Improving juror sensitivity to specific eyewitness factors: judicial instructions fail the test

Abstract

A recent New Jersey Supreme Court decision recognized the difficulty jurors have with evaluating eyewitness evidence. This decision resulted in the development of instructions that highlight factors affecting identification accuracy. Research has explored the efficacy of eyewitness instructions for improving jurors’ decision-making. Jurors in these studies are typically presented with identifications that manipulate multiple witnessing and identification conditions simultaneously, making it difficult to ascertain whether instructions help jurors evaluate any one eyewitness factor. We conducted two experiments to examine how jurors evaluate eight individual eyewitness factors with and without instructions. Across both experiments, none of the individual eyewitness factors nor instructions influenced jurors. Instructions only assisted jurors when multiple eyewitness factors were collapsed to create either extremely good or poor-quality identifications. These findings contribute to the long history of jurors remaining largely insensitive to the nuances of witnessing and identification conditions. Current safeguards may only assist jurors under limited circumstances.

In 2011, the New Jersey Supreme Court appointed a Special Master to solicit input from experts in the field of eyewitness identification and review the scientific literature. Relying in part on the Special Master’s report, the Court reached a landmark decision regarding the reliability of eyewitness evidence and how to address this evidence in court (New Jersey v Henderson, 2011). The Henderson Court believed, ‘that it is essential to educate jurors about factors that can lead to misidentifications’ (p. 303). The Court queried, ‘ . . . if even only a small number of jurors do not appreciate an important, relevant concept, why not help them understand it better with an appropriate jury charge?’ (p. 273). Thus, to improve juror comprehension of eyewitness factors, the Court ordered new jury instructions ‘informed by sound evidence on memory and eyewitness identification, which is generally accepted by the relevant scientific community’ (pp. 302–303).

If instructions serve an educational function, one would expect to see that jurors align their verdict decision with the strength of the (eyewitness) evidence, convicting when evidence is strong and acquitting when evidence is weak (called sensitivity; see Cutler, Penrod, & Dexter, 1989). However, research testing Henderson instructions suggests they are ineffective at educating jurors about eyewitness factors (e.g. Dillon, Jones, Bergold, Hui, & Penrod, 2017; Jones, Bergold, Dillon, & Penrod, 2017; Jones & Penrod, 2018). One shortcoming of these studies and others (e.g. Cutler, Dexter, & Penrod, 1990; Greene, 1988) concerns how the eyewitness factors were manipulated. Multiple eyewitness factors were manipulated simultaneously to create extremely good or poor-quality identifications. As a result, it is unclear whether jurors can evaluate any one eyewitness factor on their own or with the assistance of Henderson instructions. The current studies address this shortcoming. This nuanced approach illuminates precisely which eyewitness factors jurors are and are not capable of evaluating independently.

Witnessing and identification conditions

Eyewitness factors fall under two broad categories: estimator variables, which concern characteristics surrounding the crime (e.g. lighting), and system variables, which concern characteristics of police intervention (e.g. lineup composition; see Wells, 1978). In the current studies, we focused on how jurors evaluated eight individual eyewitness factors that have been extensively examined in prior literature and are included in the Henderson instructions (four estimator and four system variables): race, weapon focus, lighting, distance, double-blind administration, lineup type, lineup instructions and lineup composition. Witnessing conditions, such as when (a) a perpetrator is a different race than an eyewitness (Meissner & Brigham, 2001); (b) a weapon is present (Fawcett, Russell, Peace, & Christie, 2013); (c) lighting conditions are poor (Wagenaar & Van Der Schrier, 1996); and (d) the perpetrator is further away (Lampinen, Erickson, Moore, & Hittson, 2014; Lindsay, Semmler, Weber, Brewer, & Lindsay, 2008) decrease identification accuracy.

Regarding system variables, researchers recommend the use of double-blind administration to avoid investigator bias (see Cutler & Penrod, 1995; Kovera & Evelo, 2017). When a lineup administrator knows the identity of a suspect, they can influence the choices made by eyewitnesses (Charman & Quiroz, 2016; Greathouse & Kovera, 2009; Phillips, McAuliff, Kovera, & Cutler, 1999; Zimmerman, Chorn, Rhead, Evelo, & Kovera, 2017). Furthermore, they are more likely to give feedback that artificially inflates an eyewitness’s confidence (a highly influential, but potentially unreliable factor when lineup conditions are not pristine; Wixted & Wells, 2017) and alters memories for event details (Dysart, Lawson, & Rainey, 2012; see also Steblay, Wells, & Douglas, 2014 for meta-analysis on post-identification feedback). Aside from investigator bias, identification accuracy declines when (a) a single-person showup is used relative to a lineup containing multiple fillers (Steblay, Dysart, Fulero, & Lindsay, 2003); (b) police fail to inform an eyewitness that the perpetrator may or may not be present in the lineup (i.e. biased lineup instructions; Steblay, 1997); and (c) the suspect stands out from other lineup members (Fitzgerald, Price, Oriet, & Charman, 2013).

Evidence is mixed and somewhat limited on jurors’ ability to evaluate eyewitness evidence on their own (e.g. Beaudry et al., 2015; Cutler, Penrod, & Stuve, 1988; Devenport, Cutler, Stinson, & Kravitz, 2002; Modjadidi & Kovera, 2018). Estimator variables, such as cross-race identifications (Abshire & Bornstein, 2003) and weapon presence (Cutler et al., 1988; Neal, Christiansen, Bornstein, & Robicheaux, 2012), are difficult for jurors to evaluate. In contrast, jurors are sometimes able to appropriately align verdict decisions with the quality of viewing conditions, namely the combination of lighting, distance, exposure duration and disguise (Vallano, Pettalia, Pica, & Pozzulo, 2018; cf., Jones et al., 2017, for null findings regarding the combination of exposure duration, weapon presence and time delay). However, little research has examined jurors’ sensitivity to variations in lighting and distance separately.

Regarding system variables, jurors have difficulty evaluating lineup instructions (Cutler et al., 1988; Devenport & Cutler, 2004) and double-blind lineup administration (Beaudry et al., 2015; Modjadidi & Kovera, 2018; Wright, Carlucci, Evans, & Compo, 2010). However, jurors appear sensitive to lineup composition (Devenport et al., 2002; cf. Cutler et al., 1989; Cutler et al., 1988) and lineup type (showup vs. lineup) when manipulated together with lineup instructions and confirmatory feedback (Jones et al., 2017). Thus, there is little consensus on jurors’ ability to evaluate eyewitness evidence, and the findings are often obscured by the presentation of multiple eyewitness factors simultaneously. More research is needed to establish how jurors, on their own and with the assistance of safeguards, evaluate individual (as opposed to a combination of) eyewitness factors, especially lighting, distance and lineup type.

