Abstract
Lineups and photo arrays are often presented to witnesses by police officers who know which lineup member is the suspect (single-blind lineup administration) rather than by officers who do not know (double-blind administration). Administrators who are not blind to which lineup member is the suspect are more likely than blind administrators to emit behavioral cues that steer witnesses toward choosing the suspect and away from choosing fillers (i.e., a lineup member who is not the suspect). Moreover, nonblind administrators may provide confirmatory feedback to witnesses who identify the suspect, increasing their confidence in the accuracy of their identification and weakening the correlation between witness confidence and accuracy. Nonblind administrators are also more likely to interpret witnesses’ tentative statements about a suspect than about a filler as a positive identification. Because of these findings that single-blind administration biases identifications against suspects, even when they are innocent, evidence-based recommendations for best practices in the collection of eyewitness-identification evidence call for the use of double-blind lineup-administration procedures.
A police officer presents photos of men to a witness to see whether she can identify the man who shot her boyfriend. When the witness sees the third photo, she says, “Yes, that’s him.” She also indicates that the fourth photo might depict the culprit. As the witness states that the person pictured in the next to last photo also looks like the culprit, the officer shakes his head from side to side. The witness then says, “Well, you said I had to say yes or no, so I don’t want to say he kind of is the shooter.” The officer shakes his head up and down and reaches for the photo she is examining. “So, you identified photo #3,” states the officer. The witness responds that she is not certain and needs to see the photos again. The officer replies that he cannot show her a single photo but will have to show her all the photos again. When he shows her the third photo, she identifies the man in the photo as the culprit. The officer completes the paperwork without showing her the remaining photos and records that she has confidently identified the person in the third photograph, the police suspect, as the shooter. Was this identification based solely on the witness’s memory of the culprit, or did the police officer, who knew that the suspect was depicted in the third photo, play a role in the witness’s identification of the suspect and her confidence in that identification?
A lineup (or photo array, which we will use interchangeably) provides a test of the police officer’s theory that the suspect is indeed the culprit of a crime rather than innocent. Because a lineup is a hypothesis-testing enterprise, the validity of a lineup can be evaluated using the same principles that one would use to evaluate the validity of an experiment (Wells & Luus, 1990). To test their theory, detectives hypothesize that a witness who has seen the culprit will identify the suspect as the culprit. A detective creates a lineup to test this hypothesis, consisting of the suspect paired with people who are known to be innocent (i.e., fillers). The detective (the experimenter) instructs the witness about what to do and follows the protocol for showing the lineup members to the witness (the procedure). The witness provides the detective with an identification decision and often the confidence with which that decision was made (dependent measures). Note that the witness may identify the suspect, which could be a correct (i.e., a true positive) or mistaken (i.e., a false positive) identification or a filler (which is a known error). The witness may also state that the culprit is not present in the lineup, which is either a false-negative or true-negative decision depending on whether the culprit is present or absent in the lineup. The officer then draws an inference from the data and decides whether to charge the suspect, seek additional evidence, or both.
Lineups have stronger validity to the extent that one can rule out alternative explanations for the identification decision. In particular, it is desirable to rule out any possibility that the witness chose the suspect because the lineup procedure was suggestive rather than because the suspect matched the witness’s memory of the culprit. Many recommendations for reforming police lineup procedures are intended to rule out alternative explanations for a witness identifying a suspect. One such recommendation is that lineups be conducted by an administrator who does not know which lineup member is the suspect (i.e., double-blind administration; Wells et al., 1998, 2020).
The opening description of a photo-array administration performed in an actual case illustrates some of the problems that can arise when officers fail to use double-blind procedures. Single-blind administration occurs when the police officer conducting the identification procedure knows which lineup member is the suspect and which lineup members are fillers. This knowledge causes administrators, either consciously or unconsciously, to behave in ways that communicate to the witness which lineup member is the suspect.
