Eyewitness Confidence and Accuracy: When Confidence Is Informative
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Eyewitness confidence can be informative about identification accuracy when it is measured immediately after a witness makes an initial identification from a fair, properly administered lineup, before feedback or other information changes the witness's judgment. Under these conditions, research with adult eyewitnesses generally finds that higher initial confidence is associated with a greater likelihood that a suspect identification is correct. The strength of that relationship depends on the testing procedure, the population studied, and how the results are analyzed. A confident identification remains evidence to evaluate, never a guarantee of accuracy or a legal finding of guilt. The major synthesis by Wixted and Wells (2017) and the National Academies' eyewitness identification report explain why the timing and quality of the first test are central.
This distinction resolves an apparent contradiction. Wrongful-conviction histories document witnesses who were highly certain at trial and mistaken about an innocent defendant. Yet carefully controlled research also finds a substantial confidence–accuracy relationship at an uncontaminated first identification. These observations concern different points in the life of an eyewitness memory. The central question is therefore not simply whether the witness sounds sure, but when that confidence was expressed, what decision it accompanied, and whether the identification procedure protected the original memory from outside influence.
What eyewitness confidence and accuracy actually mean
Eyewitness confidence is a person's subjective judgment about how likely their identification or rejection is to be correct. It is a psychological self-assessment, often expressed in words such as "I think that is the person" or as a numerical rating. Identification accuracy is the correspondence between that decision and what happened: whether the identified suspect was in fact the perpetrator. Confidence and accuracy are different variables. They can be related without being identical.
The relevant decision is also important. A witness may identify the police suspect, choose an unrelated lineup filler, or reject the lineup because the perpetrator appears absent. Each outcome carries a different interpretation. The strongest established confidence–accuracy findings concern adult witnesses who identify a suspect during an initial, properly controlled lineup. Evidence about lineup rejections, children, courtroom identifications, and repeated recognition tests has to be evaluated separately. The distinction between suspect choices and rejections was emphasized in Brewer and Wells (2006) and updated in Seale-Carlisle (2024).
A police lineup is a test of recognition memory. It is not itself a test of whether a defendant satisfies the legal elements of a criminal charge. A witness can remember a face accurately while remaining mistaken about other events, or remember relevant events while being unable to identify the person. Investigators, psychologists, lawyers, and jurors should therefore distinguish a recorded memory judgment from the legal conclusions later drawn from it.
The key finding: initial confidence is more useful than later confidence
In a landmark review, Wixted and Wells (2017) integrated experimental and field research and concluded that high-confidence suspect identifications made on an initial, uncontaminated, properly conducted test are often highly accurate. They also found that low-confidence initial suspect identifications are substantially more error-prone. Their argument explicitly depends on fair lineups, protection against administrator influence, and an immediate record of confidence.
The finding does not imply that every witness who says "100% certain" is correct. Even among confident witnesses, mistaken selections occur. A high-confidence choice from an unfair lineup can be seriously misleading. The studies evaluate patterns of results across many people; they do not provide a mechanically valid probability for a particular accused person. Any case-level conclusion requires a review of the procedure, relevant conditions, alternative evidence, and the jurisdiction's rules.
The 2025 review The Cognitive Science of Eyewitness Memory likewise stresses the significance of the first uncontaminated test. Its authors argue that an early nonidentification or low-confidence response sometimes provides meaningful evidence favoring an innocent suspect, even when later testimony becomes more confident. This is a research interpretation of memory evidence, distinct from a legal finding of innocence.
Why an eyewitness may feel confident
Recognition can produce a strong subjective signal when a face matches the memory trace. The witness may experience relatively rapid recognition and a feeling that the choice is compelling. In general, stronger memory evidence can support both greater confidence and higher accuracy, which explains why the two measures sometimes covary. Recognition, however, is not a direct reading of a preserved photograph. Similar-looking faces, selective attention, familiarity from other contexts, guessing, and exposure to information after the event can affect the decision.
Confidence also reflects the witness's judgment process. Some people are cautious about numerical ratings; others routinely use extreme expressions. A person's style of communicating certainty should not be mistaken for a universal psychometric scale. A percentage supplied by a witness is a self-report, not a certified probability that the suspect is guilty.
