laboratory studies that employ this discriminability measure is an indispensable basis for policy decisions about the best lineup procedures. The level of discriminability associated with different lineup procedures can also be used to infer their relative effectiveness at yielding accurate testimony. This approach reveals nothing, however, about the probability
107. This has long been an effective tool used in basic research on the performance of biological systems for sensation and memory. See D. M. Green & J. A. Swets, Signal Detection Theory and Psychophysics (1966).
that a specific witness identification is correct, which is of course what the trier of fact really needs to know.108
Accuracy
Probability of correctness, or “accuracy,” is distinct from discriminability and is defined in this context as the ratio of correct identifications of the perpetrator relative to all (correct or incorrect) identifications reported. Accuracy of a population of subjects is a simple thing to measure in laboratory studies because ground truth is known, and many recent studies have done so. Like discriminability, accuracy can be used to assess the diagnostic utility of evidence under specified conditions. The important question here is whether it is possible to predict accuracy in real criminal cases, where ground truth is not known. Accuracy prediction in this case must be based on other variables that have been shown to correlate with accuracy, such as DNA genotyping.
Estimator Variables
The states of estimator variables reflect viewing conditions and cognitive status of the witness. Although not controllable by the criminal justice system, these states can be quantified for a given instance and used to predict the fidelity of vision and memory. In what follows, we examine several estimator variables that are commonly associated with real criminal cases in order to illustrate the range of effects on performance as demonstrated in laboratory studies.
Weapon Focus
“Weapon focus” is a phenomenon in which a witness’s focus of visual attention is directed away from the object that may be the most informative in the long run—the face of the perpetrator—and toward the object that is perceived as most immediately threatening to life.109 As we have learned from basic research on vision (see section titled “Visual attention” above), attention is a limited resource that can be hijacked by compelling features in a visual scene. Just as a driver must focus attention on oncoming traffic at the expense of enjoying scenery along the road, a witness’s life may depend upon the ability to keep attention focused on the perpetrator’s
108. David Dunning & Liza Beth Stern, Distinguishing Accurate from Inaccurate Eyewitness Identifications via Inquiries About Decision Processes, 67 J. Personality & Soc. Psych. 818 (1994), https://doi.org/10.1037//0022-3514.67.5.818.
109. Elizabeth F. Loftus, Eyewitness Testimony (1979).
handgun at the expense of gathering precise information about the perpetrator’s face.110
To get some sense of the prevalence and magnitude of the weapon focus effect in the eyewitness context, many applied studies have manipulated the presence of a weapon and measured eyewitness performance.111 The consensus from well-designed studies is precisely what we should expect from the basic science of visual attention and inattentional blindness: Brandishing a weapon limits the ability of eyewitnesses to discriminate between perpetrators and innocent suspects,112 and markedly impairs identification accuracy.113
Laboratory studies of weapon focus may significantly underestimate the magnitude of the effect in actual crimes, since the weapon has little threat value in a laboratory setting. Some efforts have been made to address this concern by correlating the presence of a weapon in actual crimes to identification accuracy,114 but interpretation of these analyses is hampered by lack of experimental control over other variables and questionable assumptions of ground truth.115
Finally, because the weapon focus effect is an attention-based phenomenon, it bears on the utility of the Manson assertion that eyewitness reliability (the probability that the testimony is correct) can be predicted by “the witness’ degree of attention.” As reviewed above, basic research demonstrates that there are two components to attention: magnitude (or “degree”) and directional focus. Magnitude of attention is indeed high under conditions in which a weapon is present, but the directional focus of attention in such cases precludes accurate identification of the perpetrator, because the eyewitness’s attention is directed at the weapon and not the perpetrator’s face.
Cross-Race Effect
The cross-race effect is a well-known phenomenon of visual cognition and perceptual learning, manifested as better performance on a task that involves
110. Loftus et al., supra note 62.
111. Nancy Mehrkens Steblay, A Meta-Analytic Review of the Weapon Focus Effect, 16 Law & Hum. Behav. 413 (1992), https://doi.org/10.1007/bf02352267; Jonathan M. Fawcett et al., Of Guns and Geese: A Meta-Analytic Review of the ‘Weapon Focus’ Literature, 19 Psych., Crim. & Law 35 (2011), https://doi.org/10.1080/1068316x.2011.599325.
112. Curt A. Carlson & Maria A. Carlson, An Evaluation of Lineup Presentation, Weapon Presence, and a Distinctive Feature Using ROC Analysis, 3 J. Applied Rsch. Memory & Cognition 45 (2014), https://doi.org/10.1016/j.jarmac.2014.03.004.
113. Curt A. Carlson & Maria A. Carlson, A Distinctiveness-Driven Reversal of the Weapon-Focus Effect, 6(1) Applied Psych. Crim. Just. 82 (2012), https://doi.org/10.1037/h0101806.
114. Fawcett et al., supra note 111, at 1–32.
115. Identifying the Culprit, supra note 12.
discriminating between faces of one’s own race relative to faces of a different race.116 The effect has been studied extensively in basic research on vision and memory (see sections titled “How Vision Works” and “How Memory Works” above) and is, in part, a by-product of categorical perception. The strength and context generality of this perceptual phenomenon offer good reasons to suspect that it will adversely impact the accuracy of eyewitness identification. Many laboratory studies of eyewitness identification over the past 50 years have confirmed this: The ability to discriminate between the culprit and an innocent suspect suffers and the accuracy of identifications is significantly impaired under conditions in which the eyewitness is of a different race from the lineup participants.117 This impaired eyewitness recognition ability has also been demonstrated experimentally in noncriminal real-world contexts,118 suggesting that it is not a laboratory artifact. Moreover, about half of all persons exonerated based on postconviction DNA testing were convicted, in part, based on mistaken cross-race identifications.119 Applied eyewitness studies, together with extensive basic research on this topic, demonstrate that eyewitness testimony employed by police investigators and courts is, on average, less likely to be accurate if the witness and the suspect are of different races.
Viewing Distance, Lighting, and Exposure Duration
These three factors routinely affect the fidelity of visual experience, as revealed by basic research on the operating characteristics of human vision (see section titled “How Vision Works” above). In simple terms that comport with normal human experience, far viewing distances, dim lighting, and short viewing durations all limit the amount of visual information available for object recognition.
116. McGugin et al., supra note 15; Wong et al., supra note 15.
117. Christian A. Meissner & John C. Brigham, Thirty Years of Investigating the Own-Race Bias in Memory for Faces: A Meta-Analytic Review, 7 Psych., Pub. Pol’y & L. 3 (2001), https://doi.org/10.1037//1076-8971.7.1.3; Jungwon Lee & Steven D. Penrod, Three-Level Meta-Analysis of the Other-Race Bias in Facial Identification, 36 Applied Cognitive Psych. 1106 (2022), https://doi.org/10.1002/acp.3997; Jacqueline Katzman & Margaret Bull Kovera, Potential Causes of Racial Disparities in Wrongful Convictions Based on Mistaken Identifications: Own-Race Bias and Differences in Evidence-Based Suspicion, 47 Law & Hum. Behav. 23 (2023), https://doi.org/10.1037/lhb0000503.supp; Rachel A. Wilson & Tammy L. Sonnentag, The Effect of Race of the Perpetrator and Misinformation on Eyewitness Accuracy and Confidence, 26 Mod. Psych. Studs. 8 (2021); Jesse N. Rothweiler, Kerri A. Goodwin & Jeff Kukucka, Presence of Administrators Differentially Impacts Eyewitness Discriminability for Same–And Other–Race Identifications, 34 Applied Cognitive Psych. 1530 (2020), https://doi.org/10.1002/acp.3733; Margaret Bull Kovera & Andrew J. Evelo, Eyewitness Identification in its Social Context, 10 J. Applied Rsch. Memory & Cognition 313 (2021), https://doi.org/10.1016/j.jarmac.2021.04.003.
118. John C. Brigham et al., Accuracy of Eyewitness Identification in a Field Setting, 42 J. Pers. & Soc. Psych. 673 (1982), https://doi.org/10.1037//0022-3514.42.4.673.
119. Garrett, supra note 2 at 73.
Unsurprisingly, they all affect eyewitness performance in predictable ways, as revealed by applied studies of eyewitness identification.
For example, one study found that it is harder to identify faces at a distance because of limits on spatial acuity.120 Numerous other applied eyewitness studies similarly support the conclusion that as viewing distance increases, eyewitness performance worsens.121 One study made the claim that eyewitness accuracy declines by approximately half when viewing distance increases from 3 to 20 meters.122 The quantitative characteristics of this relationship (and those drawn simply from basic research on vision) suggest that reliable measures of viewing distance in a given case can help predict the accuracy of an identification.
There is also a large basic science literature on visibility of objects as a function of ambient illumination (see Figure 1), which naturally suggests that eyewitness performance should decline as lighting dims. Indeed it does.123 Other studies have looked at the combined effects of distance and lighting, with one report offering a rule of thumb: Eyewitness performance is acceptable only if lighting exceeds 15 lux (roughly twilight) and distance is less than 15 meters.124 Findings from these applied studies, in conjunction with known operating characteristics of human vision, can support quantitative estimates of the accuracy of eyewitness reports.
Finally, basic sciences have long established that duration of exposure (viewing time) to a visual stimulus is correlated with information gain.125 Much of the impaired performance associated with short viewing durations is simply due to
120. Geoffrey R. Loftus & Erin M. Harley, Why Is It Easier to Identify Someone Close than Far Away?, 12 Psychonomic Bull. & Rev. 43 (2005), https://doi.org/10.3758/bf03196348.
121. Robert F. Lockamyeir et al., The Effect of Viewing Distance on Empirical Discriminability and the Confidence–Accuracy Relationship for Eyewitness Identification, 34 Applied Cognitive Psych. 1047 (2020), https://doi.org/10.1002/acp.3683; James Michael Lampinen et al., Effects of Distance on Face Recognition: Implications for Eyewitness Identification, 21 Psychonomic Bull. & Rev. 1489 (2014), https://doi.org/10.3758/s13423-014-0641-2; C.L. Lindsay et al., How Variations in Distance Affect Eyewitness Reports and Identification Accuracy, 32 Law & Hum. Behav. 526 (2008), https://doi.org/10.1007/s10979-008-9128-x; Thomas J. Nyman et al., The Distance Threshold of Reliable Eyewitness Identification, 43 Law & Hum. Behav. 527 (2019), https://doi.org/10.1037/lhb0000342.supp; Thomas J. Nyman et al., A Stab in the Dark: The Distance Threshold of Target Identification in Low Light, 6 Cogent Psych. 1632047 (2019), https://doi.org/10.1080/23311908.2019.1632047.
122. Lockamyeir et al., supra note 63.
123. Lisa DiNardo & David Rainey, The Effects of Illumination Level and Exposure Time on Facial Recognition, 41 Psych. Rec. 329 (1991), https://doi.org/10.1007/bf03395115; Marloes De Jong et al., Familiar Face Recognition as a Function of Distance and Illumination: A Practical Tool for Use in the Courtroom, 11 Psych., Crim. & L. 87 (2005), https://doi.org/10.1080/10683160410001715123.
124. Willem A. Wagenaar & Juliette H. Van Der Schrier, Face recognition as a function of distance and illumination: A practical tool for use in the courtroom, 2 Psych., Crim. & L. 321 (1996), https://doi.org/10.1080/10683169608409787.
125. George Sperling, The Information Available in Brief Visual Presentations, 74 Psych. Monographs: Gen. & Applied 1 (1960), https://doi.org/10.1037/h0093759.
loss of acuity and reduced contrast sensitivity.126 Sufficiently long viewing times, by contrast, equip observers with the information needed to make better decisions about what they have seen.127 Viewing durations are, of course, not under experimental control in actual eyewitness cases and, regrettably, they are not often recorded with precision,128 but one report estimates that over a third of all eyewitness reports are based on viewing durations of less than one minute.129 Numerous laboratory studies of eyewitness performance have manipulated viewing time and found, as expected, that longer viewing times are associated with greater eyewitness accuracy,130 although high-confidence identifications are generally highly accurate for both short and long exposure durations,131 an issue we take up in the section titled “Confidence” below.
126. Jacob Nachmias, Effect of Exposure Duration on Visual Contrast Sensitivity with Square-Wave Gratings, 57 J. Opt. Soc. Am. 421 (1967), https://doi.org/10.1364/josa.57.000421; J. Jason McAnany, The Effect of Exposure Duration on Visual Acuity for Letter Optotypes and Gratings, 105 Vision Rsch. 86 (2014), https://doi.org/10.1016/j.visres.2014.08.024.
127. Radoslaw Martin Cichy, Dimitrios Pantazis & Aude Oliva, Resolving Human Object Recognition in Space and Time, 17 Nature Neurosci. 455 (2014), https://doi.org/10.1038/nn.3635.
128. An eyewitness may be the only source of information regarding the duration of exposure to the perpetrator during a crime. Evidence suggests that these estimates are rarely accurate. Holly L. Gasper, Michael M. Roy & Heather D. Flowe, Improving Time Estimation in Witness Memory, 10 Frontiers Psych. 1452 (2019), https://doi.org/10.3389/fpsyg.2019.01452; A. Daniel Yarmey & Meagan J. Yarmey, Eyewitness Recall and Duration Estimates in Field Settings, 27 J. Applied Soc. Psych. 330 (1997), https://doi.org/10.1111/j.1559-1816.1997.tb00635.x.
129. Tim Valentine, Alan Pickering & Stephen Darling, Characteristics of Eyewitness Identification That Predict the Outcome of Real Lineups, 17 Applied Cognitive Psych. 969 (2003), https://doi.org/10.1002/acp.939.
130. J. Kirkland Reynolds & Kathy Pezdek, Face Recognition Memory: The Effects of Exposure Duration and Encoding Instruction, 6 Applied Cognitive Psych. 279 (1992), https://doi.org/10.1002/acp.2350060402; Amina Memon, Lorraine Hope & Ray Bull, Exposure Duration: Effects on Eyewitness Accuracy and Confidence, 94 Brit. J. Psych. 339 (2003), https://doi.org/10.1348/000712603767876262; Matthew A. Palmer et al., The Confidence-Accuracy Relationship for Eyewitness Identification Decisions: Effects of Exposure Duration, Retention Interval, and Divided Attention, 19 J. Exper. Psych.: Applied 55 (2013), https://doi.org/10.1037/a0031602; Rolando N. Carol & Nadja Schreiber Compo, The Effect of Encoding Duration on Implicit and Explicit Eyewitness Memory, 61 Consciousness & Cognition 117 (2018), https://doi.org/10.1016/j.concog.2018.02.004; Brian H. Bornstein et al., Effects of Exposure Time and Cognitive Operations on Facial Identification Accuracy: A Meta-Analysis of Two Variables Associated with Initial Memory Strength, 18 Psych., Crim. & L. 473 (2012), https://doi.org/10.1080/1068316x.2010.508458; Neil Brewer, Michael Gordon & Nigel Bond, Effect of Photoarray Exposure Duration on Eyewitness Identification Accuracy and Processing Strategy, 6 Psych., Crim. & L. 21 (2000), https://doi.org/10.1080/10683160008410829.
131. Carolyn Semmler et al., The Role of Estimator Variables in Eyewitness Identification, 24 J. Exper. Psych.: Applied 400 (2018), https://doi.org/10.1037/xap0000157; Matthew A. Palmer et al., The Confidence-Accuracy Relationship for Eyewitness Identification Decisions: Effects of Exposure Duration, Retention Interval, and Divided Attention, 19 J. Exper. Psych.: Applied 55 (2013), https://doi.org/10.1037/a0031602.supp.
The voluminous body of data on exposure duration thus demonstrates that, when reliable information about viewing times is available, this information can be used to infer the probability that eyewitness testimony is accurate.
Multiple Perpetrators
Crimes are often carried out by more than one person. Evidence indicates that the number of multiperpetrator homicides, for example, has been on the rise for decades and is estimated to exceed 20% of all murders.132 Explanations point to growing gang violence, peer pressure, and the perception of distributed responsibility (“de-individuation” of any single perpetrator)—a form of groupthink that licenses behaviors that individuals acting alone would view as impermissible.
For our purposes, the important empirical question is: How well can a witness be expected to remember one or more faces under these circumstances? This is a divided attention problem, much like the weapon focus problem addressed above, in which limited attentional resources must be distributed across multiple targets. While focused attention generally improves the fidelity of vision and memory for an attended stimulus, considerable evidence from basic science demonstrates what comports with everyday human experience: Detection, discriminability, and recognition of visual targets is impaired when attention must be subdivided between them.133 A number of laboratory studies have explored the consequences of this phenomenon for eyewitness identification.134 As expected (and mirroring the weapon focus effect on divided attention), identification accuracy is consistently poorer when multiple perpetrators are
132. Lockamyeir et al., supra note 63; Nortje et al., supra note 63, at 348–82.
133. Edward Awh & Harold Pashler, Evidence for Split Attentional Foci, 26 J. Exper. Psych. Hum. Perception & Performance 834 (2000), https://doi.org/10.1037//0096-1523.26.2.834; Arthur F. Kramer & Sowon Hahn, Splitting the Beam: Distribution of Attention over Noncontiguous Regions of the Visual Field, 6 Psych. Sci. 381 (1995), https://doi.org/10.1111/j.1467-9280.1995.tb00530.x; Won Mok Shim, George A. Alvarez & Yuhong V. Jiang, Spatial Separation Between Targets Constrains Maintenance of Attention on Multiple Objects, 15 Psychonomic Bull. & Rev. 390 (2008), https://doi.org/10.3758/pbr.15.2.390; Steven Yantis, Multi-Element Visual Tracking: Attention and Perceptual Organization, 24 Cognitive Psych. 295 (1992), https://doi.org/10.1016/0010-0285(92)90010-y; Amelia H. Harrison, Sam Ling & Joshua J. Foster, The Cost of Divided Attention for Detection of Simple Visual Features Primarily Reflects Limits in Post-Perceptual Processing, 85 Attention, Perception, & Psychophysics 377 (2023), https://doi.org/10.3758/s13414-022-02547-7; Harold Pashler, Attention and Visual Perception: Analyzing Divided Attention, in Visual Cognition: An Invitation to Cognitive Science 71 (2d ed. 1995), https://doi.org/10.7551/mitpress/3965.003.0005; M. Corbetta et al., Selective and Divided Attention During Visual Discriminations of Shape, Color, and Speed: Functional Anatomy by Positron Emission Tomography, 11 J. Neurosci. 2383 (1991), https://doi.org/10.1523/jneurosci.11-08-02383.1991; Nilli Lavie, Distracted and Confused? Selective Attention Under Load, 9 Trends Cognitive Sci. 75 (2005), https://doi.org/10.1016/j.tics.2004.12.004.