Prevalence of Blind Lineup Administration
When double-blind administration of identification procedures was first recommended (Wells et al., 1998), there were no jurisdictions routinely using the procedure. Since then, the use of blind administrators has increased yet is far from universal. According to survey results from a national sample of more than 1,300 law-enforcement agencies, 31% used blind administrators to conduct photo arrays, and only 7.9% used blind administrators to conduct in-person live lineups (Police Executive Research Forum, 2013). Thus, even 15 years after the publication of the initial guidelines recommending the use of blind administrators (Wells et al., 1998), police officers often knew which lineup member was the suspect when administering an identification procedure. Half of states in the United States still have no statewide requirements for the use of blind lineup administration (Kovera & Evelo, 2017).
Double-Blind Administration Reduces Identifications of Suspects
Not only was the recommendation for double-blind lineup identification procedures made before there was much use of the procedure, but also, there had been no empirical studies examining the effects of such procedures on eyewitness accuracy, confidence, or any other variable. There was research examining the role that interpersonal expectations played in self-fulfilling prophecies—which occur when an individual’s expectations for another’s behavior cause that behavior to occur (Rosenthal, 2002). That is, a person’s expectations about a target person can change how that person interacts—often without awareness—with the target. Those behavioral changes, in turn, produce the very behavior that is expected of the target. However, none of those studies directly tested whether the expectations held by lineup administrators (e.g., the suspect is the culprit) influenced how they conducted an identification procedure and whether any differences in the behavior of nonblind administrators influenced witnesses’ choices from a lineup or their confidence in the accuracy of that choice. More than 20 years since double-blind administration was first recommended, we now have direct evidence for the superiority of double-blind identification procedures for generating witness identifications (Kovera & Evelo, 2017).
Researchers have used three simulation paradigms to test the effects of nonblind administrators on witnesses’ identification decisions (Kovera & Evelo, 2017) with community members or undergraduate students serving as participants. Because administrators have the expectation that the witness will identify the suspect, the cue-disruption paradigm assumes that nonblind administrators will send cues, wittingly or unwittingly, that signify which lineup member is the suspect, but only if conditions permit. In the one experiment using this paradigm, experimenters, who knew which lineup member was the suspect and were told to get as many identifications of that suspect as possible, administered lineups to witnesses. Half of the time, experimenters were forced to stand behind the witnesses while they made their decisions, disrupting the ability of witnesses to receive any nonverbal cues that the administrator might be sending (Haw & Fisher, 2004). By disrupting the ability of the administrators to cue the witnesses to the suspect’s identity, witnesses were less likely to identify suspects when the administrator was behind the witnesses than in front of them.
Like the cue-disruption paradigm, the steering paradigm also assumes that knowing which lineup member is the suspect will influence the administrator to emit behaviors that steer witnesses toward identifying the suspect. In this paradigm, the lineup administrator is a confederate of the experimenters who follows a script when interacting with the witnesses. To simulate a double-blind administration, the confederate does not interact with the witness after delivering instructions about the task. To simulate steering, the confederate encourages witnesses to make either an identification or specifically an identification of the suspect (Clark, Brower, Rosenthal, Hicks, & Moreland, 2013; Clark, Marshall, & Rosenthal, 2009; Rhead, Rodriguez, Korobeynikov, Yip, & Kovera, 2015). Steering toward the suspect increases the rate at which witnesses identify the suspect (Clark et al., 2013; Rhead et al., 2015).
The double-blind paradigm examines the complete cycle of interpersonal expectancy effects by employing two participants for each experimental session: one acting as an administrator and another acting as a witness. The researchers manipulate whether participant administrators know which lineup member is the suspect and examine the effect of the manipulation on participant witnesses’ identifications of the suspect. In addition, the behaviors of the administrators are coded for type (e.g., pointing at a photo, removing a photo from sight, asking a question about a specific photo; Greathouse & Kovera, 2009) and whether the behavior was directed at the photo of the suspect or one of the fillers (Charman & Quiroz, 2016; Zimmerman, Chorn, Rhead, Evelo, & Kovera, 2017). Nonblind administrators exhibited more behaviors such as smiling when a witness looked at a photo of a suspect (Charman & Quiroz, 2016), pointing or drawing attention to the suspect (Zimmerman et al., 2017), or asking about the photo of the suspect (Zimmerman et al., 2017) than did blind administrators. As a result, witnesses were more likely to identify the suspect when the administrators knew which lineup member was the suspect (Charman & Quiroz, 2016; Greathouse & Kovera, 2009; Zimmerman et al., 2017).