What researchers mean by pristine testing conditions
The term "pristine" refers to a set of protections for the first memory test. A fair lineup gives the suspect no inappropriate prominence; the administrator cannot knowingly or inadvertently cue the expected answer; the witness is informed that the perpetrator might be absent; the identification decision and confidence are documented immediately; and feedback or repeated testing has not contaminated the result. The 2020 American Psychology–Law Society recommendations provide a detailed professional framework.
No real investigation can recreate an ideal laboratory. Police departments vary in training, recording practice, witness instructions, and the documentation available for later review. When a procedure departs from recommended safeguards, the appropriate response is to examine the likely consequence for the evidence rather than to transfer accuracy estimates from pristine experiments without adjustment.
Initial identification, subsequent identification, and courtroom confidence
An initial identification records the earliest relevant decision, commonly from a photograph array or live lineup. Later exposure to photographs, a suspect's image in news coverage, identification-related discussion, interviews, and preparation for trial can affect a witness's familiarity and certainty. Repeated selection of the same person can make that person feel increasingly familiar even if the first selection was uncertain.
At trial, a witness may confidently point to the defendant seated in an unmistakably suggestive setting. Such a courtroom statement can be compelling to observers, but its confidence is generally less diagnostic of the accuracy of the original memory than a contemporaneous, well-documented initial confidence statement. This is a central reason the National Research Council (2014) recommended documenting the witness's confidence verbatim at the first identification.
A trial transcript that contains only a firm declaration of certainty leaves an essential evidence question unresolved: what did the witness say before receiving confirmation, seeing the defendant repeatedly, or hearing details of the investigation? Investigators and courts gain information by preserving the complete sequence, including hesitation, nonselection, changing decisions, and prior expressions of uncertainty.
How feedback can inflate confidence
Postidentification feedback is one of the clearest experimentally demonstrated mechanisms that can change retrospective eyewitness reports. If an administrator indicates that the witness selected the expected person, the witness may later remember having been more certain from the beginning. Feedback can also affect recalled viewing quality, attention, and decision speed. Wells and Bradfield (1998) demonstrated these effects, and Bradfield, Wells, and Olson (2002) showed how confirming feedback damages the relationship between retrospective certainty and actual identification accuracy.
The influence does not require a dramatic statement. Reactions, gestures, the order of questions, or comments about an investigation can convey expectations. Blind or blinded administration reduces the opportunity for this influence. Recording the witness's own words before any feedback preserves information that later questioning cannot reliably reconstruct.
The risk extends to confidence expressed after other witnesses have discussed the incident. Agreement among witnesses may reflect common observation, shared postevent information, or both. A later increase in confidence establishes that the witness now feels more certain; it does not establish that the original perception became more accurate.
Confidence–accuracy calibration: what the statistics can and cannot say
Confidence–accuracy calibration examines how observed accuracy changes across groups of identification decisions made with different reported confidence levels. A confidence–accuracy characteristic, or CAC, curve commonly displays the proportion of correct suspect identifications at several confidence bands. In well-controlled studies with adult witnesses, those curves often rise as confidence increases. The approach directly examines the evidence value of a suspect identification made at a specified confidence level.
Researchers also use receiver operating characteristic, or ROC, analysis. ROC methods examine the tradeoff between correct suspect identifications and mistaken suspect identifications across decision thresholds and are especially useful when comparing identification procedures. CAC and ROC address related but different research questions. A procedure that improves discrimination and a procedure that changes willingness to choose are not necessarily the same. This distinction is discussed in the peer-reviewed methodological article Receiver Operating Characteristic Analysis and Confidence–Accuracy Characteristic Analysis.
The statistical unit matters. Some early calibration studies counted all selected lineup faces as identification decisions, whereas forensic decision-makers are often specifically interested in selection of the designated police suspect. Accuracy of all lineup choices and accuracy of suspect identifications are therefore different quantities. The 2017 synthesis showed how analyses that mix these outcomes may obscure the relationship of direct legal interest.
An illustrative probability calculation
Suppose a hypothetical identification procedure correctly identifies a guilty suspect 80% of the time and mistakenly identifies an innocent suspect 5% of the time. If, in the investigated population, only 20% of the police suspects actually committed the offense, then among 100 comparable cases we would expect 16 correct suspect identifications and 4 mistaken suspect identifications. The proportion of correct suspect identifications among those 20 identifications would be 80%. If 80% of suspects were guilty instead, the same conditional identification rates would produce 64 correct and 1 mistaken suspect identification, or about 98.5% accuracy among identifications. These numbers are invented solely to illustrate how the prevalence of guilty suspects changes the interpretation of the same test characteristics.