134. Clifford & Hollin, supra note 63; Hobson & Wilcock, supra note 63; Lockamyeir et al., supra note 63.
witnessed during a crime. For the purposes of the courts, it follows that eyewitness testimony under these conditions is less likely to be correct.
Facial Recognition Ability
Basic research shows (see section titled “Categorical perception” above) that one feature of categorical visual perception is category-specific expertise. While humans have a highly specialized system for face processing,135 face recognition ability varies significantly across the population, from “super recognizers” at one extreme to “face-blind people” at the other (see Figure 12).136 Chad Dodson and colleagues hypothesized that knowledge of the facial recognition ability of a witness might provide insight into the probative value of an identification by that witness.137 Results from this laboratory study reveal that a high-confidence identification is far more likely to be in error if the witness scores poorly on a test of facial recognition ability. These findings indicate that a simple test of face recognition ability can provide factfinders with valuable insight into the accuracy of high confidence identifications. We would not hesitate to conduct a visual acuity test to explore whether poor eyesight could have affected an eyewitness identification. Similarly, testing the face recognition ability of an eyewitness could become a standard procedure for law enforcement and for courts, which serves the purpose of inferring identification accuracy.
Retention Interval
As noted above (see section titled “Memory Storage”), many decades of basic research on memory have revealed that stored information is lost with the passage of time, as evidenced by progressive performance deterioration on standard visual discrimination and recognition tasks (see Figure 11). This memory loss is most rapid immediately after acquisition, when consolidation into long-term memory is not yet complete. After this initial period, consolidation continues but resulting long-term memories are readily modified by exposure to new information, the likelihood of which naturally increases with time. A witness’s inevitable interactions with law enforcement and legal counsel, not to mention communications from journalists, family, and friends, have the potential to significantly modify the witness’s memory of faces encountered and other details of
135. Winrich Freiwald, Bradley Duchaine & Galit Yovel, Face Processing Systems: From Neurons to Real World Social Perception, 39 Ann. Rev. Neurosci. 325 (2016), https://doi.org/10.1146/annurev-neuro-070815-013934.
136. Duchaine & Nakayama, supra note 101.
137. Grabman et al., supra note 103, at 233–43.
the crime scene.138 Thus, the fidelity of retrieved events—and the accuracy of identification—is likely to be greater when retrieval occurs closer to the time of the witnessed events.
These concerns about the completeness and stability of long-term memories have motivated a large number of applied studies of performance on eyewitness identification as a function of retention interval, defined as the length of time from the witnessed event to the lineup. A meta-analysis revealed that performance declines significantly as retention interval increases.139 Moreover, the authors maintain that the rate of memory loss, modeled as a power law,140 can be used to estimate the probability that an identification is correct: “[N]ot only can the percentage of remaining memory strength be determined for any retention interval, but this strength estimate can be translated into an estimated probability of being correct on a fair lineup of a specified size.”141
The initial strength of an encoded face memory is, of course, modulated by other estimator variables associated with witnessing the crime, such as cognitive state (e.g., attention and emotion), viewing conditions (e.g., distance, lighting, and exposure duration), and the cross-race effect.142 Nonetheless, there are strategies for estimating initial memory strength, and one study makes the promising claim that “we can now offer the judge or juror an estimate of what proportion of memory strength, regardless of its initial value, remained at the time the eyewitness’s memory for the perpetrator was tested.”143 This predictive approach could provide a quantitative foundation for the fifth Manson criterion for accuracy prediction: “the length of time between the crime and the confrontation.”
Stress and Emotion
As noted above in our review of the basic sciences of vision and memory (see sections titled “Visual attention” and “The effect of emotion on memory storage” above), considerable empirical evidence demonstrates that stress and high states of
138. Maria S. Zaragoza & Sean M. Lane, Source Misattributions and the Suggestibility of Eyewitness Memory, 20 J. Exper. Psych.: Learning, Memory & Cognition 934 (1994), https://doi.org/10.1037//0278-7393.20.4.934; William C. Thompson, et al., What Did the Janitor Do? Suggestive Interviewing and the Accuracy of Children’s Accounts, 21 Law & Hum. Behav. 405 (1997), https://doi.org/10.1023/a:1024859219764; D. Stephen Lindsay & Marcia K. Johnson, The Eyewitness Suggestibility Effect and Memory for Source, 17 Memory & Cognition 349 (1989), https://doi.org/10.3758/bf03198473.
139. Deffenbacher et al. supra note 26, at 139.
140. Wickelgren, supra note 85; Wixted & Ebbesen, supra note 85.
141. Deffenbacher et al., supra note 26.
142. Melissa A. Pigott, John C. Brigham & Robert K. Bothwell, A Field Study on the Relationship Between Quality of Eyewitnesses’ Descriptions and Identification Accuracy, 17 J. Police Sci. & Admin. 84 (1990), https://perma.cc/9GXP-8DWH.
143. Deffenbacher et al., supra note 26.
emotion can influence the allocation of visual attention as well as the encoding and retrieval of memories. Witnessing a crime, and the subsequent identification event, frequently elicit stress and often trigger high levels of emotion (commonly fear)—particularly in cases where the witness is a victim, a weapon is involved, or there is serious threat to life or limb. It naturally follows that these altered states of the observer may affect the content and validity of eyewitness reports.
The effects of stress and emotion on eyewitness performance have been difficult to study, as levels that often accompany eyewitness events in the real world are extremely difficult to replicate in a laboratory setting, in part because of the obvious pretense and the fact that regulations prohibit induction of high stress levels in human subjects. Some studies have, nonetheless, reportedly achieved moderate levels of stress and fear using unusual techniques, which generally involve manipulations that are ancillary to the identification test itself. One experiment involved subjecting active-duty military personnel to a mock POW camp (a valid part of military survival school training), which included aggressive interrogation and physical isolation-related stress.144 The study found that memories acquired during stressful events are highly vulnerable to modification by exposure to postevent misinformation, even in individuals whose level of training and experience might be considered relatively immune to such influences. Other studies of this sort have attempted to elicit stress and fear by placing subjects serving as witnesses in a creepy environment (the Horror Labyrinth of the London Dungeon)145 or administering vaccine injections to children who were serving as witnesses.146
Another approach in laboratory studies has been to independently confirm that “putative manipulations of stress/anxiety or perceived level of violence had been successful.”147 One meta-analysis limited to studies that obtained such confirmation (using standard physiological measures, such as breathing, and heart rate) found, as expected, that confirmed states of stress were correlated with impairments of identification accuracy and impaired ability to identify key characteristics of a witnessed individual’s face (e.g., hair length, hair color, eye color, shape of face, presence of facial hair). As emphasized in a more recent review, “ensuring valid stress inductions is a prerequisite for effectively studying effects of acute stress on memory performance.”148
144. C.A. Morgan III et al., Misinformation Can Influence Memory for Recently Experienced Highly Stressful Events, 36 Int’l J.L. & Psychiatry 11 (2013), https://doi.org/10.1016/j.ijlp.2012.11.002.
145. Tim Valentine & Jan Mesout, Eyewitness Identification Under Stress in the London Dungeon, 23 Applied Cognitive Psych. 151 (2009), https://doi.org/10.1002/acp.1510.
146. Douglas P. Peters, Stress, Arousal, and Children’s Eyewitness Memory, in Memory for Everyday and Emotional Events 351 (1st ed. 2018), https://doi.org/10.4324/9781315789231-15.
147. Kenneth A. Deffenbacher et al., A Meta-Analytic Review of the Effects of High Stress on Eyewitness Memory, 28 Law & Hum. Behav. 687 (2004), https://doi.org/10.1007/s10979-004-0565-x.
148. Carey Marr et al., The Effects of Acute Stress on Eyewitness Memory: An Integrative Review for Eyewitness Researchers, 29 Memory 1091 (2021), https://doi.org/10.1080/09658211.2021.1955935.
Laboratory studies have also revealed a counterintuitive relationship between accuracy and confidence under conditions of stress. Specifically, while stress clearly interferes with the accuracy of eyewitness memory, the predictive relationship between confidence and accuracy is similar across stressful and stress-free viewing conditions. (Stress-free observers are highly confident when accurate, while stressed observers are not confident when inaccurate.) This stability of the confidence–accuracy relationship implies that stressed witnesses are sufficiently aware of their cognitive state, and its potential to limit accuracy, and they “scale back their confidence judgments” accordingly.149 These findings imply that confidence reports may be useful predictors of accuracy under conditions of stress.
Despite the promise of these various approaches, it is impossible under laboratory conditions to achieve levels of stress and fear experienced by, for example, real witness–victims of rape, or a witness to the murder of a relative or friend, or a mass shooting. Nonetheless, although Manson made no mention of stress or fear in their predictive framework, the existing data argue that if stress or fear can be inferred from the circumstances of a witnessed crime, any eyewitness testimony proffered to the court will be of questionable accuracy.
Confidence
No topic in the literature on eyewitness identification has elicited as much robust discussion and research as confidence and its relationship to accuracy.150 In broad terms, this debate stems from the conflicting goals of (a) finding a measure that can predict the accuracy of an identification, and (b) avoiding the enormous personal and societal risk that such a measure might fail.
Before we review the basis for this debate, it is useful to define what we mean by confidence in this context and illustrate how it works. Confidence has a broad colloquial definition, but in the eyewitness context it refers to the perceived degree of certainty in a piece of information or a decision. Confidence normally grows when multiple sources of information paint a coherent picture.151 Many decades of basic research on confidence and the accuracy of
149. Sara D. Davis et al., Physiological Stress and Face Recognition: Differential Effects of Stress on Accuracy and the Confidence–Accuracy Relationship, 8 J. Applied Rsch. Memory & Cognition 367 (2019), https://doi.org/10.1016/j.jarmac.2019.05.006.
150. Confidence has traditionally been treated as an estimator variable in applied eyewitness research, since it is a property of the witness under the conditions of viewing. As this science has advanced, however, it has become clear that confidence as a variable is not so easily pigeonholed: It can be predictably influenced by the state of systems variables.
151. Kahneman & Tversky, supra note 77, at 237–51; Daniel Kahneman, Thinking, Fast and Slow (2011).
retrieved memories reveal that the correlation is frequently poor,152 which we can largely attribute to the natural process of smoothing over source discrepancies to create a sense of coherence, which, in turn, fosters confidence. This dissociation between confidence and accuracy has obvious implications for eyewitness testimony, and many have warned of the risks associated with the use of witness confidence statements for criminal investigation and prosecution.153
Timing of identification and conditions of confidence assessment
Applied eyewitness research has revealed nuances of the confidence–accuracy relationship. For example, the problem of confidence inflation,154 in which witnesses progressively gain greater confidence during the window of time between the initial lineup identification and the courtroom identification, has led to the recommendation that only the initial identification and confidence statements made at that time be used as evidence.155 Considerable evidence supports this
152. Loftus, supra note 109; Loftus & Pickrell, supra note 88, at 720-25; Kenneth A. Norman & Daniel L. Schacter, False Recognition in Younger and Older Adults: Exploring the Characteristics of Illusory Memories, 25 Memory & Cognition 838 (1997), https://doi.org/10.3758/bf03211328; Henry L. Roediger III & Kathleen B. McDermott, Creating False Memories: Remembering Words Not Presented in Lists, 21 J. Exper. Psych.: Learning, Memory & Cognition 803 (1995), https://doi.org/10.1037//0278-7393.21.4.803; Thomas A. Busey et al., Accounts of the Confidence–Accuracy Relation in Recognition Memory, 7 Psychonomic Bull. & Rev. 26 (2000) https://doi.org/10.3758/BF03210724; Daniel L. Schacter & Chad S. Dodson, Misattribution, False Recognition and the Sins of Memory, 356 Phil. Transactions Royal Soc. London Series B: Biological Scis. 1385 (2001), https://doi.org/10.1098/rstb.2001.0938; Hal R. Arkes, Overconfidence in Judgmental Forecasting, in Principles of Forecasting 495 (2001), https://doi.org/10.1007/978-0-306-47630-3_22; Mark Tippens Reinitz et al., Different Confidence–Accuracy Relationships for Feature-Based and Familiarity-Based Memories, 37 J. Exper. Psych.: Learning, Memory & Cognition 507 (2011), https://doi.org/10.1037/a0021961.supp; Henry L. Roediger III, John H. Wixted & K. Andrew DeSoto, The Curious Complexity between Confidence and Accuracy in Reports from Memory, in Memory and Law 84–118 (Lynn Nadel & Walter Sinnott-Armstrong eds., 2012); Oxford University Press, New York; Elizabeth F. Loftus & Rachel L. Greenspan, If I’m Certain, Is It True? Accuracy and Confidence in Eyewitness Memory, 18 Psych. Sci. Pub. Int. 1 (2017), https://doi.org/10.1177/1529100617699241.
153. Sporer et al., supra note 75; Deffenbacher, supra note 75; Reinitz et al., supra note 75; Bothwell et al., supra note 75.
154. Melissa Paiva et al., Influence of Confidence Inflation and Explanations for Changes in Confidence on Evaluations of Eyewitness Identification Accuracy, 16 Legal & Criminological Psych. 266 (2011), https://doi.org/10.1348/135532510x503340; Sporer et al., supra note 75; Deffenbacher, supra note 75; Reinitz et al., supra note 75; Bothwell et al., supra note 75; Laura Smalarz & Gary L. Wells, Do Multiple Doses of Feedback Have Cumulative Effects on Eyewitness Confidence?, 9 J. Applied Rsch., Memory & Cognition 508 (2020), https://doi.org/10.1037/h0101857; Nancy K. Steblay, Gary L. Wells & Amy Bradfield Douglass, The Eyewitness Post Identification Feedback Effect 15 Years Later: Theoretical and Policy Implications, 20 Psych., Pub. Pol’y & L. 1 (2014), https://doi.org/10.1037/law0000001; Douglass & Steblay, supra note 75.
155. Identifying the Culprit, supra note 12.
recommendation: Memory is patently malleable,156 and the very act of probing memory in a lineup procedure can actually modify that memory in a way that will lessen the accuracy of future identifications.157 Following this logic, an argument can be made that the initial identification should be the only identification obtained.158
Additional insight into this issue has come from work by John Wixted and colleagues, who argued that confidence reports made at the time of the initial identification have their greatest predictive value if they are held to a very high standard. Specifically, these investigators report that the relationship between confidence and accuracy is strong if the initial identification is made under “pristine” conditions.159 Those conditions include a properly blinded test and fair lineup construction (no participant stands out), which promote coherence of information received by the eyewitness, thus limiting the accuracy-reducing inclination to smooth over discrepancies.
Critics argue that it may be difficult to recognize pristine conditions in real casework, owing in part to limited access to situational context (e.g., what exactly did the police do?). By this argument, eyewitness confidence may place innocent suspects at great risk.160 Some disagreement remains on this point,161 but the rationale for the pristine distinction and the use of only the initial identification and confidence statement rests firmly on the cognitive science of memory and decision-making.162
156. Deborah Davis & Elizabeth F. Loftus, Eyewitness Science in the 21st Century: What Do We Know and Where Do We Go from Here?, in 1 Stevens’ Handbook of Experimental Psychology and Cognitive Neuroscience (4th ed. 2018), https://doi.org/10.1002/9781119170174.epcn116.
157. Nancy K. Steblay, Robert W. Tix & Samantha L. Benson, Double Exposure: The Effects of Repeated Identification Lineups on Eyewitness Accuracy, 27 Applied Cognitive Psych. 644 (2013), https://doi.org/10.1002/acp.2944; Nancy K. Steblay & Jennifer E. Dysart, Repeated Eyewitness Identification Procedures with the Same Suspect, 5 J. Applied Rsch. Memory & Cognition 284 (2016), https://doi.org/10.1016/j.jarmac.2016.06.010.
158. John T. Wixted et al., Test a Witness’s Memory of a Suspect Only Once, 22 Psych. Sci. Pub. Int. 1S (2021), https://doi.org/10.1177/15291006211026259.
159. Id.; Wixted & Wells, supra note 25; Semmler et al., supra note 131.
160. Shari R. Berkowitz & Steven J. Frenda, Rethinking the Confident Eyewitness: A Reply to Wixted, Mickes, and Fisher, 13 Persps. Psych. Sci. 336 (2018), https://doi.org/10.1177/1745691617751883; Shari R. Berkowitz et al., Convicting with Confidence? Why We Should Not Over-Rely on Eyewitness Confidence, 30 Memory 10 (2022), https://doi.org/10.1080/09658211.2020.1849308; Shari R. Berkowitz et al., Eyewitness Confidence May Not be Ready for the Courts: A Reply to Wixted Et Al., 30 Memory 75 (2022), https://doi.org/10.1080/09658211.2021.1952271; James D. Sauer, Matthew A. Palmer & Neil Brewer, Pitfalls in Using Eyewitness Confidence to Diagnose the Accuracy of an Individual Identification Decision, 25 Psych., Pub. Pol’y & L. 147 (2019), https://doi.org/10.1037/law0000203.
161. Travis M. Seale-Carlisle et al., Confidence and Response Time as Indicators of Eyewitness Identification Accuracy in the Lab and in the Real World, 84 J. Applied Rsch. Memory & Cognition 420 (2019), https://doi.org/10.1016/j.jarmac.2019.09.003; Wixted & Wells, supra note 25.
162. Semmler et al., supra note 131.
System Variables
System variables are causal states that are potentially under the control of the criminal justice system, since they bear on activities that occur after the witnessed events. These include (1) communication between law enforcement and the witness, (2) communication between the witness and the larger community (legal defense, media, family and friends, etc.), (3) the practice of “evidence-based suspicion,” (4) prior exposure to images of potential suspects, (5) the type of procedure used for identification, and (6) the choice of lineup fillers (lineup participants known to be innocent).
