EXECUTIVE SUMMARY

I. Expertise of Witness

I am …

I have been asked by counsel to examine portions of the [NAME] case file in order to identify factors that could have contributed to a misidentification by the witnesses in this case and about which I would be prepared to testify at the trial of defendant. In my report I note the following:

II. The Problem of Eyewitness Identification Errors (p. 8)

Legal scholars have studied the causes of mistaken identification in over 1,000 criminal cases and conclude that the single leading contributing cause of mistaken conviction established by DNA is erroneous eyewitness identifications. In the first-200 DNA exculpation cases the leading cause of the wrongful convictions was erroneous eyewitness identification--which occurred in 79 percent of the cases and more than a third of the misidentification cases had two or more witnesses who made misidentifications (13% had three or more misidentifications). The prominence of mistaken IDs in DNA cases continues to this day with mistaken eyewitness identification contributing to 72 percent of the 337 wrongful convictions established by post-conviction DNA testing. http://www.innocenceproject.org/causes-wrongful-conviction

III. The Growing Body of Scientific Research on Eyewitness Reliability (pp. 8-9)

The research on eyewitness reliability is of relatively recent vintage. The volume of research has generally more than doubled each decade since the 1960s.

IV. The Need for Expert Testimony (pp. 9-13)

Cross-examination is sometimes hailed as a powerful engine for determining truth, but is this true in eyewitness cases? Effective cross-examination requires that juries are sensitive to the factors that influence eyewitness identification accuracy. Desmarais & Read (2010) review studies of lay knowledge about factors influencing eyewitness performance and report it is common to find that 50-65% of respondents chose answers similar to those of experts. While this performance may look relatively good it generally indicates there are significant disagreements among jurors and, to some extent, the seeming agreement with experts likely arises from “good” guessing by laypersons. There is, in addition, evidence that cross-examination may impair juror assessments of eyewitness evidence by reducing the accuracy of eyewitness testimony (Valentine & Maras, 2011).

Most relevant to the question of juror sensitivity are studies which attempt to simulate the jury's actual task of evaluating eyewitness identifications. Studies of mock-jurors' abilities to differentiate between accurate and inaccurate eyewitnesses converge on a dismal conclusion about jurors' abilities. Jurors overestimate the accuracy of identifications (there are more convictions than there are accurate identifications), jurors fail to distinguish accurate from inaccurate eyewitnesses, jurors tend to ignore viewing conditions that are known to predict identification accuracy and instead base their decisions in part on eyewitness memory for peripheral details and witness confidence--both of which tend to be poor predictors of identification accuracy. Across cases in which a number of forms of evidence have been systematically manipulated (e.g. disguise, weapon focus, level of violence, retention interval, instruction bias, foil bias, and witness confidence) Only confidence influenced verdicts despite the fact that studies reveal that eyewitness confidence and identification accuracy are only modestly related.

1

V. Eyewitness Expert Testimony Helps Jurors (pp. 14-17)

Studies including Cutler, Dexter & Penrod, (1989) and Cutler, Penrod & Dexter, (1989) have shown that expert testimony can increase juror sensitivity to variations in trial evidence such as those noted above. Without such testimony, jurors reveal poor knowledge about eyewitness problems.

VI. Research Findings and Expert Testimony Meet Admissibility Criteria (pp. 18-21)

The eyewitness research community (which includes several researchers who have dual training in law and psychology) has been sensitive to the question of the admissibility of testimony about eyewitness research findings. Does this research meet the scientific standards required by the courts? One can address this question from multiple perspectives. Under the traditional Frye standard the relevant question would be whether the testimony is generally accepted within the relevant scientific community Under the Federal Rules of Evidence and Daubert considerations such as whether: (a) the expert is qualified; (b) the testimony assists the trier of fact; (c) the expert's testimony is sufficiently reliable, (d) the materials about which the expert testifies are the product of the scientific method (including falsifiable theories, peer-review).

Virtually all of the empirical eyewitness research conducted by psychologists makes use of standard experimental methods employed in all the experimental sciences. A hallmark of this research is the falsifiability/testability of research hypotheses that are tested using experimental research methods in which the variables or processes examined by researchers are carefully controlled by the researchers in order to assess their causal effects on outcome variables (such as identification accuracy). Use of appropriate research methods is an essential requirement for publication in peer-reviewed scientific journals across all scientific disciplines and psychology is no exception. The results and conclusions summarized below are the products of precisely the methods underscored by the Supreme Court in Daubert.

There is a consensus on the content of expert testimony. Kassin, Tubb, Hosch, & Memon (2001) extended prior research with experts and found levels of agreement among 64 eyewitness experts. Recently Daftary-Kapur (2010, 2013) and Penrod surveyed 71 leading eyewitness researcher (on average the experts had published 18.51 scientific journal articles, 2.9 books, 8.3 book chapters, 3.7 law review articles, 5.3 magazine articles, and 3.0 newsletter articles. Experts had testified in over 1600 cases. They found even higher levels of agreement among experts.

VII. Subject Matters for Expert Testimony in the Present Case (pp. 17-20)

Face Recognition is Far from Perfect Even Under Optimal Testing Conditions. Megreya & Burton (2008) conducted a study in which participants were shown a live person for 30 seconds and were then tested, from memory, on a 10-person photoarray—when the target was in the array he was identified by 70% of the participants and a non-target person was identified by 10% of participants. When the target was not in the array, 20.5% misidentified someone else. When participants were shown the live person and the photoarray together (a matching task rather than a memory test), the target was identified by 66.9% of the participants and a non-target person was identified by 15% of participants. When the target was not in the array, 37.8% misidentified someone else.

Field Experiments Testing Witness Memory Show High Rates of Witness Identification Errors. One relevant source of data pertaining to accuracy rates of actual eyewitness identifications emerges from realistic field studies of eyewitness identification. Across several studies reviewed below, the average correct identification-rate from presentations which included the target person was 41.8%. Thus, nearly 60%

2

of witnesses failed to identify the target when he was present. Unfortunately, the false identification rate of innocent foils was nearly as high as the rate of guilty-target identifications--35.8%. In short, identification errors were rampant (Valentine, 2008).

Studies of Real Witness Performance Using Police Archives Show High Rates of Identification Errors. Similarly low accuracy rates from actual eyewitness identifications emerge from studies of actual witnesses. In these studies it is not known whether the suspect is the actual perpetrator, but it is still possible to gauge the rate of inaccurate identifications of foils—the known-innocents placed in arrays along with the suspects. The best data come from studies in the United Kingdom—the results from nearly 17,000 actual eyewitnesses shown that nearly 40% of positive identifications were identifications of an innocent foil—which underscores that many witnesses are willing to guess and they make errors at a high rate. See, e.g., Valentine, 2008.

The “Pleading Effect” Elevates the Portion of Trial Defendants Who Have Been Misidentified. Research indicates people charged with crimes plead guilty in a high percentage of cases—(Liptak [2011] reports the rate was 97% in federal courts in 2010 and 94% in state courts in 2006). In contrast Garrett (2008) reports that just 4% of the defendants in the 200 DNA-exonerated cases he studied entered a guilty plea. The University of Michigan Exoneration Registry--with over 1700 cases as of late 2015--indicates 13% of their defendants pled guilty http://www.law.umich.edu/special/exoneration/Pages/Recent-Findings.aspx. An archival analysis (Kellstrand, 2006) of cases with both DNA and eyewitness identification evidence in the San Diego County District Attorney’s Office, reported that the individuals suspected of crimes were innocent in only 5% of cases. A little algebra demonstrates that an overall 95% plea rate combined with a 5% eyewitness error rate and a13% plea rate by innocent defendants results in eyewitness trials in which in a majority of defendants are innocent!

VIII. Witnessing and Crime Factors that Influence the Accuracy of Identifications (pp. 20-26)

Weapon Focus. Weapon focus refers to the attention witnesses give to a perpetrator’s weapon during a crime. It is expected that the attention the witness focuses on a weapon will reduce their ability to later recognize the perpetrator. Researchers have assessed eyewitness accuracy in an attempt to assess the effects of weapon-focus effects on memory; the relevant studies were reviewed in a meta-analysis by Steblay (1992). The meta-analysis included 19 studies with a total sample of 2,082 participants. The weapon-focus effect on identifications was statistically significant. A recent dissertation study at John Jay College (by an active duty Connecticut police chief John DeCarlo, 2010) illustrates the effect. DeCarlo used a videotaped robbery and found that in a no weapon condition, witnesses were able to correctly identify the target 78% of the time. When a weapon was implied by the perpetrator waving his hands around in his pocket, accuracy dropped to 55% and when a weapon was actually shown, accuracy dropped to 33%. When the perpetrator was absent from the lineup the correct rejection was 89% for the no weapon condition, but when a weapon was implied, accuracy dropped to 76% and when a weapon was actually shown, correct rejections dropped to 65%. A new meta-analysis (Fawcett, et al., 2013) confirms these conclusions using data from 47 comparisons and further notes that the size of the effects was “unaffected by whether the event occurred in a laboratory, simulation, or real-world environment.”

Stress. Deffenbacher, et al. (2004) published a meta-analysis of stress effect studies. The meta-analysis was conducted on 27 tests of the effects of heightened stress on identification accuracy. This sample included work published between 1974 and 1997, with a total of 1727 participants involved in relevant tests of the stress-- the mean proportions correct for TP lineups under high and low stress conditions were .39 and .59, respectively (with false alarm rates of 34% and 19% respectively). The effect of stress was larger for target-present than for target-absent lineups—that is, stress particularly reduced correct identification rates. The

3

effect was twice as large for eyewitness-identification studies that simulated eyewitness conditions (e.g., staged crimes) than for studies that induced stress in other ways.

These effects are confirmed and extended in a study by Morgan et al. (2004) who examined the eyewitness capabilities of more than 500 active-duty military personnel enrolled in a survival-school program participants experienced both a high stress interrogation with real physical confrontation and a low-stress interrogation without physical confrontation. Both interrogations were 40 minutes long; they were conducted by different persons. A day after release from the camp the participants were tested on their ability to recognize the interrogators—recognition accuracy for the low-stress interrogators was as high as 76% but as low as 27% for high-stress interrogators.

A recent confirmation of these findings comes from a study by Valentine & Mesout (2008) conducted in the “Horror Labyrinth” of the London Dungeon tourist attraction. Tourists encountered a “scary person” while slowly walking around the labyrinth for approximately 7 minutes. About 45 minutes later, after they completed their tour they were tested to see if they could identify the scary person from a nine-person photo-array. Participants who were divided into two groups reflecting the 50% with the highest reported anxiety produced by the event and the 50% with the lowest scores--75% of the witnesses experiencing lower anxiety were able to identify the scary person/culprit but only 18% of those with higher anxiety.