Actual police suspect-guilt base rates are rarely known with certainty and vary by investigation and selection process. Experimental studies often deliberately vary whether the perpetrator is present; that research design does not reveal the corresponding proportion in a jurisdiction's real lineup caseload. A published accuracy percentage must therefore be read together with its study design and assumptions, as emphasized by Wixted and Wells (2017). An individual's subjective confidence score does not bypass this issue.
False positives, false negatives, and uncertainty
A false positive occurs when an innocent suspect is identified as the perpetrator. A false negative, in a simplified identification context, occurs when the actual perpetrator is available in the lineup but the witness does not identify that person. A filler identification is another informative outcome and should not be silently counted as a correct suspect selection. The apparent accuracy of a confidence band can change when investigators change the lineup, the instructions, the population, or the rules for designating an innocent suspect in experiments.
Studies report group estimates with sampling uncertainty. Small numbers of rare, high-confidence errors can yield unstable percentages, particularly when comparing demographic subgroups or unusual circumstances. Courts should avoid treating a highly precise-looking number as a case-specific probability unless its underlying data, denominators, uncertainty, and relevance have been carefully examined.
Why lineup fairness and suspect selection matter
A lineup is unfair when the suspect stands out in a way that encourages selection for reasons beyond the witness's original memory. If the witness described a person with particular features but the lineup contains only one person fitting that description, selecting that person provides less diagnostic evidence. Distinctive clothing, image quality, cropping, age, or other non-memory cues can create the same problem. Even a highly confident selection can be compromised by an unfair array.
Proper filler selection and nondirective instructions matter because the witness should be comparing memory with a fair set of alternatives, including the possibility that the perpetrator is not there. An administrator who knows which photograph is the suspect may unintentionally influence a decision. Double-blind or effectively blinded procedures aim to protect the witness and the investigation. These recommendations are supported by the National Academies consensus report, the American Psychology–Law Society policy paper, and the U.S. Department of Justice's federal photo-array guidance.
Suspect selection before the lineup is also a separate issue. A suspect nominated because independent evidence points toward the person presents a different prior evidentiary context from a suspect chosen largely because of an ambiguous description. A witness's identification must not be used to conceal how the suspect entered the procedure. The stronger the opportunity for suggestion or suspect bias, the more cautious the inference drawn from confidence must be.
Estimator variables: viewing conditions and witness characteristics
Estimator variables are factors present during the original event or subsequent memory interval that investigators cannot retroactively control: exposure time, distance, lighting, viewing angle, stress, attention, disguises, weapon presence, the delay before the first test, and cross-race identification conditions. These variables can influence memory strength and identification performance. They do not all act in the same way, and no single factor generates a valid automatic probability for a particular case.
Research suggests that high initial confidence can remain informative even when certain witnessing conditions are less favorable, but the inference needs population-specific evidence and careful evaluation of study conditions. Scientists continue to investigate when a high-confidence decision retains its predictive usefulness and when suboptimal conditions or biased procedures meaningfully change calibration. The 2017 synthesis and 2025 cognitive-science review provide the central framework, while the 2024 study by Lebensfeld and Smalarz illustrates continuing questions about how observers interpret confidence when they learn that viewing conditions were poor.
Cross-race recognition research documents group-level differences under some experimental conditions, and recent studies continue to examine whether confidence calibration changes across witness and target groups. Individual race, ethnicity, or demographic membership is not a substitute for evaluating the actual observation and identification procedure. Researchers must examine sampling, face familiarity, lineup construction, and fairness rather than use demographic categories as shortcuts to an individual finding.
Likewise, a witness who experienced fear, high arousal, or distraction during an incident may have experienced genuine and vivid emotion without preserving every face detail accurately. A vivid memory and an accurate person identification are related questions that require different evidence.
Children, vulnerable witnesses, and other populations
Much of the strongest confidence–accuracy evidence concerns adults in controlled identification tasks. Children's recognition performance, confidence expression, suggestibility, and interpretation of instructions vary with age, developmental conditions, and interviewing practice. Findings established in adult lineups should not automatically be applied to children or witnesses with cognitive, communication, or memory difficulties.
A witness may also face language barriers, pressure from authority figures, accessibility problems, or a need for adapted questioning. The appropriate scientific question is whether a method has been validated for the relevant population and situation. Practitioners should document accommodations, avoid leading phrasing, and consider how questions themselves might change a person's reported confidence.