Eyewitness instructions

Various forms of eyewitness instructions generally do not affect juror decision making or result in skepticism (see Cutler et al., 1990; Greene, 1988; Leippe & Eisenstadt, 2009). For example, the Telfaire instructions comprise five factors based on previous case law: witness’s opportunity to view the perpetrator, degree of attention, accuracy of prior description of criminal, certainty and the time between the crime and identification (United States v Telfaire, 1972). Of concern, opportunity to view the perpetrator, degree of attention and certainty may be compromised by post-identification feedback (Wells & Quinlivan, 2009). Additionally, several relevant factors are missing in Telfaire that are detailed in the Henderson instructions (e.g. lineup instructions, weapon focus). A series of studies suggest that Telfaire instructions had no impact on jurors’ decisions or made jurors skeptical of all eyewitness identifications, by reducing the likelihood of conviction even when an eyewitness witnessed the crime under good conditions (i.e. bright lighting, close proximity and an unobstructed view; Cutler et al., 1990; Greene, 1988).

Based on research presented to a Special Master, the Henderson Court concluded that, ‘ . . . people do not intuitively understand all of the relevant scientific findings [concerning eyewitness identification]. As a result, there is a need to promote greater juror understanding of those issues’ (New Jersey v Henderson, 2011, p. 274). The Court’s solution, Henderson instructions, improves upon past iterations of instructions by reviewing the three stages of memory and how a variety of factors can affect eyewitness identification accuracy and reliability (see New Jersey Supreme Court, 2012, for full instruction). Despite this, the results of studies on Henderson instructions are consistent with those on Telfaire instructions. In some studies, Henderson instructions did not influence juror verdicts at all (Jones et al., 2017; Jones & Penrod, 2018). Others found that mock jurors were less likely to convict when they received Henderson instructions than when they did not (Dillon et al., 2017; Papailiou, Yokum, & Robertson, 2015). This reduced likelihood of conviction occurred even when witnessing and identification conditions were optimal (e.g. long exposure to perpetrator, unbiased lineup instructions). In other words, the instructions made jurors skeptical of all eyewitness identifications, without increasing their ability to distinguish between reliable and unreliable conditions for eyewitness identifications.

In contrast, a recent study of modified Henderson instructions resulted in sensitivity to the strength of eyewitness evidence (see Appendix; Jones & Penrod, 2018). This version of Henderson instructions was simplified (reduced Flesch–Kincaid grade level from 15 to 12), included a general list of factors that were not case-specific and asked jurors to reflect on the quality of each factor (i.e. ‘Ask yourself: Was the eyewitness stressed? Did this enhance or impair her accuracy?’; see also Pawlenko, Safer, Wise, & Holfeld, 2013). Jurors exposed to the modified instructions were more likely to convict when eyewitness evidence was strong and less likely to convict when eyewitness evidence was weak, a pattern not observed among jurors who received the original Henderson instructions (Jones & Penrod, 2018). Thus, this modified version of Henderson instructions appears promising for educating jurors as envisioned by the Henderson Court. Nevertheless, the effectiveness of modified Henderson instructions cannot be determined from one study or one set of eyewitness circumstances.

Current experiments

In all extant studies testing eyewitness instructions, the quality of multiple eyewitness factors was manipulated simultaneously resulting in identifications made under extremely good or extremely poor conditions (e.g. Cutler et al., 1990; Greene, 1988; Papailiou et al., 2015; Pawlenko et al., 2013). Only two studies have looked at estimator and system variables separately in the context of testing instruction effectiveness. In both cases, these researchers manipulated a group of estimator variables and a group of system variables, which resulted in some identifications that had mixed witnessing and identification quality (Dillon et al., 2017; Jones et al., 2017). However, prior work has been unable to determine jurors’ ability to evaluate individual eyewitness factors with the assistance of eyewitness instructions.

The current research addresses this limitation and investigates which variables jurors can best comprehend. By independently manipulating several eyewitness factors, the current experiments test how jurors evaluate identifications made under varying witnessing and identification conditions with and without the aid of Henderson instructions. These variations better reflect actual eyewitness evidence (see www.innocenceproject.org for examples) and allows for a more nuanced examination of how the quality of each individual eyewitness factor affects the decision-making process of jurors.

The results from prior studies on Henderson instructions demonstrate that, overall, the instructions do not provide an adequate framework for jurors to evaluate eyewitness evidence. The goal of the current research is to determine which variables jurors are sensitive to on their own and with the intervention of judicial instructions. Given the recent findings in Jones and Penrod (2018), we hypothesize that modified Henderson instructions will sensitize jurors to variations in identification quality. We explore whether this interaction varies by type of eyewitness factor.

Experiment 1

Method

Participants

We recruited 350 U.S. jury-eligible community members via Amazon Mechanical Turk (mTurk.com). Amazon Mechanical Turkers had a 95% or higher approval rating and at least 100 approved human intelligence tasks (HITs). Half of participants were women (50.1%), and the majority were Caucasian (69.6%), followed by African American (13.2%) and Asian (10%), and the remaining indicated they were American Indian, Native Hawaiian, or two or more races (6%). Participants ranged in age from 18–73 years (M = 36.38, SD = 11.30).

Design and manipulations

Design

Participants were randomly assigned to one condition in a 2 (eyewitness instructions: none vs. modified Henderson) × 2 (lighting: good vs. poor) × 2 (distance: close vs. far) × 2 (lineup administration: double-blind vs. non-blind) × 2 (lineup composition: fair vs. unfair) between-subjects factorial design. An a priori power analysis (80% power) suggested a sample size of 350 participants would allow us to detect a small effect (f = .15).

Manipulations

Participants read a trial transcript that contained all manipulations. The prosecution claimed the defendant committed robbery. Participants read opening and closing statements from both the prosecution and defense, as well as direct and cross examinations of the lead detective, victim and defense alibi, and reasonable doubt and burden of proof instructions. The trial transcript varied from 3759 to 3821 words. Participants randomly assigned to receive modified Henderson instructions read standardized instructions that covered 15 eyewitness factors at the end of the trial (see Appendix). In addition, we varied the quality of four eyewitness factors covered in the instructions. Participants learned of this evidence during the detective’s and victim’s testimonies. Lighting conditions were either good or poor (i.e. the parking lot where the crime occurred was described as well-lit or very dark). The distance of the event from the eyewitness was either long or short: the eyewitness estimated the perpetrator was less than a few feet away or several feet away. The lineup administrator either did or did not know who the suspect was in the lineup. Finally, the composition of the lineup varied in whether the suspect stood out relative to other lineup members.