Tracking the Identity of Multiple Perpetrators. Just as a weapon can divide the attention of a witness, multiple perpetrators can divide their attention and identification accuracy and accuracy of descriptions have been shown to decline as the number of perpetrators increases (Clifford and Hollin, 1981; Fahsing, Ask, and Granhag, 2004; Megreya, et al 2006; Shepherd, 1983). This impairment effect was clearly illustrated in the dissertation research by DeCarlo (2010) who showed his witnesses a videotaped robbery involving one or two perpetrators. When the perpetrator was present in the lineup, the single perpetrator condition produced a correct identification rate of 55% versus 33% in the multiple perpetrator condition. In the multiple perpetrator condition, 82% of witnesses correctly rejected perpetrator-absent arrays versus 92% in the single perpetrator condition.

Exposure Duration and Time Estimation. The time available for viewing a perpetrator is positively associated with the witness's ability to subsequently identify him. Shapiro and Penrod's (1986) study of more than 100 experiments showed that exposure time was a reliable predictor of accuracy. However, it should be noted that even a long exposure is no guarantee of accuracy—in the Morgan, et al. (2004) study the interrogations of soldiers whose memory was later tested lasted 40 minutes and yet the rate of correct identifications among high-stress soldiers ranged from only 27 to 49% when targets were present and foil identification rates approached 50% using standard lineups and photospreads.

Although we often have to depend on witness estimates of viewing times, this can be problematic--in his 2000 study Yarmey found that witnesses to non-routine events were, on average, overestimated by between 25% (for an event lasting 13 minutes) and more than 100% (events lasting 24 seconds or less). An event lasting 46 seconds was overestimated by 67% and events lasting one minute 20 seconds and two minutes 58 seconds were over-estimated by 40%.

Obscured Views of Faces/Disguise. Regarding obscured views of faces, there is research (Mansour, et al., 2012) on the effects of stockings pulled over the head (and the effects of a variety of other disguises). In the stocking-view as compared to no-stocking views, correct identifications dropped from 80% to 55% and false identifications rose from 21% to 36%.

4

Full face masks stockings, hats and hoods can be quite effective in diminishing the facial feature cues that are necessary for recognition. Cutler and colleagues (Cutler, Penrod, & Martens, 1987a, 1987b; Cutler et al., 1986; O'Rourke et al., 1989) examined the effects of masking a target's hair and hairline cues on subsequent identification accuracy. In these experiments participants viewed a videotaped liquor store robbery and later attempted an identification from a videotaped lineup. In half of the robberies the robber wore a knit pullover cap that covered his hair and hairline. In the other half the robber did not wear a hat. The robber was less accurately identified when he was disguised. For example, in one of the experiments (Cutler et al., 1987a) 45% of the participants gave correct judgments on the lineup test if the robber wore no hat during the robbery, but only 27% gave a correct judgment if the robber wore the hat during the robbery.

More extreme disguises can produce even more substantial impairment. This is illustrated in a study by Fisher and Cox (1995) where celebrities, known to the research participants, were identified by only 12% of participants when shown only the celebrities’ eyes (mouths alone produced recognition by 3% of participants). Of course it is important to emphasize that these participants were endeavoring to identify persons known to them through multiple exposures over extended periods of time—not the faces of strangers seen briefly.

Identifications of Voice and Body. How much do voice and body add to the identification picture? Pryke, et al. (2004) had witnesses attempt identifications based on face, body, or voice. The target was presented to participants and spoke for approximately 1 minute. Participants then attempted identification from faces, then voices , then from a lineup of bodies from the shoulder down. Correct face identifications were obtained from 74% of participants in the target present condition and the most-selected innocent face was selected by 20% of participants in the absent condition.. For the body lineup, 53% of participants correctly identified the target and 81% falsely identified the most-selected innocent person (39% of choices guilty). For the voice lineup, 40% of participants correctly identified the target and 35% falsely identified the most-selected innocent person (53% guilty). Pryke, et al. also note that although nearly 75% of witnesses successfully chose the target from the facial lineup, only 55% of witnesses selected the target from the facial lineup and either or both of the voice and body lineups.

Yarmey (2007) notes that longer opportunities to listen to a speaker yield greater the accuracy of identification. In Yarmey (1991) longer exposures produced higher correct identification rates (24% accuracy in a 6-voice lineup with an average of 3.2 minutes exposure, 30% with 4.3 minutes and 48% with 7.8 minutes). Mistaken identification rates in target-absent voice lineups were relatively high, averaging 48% across exposure times. Yarmey and Matthys similarly found relatively poor accuracy with speech sample durations of 36 seconds or less (31%) compared to samples of 2 minutes or longer (55%).

(Mis-)Recognition of Familiar Faces. Research shows that, counter-intuitively, identifications of familiar faces - as with identifications of unfamiliar faces - are influenced by a host of variables, including interaction time, contextual information, expectation, postevent information, and own-race bias. ... these variables simultaneously increase the probability that an individual will identify a given face as familiar and inflate the individual's confidence in the identification, regardless of whether the face in fact is familiar or the identification accurate.

A particularly relevant study of identifications of non-strangers is by Pezdek & Stolzenberg (2014). They recruited sophomores (N 139) from two small private high schools who viewed forty yearbook pictures including (a) twenty students who graduated from their school as seniors (fourth year) when the research participants were freshmen (first year students) (familiar faces—with whom the participants had overlapped for a full school year) and (b) twenty unfamiliar individuals drawn from the senior class of the other high

5

school. Thus, familiar faces at one high school served as unfamiliar faces at the other high school. Participants were asked to indicate whether each face was ‘familiar.’ Individuals’ familiarity judgments were only modestly related to prior contact--accuracy was fairly low (the correct identification rate for familiar faces was 42% while the false alarm rate for unfamiliar faces was 23%. This meant that over one-third of the identifications of “familiar” people were actually misidentifications of strangers.

Witness Intoxication. Dysart, et al (2002) note that popular belief is that, due to mental impairment, intoxicated witnesses are less accurate than sober witnesses. However, one theory concerning “alcohol myopia” posits that, relative to intoxicated witnesses, sober witnesses will encode more information/cues about the perpetrator that will facilitate correct rejections in target-absent procedures. Intoxicated witnesses are likely to encode only salient cues from the target and erroneous identifications will result where more subtle cues would have indicated that the suspect was not the target. Dysart, et al., as predicted, found that in the target-present showup condition, blood alcohol level was not significantly related to correct identification rate, but in the target-absent condition, higher blood alcohol levels were associated with a higher likelihood (52%) of a false identification than were lower blood alcohol levels (22%). Similarly, in a dissertations study by King (2006) participants were shown 6 target photos and were tested for recognition of 2 target faces at 1 hour post exposure, 6-hrs post exposure, and 24-hrs post exposure. The alcohol group produced an increased number of false identifications in target absent lineups while there were no significant differences in performance when the target was present.

Unconscious Transference. This term is applied when there is a “transfer of one person’s identity to that of another person from a different setting, time, or context” (Read et al., 1990, p. 3). Williams (1963) originally coined the phrase in reference an English murder case, in which an innocent person may have been executed. Williams indicates that one of the eyewitnesses who identified the defendant had briefly seen him before the crime and may have unconsciously transferred the identity of the actual perpetrator to the defendant. A classic study of the confusion of roles was conducted by Buckhout (1974) who staged a mock assault before a group who were later asked to pick the assailant from a group of six photographs including the assailant, an innocent bystander to the assault and four fillers, 40% were able to select the assailant and 25% selected the person who had been at the scene as an innocent. Read et al. (1990) had participants view an event with or without a bystander and had them attempt to identify a target person from a photoarray containing a bystander (rather than the target) and four unfamiliar foils. Twenty-five percent of the witnesess who had seen the bystander misidentified the bystander, as compared with just 12 percent of the control subjects

In their 2006 meta-analysis Deffenbacher, Bornstein and Penrod reviewed eight studies using these types of designs and found that the effect is a reliable one with an overall transference misidentification rate of 27% for previously-seen bystanders versus 17% for the same individuals when they had not been viewed as bystanders. dropped from 80% to 55% and false identifications rose from 21% to 36%.

Cross-Race Identification. Research on cross-race identification impairment began forty years ago and has included various mixes of Caucasian, Asian, Hispanic, Black, and middle-eastern witnesses. Meissner & Brigham (2001) have reviewed research on the problems of what have sometimes been called other-race or cross-race identifications or own-race bias (ORB). Meissner and Brigham analyzed data from 39 research articles, with 91 independent samples involving nearly 5,000 witness participants. Measures of correct identifications and false alarm rates, and aggregate measures of discrimination accuracy and response criterion were examined. Overall, they reported that when the perpetrator is present the ratio of correct to incorrect identifications was 40% higher for same-race identifications. The ratio of mistaken identifications to correct rejections in target-absent arrays was 56% greater for other-race identifications. Overall, the ratio

6

of correct to incorrect identifications was more than 2.2 greater for own-race faces as compared with performance on other-race faces.

Police versus Civilian Witnesses. Although there indications that police are better at remembering critical details in crime situations, the evidence also indicates police are no better or just slightly better than civilians when making person identifications--four out of five comparisons I am aware of yielded no statistically-significant advantage for police.  However, if you average performance across those studies, there is a slight advantage for police in making correct identification when the perpetrator is present (police 8% better) and a slight advantage for police when the perpetrator is absent (police 5% better). (Christianson, et al., 1998--2 comparisons; DeCarlo, 2010—2 comparisons; Lindholm, et al., 1998).

IX. Aspects of Identification Procedures that Affect Identification Accuracy (pp. 26-33)

Lineup Instruction Bias. Instructions given to an eyewitness prior to an identification test can vary in their degree of suggestiveness. Suggestive instructions strongly convey to the eyewitness the impression that the suspect is in fact in the photoarray or lineup, thereby increasing the likelihood that the eyewitness will make a positive--though not necessarily correct--identification. How can instructions convey this message?