The same principle applies across jurisdictions and cultural contexts. Laboratory lineups using university participants and standardized images do not automatically generalize to every legal system, cultural setting, investigative practice, or witness population. Where direct evidence is scarce, uncertainty should remain visible.
Identifications, lineup rejections, and uncertainty
An identification of a suspect and a rejection of an entire lineup do not communicate identical kinds of information. Correctly rejecting a lineup when the perpetrator is absent is a potentially valuable outcome, but a witness can also reject a lineup in which the perpetrator is present. Research has found that ordinary confidence ratings are often less diagnostic for lineup rejections than for suspect selections.
A 2024 experimental investigation by Seale-Carlisle examined ways to improve the diagnostic value of lineup rejections and found that changes to the procedure and the format of confidence reporting could reveal additional information. Those findings are important research developments, but experimental innovations should be distinguished from widely adopted, validated police protocols.
A witness's statement that they cannot identify anyone should be preserved rather than recast as an obstacle to getting a result. Pressing for a positive identification risks discarding an informative uncertainty signal and introducing contamination. The original selection, rejection, or inability to decide remains part of the evidence.
Confidence reported before the lineup is a different construct
A person asked before viewing a lineup, "How good do you think your memory is?" makes a prediction about future performance. A person who has just selected a face and is asked, "How sure are you that this is the person?" evaluates a decision already made. These are predictive and postdictive confidence judgments, respectively, and they should not be assumed to have the same relation to accuracy.
In two experiments involving face recognition, Nguyen, Abed, and Pezdek (2018) found that confidence expressed after a recognition decision was a substantially stronger indicator of performance than predictions made before the test. This distinction helps explain why generic claims such as "I have a great memory for faces" are insufficient substitutes for an immediate confidence statement tied to a specific identification decision.
What field studies add, and what they cannot establish
Experimental mock-crime studies have a central advantage: researchers know whether the perpetrator is in each lineup and whether a particular selection is correct. They can measure error rates under controlled conditions and compare procedural variants. Their limitation is that even carefully designed scenarios differ from the emotional, social, and investigative complexity of real crimes.
Police field studies contribute ecological relevance by documenting how real witnesses make decisions in actual cases. However, investigators typically do not independently know whether a suspect is truly guilty at the time of the lineup. Field estimates of identification accuracy may therefore require modeling, corroborating evidence, or assumptions about guilt prevalence. The 2017 research synthesis and the 2025 review by Mickes, Wilson, and Wixted explain how laboratory and field findings can inform one another while retaining these differences.
A 2026 peer-reviewed study of live showups reports an association between confidence and estimated accuracy, but its field component uses an evidentiary-strength measure as a proxy for ground truth. That is useful preliminary information, not the same as independently verifying each identification's correctness. Conclusions from one-person showups also should not be automatically transferred to multi-person lineups. Researchers need direct replication and clear measurement of error before generalizing.
How police should document confidence
The first identification procedure should be documented as a complete event. The National Academies specifically recommends recording the witness's confidence in the witness's own words at the first identification and video recording identification procedures. The American Psychology–Law Society recommendations likewise emphasize appropriate administration and preservation of the response.
The strongest record contains the witness's initial description and exposure history; the lineup's exact composition and administration method; the instructions; every identification, filler choice, or rejection; the immediate confidence statement before feedback; and the relevant audio or video. Recording pauses and corrections helps later reviewers distinguish the actual sequence from a polished narrative reconstructed after the fact.
An interviewer may request clarification when a confidence phrase is ambiguous, but should not steer the answer toward greater certainty. A witness's own expression such as "I think it's number four, but I'm not sure" should remain available even if a later summary uses a numerical category. Standardized scales may support comparison in research, but the original language carries contextual information that an investigator-created label may lose.
How lawyers, judges, and jurors should use confidence evidence
For legal fact-finders, confidence is one component of the identification evidence. The most useful questions concern when it was first recorded, which decision it accompanied, how the lineup was administered, and what happened before the witness later testified. Accurate interpretation also requires attention to other independent evidence and to the possibility of investigative contamination.
When a witness's initial selection was hesitant but the same witness later appears certain, a fact-finder should receive the documented history of that change. When initial confidence was high and the test was carefully protected, research supports treating that confidence as potentially meaningful evidence. In both circumstances, confidence is still an evidentiary observation and not a legal verdict or a validated substitute for corroboration.