The empirical findings and recommendations derived from research on system variables are of direct relevance to, and advisory of, police practices. That is, they can be used to improve eyewitness identification performance. Knowledge of the states of these variables for a given case, however, can also be useful to the courts in estimating the probability that eyewitness testimony gained by law enforcement is accurate.
Communication Between Law Enforcement Personnel and Witness
Statements that convey prior probability of suspicion
Police commonly have information that could influence both the accuracy of a witness’s identification and the confidence expressed in the identification. For example, a police officer might first report to the witness/victim of a carjacking that a suspect was apprehended in the stolen vehicle and then ask the witness to view a lineup with that suspect in it. This might seem innocuous on the surface and could be viewed as reinforcing information for the witness, but it establishes a significant prior probability that the perpetrator is in the lineup. Inferences drawn from such priors can readily bias perception, choice, and action under conditions of uncertainty.163 The result is that the witness may actually perceive the perpetrator in a target-absent lineup and may unwittingly adopt a lower criterion for reporting an identification even when recognition memory is poor.
Communication during lineups: The importance of blinded lineup administration
In cases where the administrator of a lineup knows the status of lineup participants, they are in a position to inadvertently communicate that information to a
163. Albright, supra note 36.
witness. The witness, commonly eager to be of assistance, can easily pick up unintended signals (e.g., direction of gaze, body posture, facial expressions), which may be highly suggestive and have the effect of biasing the outcome—yielding a larger fraction of misidentifications and inflation of witness confidence. Generally, the solution to these biasing effects is to limit communication between the witness and people with relevant knowledge.164
Many applied eyewitness studies support this information-control approach and demonstrate that the problem of inadvertent communication during a lineup can be eliminated by the practice of “double blinding.”165 This simple and effective tool is a pillar of the scientific method.166 In an experimental test of a novel drug, for example, a subject is prevented (hence “blinded” or “single blinded”) from knowing the experimental condition that they have been assigned to. Furthermore, the scientist who measures the effect of the drug is also prevented (“double blinded”) from knowing the experimental condition associated with that measure, all of which precludes bias based on prior knowledge. Similarly, in a lineup, the witness is naturally blinded to knowledge of the “conditions,” but the risk of bias can be reduced if the lineup administrator has no knowledge to convey—that is, the test is double blinded. This simple practice is highly effective at improving eyewitness accuracy, and the risks of not conducting double-blinded lineups are clear. For all of these reasons, failure to do so runs afoul of
164. Ryann M. Haw & Ronald P. Fisher, Effects of Administrator-Witness Contact on Eyewitness Identification Accuracy, 89 J. Applied Psych. 1106 (2004), https://doi.org/10.1037/0021-9010.89.6.1106; Robert Rosenthal, Covert Communication in Classrooms, Clinics, Courtrooms, and Cubicles, 57 Am. Psych. 839 (2002), https://doi.org/10.1037//0003-066x.57.11.839.
165. Steven E. Clark, Tanya E. Marshall & Robert Rosenthal, Lineup Administrator Influences on Eyewitness Identification Decisions, 15 J. Exper. Psych.: Applied 63 (2009), https://doi.org/10.1037/a0015185; Margaret Bull Kovera & Andrew J. Evelo, The Case for Double-Blind Lineup Administration, 23 Psych., Pub. Pol’y & L. 421 (2017), https://doi.org/10.1037/law0000139; Margaret Bull Kovera & Andrew J. Evelo, Improving Eyewitness-Identification Evidence Through Double-Blind Lineup Administration, 29 Current Directions Psych. Sci. 563 (2020), https://doi.org/10.1177/0963721420969366; Steve D. Charman & Vanessa Quiroz, Blind Sequential Lineup Administration Reduces Both False Identifications and Confidence in Those Identifications, 40 Law & Hum. Behav. 477 (2016), https://doi.org/10.1037/lhb0000197; Dario N. Rodriguez & Melissa A. Berry, Eyewitness Science and the Call for Double-Blind Lineup Administration, 2013 J. Criminology 530523 (2013), https://doi.org/10.1155/2013/530523; Jacqueline L. Austin et al., Double-Blind Lineup Administration: Effects of Administrator Knowledge on Eyewitness Decisions, in 139 Reform of Eyewitness Identification Procedures (2013), https://doi.org/10.1037/14094-007; Arthur H. Perlini & Andrew D. Silvaggio, Eyewitness Misidentification: Single vs. Double-Blind Comparison of Photospread Administration, 100 Psych. Reps. 247 (2007), https://doi.org/10.2466/pr0.100.1.247-256.
166. Michael Stolberg, Inventing the Randomized Double-Blind Trial: The Nuremberg Salt Test of 1835, 99 J. Royal Soc. Med. 642 (2006), https://doi.org/10.1177/014107680609901216; Lorraine Daston, Scientific Error and the Ethos of Belief, 72 Soc. Rsch.: Int’l Q. 1 (2005), https://doi.org/10.1353/sor.2005.0016; Paul J. Karanicolas, Forough Farrokhyar & Mohit Bhandari, Practical Tips for Surgical Research: Blinding: Who, What, When, Why, How?, 53 Can. J. Surg. 345 (2010); Gary L. Wells & C.A. Elizabeth Luus, Police Lineups as Experiments: Social Methodology as a Framework for Properly Conducted Lineups, 16 Pers. & Soc. Psych. Bull. 106 (1990), https://doi.org/10.1177/0146167290161008.
best practices,167 recommendations of the U.S. Department of Justice,168 and the International Association of Chiefs of Police.169 In a court of law, lineup blinding status is naturally a useful predictor of eyewitness accuracy.
Video recording of lineup procedures
In view of the risk that some forms of communication between the lineup administrator(s) and witness can bias the identification, many have called for video recording of the entire procedure.170 One may not expect that doing so will change the behavior of the people involved, but the recording serves as stable evidence to confirm—to the court, to the prosecution, and to the defense—whether the procedure was done in such a way to avoid a biased outcome. It also documents who was present, what the witness actually said about an identification and their level of confidence, and other behavioral characteristics (e.g., time to decide) that may bear on the accuracy of identifications. The absence of such video recordings, conversely, limits inferences about the accuracy of eyewitness testimony.
Postidentification feedback
Postidentification feedback from law enforcement can also influence a witness’s confidence in their decision.171 Applied studies have shown that simple reinforcing statements, even without specifics—“Nice work!”—can increase or decrease witness confidence, thus distorting any relationship to accuracy (see section titled “Timing of identification and conditions of confidence assessment” above).172
167. Identifying the Culprit, supra note 12.
168. U.S. Dep’t of Just., Eyewitness Identification, supra note 7.
169. IACP (International Association of Chiefs of Police) Law Enforcement Policy Center, Eyewitness Identification: Model Policy Concepts & Issues (2016), https://perma.cc/3YDT-8A77.
170. Identifying the Culprit, supra note 12; Wells et al., Policy and Procedure Recommendations (2020), supra note 27, at 3–36. U.S. Dep’t of Just., Eyewitness Identification, supra note 7.
171. Douglass & Steblay, supra note 75; Mitchell L. Eisen et al., Does Anyone Else Look Familiar? Influencing Identification Decisions by Asking Witnesses to Re-Examine the Lineup, 42 Law & Hum. Behav. 306 (2018), https://doi.org/10.1037/lhb0000291.supp.
172. Carolyn Semmler, Neil Brewer & Gary L. Wells, Effects of Postidentification Feedback on Eyewitness Identification and Nonidentification Confidence, 89 J. Applied Psych. 334 (2004), https://doi.org/10.1037/0021-9010.89.2.334; Gary L. Wells & Amy L. Bradfield, “Good, You Identified the Suspect”: Feedback to Eyewitnesses Distorts Their Reports of the Witnessing Experience, 83 J. Applied Psych. 360 (1998), https://doi.org/10.1037//0021-9010.83.3.360; Amy L. Bradfield, Gary L. Wells & Elizabeth A. Olson, The Damaging Effect of Confirming Feedback on the Relation Between Eyewitness Certainty and Identification Accuracy, 87 J. Applied Psych. 112 (2002), https://doi.org/10.1037//0021-9010.87.1.112; Nancy K. Steblay, Gary L. Wells & Amy Bradfield Douglass, The Eyewitness Post
Very high or very low confidence expressed by a witness can, in turn, be very compelling to the trier of fact, regardless of whether the identification is correct.173
Communication Between Community and Witness: Social Influences on Eyewitness Performance
As we have seen, memory is malleable and the criteria that people use for momentous decisions can change over time. People are highly social creatures—eager for social reinforcement and easily swayed by pressure for social conformity—and it should come as no surprise that in the eyewitness context (and more generally), the expectations, real or inferred, of other parties (police, attorneys, news media, family, colleagues and comrades, social media followers) can have a profound influence over what we remember, the decisions we make, and the confidence we express. There are two important features to this potential for bias: the social utility of misinformation and the scale of social networks.
The social utility of misinformation
Social influences can readily modify an individual’s memory of events. New information received from one’s social networks may be unvetted “misinformation,” which is seen as widely accepted by others and thus not held to rigorous standards. People have a regrettable affinity for such information,174 in part because knowing things, regardless of whether they are true, has social value. The end result is contamination of the original content of memory, such that decisions are less accurate and confidence in them becomes distorted. In the
Identification Feedback Effect 15 Years Later: Theoretical and Policy Implications, 20 Psych., Pub. Pol’y & L. 1 (2014), https://doi.org/10.1037/law0000001.
173. Steven D. Penrod & Brian L. Cutler, Eyewitness Expert Testimony and Jury Decisionmaking, 52 Law & Contemp. Probs. 43 (1989), https://doi.org/10.2307/1191907; Neil Brewer & Anne Burke, Effects of Testimonial Inconsistencies and Eyewitness Confidence on Mock-Juror Judgments, 26 Law & Hum. Behav. 353 (2002), https://doi.org/10.1023/a:1015380522722; James D. Sauer, Matthew A. Palmer & Neil Brewer, Mock-Juror Evaluations of Traditional and Ratings-Based Eyewitness Identification Evidence, 41 Law & Hum. Behav. 375 (2017), https://doi.org/10.1037/lhb0000235.supp; Crystal R. Slane & Chad S. Dodson, Eyewitness Confidence and Mock Juror Decisions of Guilt: A Meta-Analytic Review, 46 Law & Hum. Behav. 45 (2022), https://doi.org/10.1037/lhb0000481.supp.
174. Ilan Yaniv & Dean P. Foster, Graininess of Judgment Under Uncertainty: An Accuracy-Informativeness Trade-Off, 124 J. Exper. Psych.: Gen. 424 (1995), https://doi.org/10.1037/0096-3445.124.4.424; Rakefet Ackerman & Morris Goldsmith, Control over Grain Size in Memory Reporting—With and Without Satisficing Knowledge, 34 J. Exper. Psych.: Learning, Memory & Cognition 1224 (2008), https://doi.org/10.1037/a0012938.
extreme, people can hold and act upon “false memories” of things that never happened.175
Applied studies have explored the influence of postwitnessing information gain on eyewitness accuracy and confidence. The most straightforward of these effects is incorporation of new information into the preexisting memory of the crime scene, without awareness (a form of source memory failure).176 Acquired knowledge of the eyewitness reports of others, for example, can bring about both a change in a witness’s memory and confidence.177 Perhaps more pernicious are influences from news reports, social media, and colleagues: Eyewitnesses in laboratory studies have been shown to readily modify their own memories, without awareness, to reflect information contained in news reports or implied by interviewer questions.178
A related line of research has focused on the effect of postwitnessing information on the inclination of a witness to be accurate versus informative.179 Witnesses who are exposed to media reports that either promote the fidelity of eyewitness reports or denigrate it adopt a practice that comports.180 Those who learn that eyewitness testimony is beneficial to criminal justice are inclined to be highly informative with a lower criterion for accuracy. By contrast, those who learn that eyewitness testimony is of questionable value are more circumspect and inclined toward accuracy.
175. Loftus, supra note 109; Loftus, supra note 88.
176. Lindsay & Johnson, supra note 138, at 349–58.
177. C. A. Elizabeth Luus & Gary L. Wells, The Malleability of Eyewitness Confidence: Co-Witness and Perseverance Effects, 79 J. Applied Psych. 714 (1994), https://doi.org/10.1037//0021-9010.79.5.714; Mitchell L. Eisen et al., “I Think He Had a Tattoo on His Neck”: How Co-Witness Discussions About a Perpetrator’s Description Can Affect Eyewitness Identification Decisions, 6 J. Applied Rsch. Memory & Cognition 274 (2017), https://doi.org/10.1016/j.jarmac.2017.01.009; Anders Granhag Pär et al., Social Influence on Eyewitness Memory, in Forensic Psychology in Context: Nordic and International Approaches 139 (2010), https://doi.org/10.4324/9781315094038-8.
178. Elizabeth F. Loftus & Guido Zanni, Eyewitness Testimony: The Influence of the Wording of a Question, 5 Bull. Psychonomic Soc. 86 (1975), https://doi.org/10.3758/bf03336715; Elizabeth F. Loftus, Leading Questions and the Eyewitness Report, 7 Cognitive Psych. 560 (1975), https://doi.org/10.1016/0010-0285(75)90023-7; Elizabeth F. Loftus, David G. Miller & Helen J. Burns, Semantic Integration of Verbal Information into a Visual Memory, 4 J. Exper. Psych.: Hum. Learning & Memory 19 (1978), https://doi.org/10.1037/0278-7393.4.1.19; Robin Blom & Kuo-Ting Huang, Eyewitness Memory Contamination Through Misleading Questions by Reporters, 42 News Rsch. J. 346 (2021), https://doi.org/10.1177/07395329211030628.
179. Morris Goldsmith, Asher Koriat & Amit Weinberg-Eliezer, Strategic Regulation of Grain Size Memory Reporting, 131 J. Exper. Psych.: Gen. 73 (2002), https://doi.org/10.1037//0096-3445.131.1.73; Nathan Weber & Neil Brewer, Eyewitness Recall: Regulation of Grain Size and the Role of Confidence, 14 J. Exper. Psych.: Applied 50 (2008), https://doi.org/10.1037/1076-898x.14.1.50.
180. Muhammad Mussaffa Butt, et. al, Eyewitness Memory in the News Can Affect the Strategic Regulation of Memory Reporting, 30 Memory 763 (2022), https://doi.org/10.1080/09658211.2020.1846750.
“Pristine” conditions for eyewitness interrogation are often sought in order to avoid these forms of social biasing,181 much in the way that restricted access to task-irrelevant information is intended to avoid bias in forensic analyses.182 Pristine is good and aspirational, of course, but in the real world it becomes very difficult to limit access to information and defuse explicit or implicit social pressures.183
The long reach of modern social networks
The social networks of yore, which may have included school and work colleagues, or friends from the neighborhood, have become overshadowed today by vast online communities. Active participation in these communities frequently involves presenting yourself for validation, but the reflection is from a “trick mirror,”184 which distorts who you are—often without your awareness—and makes powerful social demands on who you should be, what you should believe, and how you should behave.185 The rewards for conformity are addictive.186 Naturally, this form of social influence has profound implications for the validity of eyewitness identification, such that in the extreme, “some eyewitness identifications may not be governed by memory at all.”187 Social pressures from online communities may cause a witness to question what they actually saw or, in the worst case, modify their memories without awareness. The “Rashomon effect” is the archetype of this phenomenon,188 but we now see this happening on a massive scale in which scores of people collectively succumb to social distortions of reality and modify their criteria for judgment.189
181. Wixted & Wells, supra note 25.
182. Nat’l Comm. Forensic Sci., Ensuring That Forensic Analysis Is Based Upon Task-Relevant Information (2015), https://perma.cc/DT64-JLGM.
183. A National Survey of Eyewitness Identification, supra note 11.
184. Jia Tolentino, Trick Mirror: Reflections on Self-Delusion (2019).
185. Jonathan Haidt, The Anxious Generation: How the Great Rewiring of Childhood is Causing an Epidemic of Mental Illness (2024).
186. Id.
187. Margaret Bull Kovera & Andrew J. Evelo, Eyewitness Identification in its Social Context, 10 J. Applied Rsch. Memory & Cognition 313 (2021), https://doi.org/10.1016/j.jarmac.2021.04.003.
188. Rashomon Effect, Wikipedia, https://perma.cc/DYM4-8T86.
189. Dustin P. Calvillo, Justin D. Harris & Whitney C. Hawkins, Partisan Bias in False Memories for Misinformation About the 2021 U.S. Capitol Riot, 31 Memory 137 (2023), https://doi.org/10.1080/09658211.2022.2127771; Seth Oranburg, Social Media and Democracy After the Capitol Riot, or, a Cautionary Tale of the Giant Goldfish, 73 Mercer L. Rev. 591 (2021); Brenna M. Davidson & Tetsuro Kobayashi, The Effect of Message Modality on Memory for Political Disinformation: Lessons from the 2021 U.S. Capitol Riots, 132 Computs. Hum. Behav. 107241 (2022), https://doi.org/10.1016/j.chb.2022.107241.
Large-scale social influences on perception and judgment are not new; the problem is simply exacerbated by online social media. One of the most extraordinary non–social media examples in recent history is the eyewitness report by thousands of devout Portuguese Catholics that on
In the end, there may be little that can be practically done beyond basic education on critical thinking to prevent large-scale social influences on memory and judgment. But law enforcement, the judiciary, juries, and witnesses themselves must be made aware of these growing risks in order to evaluate the probability that eyewitness testimony is accurate.