Additional Benefits of Double-Blind Lineup Administration
In addition to preventing nonblind administrators from communicating to witnesses which lineup member is the suspect, double-blind lineup administration prevents lineup administrators from emitting other behaviors that could contaminate other legally relevant information. Specifically, administrators’ knowledge of the suspect can contaminate both witnesses’ reports of their confidence and administrator reporting of witness statements.
Uncontaminated estimates of confidence and self-reports of witnessing conditions
Double-blind procedures prevent administrators from providing feedback to witnesses about whether they identified the suspect, allowing for uncontaminated measurement of witnesses’ confidence in the accuracy of their identifications. Jurors’ trial decisions are influenced more heavily by variations in witness confidence than by other factors that are related to witness accuracy, including the quality of the witnessing conditions or the procedures used by police to collect the identification (Cutler, Penrod, & Dexter, 1990). However, witness confidence is malleable. Even relatively innocuous statements, such as “Good, you identified the suspect” (Wells & Bradfield, 1998), cause witnesses to inflate their reported confidence in the accuracy of their decision even when they have identified an innocent person (Steblay, Wells, & Douglass, 2014). In addition to inflating witness confidence, positive feedback causes witnesses to report having a better view of the culprit, paying more attention to the culprit’s face, and having a better basis on which to make an identification (Steblay et al., 2014). To eliminate the potential for feedback to influence witnesses’ self-reports, police should record witnesses’ reports of witnessing conditions as soon as possible after the crime was witnessed (Wells et al., 2020). Witnesses’ reports of their confidence will be more strongly related to accuracy if they are gathered by administrators who are blind to which lineup member is the suspect immediately after the identification is made (Wells et al., 2020; Wixted & Wells, 2017).
Unbiased reporting of witness decisions and statements
Knowing which lineup member is the suspect may also influence what police officers report about the identification procedure. Although current recommendations for best practice include video recording the administration of identification procedures (Modjadidi & Kovera, 2018; Wells et al., 2020), most departments do not routinely video record identifications. Police reports often provide the most contemporaneous record of what happened during an identification procedure. However, administrators’ knowledge of which lineup member is the suspect influences the likelihood that they will interpret ambiguous statements from the witness about the suspect (e.g., “It could be #3 but I’m not sure”; Charman, Matuku, & Mook, 2019, p. 1260) as an identification. Likewise, nonblind administrators were more likely to report that the witness had made an identification when a confederate witness identified the suspect rather than a filler; blind administrators were equally likely to report an identification of a suspect or a filler (Rodriguez & Berry, 2014, 2020). Thus, to promote accurate reporting of witnesses’ identification decisions, lineup administrators should be kept blind to which lineup member is the suspect.
Objections to Policies Advocating Double-Blind Administration
Some scholars have raised objections about instituting double-blind administration as policy. For example, some scholars have criticized the double-blind recommendation for conducting lineups, arguing that there are too few published studies on the topic (Clark, 2012b; Clark, Moreland, & Gronlund, 2014). Double-blind methods—using placebos to control both researcher and participant expectations—have been the gold standard in medical research for decades. However, when the double-blind recommendation was first made (Wells et al., 1998), there were no studies testing the assumption that interpersonal expectancies operated in the police station the same way they did in a doctor’s office. Now there are at least 11 published articles demonstrating the influence of nonblind administrators in the eyewitness context (Eisen, Cedré, Williams, & Jones, 2018; 10 articles reviewed by Kovera & Evelo, 2017).
Some scholars have objected to the double-blind administration because it decreases the correct identification of guilty suspects as well as the mistaken identification of innocent suspects (Clark, 2012a). They have also argued that the effect of nonblind administration is greater when the suspect is the culprit than when the suspect is innocent, resulting in a greater increase in correct than mistaken identifications, despite acknowledging that the desirable increase in correct identifications is obtained using “undesirable means” (Clark et al., 2013, p. 164). Yet a meta-analysis of the known research comparing identifications of suspects in double-blind and single-blind lineups suggests that the effects of blindness are greater when the suspect is innocent (odds ratio [OR] = 2.00) than when the suspect is guilty (OR = 1.58; Kovera & Evelo, 2017). In addition, the increase in correct identifications is obtained via legally impermissible suggestion from administrators (Wells, Steblay, & Dysart, 2012).