Malpass & Devine (1981a) staged an act of vandalism during a lecture to about 350 undergraduate students, 100 of whom were asked to identify the vandal from one of two live lineups within the next three days. Half of the eyewitnesses were given instructions suggesting the perpetrator was in the array and the remaining eyewitnesses were given an "unbiased" instruction: "The person . . . may be one of the five individuals in the lineup. It is also possible that he is not in the lineup.” Among eyewitnesses who viewed a vandal-absent lineup, 78% of those who received biased instructions made a positive identification. Of course, all of whom were incorrect. In contrast, only 33% of those who received unbiased instructions made an incorrect positive identification from the vandal-absent lineup. Thus, significantly more false identifications were obtained with biased instructions than with neutral instructions. Instruction bias research was reviewed by Steblay in a 1997 meta-analysis of 22 studies involving nearly 2600 witness-participants. She found that biased instructions were particularly harmful in target-absent lineups in which witness accuracy declined from 60% (unbiased lineups) to 35% (biased lineups).

Foil Bias. The term "functional size" (Lindsay & Wells, 1980) refers to the number of viable lineup members, or the number of lineup members who plausibly match the eyewitness's description of the crime perpetrator.. The quality and the number of foils in an array clearly influence the fairness of the array--as reflected in the tendency for witnesses to make identifications, particularly false identifications.

It is clear from research on lineup and photoarray fairness that police presentations frequently fall short of being fair. One technique for making such assessments uses the “mock witness” method in which the witness’s original description of the perpetrator is presented to 30 or more “mock witnesses” together with the photoarray or picture of the lineup in question. Each witness is asked make an identification based on the description and one can divide the number of mock witnesses by the number of suspect identifications to gauge the effective size of the array (e.g. in a fair 6-person array 5 of 30 witnesses would be expected to select the suspect: 30/5 = 6…the effective size of the array). Studies of actual arrays show that suspects are picked 2 to 3 times the rates one would expect from a fair array (Brigham et al. 1990; Brigham et al. 1999; Wells & Bradfield, 1999—see Valentine & Heaton (1999) and Py (2003) for similar findings involving British and French arrays.

7

Assessing the Probative Value of Identifications, Non-Identifications and Mis-Identifications. As early as 1980 Wells and Lindsey questioned the criminal justice system's policy of treating eyewitness identifications of suspects as highly probative while treating non-identifications (both in the form of no-choices and foil choices) as non-probative. Wells and Lindsey used a Bayesian model of information to demonstrate that if an identification of a suspect increases the probability the suspect is the perpetrator, then a nonidentification or identification of someone else must reduce that probability. Wells and Olson (2002) carried the Bayesian analysis further by examining the information gain (changes in the probability the suspect is the perpetrator) as a function of the prior probability the suspect is guilty (based on evidence prior to the identification results). Filler identifications, non-identifications and “don’t know” responses all have probative value with respect to the suspect’s innocence and sometimes filler identifications and non-identifications have more probative value (with respect to innocence) than positive identifications of the suspect have with respect to guilt.

Retention Interval. Common sense tells us that memory declines over time. Can we expect eyewitness identification accuracy to decline as the time between the crime and the identification test increases? Shapiro and Penrod included retention interval in their 1986 meta-analysis. When studies that manipulated retention interval were grouped into long versus short time delays, longer delays led to 10% fewer correct identifications and 8% more false identifications. Classic research by Egan, Pittner, and Goldstein (1977) found that the percentage of mock witnesses making false identifications increased significantly over time: 48% at 2 days, 62% at 21 days, and 93% at 56 days. The percentage of subjects making no errors declined from 45% (2 days) to 29% (21 days) to 7% (56 days).

Bornstein, et al. (2012) demonstrate that loss of eyewitness memory is most rapid early on…in the first minutes and hours. The forgetting curves/loss of memory which emerges from these studies shows that memory has typically declined by 15%-20% within 2 hours, with a further 4-5% drop in the next 10 hours. From that point forward the loss is less dramatic…though it continues—as is evident in the individual studies cited above.

Simultaneous Presentation Bias. Nearly twenty years of research indicates that sequential lineups can cut the rate of false identifications quite substantially. Steblay, Dysart & Wells (2011) report a meta-analysis with 72 tests of simultaneous and sequential lineups involving over 13,000 witnesses. In 27 studies that directly compare sequential/simultaneous using both target-present and target-absent arrays, witnesses were substantially less likely to make a false identification from a target-absent array (32% mistaken identifications) than from simultaneous arrays (54% mistaken identifications). Sequential lineups also yielded significantly fewer false identifications of an innocent suspect compared to simultaneous lineups (15% vs. 28%).

Blind Presentation. Greathouse & Kovera (2009) manipulated whether a lineup administrator had knowledge of the suspect’s identity, the type of line-up (simultaneous vs. sequential), the presence of the actual perpetrator in the line-up and the type of line-up instructions (biased vs. unbiased). When the witnesses received biased instructions and simultaneous line-ups, they were more likely to make suspect identifications in non-blind than in blind lineups. The pattern of filler and suspect identifications suggested that the increase in mistaken identifications was the result of non-blind administrators influencing those who would have, under blind conditions, made filler identifications to make suspect identifications instead. Line-up rejections did not significantly increase or decrease as a function of line-up administrator knowledge. Suspect identifications were twice as diagnostic for blind administrations as they were for non-blind administrations.

8

In a more pronounced demonstration of the effects that a non-blind administrator can have, Alberts, Duncan, Wallace and Penrod (2008) trained administrators to “steer” witnesses to make positive identifications of suspects (both guilty and innocent suspects) and avoid identifications of non-suspects (foils) and to accomplish this steering without arousing the suspicion of witnesses. Steering was highly successful: selection rates for “steered” witnesses were: 57% guilty suspects, 32% innocent suspects and 19% filler identifications versus 18% guilty suspects, 10% innocent suspects and 39% filler identifications. Witnesses were essentially unaware that they had been influenced.

X. Witness Confidence and Witness Accuracy (pp. 36-40)

The problems with respect to witnesses’ confidence about the accuracy of their identifications and the actual accuracy of those identifications are manifold:

Jurors Infer Accuracy from Confidence. There is consistent evidence to indicate that the confidence that an eyewitness expresses in his or her identification during testimony is the most powerful single determinant of whether or not observers of that testimony will believe that the eyewitness made an accurate identification

Post-Identification Confidence, if Measured Properly, is Modestly Correlated with Accuracy. Perhaps the most informative study is by Sporer, Penrod, Read & Cutler (1995) who found a confidence-accuracy correlation of .41 among (forensically-relevant) choosers. These finding suggests that witnesses who are highly confident in their identifications are somewhat more likely to be correct as compared to witnesses who display little confidence.

Witnesses are also Over-Confident. Brewer, Keast, & Rishworth (2002) found that eyewitnesses who were very confident of the accuracy of their identifications (95% certain) were only about 70%-75% correct. In a 1987 study by Fleet, Brigham, & Bothwell 75% of Ss who rated themselves as extremely confident were accurate. Sauer, Brewer & Wells (2008) report a 40% error rate among witnesses who make an identification and are 90%-100% confident and a 50% error rate among those who are 70%-80% confident. Sauer et al. (2010) found that 13% of their witnesses, when tested immediately, made a positive identification with 90-100% confidence—their error rate was 20%. 10% of witnesses to the same events--tested after one week--made an identification with 90-100% confidence—their error rate was 25%. %. A recent study by Sucic and colleagues (2015) involving identification of individuals with whom participants interacted for up to a minute showed poor calibration and over-confidence in judgments.

Confidence Malleability. In a dramatic illustration of confidence malleability, Luus & Wells (1994) used a staged-crime to secure false identifications from 136 eyewitnesses. These eyewitnesses viewed a theft in pairs and were separated shortly after the theft. After being separated false identifications were obtained from the witnesses using a photospread (the eyewitnesses were unaware they had made a false identification). After making their identifications eyewitnesses were either told nothing about the identification decision of their co-witness or were given information that their co-witness ostensibly identified the same person, identified someone else, or indicated that the culprit was not in the lineup. They were then asked for their confidence levels. There were dramatic increases in the confidence that eyewitnesses expressed in their false identifications in the condition in which they were told that their co-witness identified the same person (the average confidence on a 10 pt. scale was 8.8 versus 6.9 in the no-information control condition). The lowest confidence levels were found among witnesses who were told that the co-witness had indicated that the perpetrator was not in the array (3.6).

A 2006 meta-analysis by Douglass and Steblay of 20 studies with 2,400 identifications found that witnesses receiving feedback “expressed significantly more . . . confidence in their decision compared with participants

9

who received no feedback.” Douglass & Steblay further report that witnesses “reported that they possessed a significantly better basis for making the identification…, greater clarity of the perpetrator’s image in mind …, greater ease of identification…, and needing less time to make their ID... They also reported a better memory for strangers’ faces …. and greater trust in the memory of another witness with a similar experience.... Not surprisingly, then, they are also more willing to testify about their identification decision… (p. 863) Furthermore, Douglas et al. (2010) have shown that witnesses receiving confirming post-identification feedback are viewed by others as more accurate and confident than witnesses who have not received feedback.

Each of the factors enumerated above and discussed in more detail below are implicated by the testimony of one or more witnesses in this case. As detailed below, the advantage of having expert testimony on these matters is that the testimony will sensitize jurors to impact of these factors on eyewitness identification reliability and enable them to better assess the evidence present at trial. As shown below, in the absence of expert testimony jurors’ evaluations of eyewitness evidence are dominated by a consideration of witness confidence—a factor which is, at best, only modestly informative about witness reliability and at worst, misleading.

XI. Cumulative Impact of Threats to Reliability

I have enumerated a large number of factors which have been linked to eyewitness reliability in research and are involved in this case. A strong case can be made that these factors can cumulatively increase identification errors. There is also evidence that the studies reported here typically underestimate the effects of these factors on performance? Are the effects these factors cumulative or does the presence of one factor make the others irrelevant?

Generalizability of Effects

Shapiro & Penrod (1986) noted that in the studies in their meta-analysis overall performance levels were worse in more realistic eyewitness studies higher in the laboratory studies and that those differences were readily accounted for by systematic differences between more and less realistic studies. Lindsay and Harvie (1988) reported a similar pattern: less-realistic studies appear to systematically over-state levels of witness performance. There is other evidence (see Penrod & Bornstein, 2007 for an overview) that less realistic studies understate the impact of studied variables on witness performance. This evidence emerges from meta-analyses in which researchers have tested whether the effect of the factor that is the main focus of the meta-analysis is larger or smaller under more realistic testing conditions. There are many such demonstrations.

Steblay’s (1992) meta-analysis of the weapon focus effect similarly showed that relatively artificial simulations underestimated the magnitude of the effect. Deffenbacher, Bornstein, Penrod, and McGorty’s (2004) found that the effect of stress was twice as large for staged-crime studies than for studies manipulating stress by other means.