Evidence-Based Suspicion
The concern here is that a person may be placed as a suspect in a police lineup in the absence of any plausible evidence that the person had anything to do with the crime in question. There are no legal constraints that would prevent this practice. While there are few data that bear on the frequency with which it occurs, evidence indicates that it is not uncommon.190 In fact, a recent survey of U.S. law enforcement agencies revealed that a significant fraction operate on the belief that they need no evidence at all, or perhaps little more than a hunch, before placing a person in a lineup.191 In 2020, the American Psychology-Law Society (AP-LS) charged six experts with developing modern policy and procedure recommendations for eyewitness identifications. Prominent among the recommendations in the AP-LS Report is the law enforcement use of “evidence-based suspicion” as a minimum criterion for lineup construction:
There should be evidence-based grounds to suspect that an individual is guilty of the specific crime being investigated before including that individual in an identification procedure and that evidence should be documented in writing prior to the lineup.192
October 13, 1917, the midday “sun’s disc did not remain immobile. This was not the sparkling of a heavenly body, for it spun round on itself in a mad whirl when suddenly a clamor was heard from all the people. The sun, whirling, seemed to loosen itself from the firmament and advance threateningly upon the earth as if to crush us with its huge fiery weight.” John De Marchi, The Immaculate Heart: The True Story of Our Lady of Fátima (1952) (quote attributed to eyewitness Almeida Garrett). Astronomical records indicate that this “Miracle of the Sun” did not happen, but that is not what the witnesses remembered, reported, and incorporated into life decisions. See also Jeffrey S. Bennett, When the Sun Danced: Myth, Miracles, and Modernity in Early Twentieth-Century Portugal (2012).
190. John T. Wixted et al., Estimating the Reliability of Eyewitness Identifications from Police Lineups, 113 PNAS 304 (2016), https://doi.org/10.1073/pnas.1516814112; Bruce W. Behrman & Regina E. Richards, Suspect/Foil Identification in Actual Crimes and in the Laboratory: A Reality Monitoring Analysis, 29 Law & Hum. Behav. 279 (2005), https://doi.org/10.1007/s10979-005-3617-y.
191. Richard A. Wise, Martin A. Safer & Christina M. Maro, What U.S. Law Enforcement Officers Know and Believe About Eyewitness Factors, Eyewitness Interviews and Identification Procedures, 25 Applied Cognitive Psych. 488 (2011), https://doi.org/10.1002/acp.1717
192. Wells et al., Policy and Procedure Recommendations (2020), supra note 27.
This recommendation,193 focused on law enforcement, is founded on the obvious but often overlooked fact that the simple act of placing someone in a lineup markedly increases the likelihood that they will be accused and prosecuted for a crime. Moreover, if no preexisting evidence ties that person to the crime in question, then the practice will naturally increase the probability that an innocent person is charged. There is also the worrisome question of what underlies a decision by law enforcement to include someone in a lineup in the absence of evidence for suspicion. Hunches of this sort may stem, in part, from the same prejudices that underlie biases in police traffic stops.194 They may also arise from demographic information, including where a person lives relative to the crime scene, prior criminal history, or baseless anonymous tips, none of which rise to a level that could implicate that person, to the exclusion of others, as a legitimate suspect.
The resulting AP-LS recommendation that placement of a person in a lineup must be made only with “evidence-based suspicion” is thus a powerful strategy to avoid wrongful prosecution and conviction.195 While this is a preemptive strategy that is necessarily under the control of the police, not the courts, knowledge of the practice used in any particular case can assist the court in evaluating the probability that the eyewitness testimony is correct.
Prior Exposure and Transference Effects
One feature of memory retrieval highlighted above under basic research (see section titled “Memory Storage” above) is the phenomenon of source memory failure and interference, in which prior knowledge from an “incorrect” source has been incorporated into a retrieved memory, which then biases memory-based decisions and impairs object recognition. This can happen in the context of eyewitness identification procedures employed by police if a witness has a prior opportunity to see a face that is part of a subsequent lineup. In the simplest case, the witness may be shown a mug shot of a potential suspect prior to the actual conduct of a lineup. The mug-shot face is then familiar to the witness at the time of a lineup that includes the face. But the witness may have lost all knowledge of why the face is familiar and naturally attributes it to their memory of the witnessed crime and the perpetrator. In simple terms, the critical memory
193. The recommendation was also made earlier by Gary Wells as the “reasonable suspicion criterion.” Gary L. Wells, Eyewitness Identification: Systemic Reforms, 2006 Wis. L. Rev. 615 (2006).
194. Emma Pierson et al., A Large-Scale Analysis of Racial Disparities in Police Stops Across the United States, 4 Nature Hum. Behav. 736 (2020), https://doi.org/10.1038/s41562-020-0858-1.
195. Failure to follow this recommendation would also seem to violate Fourth Amendment protection against unreasonable searches and seizures.
has been “contaminated,” without awareness, or it has been flat-out replaced by the memory of the mug-shot photo (a phenomenon termed transference).196 The expected result, which has been well documented in applied eyewitness studies, is that identification decisions made under conditions of unconscious memory contamination or transference are frequently incorrect, and often made with inflated confidence.197 Not surprisingly, the same effects can be elicited by showing a witness multiple lineups that contain the same suspect.198
Of critical importance here is the need for the court to be aware of all prior identification procedures—a recommendation made by the NRC committee on eyewitness evidence,199 and by authors of the recent AP-LS document on eyewitness policy and procedures200—as this information bears on the probability that the eyewitness testimony is correct. We note here that the 1968 Supreme Court, Simmons v. United States, fully appreciated the concern about memory contamination from photos and the implications for accuracy of testimony.201
196. Jennifer E. Dysart et al., Mug Shot Exposure Prior to Lineup Identification: Interference, Transference, and Commitment Effects, 86 J. Applied Psych. 1280 (2001), https://doi.org/10.1037//0021-9010.86.6.1280.
197. Committee Report, Report to the Secretary of State for the Home Department of the Departmental Committee on Evidence of Identification in Criminal Cases (1976) [hereinafter Devlin Report]; Evan Brown, Kenneth A. Deffenbacher & William Sturgill, Memory for Faces and the Circumstances of Encounter, 62 J. Applied Psych. 311 (1977), https://doi.org/10.1037/e666602011-038; Kenneth A. Deffenbacher, Brian H. Bornstein & Steven D. Penrod, Mugshot Exposure Effects: Retroactive Interference, Mugshot Commitment, Source Confusion, and Unconscious Transference, 30 Law & Hum. Behav. 287 (2006), https://doi.org/10.1007/s10979-006-9008-1; Dysart et al., supra note 196; Graham Davies, John Shepherd & Hadyn Ellis, Effects of Interpolated Mugshot Exposure on Accuracy of Eyewitness Identification, 64 J. Applied Psych. 232 (1979), https://doi.org/10.1037//0021-9010.64.2.232.
198. Steblay et al., supra note 157.
199. Identifying the Culprit, supra note 12.
200. Wells et al., Policy and Procedure Recommendations (2020), supra note 27.
201. Simmons v. United States, 390 U.S. 377, 383–84 (1968) (“It must be recognized that improper employment of photographs by police may sometimes cause witnesses to err in identifying criminals. A witness may have obtained only a brief glimpse of a criminal, or may have seen him under poor conditions. Even if the police subsequently follow the most correct photographic identification procedures and show him the pictures of a number of individuals without indicating whom they suspect, there is some danger that the witness may make an incorrect identification. This danger will be increased if the police display to the witness only the picture of a single individual who generally resembles the person he saw, or if they show him the pictures of several persons among which the photograph of a single such individual recurs or is in some way emphasized. The chance of misidentification is also heightened if the police indicate to the witness that they have other evidence that one of the persons pictured committed the crime. Regardless of how the initial misidentification comes about, the witness thereafter is apt to retain in his memory the image of the photograph, rather than of the person actually seen, reducing the trustworthiness of subsequent lineup or courtroom identification.”).
Identification Procedures
Law enforcement can readily adopt procedures for how they present faces to an eyewitness. Few topics in applied eyewitness research have spawned as much scientific experimentation, debate, and impact on law enforcement practices, as how to best design these identification procedures. While the goal of these efforts has been to prospectively improve eyewitness performance to better serve the needs of criminal investigation and prosecution, knowledge of the identification procedures that were used in a given case can assist courts in retrospective evaluation of the accuracy of testimony.
Showups
In the simplest identification task, known as a “showup,” the suspect alone is presented to the witness, who responds yes or no. One common reason for conducting a showup is convenience and exigence: Shortly after arriving at the crime scene, the police apprehend a suspect fitting the witness’s description, who is then shown to the witness for confirmation. Despite the seeming benefits of this procedure, it is flawed, owing to suggestive qualities that may lower a witness’s criterion for an identification: (1) the witness’s knowledge that the suspect was apprehended nearby sets up an expectation, (2) the presentation of a sole suspect leads to an inference that the police must be quite certain, and (3) the impromptu context encourages the witness to be forthcoming with assistance. Under the circumstances, we should naturally expect that showups will increase the probability of correct identification (it’s much easier to identify your suitcase if it is the only one on the carousel), but they should also increase the probability of making an incorrect identification. Not surprisingly, applied eyewitness studies demonstrate that mistaken identifications are more common using the showup procedure than with “standard” six-person lineups,202 particularly when innocent suspects resemble the witness’s description of the perpetrator.203
Confirmatory photo identification
Police will, on occasion, display a single photograph to a witness in an effort to confirm the identity of a person of interest. Police typically limit this method
202. A. Daniel Yarmey, Meagan J. Yarmey & Linda A. Yarmey, Accuracy of Eyewitness Identifications in Showups and Lineups, 20 Law & Hum. Behav. 459 (1996), https://doi.org/10.1007/bf01498981.
203. Nancy Steblay et al., Eyewitness Accuracy Rates in Police Showup and Lineup Presentations: A Meta–Analytic Comparison, 27 Law & Hum. Behav. 523, 523–40 (2003), https://doi.org/10.1023/a:1025438223608.
to situations in which the person is previously known to or acquainted with the witness. The goal is to clarify the legal identity (birth name/government name) of an individual who is well known to a witness, but only by a street name. In such examples, a witness may know (and may have known) the perpetrator for years but may be able to identify him only by a street name, such as “Diamond.” The police will typically use this procedure before making an arrest. As for mug books and showups, the procedure is highly suggestive, as it implicates the person in the photo, which may, in turn, unconsciously contaminate or replace the witness’s memory of the face of the perpetrator.204 The practical consequence of this procedure is spelled out above in the section titled “Prior Exposure and Transference Effects,” namely reduced identification accuracy in a subsequent lineup, and artifactually elevated confidence on the part of the witness.205 The existence of this practice highlights, once again, the need for courts to have knowledge of all prior identification procedures in a given case in order to evaluate their implications for eyewitness accuracy.206
Lineups: Simultaneous versus sequential procedures
To engage discriminative abilities and limit selection bias, multiperson lineups have been the standard since at least the nineteenth century. The people in the lineup commonly include the suspect and a set of people known as “fillers,” who are known to be innocent and (ideally) appear similar to the witness’s description of the perpetrator. In laboratory studies, “target absent” (innocent suspect) lineups are also used to enable a measure of the probability of identifying the known culprit relative to the probability of identifying an innocent person. In many studies conducted since the 1970s, this target present versus target absent comparison has been the standard approach to empirically evaluate the effects of estimator and system variables on eyewitness performance (many of which are cited in the foregoing sections).
In the 1980s, the discussion of factors that affect eyewitness performance broadened to include the nature of the multiperson lineup itself. Lindsay and Wells hypothesized that simultaneous viewing of lineup participants—the standard practice of the time—was a factor that prejudiced identifications.207 In particular, they posited that visual comparisons between faces (relative comparisons) could take precedence over the comparisons that really mattered—that is,
204. Wells et al., Recommendations for Lineups and Photospreads, supra note 27, at 457–70.
205. Deffenbacher et al., Mugshot Exposure Effects, supra note 197, at 287–307.
206. Identifying the Culprit, supra note 12; Devlin Report, supra note 197, at 3.
207. R.C. Lindsay & Gary L. Wells, Improving Eyewitness Identifications from Lineups: Simultaneous Versus Sequential Lineup Presentation, 70 J. Applied Psych. 556 (1985), https://doi.org/10.1037//0021-9010.70.3.556.
those between each lineup face and the visual memory of the perpetrator (absolute comparisons), leading to a selective increase in identifications of innocent suspects. To overcome this assumed problem, Lindsay and Wells proposed a novel lineup in which participants are viewed sequentially.
Early laboratory comparisons of performance on simultaneous versus sequential lineups appeared to demonstrate a sequential advantage, based on a higher ratio of correct to incorrect identifications.208 In lockstep with this science and mindful of wrongful convictions, many jurisdictions adopted the new sequential procedure.209 The problem with this outcome originates with the measure of performance—the diagnosticity ratio (DR)—which is affected both by the eyewitness’s ability to discriminate the perpetrator from an innocent suspect and by the criterion used to make an identification decision. It could have been true that sequential lineups also improve discriminability, which would be of real value, but it is impossible to know that using a single DR measure.
There is an established procedure for evaluating discriminability independently of the decision criterion. This procedure, known as analysis of receiver operating characteristics (ROC), is drawn from signal detection theory.210 Figure 13 summarizes results from over 20 years of studies that have employed the ROC approach. Each dot represents data from a published study, identified at right. The x-axis plots a measure of the difference between discriminability measures obtained for the two lineup types, with positive values (SIM > SEQ) indicating a simultaneous advantage, and negative values (SEQ > SIM) indicating a sequential advantage. (Cohen’s d is a standard index of effect size.) Dot size reflects sample size (the number of people tested in the study); the y-axis orders studies as a function of sample size. Green dots indicate studies for which the discriminability difference for lineup types is statistically significant. Thus, contrary to initial claims for the superiority of the sequential lineup based on the DR measure, collective evidence supports a small but consistent improvement in discriminability for the simultaneous lineup relative to sequential.211 Extrapolation from the laboratory suggests that identifications made in jurisdictions that employ the sequential procedure may not be based on the best human discriminative abilities, which is a factor for consideration when assessing the probability that eyewitness testimony is accurate.
208. Id.
209. A National Survey of Eyewitness Identification Procedures, supra note 11.
210. Signal detection theory includes a framework for differentiating a person’s ability to discriminate the presence and absence of a stimulus from the criterion the person uses to make that decision. David M. Green & John A. Swets, Signal Detection Theory and Psychophysics (1966).
211. Travis M. Seale-Carlisle et al., Designing Police Lineups to Maximize Memory Performance, 25 J. Exper. Psych.: Applied 410 (2019), https://doi.org/10.1037/xap0000222; Laura Mickes & John T. Wixted, Eyewitness Memory, in 2294 Oxford Handbook of Human Memory (2022).
Source: Adapted from Laura Mickes & John T. Wixted, Eyewitness Memory, in Oxford Handbook of Human Memory (2022) and used with permission of Oxford Publishing Limited through PLSclear. Data sources: Shannon M. Andersen et al., Individual Differences Predict Eyewitness Identification Performance, 60 Personality & Individual Differences, 36 (2014), https://doi.org/10.1016/j.paid.2013.12.011; Laura Mickes et al., Receiver Operating Characteristic Analysis of Eyewitness Memory: Comparing the Diagnostic Accuracy of Simultaneous Versus Sequential Lineups, 18 J. Experimental Psych.: Applied, 361 (2012), https://doi.org/10.1037/a0030609; Julia Meisters et al., Eyewitness Identification in Simultaneous and Sequential Lineups: An Investigation of Position Effects Using Receiver Operating Characteristics, 26 Memory, 1297 (2018), https://doi.org/10.1080/09658211.2018.1464581; Charles A. Goodsell, Effects of Eyewitness Memory Encoding Strength on Sequential and Simultaneous Lineup Identifications, (unpublished manuscript) (on file with Department of Psychology, Canisius College, Buffalo, New York) (2019); Ruth Horry et al., The Effects of Allowing a Second Sequential Lineup Lap on Choosing and Probative Value, 21 Psych., Pub. Pol’y, & L., 121 (2015), https://doi.org/10.1037/law0000041; Sonja Willing et al., Presenting a Similar Foil Prior to the Suspect Reduces the Identifiability of the Perpetrator in Sequential Lineups: A ROC-Based Analysis, (manuscript under review) (2019); Travis M. Seale-Carlisle et al., Designing Police Lineups to Maximize Memory Performance, 25 J. Experimental Psych.: Applied, 410 (2019), https://doi.org/10.1037/xap0000222; Travis M. Seale-Carlisle & Laura Mickes, US Line-ups Outperform UK Line-ups, 3 Royal Soc’y Open Sci., 160300 (2016), https://doi.org/10.1098/rsos.160300; Caren Rotello et al. (unpublished data); Curt A. Carlson & Maria A. Carlson, An Evaluation of Perpetrator Distinctiveness, Weapon Presence, and Lineup Presentation Using ROC Analysis, 3 J. Applied Rsch. Memory & Cognition, 45 (2014), https://doi.org/10.1016/j.jarmac.2014.03.004; David G. Dobolyi & Chad S. Dodson, Eyewitness Confidence in Simultaneous and Sequential Lineups: A Criterion Shift Account for Sequential Mistaken Identification Overconfidence, 19 J. Experimental Psych.: Applied, 345 (2013), https://doi.org/10.1037/a0034596.
Other lineup procedures
One of the limitations of traditional lineup procedures is that eyewitnesses are asked to make two distinct contributions to the decision: (1) measurement of the similarity between memory and the face of a lineup participant, and (2) classification of the face as a match or nonmatch. An eyewitness identification manifests the classification decision, but the recipient of that decision has no direct access to the measurement upon which it was based, or to the criterion that the eyewitness used to evaluate the measurement, leaving the decision susceptible to unrecognized bias. This is notably distinct from man-made information processing devices (e.g., smartphone fingerprint and face detectors), which yield similarity measurements that are accessible and classifiable by an objective statistical procedure. A lineup procedure could, in principle, render a similarity measure but exclude the eyewitness from the classification—the latter performed instead by the lineup administrator using an objective statistical analysis. Recent laboratory studies demonstrate that this is a feasible, high-accuracy approach that could transform the utility of eyewitness lineups,212 but adoption by the criminal justice system must wait for evaluation in field studies with real casework, as we consider below.
In-court identifications
In criminal trials that are built on eyewitness testimony, a nearly ubiquitous feature is the theatrical event in which a witness is asked to identify the perpetrator in front of the jury. This practice has two consequences that adversely impact efforts by jurors to assess the probability that the witness testimony is correct.
