Eyewitnesses are often asked to indicate how confident they are in the accuracy of their identification and other testimony-relevant judgments. These reports are highly influential in evaluations of identification accuracy. Unfortunately, eyewitnesses’ confidence reports are highly malleable, easily influenced by myriad variables. The solution is to record witnesses’ confidence in their identification and report of crime details immediately, so as to preserve whatever useful information confidence provides.
A witness’s confidence in the accuracy of his or her identification is perhaps the most studied of all variables related to eyewitness decision making—for good reason. Eyewitness confidence is the most intuitively appealing variable for use in assessments of accuracy. Indeed, it is specially highlighted by the U.S. Supreme Court for use in such evaluations. This recommendation is consistent with lay assumptions of what variables predict identification accuracy: People believe that a confident witness is an accurate one. This assumption does have empirical support. Under certain circumstances, there is a strong, positive correlation between confidence and accuracy. For example, when viewing conditions are disparate, a strong confidence-accuracy relationship emerges: Witnesses who see a culprit under difficult viewing conditions are less confident compared with those who see a culprit under optimal viewing conditions. Other research confirms the existence of a useful relationship between confidence and accuracy. One meta-analysis determined that the confidence-accuracy correlation for witnesses who made choices from lineups or photo spreads was moderate (r = .41). These investigations are highly valuable in clarifying the maximum possible utility of confidence reports in assessing accuracy.
These investigations capture confidence reports under the best possible circumstances. Because no external variables have been introduced (e.g., photo-spread administrator influence), they are, in some sense, pure measures of the extent to which confidence is related to accuracy. Therefore, it is possible to think of these confidence reports as the estimator versions of this variable because they derive from factors outside the control of the justice system. For example, the system cannot ensure that a witness has a good view of the culprit. Therefore, to the extent that the quality of the witness’s view determines how confident he or she is, the justice system has no hand in a witness’s confidence.
In other ways, the justice system has a substantial role in the level of confidence an eyewitness expresses in his or her identification. This influence is driven by system variables—those variables the justice system can control. Many system variables have been implicated in confidence inflation. This influence has been demonstrated across three different categories of confidence reports: current confidence in the identification decision (e.g., How confident are you in the accuracy of your identification right now?), retrospective confidence in the identification decision (e.g., How confident were you when you made your identification?), and reports about details of the witnessed event (e.g., What kind of disguise was the culprit wearing?). As described below, the malleability of confidence in each of these three categories can be attributed to system variables.
Confidence in Reports of Crime Details
Although confidence malleability is most prominently studied in relation to identification accuracy, some researchers have focused on its malleability in the context of crime detail recollections. In one investigation, eyewitnesses questioned over the course of 5 weeks reported significantly elevated confidence levels at the end of that period, without any corresponding increase in the accuracy of their reports. The same elevation occurred when eyewitnesses were questioned over the course of 5 days. In addition, a manipulation as simple as the context in which a confidence report is given can influence the magnitude of an eyewitness’s certainty. Witnesses who give reports about crime details in public provide significantly lower confidence ratings than do those witnesses who give reports privately, even though the accuracy in both groups is equivalent. The number of times a witness is interviewed and the context of the interview are both variables under the control of the investigating officers to a certain extent.
Current Confidence in Identification Accuracy
In many crimes, many people witness the same event. Some of the crimes for which innocent people were wrongfully convicted include up to five individuals all identifying the same person. In one of the most elaborate empirical examinations of the effect of cowitnesses, witnesses saw a live staged crime in pairs. Witnesses were then separated for the identification attempt and confidence report. Finally, witnesses were randomly assigned to learn one of four types of information about their cowitness’s decision. Those who learned that their cowitness identified the same person they did or identified an implausible other reported the highest levels of confidence. Those who learned that their cowitness either identified someone else or did not make an identification all had confidence levels that were significantly lower than witnesses in a control (no cowitness information) condition. Information from a cowitness can also alter confidence in reports of crime details. In one study, witnesses’ confidence in whether an accomplice was present at the scene of a crime changed depending on their partner’s report of whether that accomplice was present. The justice system has limited control of whether cowitnesses speak to one another. At the very least, cowitnesses should be separated until each has provided an identification decision, complete report of crime details, and indicated the confidence in each judgment.
Cowitnesses are not the only source of contamination for current confidence reports. Photospread administrators have long been targeted as a potential source of influence in eyewitnesses’ decisions. Initially, concerns centered on the ability of a photospread administrator to affect an eyewitness’s choices; research does suggest this is a worthy concern. Recently, however, concerns have expanded to include the problem of administrators influencing an eyewitness’s confidence. In one demonstration of this problem, eyewitnesses attempted identifications in two conditions. In one condition, the photospread administrator knew who the suspect was. In the other condition, the photo-spread administrator did not know who the suspect was. Eyewitnesses who made identifications under the first condition reported higher confidence in their accuracy than did eyewitnesses who made identifications in the second condition. The influence inherent in this situation is easily solved by ensuring that the person administering a set of photos to an eyewitness does not know who the suspect it; the system can control whether this safeguard is adopted.