Participants completed manipulation checks regarding all five manipulated variables. In all conditions, the eyewitness indicated she was confident that she made the correct identification. We intentionally used this vague language to describe confidence (as opposed to a specific percentage, such as 90% or 100% confident) to avoid overpowering the manipulations.

Dependent variables

Participants provided dichotomous judgments of the defendant’s guilt (0 = not guilty; 1 = guilty) and indicated the likelihood that the eyewitness was correct (0–100%) and the defendant committed the crime (0–100%). Participants also rated the strength of the case presented by the prosecution (e.g. ‘The prosecution had a very strong case’; 1 = strongly disagree; 9 = strongly agree) and evidence strength by witness (eyewitness and detective; 1 = strong evidence for innocence; 9 = strong evidence for guilt).

Procedure

We posted a HIT on Amazon’s Mechanical Turk website to recruit jury-eligible community members. Participants who provided informed consent were randomly assigned to one of 32 conditions. After reading a trial transcript that varied in the presence of eyewitness instructions and the quality of four eyewitness factors, participants completed a questionnaire containing dependent measures and were debriefed. Participants were paid $1.50 for completion of the experiment, which took, on average, 24.5 min. Data and materials for both experiments are available from the first author’s Open Science Framework (OSF) page.¹

Results

Participants completed five manipulation checks. The first manipulation check for instructions consisted of a checklist of 14 items. Participants were asked which of the following topics did the judge’s instructions cover (with the option to check all that applied): exposure duration, distance, confirmatory feedback, time elapsed, lighting, eyewitness lineup instructions, cross-race, weapon focus, showup, double-blind lineup administration, lineup composition, stress, intoxication and disguise. The number of checked items were added together for each participant. In addition, participants completed manipulation checks for the lighting conditions (well-lit or dark), distance between the eyewitness and perpetrator (less than a few feet away or several feet away), who administered the eyewitness lineup (the detective or a neutral officer) and the composition of the lineup (fair or biased).

Participants who received modified Henderson instructions reported that the judicial instructions mentioned more eyewitness factors than participants who did not receive modified Henderson instructions, t(318) = −26.49, p < .001, d = 2.85, 95% confidence interval, CI [2.55, 3.15]. Additionally, we observed that most participants successfully noticed the eyewitness manipulations. Compared to those who read about poor lighting conditions, participants who read about good lighting conditions were more likely to indicate that the parking lot was well-lit (4% vs. 86%), χ²(1, N = 349) = 244.09, p < .001, φ = .84. Compared to those in the long-distance condition, participants in the short-distance condition were more likely to indicate that the perpetrator was less than a few feet away (9% vs. 80%), χ²(1, N = 349) = 182.25, p < .001, φ = .72. Compared to those in the single-blind administration condition, participants in the double-blind administration condition were more likely to indicate that a neutral officer conducted the lineup (3% vs. 83%), χ²(1, N = 349) = 247.74, p < .001, φ = .84. Finally, compared to those who read about a biased lineup, participants who read about a fair lineup were more likely to indicate that the suspect did not stand out in the lineup (9% vs. 76%), χ²(1, N = 349) = 163.39, p < .001, φ = .68. We reran analyses multiple times, restricting the sample based on the number of correct manipulation checks. We present results with all participants because results did not differ as a function of correct responses.

For the main analyses, we first performed a logistic regression analysis on our primary variable of interest – dichotomous juror verdicts. We did not observe any significant partial, χ²(5, N = 349) = 3.04, p = .69, Nagelkerke R² = .01, or interactive effects, χ²(9, N = 349) = 5.13, p = .82, Nagelkerke R² = .02 (descriptives are provided in Table 1). Next, we ran a multivariate analysis of variance (MANOVA) to test for the main and interactive effects of our manipulated variables on five outcomes of interest: likelihood of a correct identification, prosecution strength, eyewitness and detective testimony, and continuous judgments of guilt. Correlations among all variables are presented in Table 2. Multivariate effects emerged for lineup administration, F(5, 313) = 2.32, p = .04, Wilk’s λ = .96, ${\eta }_{\mathrm{p}}^2$ = .04, and lineup composition, F(5, 313) = 2.30, p = .05, Wilk’s λ = .97, ${\eta }_{\mathrm{p}}^2$ = .04. Univariate effects indicate that the detective’s testimony was more indicative of guilt when the lineup was administered single-blind (versus double-blind), F(1, 317) = 4.87, p = .03, ${\eta }_{\mathrm{p}}^2$ = .02, d = 0.22, 95% CI [.01, .43], as well as when the lineup was fair (versus biased), F(1, 317) = 6.58, p = .01, ${\eta }_{\mathrm{p}}^2$ = .02, d = 1.49, 95% CI [0.29, 2.68]. No other effects were significant.

Table 1.

Experiment 1 descriptives.

Variables	Eyewitness instructions		Lighting		Distance		Lineup administration		Lineup composition		Overall
	None M (SD)	Modified M (SD)	Dim M (SD)	Well-lit M (SD)	Far M (SD)	Near M (SD)	Single M (SD)	Double M (SD)	Biased M (SD)	Fair M (SD)
Correct ID	58.45 (29.47)	61.33 (29.17)	61.88 (29.35)	58.14 (20.90)	58.25 (29.50)	61.54 (29.79)	59.36 (29.11)	60.97 (29.93)	58.61 (29.83)	62.85 (28.75)	59.87 (29.32)
Prosecution strength	4.79 (1.76)	4.85 (1.81)	4.82 (1.76)	4.83 (1.83)	4.65 (1.77)	5.03 (1.81)	4.81 (1.77)	4.87 (1.85)	4.72 (1.88)	4.95 (1.68)	4.82 (1.79)
Eyewitness testimony	5.56 (1.39)	5.77 (1.29)	5.77 (1.25)	5.56 (1.42)	5.68 (1.35)	5.65 (1.34)	5.62 (1.38)	5.70 (1.30)	5.69 (1.25)	5.64 (1.42)	5.66 (1.34)
Detective testimony	4.85 (1.43)	5.01 (1.50)	5.05 (1.35)	4.82 (1.57)	4.97 (1.50)	4.89 (1.44)	5.09 (1.28)	4.77 (1.62)	4.73 (1.54)	5.12 (1.37)	4.93 (1.47)
Guilt (continuous)	55.57 (29.46)	59.34 (28.70)	59.33 (20.21)	55.28 (20.57)	56.03 (29.60)	58.98 (29.34)	55.94 (28.90)	59.59 (29.74)	56.20 (29.36)	59.85 (29.14)	57.43 (29.11)
% Guilty	46	45	49	42	43	49	46	44	47	44	46

Note: N = 346–349. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime.