First, the witness’s level of confidence in the identification made at trial is generally highly inflated. In the Courtney case summarized above (see section titled “An Illustrative Case”), the victim-witness reported a confidence level of 60% upon initial identification. After identifying Courtney in court at the request of the prosecution, the witness expressed their confidence as: “I will never forget what he looks like.” The section titled “Timing of identification and conditions of confidence assessment,” above, delves into the specific reasons why this happens.
Second, the inflated confidence expressed during an in-court identification is not commonly recognized as such by a lay jury. It thus becomes a very powerful form of rhetoric that conveys authority on the part of the witness and stirs the
212. Sergei Gepshtein et al., A Perceptual Scaling Approach to Eyewitness Identification, 11 Nature Commc’ns 3380 (2020), https://doi.org/10.1038/s41467-020-17194-5; Sergei Gepshtein & Thomas D. Albright, Thinking Outside the Lineup Box: Eyewitness Identification by Perceptual Scaling, 10 J. Applied Rsch. Memory & Cognition 221 (2021), https://doi.org/10.1016/j.jarmac.2021.05.001.
emotions of the jurors.213 All of which moves them toward a decision to convict.214 As Supreme Court Justice William Brennan wrote, “[T]here is almost nothing more convincing than a live human being who takes the stand, points a finger at the defendant, and says ‘That’s the one!’”215 This “convincing,” of course, is the primary reason why expert testimony or carefully worded jury instructions can be valuable, particularly when they explain the reasons why confidence is inflated.
In view of the potential for biasing the outcome, one might argue that in-court identifications should be eliminated altogether in criminal trials.216 This proposal has gained some traction. The Massachusetts Supreme Judicial Court has ruled, for example, that in-court identifications should not normally be permitted unless they are the first identification, or if the out-of-court identification was suppressed.217 The Connecticut Supreme Court has ruled to allow an in-court identification only if the witness knew the defendant before the crime, or if a prior lineup was proved to be nonsuggestive.218 These rules have exceptions, as noted, and other courts conduct a burden-shifting approach toward in-court identifications; an in-court identification could also potentially be so unduly suggestive as to amount to a violation of Manson v. Brathwaite.219 Research supports the view that in-court identifications should be handled with great caution, and that jurors must be made aware that they should not place such great weight on the in-court confidence of an eyewitness.
213. Maksymilian Del Mar, The Confluence of Rhetoric and Emotion: How the History of Rhetoric Illuminates the Theoretical Importance of Emotion, 36 Law & Literature 1 (2024), https://doi.org/10.1080/1535685x.2022.2115651; Aristotle, The Rhetoric of Aristotle: A Translation (Richard Claverhouse Jebb trans., 1909).
214. Brewer & Burke, supra note 173; Slane & Dodson, supra note 173.
215. Watkins v. Sowders, 449 U.S. 341, 353 (1981) (Brennan, J., dissenting) (citing Elizabeth F. Loftus, Eyewitness Testimony (1979)).
216. Brandon L. Garrett, Eyewitnesses and Exclusion, 65 Vand. L. Rev. 451 (2012).
217. See Commonwealth v. Johnson, 45 N.E.3d 83, 92–94 (Mass. 2016) (reasoning that “a subsequent in-court identification cannot be more reliable than the earlier out-of-court identification, given the inherent suggestiveness of in-court identifications and the passage of time”); Commonwealth v. Crayton, 21 N.E.3d 157, 169 (Mass. 2014) (“Where an eyewitness has not participated before trial in an identification procedure, we shall treat the in-court identification as an in-court showup, and shall admit it in evidence only where there is ‘good reason’ for its admission.”).
218. State v. Dickson, 141 A.3d 810, 817 (Conn. 2016) (“[I]n cases in which identity is an issue, in-court identifications that are not preceded by a successful identification in a nonsuggestive identification procedure implicate due process principles and, therefore, must be prescreened by the trial court.”). Note that this ruling has been challenged. See Tatum v. Comm’r of Correction, No. 20727, 2024 WL 3433265, at *3 (Conn. July 16, 2024).
219. See State v. Hickman, 330 P.3d 551, 568 (Or. 2014) (en banc) (“Courts considering the admissibility of first-time in-court identifications generally have placed the burden of seeking a prophylactic remedy on the defendant.”); United States v. Domina, 784 F.2d 1361, 1369 (9th Cir. 1986) (noting that district court’s denial of request for in-court lineup will only be overturned if in-court identification procedures were so suggestive and conducive to irreparable misidentification as to amount to denial of due process).
Filler Selection
Fillers are lineup participants known to be innocent, who serve as lures to challenge recognition memory. The choice of fillers has long been known to markedly influence eyewitness performance in laboratory studies.220 People recognize objects probabilistically based on the degree to which they elicit a memory signal corresponding to a particular target previously seen. It naturally follows that similar objects are more likely to elicit the same memory signal and thus have similar likelihoods of being recognized as the target. The similarity of fillers to the witness’s description of the perpetrator is thus an important variable that affects both eyewitness discriminability and accuracy.
Lineups composed of fillers who are all of roughly the same degree of similarity to the witness’s description are termed “fair.” Conversely, lineups composed of fillers that possess differing degrees of similarity to the description are termed “unfair” or “biased.” An unfair lineup, in which one filler is closer in similarity to the description of the perpetrator, reduces uncertainty by reducing the number of sensible choices, thus increasing the likelihood of correctly identifying the perpetrator, but also increasing the likelihood of misidentifying someone who looks like the perpetrator.221
Despite the well-understood and potentially disastrous consequences of unfair lineups, there have been few attempts to systematize the law enforcement process of filler selection. Published guidelines for filler selection state that lineups should be constructed to ensure that “the suspect does not unduly stand out” and lineups should “avoid using fillers that so closely resemble the suspect that a person familiar with the suspect might find it difficult to distinguish the suspect from the fillers.”222 This guidance—fillers should be similar to the suspect but not too much so—is open to interpretation and is applied by different agents in different ways. It has not been effective.223
One emerging approach would use face similarity metrics to create lineups in which fillers are of known similarity to each other and to the witness’s description of the perpetrator. A recent meta-analysis of laboratory studies adopting this approach revealed that decreasing the similarity between the suspect and fillers has the effect of increasing the probability that the suspect is identified,
220. Roy S. Malpass & Patricia G. Devine, Measuring the Fairness of Eyewitness Identification Lineups, in 81 Evaluating Witness Evidence (1983); Gary L. Wells, Michael R. Leippe & Thomas M. Ostrom, Guidelines for Empirically Assessing the Fairness of a Lineup, 3 Law & Hum. Behav. 285 (1979), https://doi.org/10.1007/bf01039807.
221. This effect is much like the consequence of a showup identification.
222. U.S. Dep’t of Just., Nat’l Inst. of Just., supra note 27.
223. Carmen A. Lucas, Neil Brewer & Matthew A. Palmer, Eyewitness Identification: The Complex Issue of Suspect-Filler Similarity, 27(2) Psych., Pub. Pol’y, & L. 151 (2021).
regardless of whether that identification is correct.224 The case of Uriah Courtney, highlighted above, illustrates the tragic consequences of this practice in the real world: Courtney’s conviction was based on two witness identifications using a lineup in which Courtney stood out from many of the fillers as an exemplar of the witness descriptions.225 By contrast, increasing the similarity between suspect and fillers increases the probability of filler identification or a lineup rejection, meaning that fewer suspects are identified overall. A more recent analysis suggests that this is a consequence of reduced discriminability between perpetrator and innocent suspect.226
A recent laboratory study adds a new wrinkle to this picture and a promising target for future investigation: Testing the counterintuitive predictions of a signal detection model, the investigators found that “choosing fillers who match the description of the perpetrator but who are otherwise dissimilar to the suspect’s appearance yielded fair lineups and enhanced eyewitness identification performance.”227 To appreciate this conclusion, suppose that the witness’s verbal description of the perpetrator included red hair and a goatee. A suspect with those features is apprehended, and all lineup fillers are selected to have the same diagnostic features (i.e., the lineup is fair). The perpetrator may have other features not mentioned by the witness, such as blue eyes. Fillers may share the blue eyes feature with the perpetrator by chance, increasing the probability that the witness will misidentify them. Thus, the lineup is more likely to yield a correct outcome if fillers are chosen without regard to features that were not part of the witness’s description of the perpetrator.
Recent evidence points to another variable associated with filler–suspect similarity that has a somewhat counterintuitive effect on eyewitness performance: the size of the face database from which fillers are selected. One might expect that having a large library of faces to draw from, such as computer databases of driver’s license photos, would make it possible to find filler faces that are more suitable for a lineup and may facilitate eyewitness performance.228 Contrary to this intuition, selection of fillers from very large databases is commonly associated with greater similarity between the suspect and lineup fillers, which has the consequence of increasing the frequency of filler picks by eyewitnesses, thus reducing accuracy.229
224. Ryan J. Fitzgerald et al., The Effect of Suspect-Filler Similarity on Eyewitness Identification Decisions: A Meta-Analysis, 19 Psych., Pub. Pol’y, & L. 151 (2013), https://doi.org/10.1037/a0030618.
225. Albright, Why Eyewitnesses Fail, supra note 14, at 7758–64.
226. Fitzgerald et al., supra note 224, at 151.
227. Melissa F. Colloff et al., Optimizing the Selection of Fillers in Police Lineups, 118 PNAS e2017292118 (2021), https://doi.org/10.1073/pnas.2017292118.
228. Fitzgerald et al., supra note 224, at 151.
229. Amanda N. Bergold & Paul Heaton, Does Filler Database Size Influence Identification Accuracy?, 42 Law & Hum. Behav. 227 (2018), https://doi.org/10.1037/lhb0000289.
With a movement toward automated systems for filler selection, their ease of use, and their adoption by the police,230 database size is a critical variable.
There are many more details to be worked out in this space of filler–suspect similarity, but on the whole, it is clear that filler selection matters greatly to eyewitness performance, and thus the approach that was taken in a given case may offer insight into the probability that the eyewitness testimony is correct.
Corroborating Variables: Machine-Based Facial Recognition
There are, of course, many other variables associated with eyewitness events that fall in neither the estimator nor system category. For police investigations and criminal trials involving an eyewitness, the most useful of these variables are sources of forensic evidence that bear on, and potentially corroborate, eyewitness identification. The forms of evidence in this category are numerous, including, but not limited to, fingerprints, tool marks, chemical/biological residues, and DNA. The utility of these forms of evidence for the courts is treated elsewhere.231
One specific and very recent form of corroborating evidence deserves mention here: automated facial recognition. New artificial intelligence (AI) technology for machine-based person recognition derived from photographs or video collected in the real world, combined with the dramatic proliferation of live cameras in public spaces, has enabled law enforcement to harvest identification evidence that may complement traditional eyewitness testimony. This is a rapidly evolving and largely unregulated field, which has considerable relevance to judicial decisions about the accuracy of eyewitness testimony from real people. We refer readers to a report from the National Academies of Sciences, Engineering, and Medicine, which details the technology and its accuracy, utility, potential applications, and privacy concerns.232
Applied Eyewitness Studies in the Field
The results of applied studies conducted in the laboratory have revealed a large number of variables that affect the ability of witnesses to discriminate between the perpetrator and innocent suspects, as well as the overall accuracy of
230. A National Survey of Eyewitness Identification Procedures, supra note 11.
231. See, e.g., Valena E. Beety, Jane Campbell Moriarty & Andrea L. Roth, Reference Guide on Forensic Feature Comparison Evidence, in this manual.
232. Nat’l Acads. Scis., Eng’g & Med., Facial Recognition Technology: Current Capabilities, Future Prospects, and Governance (2024), https://doi.org/10.17226/27397.
identifications made. These findings both encourage this empirical strategy and hold promise for reform, but they have often been criticized for lack of ecological validity.233 Specifically, the laboratory crime is a recognized pretense that cannot be expected to elicit critical states of the observer, such as high values of stress, fear, pain, and loss of attentional focus. The ultimate test of the utility of laboratory discoveries is thus successful translation to real criminal casework as evaluated by field studies.234
There are three constraints associated with achieving this translational goal. First, it requires access to a high-functioning police department with a willingness to test scientific hypotheses. Second, ground truth is unknown in real cases, which means that experimental efforts must rely on independent corroboration of truth (“extrinsic evidence”) as a means to validate the effects of the variables tested. That reliable corroboration is often difficult to come by. Third, the situational context in the field is difficult to control with the level of precision that is common in a laboratory. Real-world things sometimes happen unexpectedly, or there may simply be no record of what did happen.
Operating under these constraints, a small number of studies have tested the effects of estimator and system variables in the field. Results generally support findings from laboratory studies of effects of retention interval, cross-race effect, weapon focus, and lineup type,235 but many factors have been poorly
233. Yoojin Chae, Application of Laboratory Research on Eyewitness Testimony, 10 J. Forensic Psych. Prac. 252 (2010), https://doi.org/10.1080/15228930903550608; Graham F. Wagstaff et al., Can Laboratory Findings on Eyewitness Testimony Be Generalized to the Real World? An Archival Analysis of the Influence of Violence, Weapon Presence, and Age on Eyewitness Accuracy, 137 J. Psych. 17 (2003), https://doi.org/10.1080/00223980309600596.
234. Daniel L. Schacter et al., Policy Forum: Studying Eyewitness Investigations in the Field, 32 Law & Hum. Behav. 3 (2008), https://doi.org/10.1007/s10979-007-9093-9.
235. John C. Yuille & Judith L. Cutshall, A Case Study of Eyewitness Memory of a Crime, 71 J. Applied Psych. 291 (1986), https://doi.org/10.1037//0021-9010.71.2.291; Yarmey et al., supra note 202, at 459–77; Bruce W. Behrman & Sherrie L. Davey, Eyewitness Identification in Actual Criminal Cases: An Archival Analysis, 25 Law & Hum. Behav. 475 (2001), https://doi.org/10.1023/a:1012840831846; Amy Klobuchar, Nancy K. Mehrkens Steblay & Hilary Lindell Caligiuri, Improving Eyewitness Identifications: Hennepin County’s Blind Sequential Lineup Pilot Project, 4 Cardozo Pub. L., Pol’y & Ethics J. 381 (2006); Daniel L. Schacter et al., Policy Forum: Studying Eyewitness Investigations in the Field, 32 L. & Hum. Behav. 3 (2008), https://doi.org/10.1007/s10979-007-9093-9; Nancy K. Steblay, What We Know Now: The Evanston Illinois Field Lineups, 35 Law & Hum. Behav. 1 (2011), https://doi.org/10.1007/s10979-009-9207-7; Gary L. Wells, Nancy K. Steblay & Jennifer E. Dysart, A Test of the Simultaneous vs. Sequential Lineup Methods: An Initial Report of the AJS National Eyewitness Identification Field Studies, Am. Judicature Soc. (2011), https://perma.cc/42F5-LVL5; Gary L. Wells, Nancy K. Steblay & Jennifer E. Dysart, Double-Blind Photo Lineups Using Actual Eyewitnesses: An Experimental Test of a Sequential Versus Simultaneous Lineup Procedure, 39 Law & Hum. Behav. 1 (2015), https://doi.org/10.1037/lhb0000096; Karen L. Amendola & John T. Wixted, Comparing the Diagnostic Accuracy of Suspect Identifications Made by Actual Eyewitnesses from Simultaneous and Sequential Lineups in a Randomized Field Trial, 11 J. Exper. Crim. 263 (2015), https://doi.org/10.1007/s11292-014-9219-2; Wixted et al., supra note 190.
controlled. Moreover, in real police investigations it can be hard to assess ground truth, and cognitive biases can play a complicating role in perceptions of evidence strength. Much more needs to be done in this difficult arena before changes are made to policy and practice. It seems possible that the laboratory versus field relationship will break down where the real emotion of criminal witnessing comes into play.
Effects of Eyewitness Testimony on Juries
The identification decision made by an eyewitness is only the first stage in which human factors may drive a criminal prosecution off track. The second stage is in the communication of eyewitness testimony to the trier of fact. Recent scientific studies have explored the role that communication plays in inferring the accuracy of eyewitness testimony. Specifically, the witness is viewed as the transmitter of information, and the jury is the intended receiver of that information. While much science focuses on improving the accuracy of testimony and the ability to predict that accuracy, these efforts will be in vain if the information is not correctly received by the trier of fact.
Laypeople may have a range of misconceptions regarding how eyewitness memory functions. Further, there is a traditional practice of allowing eyewitnesses to identify a person in the courtroom (see section titled “In-court identifications” above). As Justice Sonia Sotomayor explained in dissent in Perry v. New Hampshire: “At trial, an eyewitness’ artificially inflated confidence in an identification’s accuracy complicates the jury’s task of assessing witness credibility and reliability.”236 Research has confirmed that these in-court identifications powerfully influence jurors, and in addition, that the confidence of an eyewitness in court has a particularly outsized influence on lay jurors.237 An eyewitness may appear to be highly confident in court, even if the witness was highly uncertain during the earlier eyewitness identification procedure conducted by police out of court—the phenomenon known as confidence inflation (see above section titled “Timing of identification and conditions of confidence assessment”).238 Indeed, confidence can be inflated by biasing information, including feedback that eyewitnesses receive after an identification procedure and in preparation for testimony at trial.239
236. 565 U.S. 228, 252 (2012), (Sotomayor, J., dissenting).
237. See generally, Garrett et al., supra note 33 (detailing findings of mock juror survey regarding weight given to confidence of eyewitnesses).
238. See Wells et al., Policy and Procedure Recommendations (2020), supra note 27, at 21.
239. Carl Martin Allwood, Jens Knutsonn & Pär Anders Granhag, Eyewitnesses Under Influence: How Feedback Affects the Realism in Confidence Judgements, 12 Psych. Crime & L. 25, 36 (2006), https://doi.org/10.1080/10683160512331316316.