Legal standards for admissibility, evidentiary weight, exclusion, jury instructions, and expert testimony vary across U.S. jurisdictions and internationally. Psychological studies describe memory and decision performance; statutes and judicial decisions define legal duties and consequences. The National Academies' legal-framework discussion explains how scientific and constitutional approaches to identification evidence can diverge.
In three experiments published in 2024, Lebensfeld and Smalarz found that information about poor witnessing conditions changed evaluators' judgments of high- and moderate-confidence identifications differently; expert explanations reduced but did not eliminate this pattern. This study examined observers' evaluations, not the ground-truth accuracy of the real defendants in actual cases. It demonstrates that communicating scientific information to juries is itself a subject of empirical research.
A practical framework for evaluating a particular identification
A scientifically responsible review begins with the chronological evidence record. First determine what the witness observed, how a description was collected, and whether there were intervening exposures to suspect photographs or case information. Next examine whether the lineup was fair, blinded, clearly instructed, and preserved. Then identify the exact first decision and the verbatim confidence statement made at that moment. Compare later descriptions and testimony with the initial record, documenting changes without assuming that every change has the same explanation.
The reviewer should then ask which relevant research population and procedure most closely resembles the case, whether uncertainty and alternative explanations are available, and whether the findings generalize to this witness and context. A forensic expert can explain scientific evidence and methodological limits. The trier of fact applies the governing law to the complete record.
This approach reflects the broader scope of forensic psychology and legal psychology: psychological findings help legal decision-making when the evidentiary question, measurement, and inference are made explicit.
Myths and misconceptions
Myth: A confident eyewitness must be right
High initial confidence from a fair, uncontaminated adult lineup can be meaningful, but errors still occur and experimental accuracy rates vary. Confidence expressed after feedback, repeat identification, or courtroom exposure has a different evidentiary status. No amount of personal certainty directly proves criminal responsibility.
Myth: Eyewitness confidence tells us nothing about accuracy
That blanket conclusion misses a substantial body of modern research on initial suspect identifications under appropriate conditions. It remains reasonable to question poorly documented or contaminated confidence, while also recognizing the information value of an immediate, high-quality record.
Myth: If a witness becomes more confident over time, their memory has improved
Later confidence may increase through feedback, familiarity, repeated rehearsal, and social reinforcement. Such changes require examination. They do not, by themselves, demonstrate improved accuracy or deliberate dishonesty.
Myth: A witness who is unsure should be encouraged to pick the closest face
Uncertainty and rejection are legitimate test outcomes. Pressure to choose can increase mistaken selections and compromise the interpretation of confidence. A fair procedure permits the witness to state that the perpetrator is absent or that they cannot tell.
Myth: A 90% confidence statement means a 90% chance of guilt
A witness's number is a subjective judgment, not a calibrated probability for that defendant. Statistical calibration is estimated across studies under defined conditions and depends on suspect selection, lineup construction, base rates, and other evidence. Legal guilt is determined under the applicable law.
Frequently asked questions
Is eyewitness confidence correlated with accuracy?
Often yes, especially for adult witnesses making their first suspect identification from a fair lineup under conditions that minimize suggestion and contamination. The strength of the relationship depends on the decision type, confidence timing, testing procedure, and research population. Confidence at trial should not be substituted for the first recorded confidence statement.
Can a witness be 100% confident and still be wrong?
Yes. A subjective statement of complete certainty is compatible with an identification error. The critical question is how the witness reached that degree of certainty and whether the initial memory test was adequately protected.
Why is initial confidence better than courtroom confidence?
Initial confidence is documented closer to the first recognition decision and before many common sources of feedback or repeated exposure. Courtroom confidence may incorporate later familiarity and beliefs about the investigation. A recorded initial statement allows direct comparison between the two.
What does a low-confidence initial identification mean?
It generally provides less diagnostic support for the suspect's identity than a high-confidence selection under comparable conditions. It should be carefully documented and assessed with all other evidence. It does not prove that a suspect is innocent or that a witness is dishonest.
Should police ask for a confidence percentage?
A contemporaneous confidence rating can be useful, but the witness's own words should be preserved verbatim. If a scale is used, its wording, timing, instructions, and context should be recorded so the reported number is not presented as a legal probability.
Does a double-blind lineup make eyewitness evidence accurate?
Blind administration helps prevent administrator influence. It is one safeguard among several, including lineup fairness, unbiased instructions, prompt confidence documentation, and recording. It does not remove ordinary recognition errors or guarantee a correct choice.