Cumulation of Effects

In studies were several factors are manipulated at one time it is regularly observed that they simultaneously influence witness performance. For example, Cutler, Penrod, O’Rourke and Martens (1986) found, across two studies, that disguise, biased instructions, weapon visibility, retention interval, and lineup size all influenced identification accuracy. Cutler, Penrod & Marten (1987a) found simultaneous effects of disguise,

10

retention interval and mugshot viewing. Cutler, Penrod & Marten (1987b) found simultaneous effects for disguise, weapon focus, retention interval and instructions to witnesses.

Shapiro and Penrod (1986) analyzed performance across the hundreds of experimental conditions and found that even when statistically controlling for the effects of other variables most variables still explained significant portions of variability in performance. With regard to correct identification rates, attention variables and the duration of exposure per face were positively related to hit-rate performance. The number of targets studied and retention interval accounted for significant variance. With respect to identification errors, greater attention led to fewer false alarms, the number of targets studied was positively associated with false-alarm rates, there were many more false identifications when participants were confronted with live targets as opposed to still photographs and there was a negative correlation between the false-alarm rate and number of foils.

In short, there is substantial evidence that the effects of the many factors discussed above can and do have a cumulative effect on identification performance.

11

I. Expertise of XXXX

I am …

With regard to eyewitness issues in particular:

a. I am an author or co-author of

b. I have made more than XX presentations on eyewitness research before professional psychological organizations.

c. I have made a number of presentations to other professional groups of lawyers, judges, and police concerning problems of eyewitness reliability.

d. I have regularly taught about eyewitness research in law school and psychology courses.

e. I have supervised more than XX undergraduate and masters thesis research projects and XX doctoral dissertations concerned with eyewitness memory.

f. I have been the principal investigator on XX research grants (awarded through competitive review processes) that address problems of witness reliability. XX of these grants are from the Law and Social Sciences program of the National Science Foundation and the other was from the National Institute of Justice -- the research wing of the U.S. Department of Justice. The National Institute of Justice research focused on methods to improve witness identification performance.

g. In the past few years I have testified in cases including XX

I have been asked by counsel to examine portions of the case file in order to identify factors that could have contributed to a misidentification by the witness in this case and about which I would be (and would have been) prepared to testify at the trial of defendant XXXXX. In Sections II through VI below I discuss the general problem of mistaken identifications, describe the growing body of research on factors that contribute to misidentifications, discuss research on the need for and benefits of expert testimony concerning this body of scientific research and discuss the research in relation to admissibility criteria. In sections VII through X below I have highlighted research findings relating to the factors that come into play in this case—in doing so I am not commenting on the credibility of any witness nor expressing an opinion about the reliability of any identification in this case—rather, the research findings provide background information that would be useful to a jury when it makes its assessments of eyewitness evidence and witness credibility.

II. The Problem of Eyewitness Identification Errors

Archival Studies Show that Mistaken Identification is the Primary Cause of Erroneous Convictions

Several legal scholars, beginning with Borchard (1932), have studied the causes of mistaken identification in over 1,000 criminal cases (see also Brandon & Davies, 1973; Frank & Frank, 1957; Huff, 1987; Huff, Rattner & Sagarin, 1986). Huff (1987) readily concludes, on the basis of studying the 500 cases of erroneous conviction that he identified, that the single leading cause of mistaken conviction was erroneous eyewitness identification of the defendants. He states that eyewitness error was involved in nearly 60% of the cases he studied. Rattner's (1988) analysis of 205 cases of wrongful conviction demonstrated that over 50% were attributable to mistaken identifications (also see Borchard, 1932). This rate is all the more remarkable given

12

that some commentators estimate that eyewitness identifications are prominent in only 5% of trials (Loh, 1981).

The Predominant Form of Defective Evidence in DNA Exonerations Is Eyewitness Identification.

More recently, the prominence of mistaken identifications as a source of erroneous convictions has been reaffirmed by the results of exonerations based on DNA evidence. By 1998 post-conviction DNA testing had freed 62 persons in the United States convicted by juries of crimes that they did not commit -- 8 of whom were sentenced to death. In Scheck et al.'s (2000) analysis of the first 62 cases, 52 were mistaken eyewitness identification cases with a total of 77 mistaken eyewitnesses (that is, many defendants were misidentified by more than one witness).  University of Virginia law professor Brandon L. Garrett has systematically examined the first-200 DNA exculpation cases (the total count as of April 2012 is 289 cases), in which the innocent people served an average of 12 years in prison. Garrett (2008) reports that the leading cause of the wrongful convictions was erroneous eyewitness identification--which occurred in 79 percent of the cases. In a quarter of the cases, eyewitness testimony was the only direct evidence against the defendant. The Innocence Project website indicates, as of August 2009 that half of the 179 misidentification cases (out of 239 exonerations) were based just on misidentification evidence and more than a third of the misidentification cases had 2 or more witnesses who made misidentifications (13% of the reported cases had three or more misidentifications).

III. The Growing Body of Scientific Research on Eyewitness Reliability

The research on eyewitness reliability is of relatively recent vintage. The recent vintage of this research is very important in light of the fact that most of the significant appellate cases that created precedential impediments to eyewitness expert testimony pre-date the vast bulk of the research--attempts to introduce expert psychological testimony on eyewitness memory began to flourish in the early 1970s. Evidence of these efforts can be found in appellate court decisions. In the state courts, two states, Kentucky ( Pankey v. Commonwealth, 1972) and Massachusetts (Commonwealth v. Jones, 1972), upheld their trial courts' decisions to exclude such expert testimony. Trial court exclusion of such testimony was also upheld in an early and widely-cited decision by the U.S. Court of Appeals for the Ninth Circuit (U.S. v. Amaral, 1973). The novelty of the research was one of the factors that influenced these decisions and raises the question of what has changed in the past forty years?

How recent is the scientific research? A very narrow full-text search, using the PSYCHINFO database, for peer-reviewed articles containing the phrases “eyewitness identification” or “face recognition” (searches which eliminate the tens of thousands of scientific studies on other aspects of human memory) reveals that over 2,500 articles (over 1900 with those phrases in the abstract) which meet the criterion have been published since 1930 (a full-text search on “memory” yields over 117,000 publications--87,500 with “memory” in the abstract). The eyewitness and face recognition articles had the following distribution of publication dates (with hits in abstracts shown in parentheses:

1930's—3 (1)1940's—0 (0)1950's—4 (0)1960's—4 (3)1970's—60 (37)1980's—286 (200)1990's—536 (417)2000’s—1660 (1246)

13

In short, there is now a large and growing body of scientific research directed specifically at issues relating to eyewitness reliability--a body of research that has mushroomed since 1980. Furthermore, there is also a much larger body of research (numbering in thousands of studies) on human memory generally that provides a broader and deeper empirical and theoretical context for the research that focuses specifically on eyewitness performance and the factors that influence such performance. This growth of research on factors influencing eyewitness performance—and related research on jury decisionmaking in eyewitness cases and research on the effects of expert psychological testimony on eyewitness--issues is further underscored by the number of scholarly books by psychologists that have appeared in recent years.

IV. The Need for Expert Testimony—Juror Knowledge and the Weaknesses of Cross-Examination

Cross-examination is sometimes hailed as a powerful engine for determining truth, but is this true in eyewitness cases? Effective cross-examination requires that juries are sensitive to the factors that influence eyewitness identification accuracy. Suppose that an attorney, through cross-examination, establishes that the crime perpetrator was of a different race than the eyewitness, that the perpetrator was disguised and brandished a weapon, that the eyewitness was highly intoxicated, and that the lineup test from which the suspect was identified suffered from instruction bias, foil bias and presentation bias. What good will this do if the jury does not understand how these factors are likely to influence eyewitness identification accuracy? The attorney can argue during closing argument that these factors enhance the likelihood of false identifications, but the jury may find such arguments implausible, especially if they perceive the attorney as biased and the arguments as inconsistent with common sense.

Four types of studies examine juror sensitivity to the factors that influence eyewitness identification, all of which are reviewed below. The first type are survey studies that assess lay knowledge using multiple choice questions. The second type examines the abilities of lay persons to predict the outcome of eyewitness identification experiments. The third and fourth involve simulated jury decision making experiments; the third examines the influence of discredited eyewitnesses on mock-juror decisions, and the fourth examines the influence of eyewitness evidence on juror decisions.

Survey Studies of Lay Knowledge. A number of studies, published in multiple articles, have addressed lay (i.e., juror) and professional (lawyers and judges) knowledge about the factors that influence eyewitness identification. Early studies include Deffenbacher & Loftus, 1982; McConkey & Roche, 1989; Noon & Hollin, 1987 and more recent ones include Alonzo & Lane, 2006; Benton, Ross, Bradshaw, Thomas, & Bradshaw, 2006; Lane, Groft, Roussel, & Alonzo, 2007; Mitchell & Haw, 2008; Schmechel et al., 2006. Read & Desmarais (2009) review these studies and present their own data. Across studies it is common to find that 50-65% of respondents chose answers similar to those of experts. While this performance may look relatively good it generally indicates there are significant disagreements among jurors about how factors influence eyewitness performance and, to some extent, the seeming agreement with experts likely arises from “good” guessing by laypersons. Unfortunately, picking answers on a questionnaire is very different from assessing trial evidence where jurors have to generate and apply their own evaluative criteria. As Read & Desmarais observe:

These (survey) results, however, do not tell us whether community respondents’ comprehension of the topics is of the requisite depth to appreciate conceptual distinctions made at trial by eyewitness experts. Nor do they indicate whether jurors would successfully apply their knowledge to the case at hand (cf. Schmechel et al.). In recent unpublished research, Alonzo and Lane (2006) compared survey responses (to the Kassin et al. items) of students to their answers about eyewitness case vignettes and found that survey responses to specific case-relevant features did not well predict actual decisions made by these same participants about the

14

case simulations. With the use of varied formats that centred on some similar issues, Schmechel et al. also concluded that how participants respond to one statement does not necessarily predict how they will respond to another. Indeed, on several True/False items (postevent information, unconscious transference and cross-race bias), Schmechel et al.’s participants responded in a manner highly similar to our participants but then went on to demonstrate inconsistencies on problem-solving questions that involved these same factors.