As noted below, some jurisdictions and courts have adopted jury instructions and other mechanisms to improve communication between an eyewitness or their proxy (transmitter) and a jury (receiver), to better educate jurors regarding the limitations of eyewitness testimony, including by limiting the use of in-court identifications. There is, however, evidence that several of these more recent sets of jury instructions are not particularly effective at improving jurors’ ability to discriminate between more or less reliable eyewitness identifications,240 perhaps because they do not sufficiently address the weight that jurors place on the courtroom confidence of the eyewitness, and—as highlighted above—the rhetorical impact of an in-court identification. Indeed, recent evidence shows that more pointed instructions regarding in-court confidence, and the reasons why there are dangers of relying on it, may be more effective.241 Expert witnesses, of course, may address these issues head-on, and with a richer description of the relevant scientific research.
Sources of Expertise on Matters of Eyewitness Science
The value and utility of eyewitness evidence lie in both the content of testimony (who did what to whom) and the probability that the testimony is correct. As we have emphasized throughout this text, there are both general and specific features of the process of perceiving criminal events, identifying perpetrators, and testifying to experience, that bear on the probability that testimony is accurate. Knowledge and understanding of these features and their predictive relationship to accuracy are often beyond the ken of jurors, judges, and witnesses proffering testimony. In addition to numerous court rulings on this topic (particularly Manson v. Brathwaite, but also more recent federal and state court decisions that digest and rely on scientific research in the area), there are many sources of clarifying information that can help courts decide about admissibility of eyewitness testimony and assess its probative value.
These include written documents, such as (but not limited to) the present reference guide, the 2014 NRC report on eyewitness identification,242 the 2019
240. Jones & Bergold, supra note 33, at 433–55; Berman, supra note 33; Marlee Kind Dillon et al., Henderson Instructions: Do They Enhance Evidence Evaluation, 17 J. Forensic Psych. Rsch. & Prac. 1 (2017); Garrett et al., supra note 33; Jones et al., supra note 33; Laub et al., supra note 33; Papailiou et al., supra note 33.
241. Brandon L. Garrett et al., Sensitizing Jurors Regarding Eyewitness Testimony, 12 J. Applied Rsch., Memory & Cognition 1, 141–57 (2022), https://doi.org/10.1037/mac0000035.
242. Identifying the Culprit, supra note 12.
report of the Third Circuit task force on eyewitness identifications,243 and the eyewitness policy and procedural recommendations of the AP-LS.244 The underpinning of these documents is a very large peer-reviewed scientific literature from basic and applied fields of psychology and/or cognitive and neural sciences, with particular focus on vision, memory, attention, signal detection, predictive coding, and decision-making (much of which is cited herein). In addition to these written guides, which often themselves require clarification and interpretation, decisions about admissibility of eyewitness evidence and assessment of its weight (probability that the evidence is correct) can be informed by experts with demonstrated training and proficiency in the relevant fields of science. General standards for admissibility of such experts are, of course, defined by the Supreme Court’s Daubert245 ruling and Rule 702 of the Federal Rules of Evidence, which emphasize both qualifications of the expert and the validity and reproducibility of the underlying science reported by the expert.246
On the specific topic of eyewitness evidence, expert qualifications and experience should include a published record of scholarly empirical work on phenomenology, function, and underlying mechanisms of visual perception and recognition memory, applied studies of the effects of contextual variables (viewing conditions and cognitive state of the observer) on eyewitness discriminability and accuracy, and probabilistic modeling of evidence-based behavioral choice under conditions of uncertainty. Experts with these qualifications and knowledge would commonly hold tenured or tenure-track positions at accredited universities or in nonprofit research institutes, in academic disciplines of psychology and/or cognitive and neural sciences, and statistics. Such experts can draw clear and useful (and sometimes precisely quantitative) inferences about (but not limited to) the following: (1) the effects of contextual variables on initial eyewitness collection of information about criminal events, such as atmospheric conditions (e.g., rain, fog), viewing duration, viewing distance, illuminant intensity, luminance contrast, viewing angle, image “crowding,” distracting features that hijack attention, expectations, stress, and emotional states; (2) the degradation and contamination of memories of witnessed events with the passage of time; (3) the influence of particular behavioral strategies (including introduction of uncertainty and bias) for eliciting eyewitness recognition memory (e.g., lineup type, filler selection); (4) the social influence of other parties (e.g., law enforcement officers, attorneys, news media, friends, family) on choice and confidence in
243. Report of the United States Court of Appeals for the Third Circuit Task Force on Eyewitness Identifications, supra note 29.
244. Wells et al., Policy and Procedure Recommendations (2020), supra note 27, at 3–36.
245. Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579 (1993).
246. Thomas D. Albright, A Scientist’s Take on Scientific Evidence in the Courtroom, 120 PNAS e2301839120 (2023), https://doi.org/10.1073/pnas.2301839120.
proffered testimony; (5) the utility of self-report measures (e.g., confidence) that may be presented as indicia of accuracy; and (6) the consequences of poor communication between eyewitness (and counsel) and trier of fact.
The Legal Framework for Eyewitness Evidence
The potential inaccuracy of eyewitness evidence has long been well known to judges. As the U.S. Supreme Court put it in its 1967 ruling in United States v. Wade, “The vagaries of eyewitness identification are well-known; the annals of criminal law are rife with instances of mistaken identification.”247 The traditional constitutional criminal procedure approach to such evidence has involved several related protections, including the right to counsel at an in-person lineup, as well as a due process rule requiring a deferential review of the validity of an unduly suggestive lineup. In addition, courts have conducted pretrial hearings to examine the validity of eyewitness evidence, and offered fairly basic and brief jury instructions to explain considerations relevant to eyewitness evidence to jurors. Those rules were developed in the late 1960s and 1970s, and in federal courts, the basic framework has remained largely unchanged ever since.
In the nearly five decades since, a substantial body of visual perception and eyewitness memory science, as described above, has transformed our understanding of eyewitness evidence. This research has affected police identification practices, including those of federal law enforcement agencies. In federal court, these developments have informed, for example, a move toward receptivity to expert testimony regarding the science of eyewitness vision and memory, where it would assist the jury and satisfy other requirements regarding expert testimony,248 with other courts preferring the use of jury instructions.249 More recently, a noteworthy Third Circuit Task Force on Eyewitness Evidence examined a range of recommendations regarding how to consider eyewitness evidence.250
247. United States v. Wade, 388 U.S. 218, 228 (1967).
248. See generally, United States v. Rodriguez-Felix, 450 F.3d 1117, 1124 (10th Cir. 2006), (summarizing “there has been a trend in recent years to allow such testimony” and “subject to the trial court’s careful supervision, properly conceived expert testimony may be admissible to challenge or support eyewitness evidence” when it would assist the jury); United States v. Harris, 995 F.2d 532, 534 (4th Cir. 1993) (summarizing “[u]ntil fairly recently, most, if not all, courts excluded expert psychological testimony on the validity of eyewitness identification” but describing how courts came to recognize trial court discretion to allow such testimony).
249. See, e.g., United States v. Jones, 689 F.3d 12, 20 (1st Cir. 2012) (finding it “within the district court’s province to provide this information through instructions rather than through dueling experts”).
250. Report of the Third Circuit Task Force on Eyewitness Identifications, supra note 27.
State judges have adopted new frameworks for considering admissibility that incorporate new scientific developments: through interpretation of state constitutional provisions, through evidence law, by tasking state commissions with revising jury instructions, and through the judges’ supervisory power. A series of courts have concluded, for example, “as a matter of state constitutional law, that it is appropriate to modify [due process standards] to conform to recent developments in social science and the law.”251 Courts increasingly recognize that vision is limited by uncertainty and bias, where “research further reveals that an array of variables can affect and dilute memory and lead to misidentifications.”252 State courts have adopted new jury instructions for eyewitness evidence, new standards for in-court identifications, and new admissibility standards, as well as requiring that pretrial reliability hearings (validity hearings) be conducted.253 As in federal courts, most state courts now permit, within the discretion of the trial judge, the appointment of expert witnesses on eyewitness perception and memory and may find it to be an error to exclude an eyewitness expert that would play an important role in educating the jury.254 Further, state lawmakers have been quite active in this area, enacting a series of statutes that regulate eyewitness identification procedures, and also, to a lesser extent, how judges review such evidence when procedures do not comply with these statutes.255
The sections that follow describe: (1) the U.S. Supreme Court’s constitutional rulings regarding eyewitness evidence; (2) the rulings by federal courts of appeals and lower federal courts; (3) state court rulings, including those creating new frameworks for reviewing eyewitness evidence, and those regarding expert evidence and jury instructions; and (4) state statutes regarding eyewitness evidence.
The U.S. Supreme Court
The U.S. Supreme Court first examined the constitutional criminal procedure implications of eyewitness identifications in a trilogy of cases decided in 1967. In Gilbert v. California256 and United States v. Wade,257 the Court held that, once indicted, a person has a right under the Sixth Amendment to have a defense lawyer present at an in-person live lineup.258 Further, in Wade, the Court held that a failure to provide defense counsel at a postindictment lineup does not
251. State v. Harris, 191 A.3d 119, 134 (Conn. 2018).
252. State v. Henderson, 27 A.3d 872, 247 (N.J. 2011).
253. See Thomas D. Albright & Brandon L. Garrett, The Law and Science of Eyewitness Evidence, 102 Boston U. L. Rev. 511, 591 & Appendix B (2022) (surveying state court rulings).
254. See id. at 593 & Appendix C (surveying leading state cases).
255. See id. at 580 & Appendix A (surveying state statutes).
256. Gilbert v. California, 388 U.S. 263 (1967).
257. United States v. Wade, 388 U.S. 218, 228 (1967).
258. Id. at 236–38; Gilbert, 388 U.S. at 272.
result in suppression of the evidence if the witness had an “independent source” for the identification, based on factors including the opportunity to observe the culprit at the crime scene, any discrepancies in witness descriptions, any prior identifications or failure to identify the person, and the lapse of time between the act and the lineup.259 The Court later held, in its 1973 ruling in United States v. Ash, that right to counsel does not extend to photo array procedures, which, today, police use far more than live or in-person lineups.260
In Stovall v. Denno,261 the third case in the 1967 trilogy, the Supreme Court held that the Due Process Clause also regulates eyewitness identification procedures, stating that certain procedures may be “so unnecessarily suggestive” that the identification evidence must be suppressed.262 However, the Court rejected any per se rule against the use of showup identifications and held that, based on a review of the totality of the circumstances, such identifications could be admissible despite the risks of suggestion.263 Further, in its 1968 ruling in Simmons v. United States,264 the Court emphasized the overall reliability of the eyewitness’s identification in finding no grounds for suppressing the identification under the due process theory announced in Stovall.265
In 1972, in Neil v. Biggers,266 the Court found, in a case involving a trial that occurred pre-Stovall, that even if an identification was conducted in an unnecessarily suggestive manner, a court should consider five factors in examining whether there was a “likelihood of misidentification” and a due process remedy warranted:
the opportunity of the witness to view the criminal at the time of the crime, the witness’ degree of attention, the accuracy of the witness’ prior description of the criminal, the level of certainty demonstrated by the witness at the confrontation, and the length of time between the crime and the confrontation.267
In 1977, in Manson v. Brathwaite, the Supreme Court adopted this Biggers test for all cases, as a general due process rule for consideration of eyewitness identification evidence potentially affected by undue law enforcement suggestion.268 The Court emphasized that “reliability is the linchpin in determining the admissibility of identification testimony.”269 Although the Court’s due process rule
259. Wade, 388 U.S. at 241.
260. United States v. Ash, 413 U.S. 300, 321 (1973).
261. Stovall v. Denno, 388 U.S. 293 (1967).
262. Id. at 301–02.
263. Id.
264. Simmons v. United States, 390 U.S. 377 (1968).
265. Id. at 378, 384–86.
266. Neil v. Biggers, 409 U.S. 188 (1972).
267. Id. at 199–200.
268. Manson v. Brathwaite, 432 U.S. 98, 114 (1977) (holding Biggers factors should be used to assess reliability).
269. Id.
asks whether police used suggestive identification procedures, any such suggestiveness can be excused based on a review of the five “reliability” factors set out in Biggers.270 The Court did not assign any particular weight to these factors.271
In the years since the ruling, scientific research has advanced considerably, and in ways that might inform how a court applies the test. One central concern that scientists have raised concerning this test is that it could permit judges to excuse suggestive eyewitness identification procedures, based on an assumption that other indicia of reliability might counteract the influence of police suggestion.272 The NRC report emphasized:
[T]he test treats factors such as the confidence of a witness as independent markers of reliability when, in fact, it is now well established that confidence judgments may vary over time and can be powerfully swayed by many factors.273
However, in applying the test, the Biggers factors need not be considered as “independent,” and judges have discretion regarding what weight to attach to them. Thus, the application of the Manson test could be informed by more recent scientific research.
Subsequent rulings by the Supreme Court have not altered the test adopted in Manson, but they have added further discretion and also limitations to the inquiry. In Watkins v. Sowders, the Supreme Court held that an evidentiary hearing regarding identification testimony, outside the presence of the jury, need not be conducted in all cases, but remains within the discretion of the trial judge.274 The Court noted that it may often be advisable to do so outside the presence of the jury but that in some situations it would not be needed, and cross-examination of the eyewitness can preserve a defendant’s due process rights.275
Most recently, the Supreme Court held in Perry v. New Hampshire that when eyewitness misidentifications are not due to intentional police action, the Due Process Clause does not apply.276 In that case, the suspect was detained near the
270. Id.
271. Id. at 116 (stating only that Biggers factors are “for the jury to weigh”).
272. Wells & Quinlivan, supra note 20, at 16 (condemning federal courts’ application of Manson test to find that identifications were reliable even when surrounding procedures were “highly suggestive”).
273. See Identifying the Culprit, supra note 12.
274. Watkins v. Sowders, 449 U.S. 341, 349 (1981).
275. Id. For examples of lower federal courts emphasizing that cross-examination sufficiently addressed reliability concerns, see, e.g., United States v. Maloney, 513 F.3d 350, 356 (3d Cir. 2008); United States v. Saint Louis, 889 F.3d 145, 154–55 (4th Cir. 2018); Amador v. Quarterman, 458 F.3d 397, 415 (5th Cir. 2006); United States v. Stokes, 631 F.3d 802, 807 (6th Cir. 2011).
276. Perry v. New Hampshire, 565 U.S. 228, 245 (2012) (“The fallibility of eyewitness evidence does not, without the taint of improper state conduct, warrant a due process rule requiring a trial court to screen such evidence for reliability before allowing the jury to assess its creditworthiness.”).
crime scene and the eyewitness looked out of her apartment, saw him there, and made an identification.277 The majority did note that they “do not doubt either the importance or the fallibility of eyewitness identifications” but maintain that state evidence law and safeguards such as expert testimony and jury instructions should be relied on to ensure accurate presentation of evidence.278
Lower Federal Court Rulings
In 1972, in United States v. Telfaire, the Court of Appeals for the District of Columbia Circuit crafted particularly influential jury instructions regarding eyewitness evidence.279 The instructions explain that jurors “must consider the credibility of each identification witness in the same way as any other witness, consider whether he is truthful, and consider whether he had the capacity and opportunity to make a reliable observation on the matter covered in his testimony.” The instructions ask jurors to: examine the “circumstances under which the identification was made”; “scrutinize the identification with great care”; “consider the length of time that lapsed between the occurrence of the crime and the next opportunity of the witness to see the defendant”; and consider any instances in which the witness failed to identify the defendant.
The instructions are fairly brief and generically touch on certain factors that can be relevant to assessing the reliability of eyewitness evidence. The Telfaire ruling was a step forward at the time, encouraging judges to provide additional context regarding eyewitness evidence. To be sure, the ruling predates the body of modern research on the subject, and as we will describe, courts have more recently offered instructions explicitly grounded in scientific research. While most federal courts use the Telfaire instructions,280 other federal courts adopt a “flexible approach” that provides district courts with discretion whether to use eyewitness jury instructions should they conclude, based on “strong reliability” using the Biggers/Manson factors, that no such instruction is necessary.281 Thus, those federal courts rely on the factors set out in Manson v. Brathwaite in deciding whether to give the jury the Telfaire jury instructions.282 Some federal courts
277. Id.
278. Id.
279. United States v. Telfaire, 469 F.2d 552 (D.C. Cir. 1972) (per curiam).
280. United States v. Anderson, 739 F.2d 1254, 1258 (7th Cir. 1984); see also United States v. Greene, 591 F.2d 471, 475 (8th Cir. 1979).
281. See, e.g., United States v. Luis, 835 F.2d 37, 41 (2d Cir. 1987) (“We believe this flexible approach remains the better course because it avoids imposing rigid requirements on trial courts under the threat that failure to give the requested charge will later be grounds for automatic reversal.”).
282. Telfaire, 469 F.2d at 558–59 (providing model jury special instructions on identification).
have also modestly supplemented the Telfaire instructions to include additional factors, such as whether it was a cross-racial identification.283
On the question of admissibility of expert evidence regarding eyewitness memory, federal courts, like state courts, have in recent years become more receptive to the use of such experts.284 Thus, in 1999, the Eighth Circuit explained it was “especially hesitant” to find it an abuse of discretion not to admit eyewitness expert testimony unless the case rests “exclusively on uncorroborated eyewitness testimony.”285 Other courts, however, emphasized the usefulness of eyewitness expert testimony and found it at times an abuse of discretion not to permit the defense such an expert.286 Pre-Daubert, the Eleventh Circuit was the only federal court of appeals that had adopted a per se rule that eyewitness expert evidence would not be permitted.287 The Eleventh Circuit still maintains that approach, while every other court of appeals now permits such experts, subject to the discretion of the district judge.288
More recent federal rulings regarding experts on eyewitness evidence have relied more heavily on modern scientific research to explain the need to provide such research to the jury. For example, the Second Circuit, in its 2021 ruling in United States v. Nolan, found defense counsel per se ineffective for failing to call an expert witness on eyewitness evidence. The panel explained: “Strickland ordinarily does not require defense counsel to call any particular witness,” but given
283. Comm. on Model Crim. Jury Instructions for the Third Circuit, Model Criminal Jury Instructions § 4.15 (2017).
284. For an overview, see Lauren Tallent, Through the Lens of Federal Evidence Rule 403: An Examination of Eyewitness Identification Expert Testimony Admissibility in the Federal Circuit Courts, 68 Wash. & Lee L. Rev. 765, 777–78 (2011).