Does confidence work the same for lineup rejections?
No. The confidence–accuracy relationship for rejecting a lineup is generally less straightforward than for a positive suspect identification. Recent research explores better procedures and confidence measures for interpreting rejections.
Can a highly confident identification alone establish guilt?
A psychological finding about identification accuracy does not determine the legal sufficiency of evidence in any jurisdiction. Legal determinations require the governing evidentiary and procedural standards and the whole record.
Do stress, poor lighting, and cross-race conditions always invalidate confidence?
These conditions can affect memory, accuracy, or the generalizability of an estimate. Their effect is not uniform across all witnesses and situations. Review the specific observation, testing procedure, population evidence, and uncertainty.
What if the police did not record the first confidence statement?
The loss limits a later reviewer's ability to separate contemporaneous memory confidence from later influences. Subsequent testimony cannot fully reconstruct the missing first response. Investigators and courts should document what is known and where evidence is unavailable.
Conclusion
Eyewitness confidence and accuracy have a conditional, scientifically meaningful relationship. The best-supported use of confidence concerns an adult witness's initial, uncontaminated suspect identification from a fair and properly administered lineup, with certainty recorded immediately in the witness's own words. Later confidence, contaminated testing, unfair procedures, different decision types, and unvalidated population transfers can substantially weaken the inference. The practical implication is to preserve the first test of memory, examine the full sequence of events, and interpret confidence as one carefully bounded piece of evidence within the legal process.
Related Articles
References
Bradfield, A. L., Wells, G. L., & Olson, E. A. (2002). The damaging effect of confirming feedback on the relation between eyewitness certainty and identification accuracy. Journal of Applied Psychology, 87(1), 112–120. https://doi.org/10.1037/0021-9010.87.1.112.
Brewer, N., & Wells, G. L. (2006). The confidence–accuracy relationship in eyewitness identification: Effects of lineup instructions, foil similarity, and target-absent base rates. Journal of Experimental Psychology: Applied, 12(1), 11–30. https://doi.org/10.1037/1076-898X.12.1.11.
Lebensfeld, T. C., & Smalarz, L. (2024). Witnessing-condition information differentially affects evaluations of high- and moderate-confidence eyewitness identifications. Cognition, 250, 105841. https://doi.org/10.1016/j.cognition.2024.105841.
Mickes, L., Amendola, K. L., Chen, X., Valdovinos Olson, M., Carlson, C. A., & Gronlund, S. (2026). Predicting accuracy in eyewitness showups: Confidence and response time in the laboratory, confidence in the field. Cognitive Research: Principles and Implications. https://doi.org/10.1186/s41235-026-00734-w.
Mickes, L., Wilson, B. M., & Wixted, J. T. (2025). The cognitive science of eyewitness memory. Trends in Cognitive Sciences, 29(7), 655–668. https://doi.org/10.1016/j.tics.2025.01.008.
National Research Council. (2014). Identifying the culprit: Assessing eyewitness identification. National Academies Press. https://doi.org/10.17226/18891.
Nguyen, T. B., Abed, E., & Pezdek, K. (2018). Postdictive confidence (but not predictive confidence) predicts eyewitness memory accuracy. Cognitive Research: Principles and Implications, 3, 32. https://doi.org/10.1186/s41235-018-0125-4.
Seale-Carlisle, T. M. (2024). Improving the diagnostic value of lineup rejections. Cognition, 252, 105917. https://doi.org/10.1016/j.cognition.2024.105917.
U.S. Department of Justice. (2017). Justice Department announces department-wide procedures for eyewitness identification. Official DOJ record.
Wells, G. L., & Bradfield, A. L. (1998). "Good, you identified the suspect": Feedback to eyewitnesses distorts their reports of the witnessing experience. Journal of Applied Psychology, 83(3), 360–376. https://doi.org/10.1037/0021-9010.83.3.360.
Wells, G. L., Kovera, M. B., Douglass, A. B., Brewer, N., Meissner, C. A., & Wixted, J. T. (2020). Policy and procedure recommendations for the collection and preservation of eyewitness identification evidence. Law and Human Behavior, 44(1), 3–36. https://doi.org/10.1037/lhb0000359.
Wixted, J. T., & Wells, G. L. (2017). The relationship between eyewitness confidence and identification accuracy: A new synthesis. Psychological Science in the Public Interest, 18(1), 10–65. https://doi.org/10.1177/1529100616686966.