Finally, some authors have used signal detection frameworks to suggest that—through their behavior—nonblind administrators provide witnesses with information that affects how liberally they set their criterion for choosing a member of the lineup (i.e., how certain they must be to make an identification; Mickes & Gronlund, 2017) but not discriminability of the suspect from the fillers. However, conditions that increase identifications of the suspect also increase administrator behaviors that call attention to the suspect and away from fillers (Zimmerman et al., 2017). Increases in witness identifications of suspects are accompanied by losses in filler identifications, not rejections (Eisen et al., 2018; Greathouse & Kovera, 2009; Kovera & Evelo, 2017). This phenomenon has been termed the filler-to-suspect shift (Kovera & Evelo, 2017). In sum, evidence is mounting that nonblind-administrator influence occurs through a transfer of information—information about which lineup member is the suspect—from the administrator to the witness and not shifts in criterion for choosing or changes in discriminability.
Conclusion
Before psychological scientists recommended the use of double-blind lineups, no police departments were using them. Now double-blind lineups are used in at least half of the states in the United States. This growth in the use of double-blind lineup procedures is directly attributable to the research and advocacy of psychological scientists. There are now substantial data demonstrating that administrators’ knowledge of which lineup member is the suspect causes them to emit behaviors that are directed toward the suspect and subsequently influence the choices that witnesses make from the lineups. The positive feedback that witnesses receive from nonblind administrators can increase their confidence, attenuating the relationship between witness confidence and witness accuracy. New studies on the effects of nonblind administrators’ accuracy in reporting witness choices provide further evidence to support the recommendation for blind lineup administration.
Indeed, the authors of a new scientific review paper, endorsed by the American Psychology-Law Society (AP-LS), noted that the evidence undergirding the double-blind lineup recommendation had grown since the last AP-LS scientific review paper and retained the recommendation with slight modification (Wells et al., 2020). Recognizing that the double-blind procedure may be difficult to implement in small departments in which everyone is knowledgeable about everyone else’s cases, the authors included an option for conducting a blinded rather than blind procedure in which the lineup members’ photographs are placed in an envelope along with instructions so that witnesses can self-administer the lineup and record their decision and their confidence in that decision while alone in a room. The recommendation also cautions that the blind administrator should be the only person in the room with the witness to ensure that there is no one in the room who knows which lineup member is the suspect. Finally, the new recommendations caution that in cases with multiple witnesses, there must be a new blind administrator for each witness. Even blind administrators can steer witnesses toward choosing the suspect if they are administering a lineup containing the same suspect to a second witness because the administration of the lineup to the first witness provides information about which lineup member is the suspect to the administrator (Douglass, Smith, & Fraser-Thill, 2005; McCallum & Brewer, 2018).
Concerns about the effects of social-influence processes on the accuracy of eyewitness identifications, in combination with decreasing costs associated with computer administration of identification procedures even in the field, may result in the eventual removal of police officers from the identification procedure altogether. Until then, the best practice for collecting eyewitness-identification evidence is for the procedure to be conducted by an administrator who does not know which lineup member is the suspect.
Recommended Reading
Kovera, M. B., & Evelo, A. J. (2017). (See References). The most comprehensive review article on the topic of doubleblind lineup administration.
Steblay, N., Wells, G. L., & Douglass, A. B. (2014). (See References). A meta-analysis (i.e., quantitative review) of the research on the effects of postidentification feedback on witness confidence and self-reports of viewing conditions.
Wells, G. L., Kovera, M. B., Douglass, A. B., Brewer, N., Meissner, C. A., & Wixted, J. (2020). (See References). The American Psychology-Law Society’s scientific review article describing nine recommended best practices for collecting eyewitness-identification evidence, including double-blind administration.