285. See United States v. Villiard, 186 F.3d 893, 895 (8th Cir. 1999).
286. See, e.g., United States v. Rodriguez–Felix, 450 F.3d 1117, 1125 (10th Cir. 2006) (“[A]n expert’s testimony describing how certain factors, falling outside a typical juror’s experience, may affect an eyewitness’s identification is the very type of scientific knowledge to which Daubert’s relevance prong is addressed.”); United States v. Stevens, 935 F.2d 1380, 1400 (3d Cir. 1991) (reversing where misidentification was the key issue at trial and expert’s proposed testimony was outside the realm of typical juror knowledge); United States v. Moore, 786 F.2d 1308, 1313 (5th Cir. 1986) (“In some cases casual eyewitness testimony may make the entire difference between a finding of guilt or innocence. In such a case expert eyewitness identification testimony may be critical.”); United States v. Smith, 736 F.2d 1103, 1106 (6th Cir. 1984) (recognizing that expert testimony on eyewitness identification “might have refuted [jurors’] otherwise common assumptions about the reliability of eyewitness identification”); see also Bell v. Miller, 500 F.3d 149, 155–56 (2d Cir. 2007); United States v. Brownlee, 454 F.3d 131, 144 (3d Cir. 2006); Ferensic v. Birkett, 501 F.3d 469, 477–80 (6th Cir. 2007).
287. United States v. Holloway, 971 F.2d 675, 679 (11th Cir. 1992).
288. United States v. Owens, 682 F.3d 1358, 1359 (11th Cir. 2012) (memorandum) (Barkett, J., dissenting from denial rehearing en banc) (“Having been given the opportunity to change our court’s position that appellate courts are never permitted to review for abuse of discretion the exclusion of expert testimony regarding the reliability of eyewitness identifications, we should avail ourselves of it. That isolated position, established thirty years ago, conflicts with all of the other circuits and all but five of the states that have considered the question.”).
the centrality of the eyewitness identification evidence, the failure to call an expert was a prejudicial “fail[ure] to render reasonable professional assistance.” The Sixth Circuit explained, in a 2007 ruling finding exclusion of the defense eyewitness expert to be unconstitutional, that the significance of an expert’s testimony “cannot be overstated” because, without it, a jury has “no basis beyond defense counsel’s word to suspect the inherent unreliability” of an eyewitness identification.289
The Third Circuit Court of Appeals convened a task force on eyewitness identifications that issued a detailed report in 2019 that surveyed scientific research regarding eyewitness evidence.290 The report was not binding, but rather made recommendations to law enforcement, advocates, and courts, including recommendations regarding conducting lineups, videotaping lineups, documenting witness confidence, blind administration, as well as continuing education materials for lawyers. The task force did not make recommendations concerning amending jury instructions, noting that a separate Third Circuit committee considers model jury instructions for the court. Those instructions already include recommendations for a range of additions to the Telfaire instructions.291
State Court Rulings
Several state courts have modified the federal due process rule, relying on the research that has developed in the intervening decades since Manson. Several state courts have adopted different factors in a “refinement” of the Manson and Biggers factors.292 Additional state courts, however, have rejected the Manson test, substituting a new framework based on scientific research.293 Still additional states have adopted changes to jury instructions and expert evidence regarding eyewitnesses.
289. Ferensic, 501 F.3d at 477–80.
290. Report of the United States Court of Appeals for the Third Circuit Task Force on Eyewitness Identifications, supra note 27.
291. Comm. on Model Crim. Jury Instructions for the Third Circuit, Model Criminal Jury Instructions § 4.15 (2017), supra note 283.
292. See, e.g., State v. Ramirez, 817 P.2d 774, 780–81 (Utah 1991) (altering three “reliability” factors to focus on effects of suggestion), abrogated by State v. Lujan, 459 P.3d 992 (2020); State v. Marquez, 967 A.2d 56, 69–71 (Conn. 2009) (adopting detailed criteria for assessing suggestion); Brodes v. State, 614 S.E.2d 766, 771 & n.8 (Ga. 2005) (rejecting use of eyewitness certainty); State v. Hunt, 69 P.3d 571, 576 (Kan. 2003) (adopting five factor “refinement” of federal due process test); State v. Almaraz, 301 P.3d 242, 253 (Idaho 2013) (“[B]y outlining the system and estimator variables that research has convincingly shown to impact the reliability of eye-witness identification, we hope to provide guidance to lower courts applying the test. . . .”).
293. State v. Henderson, 27 A.3d 872, 877 (N.J. 2011) (finding scientific evidence shows Manson test should be revised); State v. Lawson, 291 P.3d 673, 678 (Or. 2012) (en banc) (revising Manson test in light of scientific evidence); Commonwealth v. Gomes, 22 N.E.3d 897, 909 (Mass. 2015) (finding scholarly research should inform identification instructions); Young v. State, 374 P.3d 395,
The most notable rulings have been by the Supreme Courts of Alaska (Young v. Alaska, 2016), Connecticut (State v. Guilbert, 2012), Hawaii (State v. Cabagbag, 2012), New Jersey (State v. Henderson, 2011), Oregon (State v. Lawson, 2012), Utah (State v. Clopten, 2009), and Wisconsin (State v. Dubose, 2005).294
The first prominent judicial ruling involving a wholesale reconsideration of the federal due process framework was the New Jersey Supreme Court ruling in State v. Henderson. The court-appointed special master “evaluate[d] scientific and other evidence about eyewitness identifications[,] . . . presided over a hearing that probed testimony by seven experts and produced more than 2,000 pages of transcripts along with hundreds of scientific studies,” and then issued a detailed report.295 The decision set out a framework, grounded in scientific research, in which pretrial hearings must examine eyewitness identification evidence to assess its reliability.296 A defendant must show evidence of unreliability, and the state must counter with evidence of reliability. In response, the judge may consider remedies, including jury instructions at trial.
In 2016, the Alaska Supreme Court revised the Manson test and adopted a more detailed framework, asking judges to focus on a range of variables that affect the reliability of eyewitness evidence, as in the Henderson decision, and adjudicating these questions pretrial, including with the benefit of expert testimony to explain how those variables may apply to different factual situations.297 If the defendant meets this burden, “the trial court should suppress the evidence—both the pretrial identification and any subsequent in-court identification by the witness.” If the defendant does not, however, then “the court should admit the evidence and provide the jury with an instruction appropriate to the context of the case.” The approach, thus, resembles the type of functional framework adopted in New Jersey. In 2018, in State v. Harris, the Connecticut Supreme Court adopted a framework much like that in New Jersey (noting the overlap with that approach), holding expert testimony may be valuable, and “it may be appropriate for the trial court to craft jury instructions to assist the jury in its consideration of this issue.”298
427 (Alaska 2016) (urging trial courts to incorporate evolving scientific understanding to supplement test variables).
294. See, e.g., Young, 374 P.3d at 427; State v. Guilbert, 49 A.3d 705 (Conn. 2012); State v. Cabagbag, 277 P.3d 1027, 1035–38 (Haw. 2012) (holding that when identification is central issue, court must, at defendant’s request, give jury instruction about factors affecting reliability); Henderson, 27 A.3d 872; State v. Lawson, 291 P.3d 673 (Or. 2012); State v. Clopten, 223 P.3d 1103 (Utah 2009); State v. Dubose, 699 N.W.2d 582 (Wis. 2005). The Cabagbag ruling prompted model jury instructions in Hawaii. State v. Kaneaiakala, 450 P.3d 761, 774 (Haw. 2019) (holding that when applicable, jury must receive instructions to consider impact of suggestive procedures on identification reliability).
295. Kaneaiakala, 450 P.3d at 877.
296. Henderson, 27 A.3d. at 894–96, 925–26 (noting research shows “that memory is a constructive, dynamic, and selective process”).
297. Young, 374 P.3d at 413.
298. Harris, 191 A.3d at 134.
Adopting a slightly different approach, the Hawaii Supreme Court ruled in 2019 that “courts must, at minimum, consider any relevant factors set out in the Hawaii Standard Instructions governing eyewitness and show-up identifications, as may be amended.” Thus, the court did not set out a series of factors, as the New Jersey court did, but rather set out jury instructions that may be changed over time.299 Tracking the New Jersey focus on system variables, the court also emphasized the importance of suggestion: “[T]rial courts must also consider the effect of the suggestiveness on the reliability of the identification in determining whether it should be admitted into evidence.”
In 2015, the Massachusetts Supreme Judicial Court “review[ed] the scholarly research, analyses by other courts, amici submissions,” and the report by a Massachusetts Supreme Judicial Court Study Group on Eyewitness Identification.300 The court recommended judges provide a set of more concise jury instructions on eyewitness identification evidence. The approach similarly includes a range of research-informed factors, but adopts jury instructions that are more manageable in length.
In contrast to these approaches, which rely on a multifactored standard for admissibility and instructions for educating jurors, the Oregon Supreme Court has endorsed review of reliability of eyewitness evidence relying on a general Rule 403 analysis under the Oregon Rules of Evidence. The Oregon court identified two problems with the prior framework: The “threshold requirement of suggestiveness inhibits courts from considering evidentiary concerns” and the “inquiry fails to account for the influence of suggestion on evidence of reliability.”301
Other state courts have made more selective changes, such as rejecting eyewitness confidence as a factor to be considered when assessing the validity of an eyewitness identification,302 or requiring additional jury instructions.303
299. Kaneaiakala, 450 P.3d at 777.
300. Commonwealth v. Gomes, 22 N.E.3d 897, 900 (Mass. 2015).
301. State v. Lawson, 291 P.3d 673, 748 (Or. 2012).
302. State v. Harris, 191 A.3d 119, 134 (Conn. 2018). 191 A.3d at 134 (expanding protections for suggestiveness of eyewitness identification procedure); State v. Almaraz, 301 P.3d 242, 253 (Idaho 2013) (naming estimator and system variables to be considered in applying Manson test); People v. Adams, 423 N.E.2d 379, 383–84 (N.Y. 1981) (concluding where witness makes identification under influence of suggestive procedure and there is no independent source that suggests defendant is perpetrator, court should exclude evidence); State v. Ramirez, 817 P.2d 774, 781 (Utah 1991) (modifying Manson test to remove level of certainty as factor), abrogated by State v. Lujan, 459 P.3d 992 (2020); State v. Hunt, 69 P.3d 571, 572 (Kan. 2003) (refining Neil v. Biggers, 409 U.S. 188 (1972), approach by adding factors from Ramirez); State v. Discola, 184 A.3d 1177 (Vt. 2018) (rejecting witness certainty as factor in assessing reliability).
303. State v. Kaneaiakala, 450 P.3d 761, 777–78 (Haw. 2019) (holding when identification has been procured through suggestive procedure or when central to case, jury must be instructed to consider potential impact of suggestive procedures on reliability of identification).
In addition, several courts have limited the use of courtroom identifications.304 The Massachusetts Supreme Judicial Court and the Connecticut Supreme Court have ruled that no in-court identification is permitted if an out-of-court identification was suppressed as unduly suggestive.305 The Massachusetts Supreme Judicial Court has also ruled that first-time courtroom identifications should not normally be permitted.306 Other courts adopt a burden-shifting approach toward in-court identifications.307
State Eyewitness Evidence Statutes
Lawmakers in almost half of the states have required law enforcement adoption of best practices for eyewitness evidence. State legislation has required the study of eyewitness procedures, development of model policies, or increasingly, outright adoption of eyewitness identification practices by law enforcement. To date, twenty-five states have enacted legislation requiring law enforcement adoption of written eyewitness identification procedures.308 These statutes were often enacted to ensure uniformity in adoption of best practices. Most of those
304. For a proposal that prior identifications should be admitted, but in-court identifications and statements of confidence should not, see Brandon L. Garrett, Judging Eyewitness Evidence, 104 Judicature 30, 34–35 (2020), https://perma.cc/5ETW-QUV5.
305. See Commonwealth v. Johnson, 45 N.E.3d 83, 92–94 (Mass. 2016) (reasoning that “a subsequent in-court identification cannot be more reliable than the earlier out-of-court identification, given the inherent suggestiveness of in-court identifications and the passage of time”); State v. Dickson, 141 A.3d 810, 817 n.2 & n.3 (Conn. 2016) (“[I]n cases in which identity is an issue, in-court identifications that are not preceded by a successful identification in a nonsuggestive identification procedure implicate due process principles and, therefore, must be prescreened by the trial court.” (footnotes omitted)).
306. See Commonwealth v. Crayton, 21 N.E.3d 157, 169 (Mass. 2014) (“Where an eyewitness has not participated before trial in an identification procedure, we shall treat the in-court identification as an in-court showup, and shall admit it in evidence only where there is ‘good reason’ for its admission.”).
307. See State v. Hickman, 330 P.3d 551, 568 (Or. 2014), modified on reconsideration, 343 P.3d 634 (2015) (“Courts considering the admissibility of first-time in-court identifications generally have placed the burden of seeking a prophylactic remedy on the defendant.”); United States v. Domina, 784 F.2d 1361, 1369 (9th Cir. 1986), cert. denied, 479 U.S. 1038 (1987) (noting that district court’s denial of request for in-court lineup will only be overturned if in-court identification procedures were so suggestive and conducive to irreparable misidentification as to amount to denial of due process).
308. See Cal. Penal Code § 859.7 (West 2022); Colo. Rev. Stat. § 16-1-109 (2022); Conn. Gen. Stat. § 54-1p (2021); Fla. Stat. § 92.70 (2021); Ga. Code Ann. § 17-20-2 (2021); Hawaii Rev. Stat. 801k (2022); 725 Ill. Comp. Stat. 5/107A-2 (2021); Kan. Stat. Ann. § 22-4619 (2021); La. Code Crim. Proc. Ann. Arts. 251–53 (2021); Md. Code Ann., Pub. Safety § 3-506 (West 2021); Minn. Stat. § 626.8433 (2021); Neb. Rev. Stat. § 81-1455 (2021); Nev. Rev. Stat. § 171.1237 (2021); N.H. Rev. Stat. Ann. § 595-C:2 (2022); N.M. Stat. Ann. § 29-3B-3 (2022); N.Y. Crim. Proc. Law § 60.25 (Mckinney 2021); N.C. Gen. Stat. § 15A-284.52 (2021); Ohio Rev. Code Ann. § 2933.83 (West 2021); Okla. Stat. Tit. 22, § 21 (2021); Tex. Code Crim. Proc. Ann. Art. 38.20 (West 2021); Utah Code Ann. § 77-8-4 (West 2021); Vt. Stat. Ann. Tit. 13, § 5581 (2022); Va. Code Ann. § 19.2-390.02 (2021); W. Va. Code §§ 62-1E-1 to -2 (2021); Wis. Stat. § 175.50 (2022).
states regulate the eyewitness identification procedures to be used, including blinded lineups (see above section titled “Communication during lineups: The importance of blinded lineup administration”), clear written instructions, and documenting the confidence of an eyewitness.309 In addition, seventeen states have revised their jury instructions, departing from the Telfaire model or models reciting factors from Manson v. Brathwaite.310
For example, Massachusetts developed model jury instructions regarding eyewitness identifications that include both preliminary language, to be read prior to opening statements, and final instructions.311 The preliminary language advises the jury that “[t]he mind does not work like a video recorder” and cautions that factors during and after the observed event can alter an individual’s memory of that event. The final instructions are far more detailed, describing identification as a three-stage process: perception of an event, storage of information about the event in one’s mind, and later recollection of that stored information.312 The instructions caution that a variety of factors, each of which is listed in the instructions themselves, may affect accuracy of a memory at any stage of the process. In selecting the factors to be included in its revised jury instructions, Massachusetts relied on “a near consensus [standard] in the relevant scientific community.”313 The analysis suggested five such principles (although without referring to factors that affect the validity of visual perceptual experience):
- human memory does not operate like a video recording that a person can replay to recall what happened;
- a witness’s level of confidence in an identification may not indicate its accuracy;
- high levels of stress can reduce the likelihood of making an accurate identification;
- information from other witnesses or outside sources can affect the reliability of an identification and inflate an eyewitness’s confidence in the identification; and
309. See Albright & Garrett, supra note 253, at 511, Appendix A.
310. Those states are Alaska, Connecticut, Florida, Georgia, Hawaii, Kansas, Maine, Maryland, Massachusetts, Missouri, New Jersey, New York, North Carolina, Ohio, Pennsylvania, Utah, and Virginia.
311. Mass. Sup. Jud. Ct., Model Jury Instructions on Eyewitness Identification: Preliminary/Contemporaneous Instruction 2 (2015); Mass. Sup. Jud. Ct., Model Jury Instructions on Eyewitness Identification: Model Eyewitness Identification Instruction 2 (2015) (establishing jury instruction to be given where jury heard evidence affirmatively identifying defendant and such identification is contested).
312. Model Jury Instructions on Eyewitness Identification: Model Eyewitness Identification Instruction 2 (Mass. Sup. Jud. Ct. 2015).
313. Id. at 2–10.
- viewing the same person in multiple identification procedures may increase the risk of misidentification.314
We note that there is not strong evidence that these jury instructions fully accomplish their aim to better educate jurors regarding the potential strengths and weaknesses of eyewitness evidence, including because jurors place such powerful weight on courtroom identifications and courtroom expressions of confidence.315 Indeed, research has not found that the instructions adopted in New Jersey, or the more concise Massachusetts instructions, are effective in helping jurors distinguish between high- and low-confidence initial identifications.316 More recent research suggests that instructions explaining why it is important to place weight on initial confidence, and not courtroom identifications and courtroom confidence, may help jurors make that important distinction.317
Police Practice Recommendations
In recent years, professional policing organizations, such as the IACP, Major Cities Chiefs Association, and the CALEA, have taken active roles in promoting best practices, including through model policies.318 The Third Circuit task force emphasized that results of research on eyewitness procedures must be translated to law enforcement personnel with proper guidelines for maximum effectiveness. The U.S. Department of Justice adopted a policy in 2017, applying to all federal law enforcement agencies, that requires blind administration of lineups, among other procedures.319 Twenty-six states and the federal government have voluntarily adopted such policies regulating eyewitness identifications.320 The model policies generally require blinded lineups (see above section titled “Communication during lineups: The importance of blinded lineup administration”),
314. Ralph D. Gants & Erik N. Doughty, Where Science Conflicts with Common Sense: Eyewitness Identification Reform in Massachusetts, 79 Alb. L. Rev. 1617, 1625–26 (2017) (quoting Commonwealth v. Gomes, 22 N.E.3d 897, 903, 909 (Mass. 2015)).