Table 2.

Experiment 1 correlation matrix among dependent variables.

Variable	Verdict	1	2	3	4
1. Correct ID	.70*
2. Prosecution strength	.59*	.66*
3. Eyewitness testimony	.45*	.59*	.55*
4. Detective testimony	.48*	.46*	.49*	.48*
5. Guilt (continuous)	.66*	.85*	.61*	.54*	.44*

 

Note: N = 349. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime. Verdict: 0 = not guilty, 1 = guilty.

*p < .001.

To provide a comparison with past studies, we collapsed the four eyewitness factors to create a dichotomous identification quality variable (extremely good or extremely poor). We examined the effects of this identification quality variable and instructions, as well as their interaction via logistic regression and MANOVA. For the logistic regression, the omnibus tests for the partial, χ²(2, N = 43) = 0.53, p = .77, Nagelkerke R² = .02, and interactive effects, χ²(3, N = 43) = 0.78, p = .85, Nagelkerke R² = .02, were non-significant. Similarly, we did not observe any significant main or interactive effects for any continuous outcomes examined in MANOVA, Fs(5, 35) > 0.38, p > .30, Wilk’s λ > .85, ${\eta }_{\mathrm{p}}^2$ > .05 (see Table 3).

Table 3.

Experiment 1 sensitivity tests: dependent variables as a function of eyewitness instructions and identification quality.

	Eyewitness instructions
Dependent variable	Absent M (SD)	Present M (SD)
Correct ID
Extremely poor	47.22 (30.73)	63.64 (31.66)
Extremely good	44.18 (36.02)	56.33 (34.83)
Prosecutionstrength
Extremely poor	4.33 (1.94)	4.27 (1.85)
Extremely good	4.18 (1.94)	4.92 (1.98)
Eyewitnesstestimony
Extremely poor	5.33 (1.80)	5.64 (1.43)
Extremely good	4.91 (1.76)	5.50 (1.58)
Detectivetestimony
Extremely poor	5.00 (1.66)	4.82 (1.25)
Extremely good	4.00 (1.73)	4.67 (1.97)
Guilt (continuous)
Extremely poor	42.67 (33.65)	52.46 (33.20)
Extremely good	40.82 (32.56)	54.25 (37.16)
% Guilty
Extremely poor	56	45
Extremely good	36	42

Note: N = 43. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime.

Discussion

We set out to examine which eyewitness factors jurors are sensitive to on their own or with the assistance of modified Henderson instructions. We hypothesized that modified Henderson instructions would sensitize jurors to the four eyewitness factors manipulated in this experiment: lighting, distance, lineup administration and lineup composition. However, the results suggest jurors were largely unaffected by the manipulations. Only two exceptions emerged. Jurors changed their perceptions of the detective’s testimony based on the type of lineup administration and composition. Jurors aptly noted that the detective’s testimony was more indicative of the defendant’s guilt when the lineup was fair than when it was biased. However, jurors also thought the detective’s testimony was more indicative of the defendant’s guilt when the lineup administrator knew who the suspect was in the lineup (as opposed to double-blind). This is problematic given the extensive research on single-blind administration, which can bias eyewitness memory and decisions (Charman & Quiroz, 2016; Dysart et al., 2012; Greathouse & Kovera, 2009; Phillips et al., 1999; Steblay et al., 2014; Zimmerman et al., 2017).

The findings suggest that jurors are unable to evaluate the individual eyewitness factors of lighting, distance, lineup administration and lineup composition on their own (see also Cutler et al., 1989; Cutler et al., 1988; Modjadidi & Kovera, 2018; Wright et al., 2010; cf., Devenport et al., 2002). Furthermore, unlike Jones and Penrod (2018), we observed that jurors were not sensitive to extremely good or poor identifications even when they received modified Henderson instructions. These null findings add to the existing research examining previous iterations of Henderson instructions (Jones et al., 2017; Jones & Penrod, 2018). Before drawing conclusions regarding jurors’ ability to evaluate eyewitness evidence or the effectiveness of these instructions, we conducted a second Experiment to examine four additional eyewitness factors.

Experiment 2

Method

Participants

Given the null findings in Experiment 1, we collected a larger sample and implemented more stringent recruitment criteria. We increased power (90% power) to detect a small effect (f = .15), which suggested a sample size of 469. We subsequently recruited 475 U.S. jury-eligible community members via mTurk.com. In order to participate in the experiment, Amazon Mechanical Turkers had to have Masters status, a 95% or higher approval rating, and over 1000 approved HITs. Most participants identified as female (51.6%) and Caucasian (75.5%; followed by 9.1% Other, 8.4% African American and 7% Hispanic). Participants ranged in age from 19–78 years (M = 36.09; SD = 11.59).

Design and materials

Participants were randomly assigned to one condition in a 2 (eyewitness instructions: none vs. modified Henderson) × 2 (race: cross race vs. same race) × 2 (weapon: present vs. absent) × 2 (lineup type: six-person vs. single person showup) × 2 (lineup instructions: biased vs. unbiased) between-subjects factorial design. We used the same trial transcript as that in Experiment 1 but altered the eyewitness factors. The trial transcript varied from 3579 to 3672 words.

Manipulations, measures and procedure

We randomly assigned participants to receive no eyewitness instructions or modified Henderson instructions (see Appendix). We varied the quality of four eyewitness factors. The race of perpetrator varied (either Caucasian or African American; the victim was always Caucasian). Additionally, we varied whether the perpetrator used a knife, whether the eyewitness viewed a six-person lineup or a single-person showup, and whether the lineup administrator explicitly informed the eyewitness that the perpetrator may or may not be in the lineup. As with Experiment 1, the eyewitness indicated she was confident that she made the correct identification in all versions of the trial. Participants completed manipulation checks regarding the eyewitness instructions and lineup instructions.² Additionally, participants completed the same dependent measures and followed the same procedure as those in Experiment 1. We paid subjects $1.20 for their time, which was 18 min on average.