Judges and juries appear to perform on a par: Wise and Safer (2004) tested 160 judges’ knowledge on 14 statements about factors whose effects on eyewitness accuracy are supported by strong empirical evidence. Judges averaged only 55% correct and their knowledge about these factors was unrelated to the number of years they practiced law, whether they practiced criminal law, how long they had served as a judge, and whether they served as a state or federal judge or a trial or appellate judge. The same authors (2010) compared 160 U.S. judges, 57 law students, and 121 undergraduates with respect to their knowledge and beliefs about factors affecting eyewitness accuracy and found judges (55%) no more knowledgeable than undergraduates (58%), and both groups were less knowledgeable than law students (66%) (after controlling for guessing the percentages correct were 10% for judges, 16% for undergraduates and 31% for law students). For all groups, increased knowledge about factors influencing eyewitness performance actors was associated with beliefs that might reduce wrongful convictions. All three groups underestimated what potential jurors know about eyewitness testimony. Wise and Safer concluded that: “The results suggest that increasing judges’ knowledge of eyewitness testimony might help them to reduce wrongful convictions and to more accurately assess when eyewitness experts are needed.”

There is other evidence that laypersons are far from adept at evaluating eyewitness performance.

Prediction Studies. In prediction studies, subjects are provided with descriptions of the methodology used in eyewitness identification experiments and are asked to predict the results. If subjects are sensitive to the factors that influence identification accuracy, they should be reasonably accurate at predicting study outcomes. However, Kassin's (1979) found that his research participants were not sensitive to the influence of crime-seriousness on identification accuracy nor to the absolute level of the identification accuracy-rates. Wells (1984) reported several prediction studies. In one, subjects believed confidence was strongly related to accuracy and the second demonstrated that subjects were not sensitive to the influence of one factor that clearly contributes to the suggestiveness of identification procedures: instructional bias.

Brigham and Bothwell conducted a prediction study with a random sample of 90 community members from Leon County, Florida, all of whom were registered to vote and therefore eligible jurors. Brigham and Bothwell's results further reinforced the findings of Kassin and Wells and indicate that prospective jurors overestimate the accuracy of eyewitness identifications.

Mock-Jury Studies of Juror Decision Making. Most relevant to the question of juror sensitivity are studies which attempt to simulate the jury's actual task of evaluating eyewitness identifications. Researchers have taken several approaches to studying juror sensitivity. I have identified three distinct approaches. The first examines prospective jurors' abilities to discriminate between accurate and inaccurate eyewitnesses. The second examines the influence of credible and discredited eyewitnesses on mock-juror decisions. The third examines mock-juror sensitivity to the factors that influence identification accuracy. Throughout these sets studies the judgments of hundreds of prospective jurors (and some experienced jurors) are examined in response to a wide variety of simulated cases.

Mock Jury Evaluations of Accurate and Inaccurate Eyewitnesses. Wells, Lindsay & Ferguson (1979) had the identification testimony of witnesses evaluated by 201 undergraduates who served as mock jurors.

15

When the questions addressed to the witnesses were non-leading, inaccurate eyewitnesses were believed by more jurors (86%) than were accurate eyewitnesses (76%). In contrast, when the questions were leading, accurate eyewitnesses were believed by more jurors (84%) than were inaccurate eyewitnesses (73%). Among jurors exposed to non-leading cross-examination, 76% correctly identified accurate eyewitnesses but only 14% correctly identified inaccurate eyewitnesses.

Among jurors exposed to leading cross-examination, 84% correctly identified accurate eyewitnesses but only 27% correctly identified inaccurate eyewitnesses. Wells et al. also found that the confidence of the eyewitness in his/her identification accuracy correlated significantly (r = .53) with whether or not the juror believed the eyewitness but nonsignificantly (r = .05) with the actual accuracy of the witness’s decision. In other words, jurors were more likely to believe confident eyewitnesses but confident eyewitnesses were no more likely to be accurate than less confident eyewitnesses.

A partial replication of this study by Lindsay, Wells, & O'Connor (1989) tested whether cross-examination of eyewitnesses conducted by experienced attorneys (versus inexperienced attorneys) might aid jurors in differentiating between accurate and inaccurate witnesses. The experienced attorneys (16 attorneys at least five years past their admission to practice, M = 12 years) and (16) inexperienced senior law students examined and cross-examined witnesses to a staged crime. The direct examinations generally ranged from 5 to 15 minutes and the cross-examinations from 3 to 35 minutes. Mock jurors' beliefs in eyewitness accuracy (as indexed by verdicts) were unrelated to witness accuracy--jurors could not differentiate accurate from inaccurate witnesses. Greater attorney experience did not aid jurors in making the differentiation, although jurors perceived a variety of differences in the performances of experienced versus inexperienced attorneys. In this instance, perceived witness confidence was significantly correlated with verdicts (r= .29) although witness self-rated confidence was not (r= -.07).

Wells & Leippe (1981) found that the accuracy of eyewitness recall of details was significantly correlated (r = -.56) with juror belief in the eyewitness identification. In other words, the more peripheral details recalled incorrectly, the less likely the identification was believed by the jurors. In short, to a substantial degree mock-jurors evaluated identification testimony on the basis of witness memory for peripheral details. Examination that underscored errors in memory for peripheral details significantly weakened the credibility of accurate witnesses but did not reveal the inaccurate witnesses. This unfortunate set of results is further compounded by the fact that, in Wells and Leippe's study, eyewitnesses’ memory for peripheral details was inversely associated with identification accuracy!

Lindsay, Wells & Rumpel (1981) also examined mock jurors' abilities to discriminate accurate from inaccurate eyewitnesses. Thefts were staged before 108 undergraduates, each assigned to one of three viewing conditions designed to produce low, moderate and high levels of identification accuracy. Eyewitnesses later attempted to identify the thief from six-person photoarrays. A sample of eyewitnesses who made positive identifications were then cross-examined while being videotaped. The viewing condition manipulation was successful: of the eyewitnesses who made a positive identification, 33% in the low accuracy, 50% in the moderate accuracy and 74% in the high accuracy conditions were correct. These percentages differed significantly from one another.

The videotaped cross-examinations of eight accurate and eight inaccurate eyewitnesses from each viewing condition were then shown to 96 undergraduates--77% of confident witnesses were believed, versus 59% of low confidence witnesses. Overall, 62% of the low accuracy condition witnesses were believed, 66% of the moderate accuracy condition witnesses were believed and 77% of the high accuracy condition witnesses were believed. Thus, jurors unfortunately thought more witnesses were correct than was merited by the performance of the witnesses in the different witnessing conditions . Furthermore, the jurors gave the

16

witnesses much more credence than was merited by the levels of accuracy of the witnesses they actually viewed--half the eyewitnesses selected from each witnessing condition had actually made correct identifications and perfect performance by jurors would thus have produced 50% belief rates for each of the three conditions.

Witness confidence and witnessing condition also interacted: among eyewitnesses with high confidence, viewing condition had a trivial influence on juror beliefs: 76% of eyewitnesses in the low accuracy condition, 76% in the moderate accuracy condition and 78% in the high accuracy condition were believed. In contrast, among eyewitnesses with low confidence, viewing condition had an impact on juror beliefs. The corresponding percentages of eyewitnesses believed were: 47%, 54% and 76%, respectively. Thus, jurors ignored witnessing conditions when the witness was very confident, but gave the witnessing conditions consideration when the witness was not highly confident. Jurors appear to suspend their critical powers when confronted by confident witnesses.

Although the jurors were somewhat sensitive to witnessing conditions, they were not more accurate in their overall assessments of witnesses across the three witnessing conditions--accuracy levels (collapsing across correct and incorrect identifications) were roughly 50% in all three conditions. What happened was that jurors generated a different mix of errors across conditions. In fact, the “improvements” in identifying correct eyewitnesses were fully offset by reduced levels of accuracy in picking out witnesses who made incorrect identifications (39%, 34%, and 25% respectively).

There are two unfortunate aspects to these results. First, to the extent the criminal justice system differentially eliminates witnesses low in confidence from the prosecution process (because, for example, prosecutors are hesitant to proceed with cases based on identifications by less confident witnesses) these results suggest that jurors may commonly find themselves in a situation where they are inclined to rely on (high) witness confidence and not critically examine the circumstances under which an identification was made. Second, Lindsay, et al. found only a weak relationship between witness confidence and witness accuracy (r =.26), thus the jurors in this study were relying on less than fully diagnostic information when using confidence to gauge witness accuracy.

Lindsay, Wells & O'Connor (1989) conducted an experiment to examine whether the findings of their previous research would generalize in a more realistic trial situation. Simulated trials were shown to 178 undergraduates, each of whom viewed one taped trial. Mock jurors rendered verdicts and answered other questions about the trials. The conviction-rate did not differ significantly as a function of accuracy of the eyewitness (jurors could not differentiate accurate from inaccurate eyewitnesses): guilty verdicts were rendered by 68% of subjects exposed to eyewitnesses who made correct identifications and 70% of subjects exposed to eyewitnesses who made false identifications. The degree of experience of attorneys did not significantly influence verdict nor did experience interact with accuracy of eyewitness in the prediction of verdict. The studies of mock-jurors' abilities to differentiate between accurate and inaccurate eyewitnesses converge on a dismal conclusion about jurors' abilities. Jurors overestimate the accuracy of identifications (there are more convictions than there are accurate identifications), jurors fail to distinguish accurate from inaccurate eyewitnesses, jurors tend to ignore viewing conditions that are known to predict identification accuracy and instead based their decisions in part on eyewitness memory for peripheral details and witness confidence--both of which tend to be poor predictors of identification accuracy.

17

Mock Jury Studies of the Factors That Influence Juror Decisions. A number of studies have examined variations in trial evidence that might affect juror assessments of eyewitness identifications. Lindsay, Lim, Marando and Cully (1986) found that in the presence of conflicting eyewitness identifications, alibis influenced verdicts, but witness consistency and witness viewing time did not .

Bell & Loftus (1989) found that detail of testimony influenced subjects' verdicts, but given that eyewitness research (Cutler, Penrod & Martens, 1987a; Wells & Leippe, 1981) shows that memory for physical characteristics and peripheral details does not predict eyewitness identification accuracy, these data provide further evidence that jurors are insensitive to some of the factors that influence eyewitness identification accuracy and inappropriately sensitive to factors that are not diagnostic of eyewitness accuracy.