315. See generally, Garrett et al., supra note 33.
316. Id.
317. Garrett et al., supra note 34.
318. See U.S. Dep’t of Just./Int’l Ass’n of Chiefs of Police, National Summit on Wrongful Convictions: Building A Systemic Approach to Prevent Wrongful Convictions 13–14 (2013), https://perma.cc/ML3T-2ZJE; Comm’n on Accreditation for L. Enf’t, CALEA L. Enf’t Agency Standards § 42.2.11 (2010); Int’l Ass’n of Chiefs of Police Pol’y Ctr., Model Pol’y: Eyewitness Identification 1–4 (2016).
319. Procedures for Conducting Photo Arrays, supra note 7.
320. The states are Arkansas, Colorado, Connecticut, Delaware, Florida, Georgia, Kansas, Kentucky, Louisiana, Maine, Maryland, Michigan, Minnesota, Montana, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, Rhode Island, Utah, Vermont, Virginia, Washington, and Wisconsin.
clear written instructions, and documenting confidence; many recommend videotaping. Policing organizations have incorporated research recommendations into lineup policies and training; federal support has been important to such efforts.321
National organizations have also set out recommendations for law enforcement regarding incorporation of research into eyewitness identification procedures. The American Law Institute, as part of its Principles of Policing project, adopted principles for law enforcement concerning eyewitness identifications.322 In 2020, the AP-LS released a scientific review paper summarizing research in the field and making new recommendations.323
Conclusion
The evolving law of eyewitness evidence shows how science can play a pivotal role in our legal system. While constitutional precedent can change slowly, its practical application can be informed by scientific research. Evidentiary practices, like appointment of experts, pretrial hearings and rulings, and jury instructions, can also be informed by scientific research. We have experienced a paradigm shift: Almost every jurisdiction—through judicial rulings, legislation, model policy, and police policy—has embraced scientific research regarding eyewitness evidence. These approaches aim to more accurately collect evidence from eyewitnesses, and to prevent contamination of that evidence through suggestion and other forms of bias.
321. These efforts began in the late 1990s with the National Institute of Justice’s Technical Working Group for Eyewitness Evidence. See U.S. Dep’t of Just., Nat’l Inst. of Just., Eyewitness Evidence: A Guide For Law Enforcement, supra note 27 (putting forth recommendations to improve procedures for collection and preservation of eyewitness evidence within criminal justice system).
322. Principles of the Law: Policing ch. 10 (A.L.I. 2022).
323. Wells et al., Policy and Procedure Recommendations (2020), supra note 27, at 3.
Glossary of Terms
accuracy. The probability that the identification is of the perpetrator. Estimated by predictive variables in real casework and commonly measured in known source tests as the number of correct identifications relative to the total number of identifications (correct + incorrect).
attentional hijacking. The condition in which a highly salient stimulus directs the focus of attention away from an intended target of attention.
bias. To evaluate something in a prejudicial manner. The meaning here is not pejorative: It refers to prior knowledge or experience that is used to infer the meaning of an incomplete or ambiguous sensory event.
binary classification. A cognitive process in which the individual must decide between two discrete alternatives, as in the lineup participant is a match or a nonmatch to memory of the perpetrator of a crime.
blinded test. (Also known as “single-blinded test.”) A classification task in which the witness is prevented (hence “blinded”) from knowing the status of lineup participants. See double-blinded test.
categorical decision. A decision in favor of one of a set of discrete alternatives, such as classification of a lineup participant as a match or a non-match.
categorical perception. The experience in which different sensory stimuli of the same category are perceived similarly, which entails loss of sensitivity to individual sensory differences within a category.
center of gaze. The region of visual space corresponding to the portion of the retina with the highest density of photoreceptors; the region with the highest visual acuity.
color blindness. A genetically determined condition in which an individual has fewer than three types of color-specific photoreceptors, which limits the range of discriminable colors.
confidence. A cognitive state and expression that reflects perceived certainty of a decision, as in perceived certainty that one’s identification of the perpetrator is correct.
confidence inflation. The increase in perceived certainty in a decision that follows from reinforcement and perceived corroboration by others.
contextual variables. Variables associated with a signal event, as in environmental, perceptual, or cognitive states that are correlated with the experience of witnessing a crime and identifying a perpetrator. These variables may be useful in assessing the accuracy of the identification.
contrast sensitivity. A quantified measure of an observer’s sensitivity to a difference between two sensory stimuli, such as different intensities of light, which bears on the ability to detect and discriminate patterns.
cross-race effect. A well-established perceptual phenomenon in which the ability of an observer to distinguish between faces of members of their own race exceeds the ability to distinguish between faces of members of a different race. The effect is believed to be a consequence of perceptual learning resulting from predominant exposure of individuals to same-race faces during critical stages of development. It has the obvious potential to yield differential error rates in same-race versus cross-race eyewitness identifications.
decision criterion. The threshold value of a decision variable that an observer uses to distinguish two different classifications of a sensory stimulus, as in the specific degree of perceptual similarity between a lineup face and a face stored in memory that distinguishes between a match versus a nonmatch decision.
decision variable. The measured or computed variable that an observer uses to make a classification decision, as in the similarity between the appearance of a lineup face and the eyewitness’s memory of the face of the culprit.
declarative memory. Long-term memory for facts and events that can be retrieved at will and brought into conscious awareness.
degree of visual angle. The angular extent of visual space from the nodal point of the eye, which is commonly used to refer to the spatial extent of an object as optically projected onto the back of the eye. In the context of object recognition, it is used to determine the ability to resolve features in the projected image.
diagnosticity ratio (DR). The ratio of the probability of correct identification of a known culprit to the probability of incorrect identification of a known innocent suspect. Used in laboratory studies of eyewitness performance as a measure of the ability of a witness to distinguish the culprit from an innocent person. Not a true measure of discriminability, since the diagnosticity ratio is also influenced by the decision criterion employed by the eyewitness.
discriminability. Reflects the degree to which an instrument of measurement can distinguish the values of two inputs, assessed independently of the decision criterion employed to make the classification decision. In the eyewitness context, this refers to the extent to which the witness can tell the difference between the culprit and an innocent suspect.
double-blinded test. A classification task in which the witness and the administrator of the task are both “blinded” from knowing the status of lineup participants; this is modeled after an element of the scientific method designed to prevent decision bias based on prior knowledge of experimental conditions.
ecological validity. The extent to which variables, results, and conclusions from an experimental study generalize to the real world.
episodic memory. The form of declarative memory that holds information about the sequence or narrative content of memory. Eyewitness memories are of this type.
estimator variable. Contextual factors associated with the viewing conditions and perceptual/cognitive state of the witness at the time of the crime. These include environmental factors such as intensity of illumination and duration of events, as well as observer variables, such as acuity, emotion, and pain. These variables are not controllable in an eyewitness context. However, because they can predictably impair the fidelity of visual perception and memory encoding, they thus have great potential to inform the trier of fact as to the probability that an identification is accurate.
face blindness. Laboratory studies have shown that face recognition ability varies continuously across the human population. At one extreme of this distribution are “face-blind” people who are extremely poor at recognition memory for faces. The degree of face blindness can be quantified by now-standard face recognition tests. Performance on such tests is an estimator variable that can be used for quantitative prediction of the accuracy of eyewitness reports. Contrasted with face super-recognizers.
face super-recognizer. Laboratory studies have shown that face recognition ability varies continuously across the human population. At the extreme opposite end of the continuum from face-blind people are face super-recognizers, who excel at recognition memory for faces. The degree of performance can be quantified by now-standard face recognition tests, which can serve as quantitative predictors of the accuracy of eyewitness reports.
fair lineup. A lineup composed of fillers who all bear a similar perceptual relationship to the witness’s reported description of the culprit. Fair lineups are less likely to elicit biased identifications and wrongful convictions of people who stand out from the crowd based on similarity to the witness’s description.
gaze direction. The point in visual space that is the momentary target of the center of gaze, which possesses the highest visual acuity.
illusory conjunctions. A perceptual phenomenon in which briefly viewed attributes of a scene become disjoined from the objects they originate from and conjoined with others, as in a red letter D and a blue number 2 that are perceived as a blue D and a red 2.
inattentional blindness. Failure to perceive an otherwise prominent visual stimulus that results from lack of attentional focus directed to the stimulus.
lineup type. The specific type of visual presentation and behavioral task used to test an eyewitness’s ability to identify the culprit.
long-term memory. Memories that have been consolidated into long-term storage, where they are less subject to forgetting but still modifiable by the acquisition of new information.
luminance contrast. A measure of the contrasting amounts of light reflected (i.e., luminance) from two spatially adjacent features of a visual scene, which is a primary determinant of an observer’s ability to perceptually distinguish the features.
memory consolidation. The transfer of perceptual or cognitive information from labile short-term storage into more stable long-term storage, which is a time-dependent process mediated by changes in neuronal connectivity.
memory encoding. The transfer of perceptual or cognitive experiences into a form storable by the brain, which underlies the ability to learn from those experiences.
memory plasticity. Memory is modifiable over time, to a high degree during acquisition and to a lesser degree once consolidated. Memory systems are more plastic during critical stages of human development, but retain some plasticity throughout life.
memory retention interval. The amount of time that passes between initial acquisition of a new memory and a subsequent test of the fidelity of that memory. Within that retention window, memories predictably fade because of forgetting and can be modified by intrusion of new information.
memory retrieval. The process of bringing a stored piece of information back to a state where it can be used to guide decisions and behavior. Retrieval is commonly elicited by an external sensory stimulus or by an internal association.
memory storage. The state in which memories are maintained for the long term and accessible for retrieval at a later time. Memories are less subject to forgetting during this period but can still be modified by acquisition of new information.
misdirection. Direction of the focus of visual attention away from the location of a behaviorally relevant object or feature. A prominent tool used in magical tricks that demonstrates the easy fallibility of visual perceptual experiences and the inappropriate assertion of confidence in such experiences. This phenomenon underlies the “weapon focus” effect that impairs accuracy of eyewitness identification.
myopia. A visual pathology in which the optics of the eye are misaligned, such that the projected pattern of light on the back surface of the eye is not well focused. Although correctable with external lenses or corneal surgery, this pathology is common enough in the human population, to varying degrees, that it could easily limit the accuracy of eyewitness identification.
noise. Random or irrelevant elements that interfere with detection of coherent and informative signals. In human sensory systems, this interference is frequent and can take the form of distortion of signals at the sensory periphery, such as optical impairment, or by diffraction of light by fluid in the eye. Other sources
of noise include occluding elements of a visual scene or attentional distractions. The result is reduced fidelity of sensation and perceptual uncertainty.
object surface reflectance. The amplitude and chromatic spectrum of white light reflected from the surface of an object. This reflectance is a physical property of the surface itself, which interacts with the amplitude and chromatic spectrum of light illuminating the scene.
operating characteristic. A set of parameters that define the performance of an instrument based on empirical assessment and knowledge of its underlying mechanism. Generally, these describe performance in terms of the validity and reliability of the output for a given type of input. Quantitative performance descriptions of this sort offer a useful systems approach to understanding the abilities and weaknesses of eyewitness reports.
perception. The process by which evanescent sensations are linked to environmental cause and made enduring and coherent through the assignment of meaning, utility, and value.
perceptual improvisation. Unconscious search through memory for information (probabilities of the accuracy of previously acquired knowledge, beliefs, and hypotheses) that enables rapid identification, recognition, and spatial understanding of novel and otherwise unrecognizable stimuli.
perceptual learning. Long-term experience-dependent changes in the way that people perceive sensory stimuli, which can enhance or distort perceptual experience. Apropos of eyewitness identification, the cross-race effect, which comes about through extended selective exposure to faces of people of one’s own race, is a consequence of perceptual learning.
priors. Shorthand for “prior probability,” which refers to the probability of accuracy for knowledge, beliefs, or hypotheses about the world that an observer holds prior to a signal event, such as witnessing a crime. Used synonymously with prejudices or biases, priors enable an observer to unconsciously fill in blanks in the presence of perceptual uncertainty.
receiver operating characteristic (ROC). A standard modeling approach to the evaluation of performance on a classification task, drawn from signal detection theory. In the eyewitness context, the ROC is generally presented as a curve that plots the probability of correct detection of a target (the culprit) against the probability of incorrect detection of a non-target (an innocent suspect). The area under the ROC curve is a single-value measure of the ability to discriminate the culprit from an innocent suspect, which has the advantage of being independent of the decision criterion used by the eyewitness. Use of this approach has transformed analysis of data from laboratory studies of factors that affect eyewitness performance, in part because it yields a true measure of discriminability.
recognition memory. The ability to determine whether a present sensory stimulus (a “cue”) is the same as a stimulus previously seen and memorized, and subsequently retrieved from memory. In the eyewitness context, a positive outcome from this similarity assessment is the event that leads to an assertion of identification.
reliability. A property of an instrument of measurement (e.g., an eyewitness) that reflects the degree to which it yields the same result every time it is used to measure the same thing. Often used inappropriately to refer to the validity of an instrument of measurement.
retina. The sheet of neuronal tissue that lines the back of the eye. Light reflected from surfaces in the visual environment passes through the lens of the eye and is projected onto the retina, which includes the photoreceptors that are responsible for phototransduction (conversion of luminous energy into neuronal energy). Numerous optical and neuronal noise sources can create visual uncertainty at this stage of processing, undermining the accuracy of perceptual experience.
semantic memory. The form of declarative memory that holds information about meanings of experiences and acquired knowledge of the world.
sensation. The consequence of spatially and temporally patterned stimulation of biological sensory receptors. Antecedent to and contrasted with perception, through which sensations are linked to environmental cause.
sequential lineup. A lineup identification procedure in which each face in the lineup is presented to the witness in isolation, and the complete set is presented in a temporal sequence. Believed to promote absolute lineup comparisons.
short-term memory (working memory). The memory system that has limited capacity for volume of content and duration of storage, but where content is actively maintained and readily available for use to guide decisions and actions (i.e., it is “working” memory). Short-term memory is highly labile storage, subject to distortion or loss in the presence of distractions. Extended maintenance of content requires consolidation into long-term memory storage.
signal detection. From signal detection theory, which is a theoretical framework for understanding the processes and consequences of decision-making in the presence of uncertainty. Used for decades as an efficient means to analyze behavioral responses to sensory stimulation.
simultaneous lineup. A lineup identification procedure in which all faces in the lineup are presented to the witness at the same time.
source memory failure. Loss of the ability to remember how one acquired a specific memory, while the memory itself is intact. A common consequence of this failure is that the source of the memory is wrongly attributed based on priors and may incorrectly modify other memories. This is a memory
system analogue of the introduction of biases to resolve uncertainty in visual perception.
system variables. Procedures and states that may influence the accuracy of an eyewitness report and are potentially under the control of the criminal justice system, since they bear on activities that occur after the witnessed events. Includes lineup type and blinded versus nonblinded lineup administration.
target absent lineup. A type of lineup procedure that does not include the culprit (“target”) as one of the lineup participants. Commonly refers to an empirical practice in laboratory studies of eyewitness performance, which is used to assess the probability that witnesses mistakenly identify an innocent suspect. Results (probabilities of incorrect identifications) are compared to those obtained using a target present lineup. The ratio of correct to incorrect identification probabilities is the diagnosticity ratio.
target present lineup. A type of lineup procedure that includes the culprit (“target”) as one of the lineup participants. Commonly refers to an empirical practice in laboratory studies of eyewitness performance, which is used to assess the probability that witnesses correctly identify the culprit. Results (probabilities of correct identifications) are compared to those obtained using a target absent lineup. The ratio of correct to incorrect identification probabilities is the diagnosticity ratio.
task irrelevant information. Contextual information associated with a visual or cognitive task that is not required to solve the task but is at risk of biasing the outcome. Apropos of eyewitness identification, task irrelevant information can refer to reports from journalists, family, friends, and legal counsel that may artifactually influence a witness’s degree of certainty in their identification. Contrasted with task relevant information, which is needed to solve the task.
uncertainty. In the eyewitness context, the state of being uncertain about a measured piece of sensory information, as a result of noise and ambiguity. The common consequence is that uncertainty is resolved by completion of perceptual experience based on prior knowledge or beliefs.
unfair (biased) lineup. A lineup in which the chosen fillers possess different degrees of similarity of appearance to the witness’s description of the culprit. Unfair lineups can bias the choices made by a witness because some faces stand out from the crowd, which may lead to wrongful investigation, prosecution, and conviction.
validity. The property of an instrument of measurement that reflects the degree to which it yields the result that it was intended to measure. In the eyewitness context, this generally refers to the measured probability that eyewitnesses will correctly identify the culprit, which is typically obtained from laboratory studies in which ground truth is known.
visual acuity. A measure of the clarity and sharpness of vision, empirically assessed by the threshold ability to resolve points in an image at varying degrees of angular separation. A measure of an eyewitness’s visual acuity is essential to predicting the accuracy of an identification.
visual crowding. Reduction of the probability of correctly detecting and perceiving a target stimulus when it is viewed in a dense arrangement of other stimuli. In the eyewitness context, this is manifested as increasing limits on the ability to correctly perceive the actors and events of a crime as a function of the density of actors and objects in the scene.
visual performance. A generic term for the quantifiable ability of an observer to accurately detect and discriminate visual stimuli.
visual search. A well-studied perceptual phenomenon in which the amount of time it takes to detect a specific object amid a set of related objects is dependent on the presence or absence of distinctive features. In the presence of distinctive features, such as a unique color, immediate perceptual “pop-out” occurs. If, on the other hand, the object is defined by a unique conjunction of features (for example, a single red letter T in a field of green Ts and both red and green Ls), time to target object detection is predictably delayed because the visual system must “search” serially through the scene until it is encountered. This attentional delay has the potential to markedly increase uncertainty in time-limited eyewitness events.
This page intentionally left blank.