Results

Participants completed two manipulation checks. First, participants completed a checklist of nine items regarding the contents of instructions. Participants were asked which of the following topics did the judge’s instructions cover (with the option to check all that applied): exposure duration, distance, confirmatory feedback, time elapsed, lighting, lineup instructions, cross-race, weapon focus and showup. The number of checked items were added together for each participant. In addition, participants indicated whether the police informed the victim that the suspect may or may not be present in the identification procedure.

Participants who received modified Henderson instructions reported that the judicial instructions included more eyewitness factors than did participants who did not receive modified Henderson instructions, t(394) = −15.55, p < .001, d = 1.36, 95% CI [1.16, 1.56]. Additionally, compared to participants in the biased lineup instruction condition, participants in the unbiased lineup instruction condition were more likely to respond that the police gave unbiased lineup instructions (39% vs. 88%), χ²(1, N = 444) = 115.17, p < .001, φ = .51. As with Experiment 1, most participants successfully passed manipulation checks. Results did not differ as a function of restricting the sample to those who passed manipulation checks. Thus, we retained all participants for subsequent analyses.

For the main analyses, we first examined the effects of each manipulated variable and all interactions with eyewitness instructions on verdict using logistic regression analysis (descriptives are provided in Table 4). The omnibus tests indicate no significant partial, χ²(5, N = 475) = 3.53, p = .62, Nagelkerke R² = .01, or interactive effects, χ²(9, N = 475) = 6.64, p = .68, Nagelkerke R² = .02. Next, we ran a MANOVA to test for the main and interactive effects of our manipulated variables on five outcomes of interest: likelihood of a correct identification, prosecution strength, eyewitness and detective testimony, and continuous judgments of guilt. Correlations among all variables are presented in Table 5. No multivariate effects were significant, Fs(5, 432) > 0.26, ps > .09, Wilk’s λ > .98, ${\eta }_{\mathrm{p}}^2$ = .01.

Table 4.

Experiment 2 descriptives.

Variables	Eyewitness instructions		Race		Weapon		Lineup type		Lineup instructions		Overall M (SD)
	None M (SD)	Modified M (SD)	Cross-race M (SD)	Same race M (SD)	Present M (SD)	Absent M (SD)	Single M (SD)	Six Person M (SD)	Biased M (SD)	Unbiased M (SD)
Correct ID	65.59 (24.66)	60.14 (25.80)	62.90 (24.86)	63.13 (25.86)	63.51 (25.71)	62.45 (24.94)	60.92 (26.05)	65.22 (24.40)	62.38 (25.75)	63.67 (24.91)	63.01 (25.33)
Prosecution strength	6.06 (2.27)	5.84 (2.14)	5.86 (2.18)	6.06 (2.24)	6.05 (2.08)	5.85 (2.34)	5.80 (2.19)	6.12 (2.22)	6.10 (2.24)	5.82 (2.18)	5.96 (2.21)
Eyewitness testimony	7.00 (1.78)	6.86 (1.68)	6.87 (1.77)	7.00 (1.70)	7.04 (1.71)	6.81 (1.76)	6.81 (1.77)	7.06 (1.69)	7.01 (1.70)	6.85 (1.77)	6.93 (1.74)
Detective testimony	6.18 (1.84)	5.81 (1.90)	6.00 (1.89)	6.01 (1.87)	5.97 (1.93)	6.04 (1.82)	5.90 (1.98)	6.11 (1.76)	5.93 (1.83)	6.09 (1.93)	6.00 (1.88)
Guilt (continuous)	64.91 (25.13)	61.05 (25.07)	64.05 (23.89)	62.07 (26.40)	63.42 (25.60)	62.71 (24.70)	61.34 (25.38)	64.92 (24.83)	62.41 (25.29)	63.78 (25.04)	63.08 (25.15)
% Guilty	52	49	50	52	51	51	47	54	53	79	51

Note: N = 471–475. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime.

Table 5.

Experiment 2 correlation matrix among dependent variables.

Variable	Verdict	1	2	3	4
1. Correct ID	.63*
2. Prosecution strength	.55*	.61*
3. Eyewitness testimony	.37*	.56*	.53*
4. Detective testimony	.41*	.47*	.48*	.44*
5. Guilt (continuous)	.63*	.85*	.62*	.37*	.41*

Note: N = 471–475. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime. Verdict: 0 = not guilty, 1 = guilty.

*p < .001.

We again evaluated whether jurors are sensitive to identifications made under extremely good or poor conditions. We created a single identification quality variable by collapsing the four eyewitness factors. The logistic regression analysis indicated that the omnibus test with the partial effects of instructions and identification quality along with its interaction on verdict was significant, χ²(3, N = 34) = 7.30, p = .007, Nagelkerke R² = .27. In support of our hypothesis, there was an interaction between instructions and identification quality, Wald χ²(1, N = 34) = 6.47, p = .01, Exp(B) = 56.25. When jurors received modified Henderson instructions, 71% of jurors convicted when identification quality was good, and 25% convicted when identification quality was poor. However, without instructions, 75% of jurors convicted when identification quality was poor, while 29% convicted when identification quality was good (see Table 6).

Table 6.

Experiment 2 sensitivity tests: dependent variables as a function of eyewitness instructions and identification quality.

Dependent variable	Eyewitness instructions
	Absent M (SD)	Present M (SD)
Correct ID
Extremely poor	71.75 (22.62)	51.25 (27.89)
Extremely good	64.00 (16.64)	75.33 (21.27)
Prosecutionstrength
Extremely poor	7.25 (1.60)	6.00 (1.69)
Extremely good	5.29 (2.81)	6.83 (2.64)
Eyewitnesstestimony*
Extremely poor	7.67 (1.30)	6.13 (1.13)
Extremely good	5.57 (1.62)	7.83 (1.17)
Detectivetestimony
Extremely poor	6.50 (2.36)	5.25 (2.12)
Extremely good	6.00 (1.63)	6.33 (1.97)
Guilt (continuous)
Extremely poor	68.25 (23.53)	62.88 (20.87)
Extremely good	58.14 (25.20)	73.83 (24.44)
% Guilty*
Extremely poor	75	25
Extremely good	29	71

 

Note: N = 34. Correct ID: 0–100% likelihood of a correct identification. Prosecution strength: 9-point Likert scale with higher values = more agreement that the prosecution had a strong case. Eyewitness and detective testimony: 9-point Likert scale with higher values = more evidence of guilt. Guilt (continuous): 0–100% likelihood that the defendant committed the crime.