Cutler, Penrod & Stuve, 1988; Cutler, Penrod & Dexter, 1990) also conducted a mock jury experiments to examine the factors that jurors use to evaluate eyewitness identification evidence. A simulated trial was shown to 321 University of Wisconsin undergraduates and 129 former jurors from Dane County, Wisconsin. The trial concerned a defendant accused of the armed robbery of a liquor store . A number of forms of evidence were systematically manipulated in these trials (e.g. disguise, weapon focus, level of violence, retention interval, instruction bias, foil bias, and witness confidence). Only confidence influenced verdicts despite the fact that other studies reveal that eyewitness confidence and identification accuracy are only modestly related (e.g., Bothwell, Deffenbacher, & Brigham, 1987). Detailed analyses of juror memory for the evidence (see Cutler, Penrod & Stuve, 1988) indicates that the subjects paid attention to the testimony and recalled it with high accuracy rates, hence, lack of attention and poor memory cannot explain the null effects of eyewitness evidence on mock-jurors' decisions. In addition, this research indicates that the judgment processes of eligible and experienced jurors are comparable to those of college student subjects and supports the generalizability of the research described above.

Conclusions

In sum, there are many reasons to be concerned about the quality of jury decisionmaking in eyewitness identification cases—in the absence of expert testimony.

1. Jurors apparently have difficulty reliably differentiating accurate from inaccurate eyewitnesses.2. A major source of juror unreliability is their reliance on witness confidence--which is:

a. A dubious indicator of eyewitness accuracy even when measured at the time an identification is made, and

b. Appears to be highly malleable and influenced by post-identification factors such as repeated questioning, briefings in anticipation of cross-examination, and feedback about the behavior of other witnesses. These factors cannot increase witness accuracy and are therefore likely to further reduce any relationship between witness confidence and accuracy.3. Jurors appear to overbelieve eyewitnesses.4. Jurors are not adequately sensitive to aspects of witnessing and identification conditions that are arguably better predictors of witness accuracy than is witness confidence but are sometimes sensitive to factors (such as recall of peripheral details) that are not diagnostic of witness accuracy.

V. Eyewitness Expert Testimony Helps Jurors

Some judges believe that expert testimony will confuse the jury. Others believe that it will prejudice the jury. I have tried to classify the possible effects of expert testimony and consider these effects below. Judicial impressions about the possible effects of expert testimony are somewhat speculative, although the issues raised in these speculations can, in fact, be addressed empirically. Recognition of the empirical

18

questions has prompted some researchers to bring data to bear on the question of what effects expert testimony has on jury decision making.

The Expert's Possible Effects on the Jury

Plausible effects of expert testimony can generally be classified into three categories: juror confusion, juror sensitivity, and juror skepticism.

Juror Confusion.

The first possible outcome is that expert testimony will have no effect whatsoever on the judgments of jurors. Though pessimistic, this hypothesis is really quite reasonable given that jurors have been shown to have difficulty in understanding and applying a variety of legal concepts at virtually every stage of the trial process (Penrod & Cutler, 1987).

Juror Sensitivity: Knowledge and Integration. If expert testimony does in fact influence juror decision making, what effect should it have? Clearly a desirable effect is to improve juror sensitivity to the factors that influence eyewitness memory (McCloskey et al., 1986; Wells, 1986). Sensitivity refers to the knowledge of how a given factor influences eyewitness memory and the ability to render decisions in accordance with that knowledge. Thus, sensitivity contains two components: knowledge and integration. Knowledge refers to awareness of the manner in which a factor influences eyewitness memory, including the direction and magnitude of the effect for a given factor. Integration, in this context, refers to the ability to render decisions that reflect knowledge.

How are knowledge and integration pertinent to the issue of juror sensitivity? It might be the case that jurors are unaware of the manner in which some factors influence eyewitness memory. For example, the survey studies described in the previous chapter revealed that jurors are insensitive to the equivocal effects of training on eyewitness identification accuracy. But they are somewhat sensitive to the influence of cross-race recognition. Even if jurors are aware of the relative effects of a given factor on eyewitness memory, the magnitude of that effect might be attenuated in the juror's integration of the evidence. In other words, the jurors' judgments might not reflect their a priori beliefs. Decision making research in a variety of psychological domains (e.g., Goldberg, 1968) shows that integration is quite difficult to achieve, even by trained experts. The mock jury studies discussed above showed that jurors are insensitive to many forms of eyewitness evidence, but it is not clear whether this insensitivity is due to lack of knowledge, poor integration skills, or some combination of the two. One plausible effect of expert testimony is that it could improve both knowledge and integration of eyewitness evidence.

Juror Skepticism. Though it is agreed that improved juror sensitivity is a desirable effect of expert testimony (McCloskey et al., 1986; Wells, 1986), there is considerable disagreement as to whether jurors should be made more skeptical of the accuracy of eyewitness identifications. There is ample evidence that eyewitness identifications are often inaccurate. As mentioned above, realistic field experiments (e.g., Brigham et al., 1982; Krafka & Penrod, 1985) show that witnesses give correct judgments on identification tests approximately 50% of the time. The prediction studies and Wells' studies of mock jurors' abilities to discriminate between accurate and inaccurate eyewitnesses provide some evidence that jurors "overbelieve" eyewitnesses, but few conclusions can be reached about the desirability of a skepticism effect.

19

Positive Effects of Expert Psychological Testimony

We have seen that jurors are largely insensitive to variations in trial evidence that ought to permit them to (at least partially) differentiate between good and poor eyewitnessing conditions (and therefore, between accurate and inaccurate eyewitnesses). The critical question is whether expert testimony enhances jurors' abilities to make those differentiations. A perhaps less critical issue is whether expert testimony simply makes jurors more skeptical of eyewitness evidence but does not produce enhanced differentiation of good versus poor witnessing conditions.

Early expert testimony research revealed some consistent findings with respect to expert testimony. Most evident is that reduced belief and fewer convictions are obtained if expert testimony is presented (Fox & Walters, 1986; Hosch et al., 1980; Loftus, 1980; Maass et al., 1985; Wells et al., 1980).

For example, in Loftus (1980, expt 1) participants read a summary of a case in which they learned some details about an attack and the circumstances surrounding an eyewitness identification. Some participants heard about a nonviolent incident, in which the victim managed to wrestle away the assailant's pistol before the assailant fled. Other participants heard a violent version of the incident, in which another person came to assist the victim. Before fleeing, the assailant fired two shots at the second person, who died instantly. A psychologist testified that there were several factors in the criminal incident known to produce difficulties for an accurate identification: cross-race, stress, presence of a weapon, and drinking. Only stress actually varied across participants.

Expert testimony led to a 12% reduction of guilty verdicts in the nonviolent case (where cross-race identification and drinking came into play), and a 25% reduction of guilty verdicts in the violent crime where heightened stress and enhanced weapon-focus came into play. Sensitivity and skepticism are both apt under these conditions.

Fox and Walters (1986) exposed 128 undergraduates to videotaped segments of eyewitness testimony and expert testimony. The witness expressed either high or low confidence. Three conditions of expert testimony were crossed with the eyewitness conditions: no expert testimony, general expert testimony, and specific expert testimony. The general expert testimony included identification accuracy rates obtained in previous experiments, general memory processes (acquisition, retention, and retrieval), and types of memory (sensory, short-term, and long-term). Specific testimony, on the other hand, consisted of similar testimony, but instead of general memory processes the expert psychologist discussed the effects of 12 specific factors that are known to influence eyewitness memory (e.g., physical factors, exposure time, retention interval, stress, weapon focus, the fairness of lineup procedures). In addition, in all expert testimony conditions the expert psychologist discussed the weak relation between confidence and accuracy. The percentages of mock jurors who believed the eyewitness in the high and low confidence conditions were 70% and 55% among those who heard no expert testimony, 50% and 18% among those who heard general expert testimony, and 30% and 5% among those who heard specific expert testimony. Participants rated the culpability of the defendant

20

on a 10-point scale. General (M = 4.81) and specific (M = 4.23) expert testimony led to significantly lower culpability ratings in comparison to the no-expert testimony condition (M = 6.07). This pattern of results is what one would expect where the expert testifies about a modest relationship between confidence and accuracy (the other manipulated variable) in addition to testifying about matters that would, quite logically, produce skepticism about the eyewitness identification.

In studies like these it is not always clear whether reduced belief is due to improved sensitivity to factors that might have impaired the witnesses ability to make a correct identification, to enhanced skepticism, or to both. The only experiments that simultaneously and independently varied witnessing factors that influence identification accuracy and the presence of expert testimony were those by Loftus (1980) and Wells et al. (1980), and both show trends toward improved sensitivity. It is possible to independently examine skepticism and sensitization effects, and it is possible to detect juror confusion, as well. Tests of sensitizing effects require that subjects be presented some combination of expert versus no-expert testimony and good versus poor witnessing conditions.

For example, in Wells & Wright (1983) mock jurors heard testimony from a witness (24 different witnesses were sampled from a larger number of eyewitnesses who were tested under poor, moderate and good witnessing conditions) and either did or did not hear expert testimony. The results (including the baserate of accuracy for witnesses in poor, moderate and good testing conditions) are shown in the following table. As in some earlier studies (Lindsay. Wells, & Rumpel, 1981; Wells & Leippe, 1981; Wells, Lindsay, & Ferguson, 1979; Wells. Lindsay, & Tousignant, 1980), participant-jurors in the no-expert conditions were as likely to believe the identification testimony of inaccurate witnesses as they were to believe accurate witnesses. With expert testimony, however, there was an average of 15% difference in the belief of accurate as compared to inaccurate witnesses. In addition, the no-expert conditions not only failed to produce significant differences in juror belief between good and poor viewing conditions the pattern of belief across viewing conditions was not systematic. The expert conditions, however, yielded systematic differences between good and poor viewing conditions.

In order to further address these issues Cutler, Dexter & Penrod, (1989 and Cutler, Penrod & Dexter, (1989) designed experiments to examine two of the three hypotheses (sensitization and skepticism) concerning the effects of expert testimony on jurors' judgments. Witnessing and identification conditions (WIC), witness confidence, and the presence of expert testimony were varied systematically, thus permitting independent tests of sensitivity and skepticism.

A videotaped trial was used. The expert testimony was very realistic and the expert was rigorously cross-examined. Three factors, each having two levels, were manipulated in the videotaped trial: witnessing and identification conditions, witness confidence, and expert testimony. In the "poor WIC" condition the eyewitness and police officer testified that the robber was disguised (i.e., wore a hat that covered his hair and hairline), the robber outwardly brandished a handgun (presumably invoking a weapon focus effect); the retention interval between the crime and the identification was 14 days, and the officer in charge of the lineup did not explicitly offer the witness the option of rejecting the lineup (suggestive lineup instructions).