Retrospective Confidence in Identification Accuracy
Malleability in retrospective confidence reports is perhaps the most problematic of the three categories, in part because this category is specifically highlighted by the U.S. Supreme Court for use in determining accuracy. The Court indicates that the relevant confidence report is from the “confrontation,” suggesting that they recognized the possibility for confidence to increase over time. Unfortunately, profound distortions in witnesses’ memories of how confident they were at the time of the identification are easy to create with postidentification feedback. Witnesses who hear that their identification was correct report remembering with greater certainty at the time of the identification compared with witnesses who heard nothing about their accuracy. Because this effect occurs for eyewitnesses who have made false identifications, the manipulation produces a set of highly confident, but wrong, eyewitnesses. As with the other system variables described above, this one has an easy solution. If confidence reports are collected immediately after an identification is made, eyewitnesses’ confidence reports will not be vulnerable to influence by the photospread administrator.
Implications of Confidence Malleability
As noted above, there is a nontrivial, positive relationship between confidence and accuracy under certain circumstances. However, because confidence is malleable, the significant relationship between the two variables can easily be compromised or even eliminated. One way in which the confidence-accuracy correlation is eliminated is by suggesting to witnesses that they prepare for cross-examination. In such a situation, witnesses who have made inaccurate identifications often inflate their confidence to the point where their confidence is indistinguishable from that of accurate witnesses. Postidentification feedback has a similarly devastating effect on the confidence-accuracy correlation: Witnesses who hear that their identification was correct report equivalent levels of confidence, regardless of whether their actual identification was accurate or inaccurate.
The implications of confidence inflation are profound because a witness’s confidence in the accuracy of his or her identification carries enormous weight in judgments of accuracy, often trumping other variables. In one set of studies, mock jurors were provided with information about 10 variables, all of which influence identification accuracy (e.g., the culprit’s disguise). None of the 10 variables influenced mock jurors’ assessments except confidence. In another experiment, jurors who participated disregarded the quality of a witness’s view, evaluating him or her positively as long as confidence was high. This reliance on confidence is unproblematic except that eyewitnesses routinely produce highly confident reports about identifications that are incorrect. Ample real-world evidence suggests that this is a significant problem. Many individuals exonerated by DNA evidence were convicted on the basis of confident eyewitness identifications.
Future Research on Confidence Malleability
Even though confidence malleability is a well-studied phenomenon, there are many unanswered questions. For example, researchers do not yet know for how long confidence is malleable. Some research suggests that postidentification feedback still influences retrospective certainty reports even when it is given 48 hours after an event is witnessed. These results are provocative—suggesting that confidence may be malleable for extended periods of time. However, because few studies include manipulations of time, the extent to which confidence is malleable is not well understood. The reason for the susceptibility of confidence to external influences is also not well understood. One contributing factor may be that confidence reports are derived from many sources. One other factor is undoubtedly the extent to which the stimulus matches the witness’s memory (i.e., ecphoric similarity). Another factor is the desire of witnesses to determine whether their judgment is correct (i.e., the desire for informational influence). In at least one study, the tendency to conform eyewitness decisions to others was highest when the witnessing conditions made identification difficult (i.e., the stimulus was in view for a very short time) and the task was important.
Remedies for Confidence Malleability
The most obvious remedy for confidence inflation is to record witnesses’ reports immediately after an identification is made or a crime is reported. This solution is appealing for three reasons. First, providing a confidence report may in fact inoculate witnesses against future inflations. In one study, witnesses who provided a private retrospective confidence report were less affected by postidentification feedback than were witnesses who did not. Second, records of confidence reports would allow defense attorneys to challenge subsequent inflation through cross-examination at trial. This is likely to be difficult, as one study demonstrated that mock jurors are resistant to attempts to undermine a witness’s confidence report by providing evidence that it has inflated over time. Finally, recording confidence is easy. It does not require specialized equipment or training. It can easily be incorporated into interviews with witnesses. Should immediate confidence reports not be recorded, another common suggestion is to introduce expert testimony on the malleability of confidence. This solution is less appealing because research has demonstrated that mock jurors are relatively insensitive to testimony impugning the correlation between confidence and accuracy. In some studies, jurors persist in using confidence reports even after being told that they are only minimally useful in assessing accuracy. Therefore, the most reasonable solution is to prevent eyewitnesses’ confidence from inflating in the first place. The best way to do this is to collect immediate records of confidence reports in both identification accuracy and crime details.
- Baron, R. S., Vandello, J. A., & Brunsman, B. (1996). The forgotten variable in conformity research: Impact of task importance on social influence. Journal of Personality and Social Psychology, 71(5), 915-927.
- Bradfield, A., & McQuiston, D. E. (2002). When does evidence of eyewitness confidence inflation affect judgments in a criminal trial? Law and Human Behavior, 28(4), 369-387.
- 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.
- Lindsay, S. D., Read, J. D., & Sharma, K. (1998). Accuracy and confidence in person identification: The relationship is strong when witnessing conditions vary widely. Psychological Science, 9(3), 215-218.
- Luus, C. A. E., & Wells, G. L. (1994). The malleability of eyewitness confidence: Co-witness and perseverance effects. Journal of Applied Psychology, 79, 714-724.
- Sporer, S., Penrod, S., Read, D., & Cutler, B. L. (1995). Choosing, confidence, and accuracy: A meta-analysis of the confidence-accuracy relation in eyewitness identification studies. Psychological Bulletin, 118, 315-327.