*p < .05.

We followed the logistic regression analysis with a MANOVA to examine all correlated outcomes. The MANOVA also indicated that there was a significant multivariate effect for the interaction between instructions and identification quality, F(5, 25) = 3.71, p = .01, Wilk’s λ = .53, ${\eta }_{\mathrm{p}}^2$ = .43, with univariate effects for perceptions of the eyewitness’s testimony, F(1, 29) = 16.17, p < .001, ${\eta }_{\mathrm{p}}^2$ = .36. Similar to dichotomous verdict decisions, when jurors received modified Henderson instructions, they perceived the eyewitness testimony as more indicative of the defendant’s guilt when identification quality was good (M = 7.83, SD = 1.17) than poor (M = 6.13, SD = 1.13), d = 1.49, 95% CI [0.29, 2.68]. Without instructions, jurors perceived the eyewitness testimony as more indicative of the defendant’s guilt when identification quality was poor (M = 7.67, SD = 1.30) than when it was good (M = 5.57, SD = 1.62), d = 1.47, 95% CI [0.43, 2.52]. While beyond traditional levels of statistical significance, means were in the same direction for all remaining outcomes: likelihood of a correct identification, prosecution strength, detective testimony and continuous judgments of guilt.

Discussion

For Experiment 2, we evaluated the effects of different eyewitness factors on jurors’ decision-making. We manipulated eyewitness identification quality by individually varying the race of the perpetrator, the presence of a weapon, lineup type and lineup instructions. Participants’ perceptions regarding the identification, evidence strength and guilt were unaffected by any of these individual eyewitness factors. In contrast to our hypothesis, the addition of modified Henderson instructions did not improve jurors’ ability to evaluate individual eyewitness factors. These null findings occurred despite increasing the sample size and participation requirements and suggest jurors continue to fail to appreciate the effects of various eyewitness factors on identification accuracy (see also Abshire & Bornstein, 2003; Cutler et al., 1988; Devenport & Cutler, 2004; Neal et al., 2012).

However, unlike Experiment 1, we observed that instructions aided jurors in evaluating more extreme identifications (see Jones & Penrod, 2018; Pawlenko et al., 2013). When we only examined identifications where all four eyewitness factors were either good or poor, an interaction between instructions and identification quality emerged that aligned with our hypothesis. Instructions enabled jurors to distinguish between the quality of these two identifications, with fewer guilty verdicts and less weight given to the eyewitness’s testimony with a poor-quality identification than with a good-quality identification. Jurors did not demonstrate this sensitivity on their own. Without instructions, jurors rendered fewer guilty verdicts and gave less weight to the eyewitness’s testimony when identification quality was good and more guilty verdicts and weight to the eyewitness’s testimony when identification quality was poor. The finding that this assumption was corrected and reversed by the modified Henderson instructions suggests some potential for the curative function of judicial instructions.

Overall, these findings suggest that jurors are hesitant to acquit a defendant when there is eyewitness evidence, even when that evidence is weak. The tipping point for jurors may only occur when an identification is totally flawed and only when they have the assistance of modified Henderson instructions. Otherwise jurors will likely give the benefit of the doubt to the eyewitness and not the defendant.

General discussion

Across two experiments, we sought to examine juror sensitivity to eight individual eyewitness factors. In Experiment 1, we evaluated lighting, distance, lineup administration and lineup composition. In Experiment 2, we evaluated cross-race, weapon presence, lineup type and lineup instructions. Furthermore, in both experiments we varied whether participants received modified Henderson instructions prior to rendering a verdict. Given recent evidence that some modifications to the Henderson instructions induced sensitivity to identifications that were extremely good or extremely poor (Jones & Penrod, 2018), we sought to examine how the instructions affected the evaluation of these eight individual eyewitness factors. Only lineup administration and composition changed perceptions of the detective’s testimony in Experiment 1. These changes did not alter verdict decisions. The results of both experiments suggest that jurors, on their own or when given instructions, did not consider the individual witnessing and identification conditions as they determined whether the identification was accurate and indicative of the defendant’s guilt.

The examination of individual eyewitness factors explores a different set of questions than much of the research on the effectiveness of eyewitness instructions. Typically, participants in these experiments evaluate eyewitness identifications made under extremely good or extremely poor conditions. Most of this research suggests that jurors are not sensitive to witnessing and/or identification conditions on their own (e.g. Cutler et al., 1990; Dillon et al., 2017, concerning system variables; Greene, 1988; Jones et al., 2017, concerning estimator variables; Papailiou et al., 2015). Studies examining individual factors have come to similar conclusions: jurors are not sensitive (as evidenced by changes in verdict) to a variety of individual eyewitness factors, such as the cross-race effect (Abshire & Bornstein, 2003), weapon presence (Cutler et al., 1989; Cutler et al., 1988), double-blind lineup administration (Beaudry et al., 2015; Modjadidi & Kovera, 2018; Wright et al., 2010), lineup composition (Cutler et al., 1989; Cutler et al., 1988; Devenport & Cutler, 2004; cf., Devenport et al., 2002), lineup instructions (Cutler et al., 1989; Cutler et al., 1988; Devenport & Cutler, 2004) and lineup size (Cutler et al., 1989; Cutler et al., 1988). In addition, this is one of the first studies to individually evaluate lighting, distance and lineup type, coming to similar conclusions regarding juror (in)sensitivity. Thus, these results add to the growing conclusion that jurors are unable to evaluate eyewitness evidence on their own.

The finding that jurors were not influenced by the problematic nature of any one eyewitness factor suggests that nuanced identifications are difficult for jurors to evaluate. To address this limitation, courts have utilized several safeguards. We focused on instructions because courts tend to favor this type of safeguard over other more expensive and time-intensive alternatives (e.g. expert testimony; Sheehan, 2011). The Henderson Court surmised that instructions would ‘guide juries about the various factors that may affect the reliability of an identification in a particular case (New Jersey v Henderson, 2011, p. 874)’. However, modified Henderson instructions did not assist in the evaluation of individual eyewitness factors for either experiment.