In the "good WIC" conditions the witness and police offered testimony that the robber was not disguised, that the handgun was hidden throughout the robbery, that the retention interval was two days, and that the lineup instructions were not suggestive. The witness testified either that she was 80% or 100% confident that she had correctly identified the robber. In conditions containing expert testimony, the expert gave the testimony described above. In conditions containing no expert testimony, the prosecuting and defense attorneys nonetheless reiterated the witnessing and test conditions surrounding the identification in order to maximize their effects on jurors' decisions in the no-expert control group. After viewing one version of the

21

trial, mock-jurors rendered verdicts and rated the credibility of the eyewitness and the strength of the prosecution's and defense's cases. Ratings were recorded on seven-point scales.

Subjects were 96 eligible and experienced jurors who were called for jury duty in Dane County, Wisconsin. All subjects were recruited by telephone within one year of having served (or having been called to serve) on a jury. Data from the 96 subjects were combined with data from a sample of 538 undergraduates. Combining these two subsamples (N = 634) allowed us to test the effects of expert testimony with maximum power and permitted us to test whether expert testimony differentially affects eligible jurors' and undergraduate jurors' decisions.

Expert testimony produced sensitizing effects on witness credibility ratings and on prosecution and defense case strength ratings. Among jurors who heard expert testimony, witness confidence produced a weaker effect on eyewitness credibility ratings (d = .17) and on defense case strength ratings (d = .02) than for jurors who heard no expert testimony (d = .50 and .26, respectively). Jurors exposed to expert testimony were less likely to use witness confidence in evaluating the credibility of the eyewitness than were jurors not exposed to expert testimony. Likewise, among jurors who heard expert testimony, WIC had a greater impact on both prosecution (d = .49) and defense (d = .56) case strength ratings than among jurors who heard no expert testimony (d = .11 and .16, respectively). Jurors exposed to expert testimony used WIC to a greater extent in evaluating the strength of the prosecution's and defense's cases than did jurors not exposed. Thus, these findings show that expert testimony sensitizes jurors to the importance of WIC and the relative lack of importance of witness confidence.

There was little evidence for expert-induced skepticism. Expert testimony produced no main effects on eyewitness credibility ratings, prosecution case strength ratings, or verdicts. Inclusion of the expert testimony did improve the defense case strength ratings. This effect was greater among eligible jurors than among undergraduate jurors.

The results of this study and earlier studies provide support for expert psychological testimony in eyewitness cases. Without such testimony, jurors reveal poor knowledge and/or integration of knowledge about eyewitness problems. Further, there is little evidence that expert testimony caused jurors to become more skeptical. The Cutler, et al. study also addressed the issue of external validity and whether research with undergraduate jurors is justified. Indeed, the study found that eligible and experienced jurors were essentially comparable to undergraduate jurors. The nonsignificant differences observed in the study should be considered in light of its design and the very large sample size (N = 634), both of which increase the likelihood of detecting such differences.

Conclusions

Taken together, these studies indicate that expert psychological testimony serves as a safeguard against mistaken identification. There is little empirical evidence that jurors are confused by the testimony or are prejudiced by it and measures of recall of expert testimony in the studies where this has been measures indicate that jurors do recall and understand the testimony. Expert testimony appears to have the beneficial effect of educating jurors about factors that influence eyewitness identification and enhancing their reliance on those factors when rendering decisions in eyewitness cases. Indeed, results indicate that expert testimony on eyewitness memory can work on the behalf of the prosecution as well as the defense

VI. Research Findings and Expert Testimony Meet Admissibility Criteria

22

Frye : The eyewitness research community (which includes several researchers who have dual training in law and psychology) has been sensitive to the question of the admissibility of testimony about eyewitness research findings. Does this research meet the scientific standards required by the courts? One can address this question from multiple perspectives. Under the traditional Frye standard the relevant question would be whether the testimony is generally accepted within the relevant scientific community (a standard which, arguably, permitted experts to achieve admissibility if they could point to other experts who agreed—even if the testimony had no scientific basis. Some of the earliest reversals/admission of expert eyewitness testimony came in Frye jurisdictions (e.g., People v. McDonald, 1984; State v. Chapple, 1983).

FRE: The Federal Rules of Evidence refined Frye: Rule 702 specifies another criterion for admissibility. Rule 702 states that expert testimony is admissible if:

(a) the expert is qualified; (b) the testimony assists the trier of fact; (c) and the expert's testimony is sufficiently reliable.

The federal rules adopted a standard that requires the expert testimony "assist" the jury. Does eyewitness research meet these standards (Frye, FRE)?

Is eyewitness research predicated on the scientific method?

The answer is an unequivocal "YES." Virtually all of the empirical eyewitness research conducted by psychologists makes use of standard experimental methods employed in all the experimental sciences. A hallmark of this research is the falsifiability/testability of research hypotheses that are tested using experimental research methods in which the variables or processes examined by researchers are carefully controlled by the researchers in order to assess their causal effects on outcome variables (such as identification accuracy). Virtually everyone who has earned a PhD in psychology in the past 40 years has been trained to use scientific research methods. Use of appropriate research methods is an essential requirement for publication in peer-reviewed scientific journals across all scientific disciplines and psychology is no exception. The results and conclusions summarized below are the products of precisely the methods underscored by the Supreme Court in Daubert.

The research studies relied upon here have typically survived two levels of peer review . Most of the research upon which the proffered testimony is predicated is the product of research supported by funding sources such as the National Science Foundation, the National Institutes of Mental Health, the National Institute of Justice, the Research Council of Canada, and equivalent agencies in Great Britain, Australia, and Germany (the vast bulk of the research has been conducted by researchers from these six countries). These national funding agencies typically subject research proposals to a review process in which anonymous evaluations are solicited from a half-dozen to as many a twenty scientific reviewers. Proposals are evaluated for soundness of research design and analysis and the contributions they are likely to make to our understanding of the processes under study. At an agency such as the Law and Social Sciences program at the National Science Foundation only one in five such proposals receive funding.

Once data collection and analyses are completed and the results are written up for publication a second round of peer review begins. Articles are submitted to a single scientific journal for peer review (in contrast to the publication process in law journals where an author could, in fact, submit the same article to any of the more than three hundred law reviews--which are overwhelmingly student edited--and wait for one of them to accept the paper). The editor of the journal will typically solicit three anonymous outside reviews of each submitted manuscript (and will also evaluate the manuscript herself). Relatively few articles are accepted for

23

publication (the rejection rate in most psychology journals is around 80%). Authors of rejected manuscripts may choose to revise their manuscripts and submit them to another journal--where the manuscript will once again go through the peer review process.

Although some original research first appears in edited scientific books and is not subjected to as rigorous a form of peer review, these chapters typically undergo review by the volume editors and therefore reflect the input of peers who were not directly involved in the research.

There is a consensus on the content of expert testimony?

Courts have traditionally required that the content of expert testimony reflect scientific principles generally accepted in the field, though the more modern trend, reflected in the Supreme Court's 1993 decision in Daubert is not to look exclusively to general acceptance but to consider the scientific validity of the procedures that have produced the knowledge represented in the expert testimony. A reliable and valid assessment of consensus in the field requires a systematic sampling of opinion.

Yarmey and Jones (1983) were the first to attempt to address the level of consensus empirically. They provided 16 eyewitness experts with hypothetical scenarios and forced-choice response format to assess their predicted outcomes. High levels of agreement were obtained on many topics.

Kassin, Ellsworth & Smith (1989) replicated and expanded significantly upon Yarmey and Jones' findings. They conducted a large-scale survey of eyewitness experts from the U.S., Canada and Europe. A total of 63 experts responded to the survey. Each respondent completed a 24-page questionnaire in which (s)he evaluated the reliability of 21 eyewitness phenomena and provided personal information concerning her or his educational background, employment, publications and experience as an eyewitness expert. The vast majority of the respondents had Ph.D.s, and the average number of relevant publications was 6.35 (most of which were in scientific journals). Most (56%) had testified as experts on eyewitness memory at least once. In total they estimated having testified on 478 occasions: 364 times for the defense in criminal cases, 29 times for the prosecution in criminal cases, 54 times for the plaintiff in civil cases and 31 times for the defendant in civil cases.

Kassin, Tubb, Hosch, & Memon (2001) replicated and extended this study using somewhat different questions and found the following levels of agreement among 64 experts (ninety-two percent had published one or more books, chapters, or articles on the psychology of eyewitness identification. On average, respondents had authored or edited 2.15 books, 6.54 chapters, 13.22 scientific journal articles, 1.42 law review articles, and 5.38 magazine or newsletter articles).

Recently Daftary-Kapur (2010) and Penrod surveyed 71 leading eyewitness researchers. The experts in the current study are even more accomplished than those in the prior studies. On average the experts had published 18.51 scientific journal articles, 2.9 books, 8.3 book chapters, 3.7 law review articles, 5.3 magazine articles, and 3.0 newsletter articles. Overall, the experts had been asked to testify at least 2,719 times (the actual number could be much higher as the survey instrument, which was administered over the internet permitted a maximum report of 100+ cases—which was treated as 100 cases for the analyses reported here). Nearly a quarter (23.5%) of the respondents who answered this question used the 100+ category. Experts had testified in over 1600 cases. Not all the participants were eager or willing to testify in court. Experts agreed to testify 66% of the time, and among those who agreed, actually testified 88% of the time.

Daftary-Kapur and Penrod asked about the following phenomena and obtained the following results:

24

Topic Statement 1. Stress Very high levels of stress impair the accuracy of eyewitness testimony.2. Weapon focus The presence of a weapon impairs an eyewitness's ability to accurately identify the perpetrator's face.3. Showups The use of a one-person showup instead of a full lineup increases the risk of misidentification.4. Lineup fairness The more members of a lineup resemble the suspect, the higher is the likelihood that identification of the suspect is accurate.5. Unconscious Eyewitnesses sometimes identify as a culprit someone they have Transference seen in another situation or context.6. Cross-race bias Eyewitnesses are more accurate when identifying members of their own race than members of other races.7. Alcoholic intoxication Alcoholic intoxication impairs an eyewitness's later ability to recall persons and events.8. Mugshot-induced bias Exposure to mug shots of a suspect increases the likelihood that the witness will later choose that suspect in a lineup.9. Presentation format Witnesses are more likely to misidentify someone when presented with a

simultaneous (as opposed to sequential) lineup.10. Elderly witnesses Elderly eyewitnesses are less accurate than are younger adults.11. Identification speed The more quickly a witness makes an identification upon seeing the lineup, the more accurate he or she is likely to be.