These null results held in Experiment 1 even when we examined extreme identifications; that is, jurors’ verdicts and related decisions did not vary when lighting, distance, lineup administration and lineup composition conditions were all poor or all good. However, in Experiment 2, we observed that modified Henderson instructions assisted jurors in evaluating two extreme identifications consisting of different eyewitness factors than those in Experiment 1: cross-race, weapon presence, lineup type and lineup instructions. Jurors were less likely to convict and rely on the eyewitness testimony with the poor identification and more likely to convict and rely on the eyewitness testimony with the good identification, but only when they received modified Henderson instructions. The difference in this finding between Experiments 1 and 2 may be due to the different eyewitness factors that were manipulated or, perhaps, because the sample was larger and more restrictive. Future research should disentangle these possibilities.

A peculiar finding of Experiment 2 was that jurors who did not have the assistance of Henderson instructions were more convinced by a poor-quality identification. Wells and Olson (2001) originally suggested that some eyewitness factors generally impair eyewitness performance, such as cross-race, weapon presence, and biased lineup instructions. Such factors do not induce the eyewitness to choose one lineup member over another or to make a choice at all. In contrast, other eyewitness factors, such as the use of a showup, can explain why the suspect was chosen. Jurors may ignore general performance impairment variables as they do not point specifically to the suspect, while suspect-bias variables should increase skepticism of the eyewitness identification (Wells & Olson, 2001; see also Wells & Loftus, 2003). In contrast to these speculations, our findings suggest that jurors may have been moved by the eyewitness’s ability to make an identification despite the poor conditions (three of the four eyewitness factors in this study were general performance impairment variables). The use of the showup may have only served to confirm their suspicions about the defendant’s guilt. Future research should examine these possible explanations to better understand how jurors evaluate individual eyewitness factors, especially those that distinguish between general performance impairment and suspect-bias variables.

Limitations

The use of a convenience sample, brief trial transcript and individual jurors reduced ecological validity. The convenience samples of mTurkers may differ in unknown ways from the population of jury-eligible adults. However, we restricted the sample to adults (18 years and older) that were U.S. citizens and placed a number of restrictions to ensure we received high-quality data (e.g. 95% approval rating and a minimum of 100 HITs; see Peer, Vosgerau, & Acquisti, 2014). We increased these restrictions in Experiment 2 (i.e. Masters status and a minimum of 1000 approved HITs). This change could explain the difference in findings between Experiments 1 and 2, though the results of the manipulation checks suggests this is unlikely. We also presented jurors with a brief trial transcript (1827–2075 words), yet our results largely overlap with those using more ecologically valid stimulus materials, such as re-enacted trials (Dillon et al., 2017; Jones et al., 2017). Furthermore, a recent meta-analysis suggests little differences in outcomes across various stimulus materials (Bornstein et al., 2017).

Finally, we relied on individual jurors because we conducted both experiments online. There are several psychological processes inherent in deliberations that are absent in studies examining juror decision-making (e.g. group recall, see Salerno & Diamond, 2010). Therefore, we are unable to determine whether the findings here, especially regarding modified Henderson instructions, would replicate with jury deliberation. However, a few studies have examined jury sensitivity with and without eyewitness instructions (Bergold, Jones, Dillon, & Penrod, in press; Greene, 1988). The findings of these studies are similar to those using individual jurors; that is, the instructions result in null or skepticism effects among juries. Given the replication crisis in psychology (see Pashler & Wagenmakers, 2012), the continued examination of jurors’ ability to evaluate eyewitness evidence improves the overall psychological study of juror decision-making and highlights the need for safeguards to effectively address jurors’ shortcomings.

Conclusions

The findings across both experiments suggest that eyewitness factors are once again ‘beyond the ken’ of jurors (Desmarais & Read, 2011). Jurors have difficulty evaluating eyewitness evidence when the quality of estimator and system variables is mixed. This finding is concerning given that this is likely to reflect the conditions underlying actual eyewitness evidence. Even when an identification is extremely poor or good due to lighting, distance, lineup administration and lineup composition conditions, jurors appear unable to adjust their verdict and other related decisions, regardless of modified Henderson instructions. However, the addition of modified Henderson instructions appears to assist jurors in evaluating extreme identifications that vary simultaneously by cross-race, weapon presence, lineup type, and lineup instructions. Future research should examine potential mechanisms to tease this finding apart.

These and other studies (e.g. Dillon et al., 2017; Jones et al., 2017; Papailiou et al., 2015) continue to demonstrate that courts should be wary of allowing unreliable eyewitness evidence to be presented to juries, as it is difficult for jurors to properly evaluate such evidence. Elevating the admissibility standards for identification evidence may also be necessary because the addition of instructions may only help under limited circumstances. Two of the recommendations by the Special Master in New Jersey v Henderson (2011) speak directly to this concern. First, the Special Master recommended requiring a pre-trial admissibility hearing to evaluate all relevant estimator and system variables in a case. Second, the Special Master recommended that the State bear the burden of establishing that an eyewitness identification is reliable. The Henderson Court did not adopt these recommendations, but accumulating research in this area suggests that these changes may be needed. Until then, we encourage future researchers to explore new avenues for educating jurors about eyewitness identifications.

Appendix. Modified Henderson instructions.

The court instructs you as follows

Eyewitness identification evidence must be scrutinized carefully. Research has shown that there are risks of making mistaken identifications. Human memory is not foolproof. Research has revealed that human memory is not like a video recording that a witness need only replay to remember what happened. Memory is far more complex.

To decide whether the identification testimony is sufficiently reliable evidence to conclude that this defendant is the person who committed the offense charged, you should consider the following factors that are related to the witness, the alleged perpetrator, and the criminal incident itself. In particular, you should consider:

1. The Witness’s Opportunity to View and Degree of Attention: In evaluating the reliability of the identification, you should assess the witness’s opportunity to view the person who committed the offense at the time of the offense and the witness’s degree of attention to the perpetrator at the time of the offense. In making this assessment you should consider the following:

   1. Stress: Even under the best viewing conditions, high levels of stress can reduce an eyewitness’s ability to recall and make an accurate identification. Therefore, you should consider a witness’s level of stress and whether that stress, if any, distracted the witness and made it harder for him or her to identify the perpetrator.
      Ask Yourself: Was the eyewitness stressed? Did this enhance or impair her accuracy?
   2. Duration: The amount of time an eyewitness has to observe an event may affect the reliability of an identification. Although there is no minimum time required to make an accurate identification, a brief or fleeting contact is less likely to produce an accurate identification than a more prolonged exposure to the perpetrator. In addition, time estimates given by witnesses may not always be accurate because witnesses tend to think events lasted longer than they actually did.
      Ask Yourself: For how long did the eyewitness view the perpetrator? Did this enhance or impair her accuracy?
   3. Weapon Focus: You should consider whether the witness saw a weapon during the incident and the duration of the crime. The presence of a weapon can distract the witness’s attention away from the perpetrator's face. As a result, the presence of a visible weapon may reduce the reliability of a subsequent. In considering this factor, you should take into account the duration of the crime because the longer the event, the more time the witness may have to adapt to the presence of the weapon and focus on other details.
      Ask Yourself: Did the perpetrator have a weapon? Did this enhance or impair her accuracy?
   4. Distance: A person is easier to identify when close by. The greater the distance between an eyewitness and a perpetrator, the higher the risk of a mistaken identification. In addition, a witness’s estimate of how far he or she was from the perpetrator may not always be accurate because people tend to have difficulty estimating distances.
      Ask Yourself: How far apart were the perpetrator and the eyewitness? Did this enhance or impair her accuracy?
   5. Lighting: Inadequate lighting can reduce the reliability of an identification. You should consider the lighting conditions present at the time of the alleged crime in this case.
      Ask Yourself: What were the lighting conditions during the crime? Did this enhance or impair her accuracy?
   6. Intoxication: The influence of alcohol can affect the reliability of an identification. An identification made by a witness under the influence of a high level of alcohol at the time of the incident tends to be less reliable than an identification by a witness who drank a small amount of alcohol or none at all.
      Ask Yourself: Was the eyewitness intoxicated? Did this enhance or impair her accuracy?
   7. Disguises/Changed Appearance: The perpetrator’s use of a disguise can affect a witness’s ability both to remember and identify the perpetrator. Disguises like hats, sunglasses, or masks can reduce the accuracy of an identification. Similarly, if facial features are altered between the time of the event and a later identification procedure, the accuracy of the identification may decrease.
      Ask Yourself: Was the perpetrator wearing a disguise or did their appearance change? Did this enhance or impair her accuracy?

2. Prior Description of Perpetrator: Consider the accuracy of any description the witness gave after observing the incident and before identifying the perpetrator. Did the prior description match the photo or person picked out later? Did the prior description provide details or was it just general in nature? Was the witness's testimony at trial consistent with, or different from, his/her prior description of the perpetrator?

   Ask Yourself: Did the eyewitness’ prior description of the perpetrator match the defendant? Does this suggest the identification is more or less accurate?

3. Time Elapsed: Memories fade with time. As a result, delays between the commission of a crime and the time an identification is made can affect the reliability of the identification. The more time that passes, the greater the possibility that a witness’s memory of a perpetrator will weaken.

   Ask Yourself: How long was it between the crime and the lineup identification? Did this enhance or impair her accuracy?

In evaluating the reliability of a witness’s identification, you should also consider whether it was the result of a suggestive procedure. In that regard, you may consider everything that was done or said by law enforcement to the witness during the identification process including:

1. Lineup Composition: A suspect should not stand out from other members of the lineup. The reason is simple: an array of similar-looking individuals forces witnesses to examine their memory. In addition, a biased lineup may inflate a witness’s confidence in the identification because the selection process seemed so easy to the witness. It is, of course, for you to determine whether the composition of the lineup had any effect on the reliability of the identification.

   Ask Yourself: Did all or nearly all of the pictures/lineup members match the eyewitness’ description of the perpetrator? Did this strengthen or weaken the evidence against the defendant?

2. Fillers: Lineups should include a number of possible choices for the witness, commonly referred to as ‘fillers.’ The greater the number of possible choices, the more likely the procedure will serve as a reliable test of the witness’s memory. A minimum of six persons or photos, each of whom could reasonably be the perpetrator, should be included in the lineup.

   Ask Yourself: How many reasonable candidates were in the lineup? Did this strengthen or weaken the evidence against the defendant?

   In determining the reliability of the identification, you should also consider whether the identification procedure was properly conducted.

   1. Double-blind: A lineup administrator who knows which person or photo in the lineup is the suspect may intentionally or unintentionally convey that knowledge to the witness. That increases the chance that the witness will identify the suspect, even if the suspect is innocent. For that reason, whenever feasible, live lineups and photo arrays should be conducted by an officer who does not know the identity of the suspect.
      Ask Yourself: Might the officer conducting the lineup do anything to suggest to the eyewitness which person or picture to choose, or not to choose? Did this strengthen or weaken the evidence against the defendant?
   2. Instructions: You should consider what was or what was not said to the witness prior to viewing a lineup or showup. Identification procedures should begin with instructions to the witness that the perpetrator may or may not be present and that the witness should not feel compelled to make an identification. The failure to give this instruction can increase the risk of misidentification. If you find that the police did or did not give this instruction to the witness, you may take this factor into account when evaluating the identification evidence.
      Ask Yourself: Did the officer conducting the lineup tell the eyewitness that the perpetrator may or may not be present in the lineup? Did this strengthen or weaken the evidence against the defendant?
   3. Confidence and Accuracy: The witness made a confidence statement at the time the identification was made. Although some research has found that highly confident witnesses are more likely to make accurate identifications, eyewitness confidence is not a perfect indicator of accuracy.
      Ask yourself: How much weight should I give the confidence statement?
   4. Feedback: Feedback occurs when police officers, or witnesses to an event who are not law enforcement officials, signal to eyewitnesses that they correctly identified the suspect. That confirmation may reduce doubt or produce a false sense of confidence in a witness. Feedback may also falsely enhance a witness’s recollection of the quality of his or her view of an event. It is for you to determine whether or not a witness’s recollection in this case was affected by feedback or whether the recollection instead reflects the witness’s accurate perception of the event.
      Ask Yourself: Did the police do anything that could inflate the eyewitness’ confidence? Does this strengthen or weaken the usefulness of the eyewitness’ confidence statement?

Keep in mind that the presence of any single factor or combination of factors is not proof that a particular witness is incorrect. Instead, you should consider the factors that I have discussed as you assess all of the circumstances of the case, including all of the testimony and documentary evidence, in determining whether a particular identification made by a witness is accurate and thus worthy of your consideration as you decide whether the State has met its burden to prove identification beyond a reasonable doubt.

Funding Statement

This work was supported by the National Science Foundation [grant number 1228497].

Ethical standards

Declaration of conflicts of interest

Angela Jones has declared no conflicts of interest

Amanda Bergold has declared no conflicts of interest

Steven Penrod has declared no conflicts of interest

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed consent

Informed consent was obtained from all individual participants included in the study