25

Other sources of information indicating consensus: One’s appreciation of the degree of consensus within the scientific community is shaped by at least two other important sources of information: the way people write about research findings and the way they talk about the findings in formal and informal settings. With respect to talking, I would note there are scores of opportunities to hear researchers talk about research findings during formal conference presentations (both talks and poster presentations) and during informal conversations at conferences. In sum, converging evidence indicates that considerable consensus does exist regarding the influence on eyewitness memory of a wide variety of factors.

VII. Subject Matters for Expert Testimony in the Present Case

In the remainder of this document I identify topics and research about which I would testify in the present case.

Low Rates of Eyewitness Identification Accuracy

How often do witnesses make mistakes? Such information can be useful to jurors in the same way that knowing the frequency of repair records of automobiles can be useful to someone buying an automobile—such numbers reveal the general risks that jurors/automobile buyers confront.

Experimental Studies Show a Pattern of High Rates of Eyewitness Identification Errors

The research of Haber & Haber (2001) provides further evidence on witness accuracy rates and also supports the conclusion that witnesses who believe that have witnessed actual crimes will perform similarly to witnesses who know they are participating in an experiment. They performed a meta-analysis of experiments in which witness-participants observed a crime or other event and then attempted to identify the perpetrator—across studies Haber & Haber found comparable levels of performance—see the Table below. In 23 studies subject-witnesses watched a video, slide presentation or movie of a crime. In 14 studies, a crime was enacted in the presence of the participants who believed a criminal event was occurring as they watched. However, later—before attempting identifications--the witnesses were told the crime had been staged. In 7 studies, the participants believed they observed a real crime and believed the lineup presentation was real.

Target Present Target AbsentHit Foil ID Miss CR Foil ID Overall

CorrectDemo-Crime/Demo-Lineup 43% 34% 24% 49% 51% 46%Real-Crime/Demo-Lineup 52% 24% 24% 48% 52% 50%Real-Crime/Real-Lineup 47% 24% 28% 53% 47% 50%

               As shown in the Table, the percentages are very close across the three groups of studies and the small differences are not significant (p> 0.05).  Overall, what we learn from these studies is that identifications for persons seen briefly, in non-stressful conditions, and attempted after brief delays, are frequently inaccurate—witnesses in target-present arrays appear to be accurate about half the time and one-third of positive identifications are erroneous identifications of a foil. In target absent arrays, the mistaken identification rate is about 50%.

Field Experiments Testing Witness Memory Show High Rates of Witness Identification Errors

26

One relevant source of data pertaining to accuracy rates of actual eyewitness identifications emerges from realistic field studies of eyewitness identification. Some researchers (Brigham, Maass, Snyder & Spaulding, 1982; Hosch & Platz, 1988; Krafka & Penrod, 1985; Pigott, Brigham & Bothwell, 1990) have attempted to reap the benefits of both laboratory experiments and realistic crime conditions by conducting well-controlled experiments in a more realistic, field setting. In these studies, data were gathered in these experiments from a total of 291 mock eyewitnesses who were administered 536 separate identification tests. The average correct identification-rate from presentations which included the target person was 41.8%. Thus, nearly 60% of witnesses failed to identify the target when he was present. Unfortunately, the false identification rate of innocent foils was nearly as high as the rate of guilty-target identifications—41%. The false identification rate of foils in target-absent presentations was assessed in two of the studies--the average was 35.8%. In short, identification errors were rampant.

What we learn from these experiments is that identifications for persons seen briefly, in non-stressful conditions, and attempted after brief delays, are frequently inaccurate. Based on the theory that customer-present photoarrays resemble the situation in which the suspect is guilty and customer-absent photoarrays represent the situation in which the suspect is innocent, only 2 out of 5 guilty persons were correctly identified and an innocent person was falsely identified on 1 of 3 target-absent arrays. In one of these studies (Pigott et al., 1990), the mock eyewitnesses were bank tellers, 70% of whom reported that they had received training for eyewitness situations.

Face Recognition is Far from Perfect Even Under Optimal Testing Conditions

What we learn from these experiments is that identifications for persons seen briefly, in non-stressful conditions, and attempted after brief delays, are frequently inaccurate. Based on the theory that customer-present photoarrays resemble the situation in which the suspect is guilty and customer-absent photoarrays represent the situation in which the suspect is innocent, only 2 out of 5 guilty persons were correctly identified and an innocent person was falsely identified on 1 of 3 target-absent arrays. In one of these studies (Pigott et al., 1990), the mock eyewitnesses were bank tellers, 70% of whom reported that they had received training for eyewitness situations.

Immediate Recognition of Faces from Memory. Is person-recognition inherently difficult or might witnesses perform better under optimal circumstances? Megreya and colleagues have conducted studies testing recognition of faces under fairly optimal conditions and found worrisomely poor performance. For example, Megreya & Burton (2006) had participant/witnesses study faces such as the following for as long as the participants needed in order to assure they could recognize the face on a test administered 2 seconds after the target face was removed (participants studied the faces 5 to 10 seconds each and were tested on a 10-person photoarray). The faces were from graduating classes of police officers with the studied-pictures and the photo-array pictures taken the same day. Pictures were shown in high-quality photographs of about 2” x 3” and were very likely all the same race as participants (the study was conducted in Scotland).

27

The following is an example of a 10-person array used with the face above. Participants were reminded that the target face may or may not be in the array.

Performance on what seems like a relatively simple and straight-forward test of face recognition proved surprisingly poor: only 59.5% of participants correctly identified the face when it was present in the array (another 9.5% erroneously picked someone else). When the target-person was not in the array, 23.5% erroneously picked someone from the array. Because half the arrays contained the target and half did not, only 64% of all selections were correct (59.5/(59.5+9.5+23.5)). Note that one might argue that the 23.5% of witnesses shown a target-absent array who nonetheless erroneously identified someone are essentially making calculated guesses. Furthermore, because participants were randomly assigned to target-absent and target-present arrays, one could further argue that a comparable 23.5% of the choosers in the target-present arrays were similarly making calculated guesses (9.5% miscalculated and picked a non-target member of the array and another 14% guessed correctly – thus, nearly 1 in 4 of the correct identifications of the target may be little more than calculated guesses … choices of the person who most-resembles the target.

Immediate Recognition of Two Faces from Memory. In a second study Megreya & Burton (2006) simply presented 2 faces sequentially (on the theory that people encounter others sequentially) and then tested recognition (for one or the other person) in similar 10-person photoarrays (with the same reminder that one of the persons may or may not be in the array. This time the correct identification rate fell to 31.2%, the false identification rate in target-present arrays was 25.8% and the false identification rate in target-absent arrays was 39.2%--just 32% of all selections were correct. As per the argument above, it would appear that nearly 40% of witnesses were making calculated guesses.

Immediate Recognition of Faces—Matching from Pictures. In a third study the same researchers allowed the participants to view the targets while they made their selections from the photoarrays (this is termed a matching task and does not involve any memory—the target and the identification-array are viewed simultaneously). In this matching task performance was also surprisingly poor: the target was selected just 70% of the time from target-present arrays (with a 12.3% false identification rate) and in target-absent arrays,

28

the false identification rate was 33.6% (just 60% of selections were correct). Note that the rate of calculated guessing was still very high even though the decisions were not based on memory.

Immediate Recognition of Faces—Matching Live Persons to Pictures from Memory. Megreya & Burton (2008) pushed the testing further: in one study participants were shown a live person for 30 seconds and were then tested, from memory, on a 10-person photoarray—when the target was in the array he was identified by 70% of the participants and a non-target person was identified by 10% of participants. When the target was not in the array, 20.5% misidentified someone else (70% of all positive choices were correct—with a calculated-guessing rate of 20.5%).

Immediate Recognition of Faces—Matching Live Persons to Pictures When participants were shown the live person and the photoarray together (a matching task rather than a memory test), the target was identified by 66.9% of the participants and a non-target person was identified by 15% of participants. When the target was not in the array, 37.8% misidentified someone else (thus, 55.9% of all positive choices were correct and for some reason the opportunity to view the target nearly doubled the calculated-guessing rate). Finally, when the target was shown together with just a single photograph and the target was in the photograph he was identified by 84.4% of the participants and when the target was not in the photograph, 15.6% misidentified that person. In this instance the calculated-guessing rate is at least 15.6% but we cannot know how many participants guessed the target.

The difficulties inherent in recognizing faces is underscored by the following set of pictures from Jenkins, et al. (2011) who pose the simple question—how many different people are shown in the photographs. The answer is given on the signature page below.

(Mis)-Identification of Non-Strangers

29

Of course the studies by Megreya & Burton (2006) reviewed above at pp. 3-5 demonstrate that identifications are not necessarily reliable even when a “witness” is able to make a direct comparison between a live person and a set of photographs and they are even worse when working from memory. Of course the individuals in the Megreya & Burton studies were strangers—how much difference would it make if the individuals had been seen before—even repeatedly and over and extended period of time--by the witnesses?

The most relevant study of identifications of non-strangers is by Pezdek & Stolzenberg (2014). They recruited sophomores (N 139) from two small private high schools who viewed forty yearbook pictures including (a) twenty students who graduated from their school as seniors (fourth year) when the research participants were freshmen (first year students) (familiar faces) and (b) twenty unfamiliar individuals drawn from the senior class of the other high school. Thus, familiar faces at one high school served as unfamiliar faces at the other high school. Participants were asked to indicate whether each face was ‘familiar.’ Individuals’ familiarity judgments were only modestly related to prior contact--accuracy was fairly low (the correct identification rate for familiar faces was 42% while the false alarm rate for unfamiliar faces was 23%. This meant that over one-third of the identifications of “familiar” people were actually misidentifications of strangers. As Pezdek & Stolzenberg put it, the results render “an eyewitness’s report of having seen a perpetrator casually in the past of limited forensic value.”

Studies of Real Witness Performance Using Police Archives Show High Rates of Identification Errors

Similarly low accuracy rates from actual eyewitness identifications emerge from studies of actual witnesses. In these studies it is not known whether the suspect is the actual perpetrator, but it is still possible to gauge the rate of inaccurate identifications of foils—the known-innocents placed in arrays along with the suspects. The best data come from studies in the United Kingdom—the results from nearly 17,000 actual eyewitnesses are shown in the table below.

UK Archival Studies –9-person arrays [row percentages]

Witnesses/Arrays

Suspect ID

Foil

ID No ID

Slater 843/302 36% 22% 42%

Wright & McDaid 1,561/616 39% 20% 41%

Valentine, et al. (VIPER) 584/295 41% 21

% 39%

Pike,et al. 8,800/xx 49% - -W Yorkshire-traditional 1,635/xx 35% - -W Yorkshire-VIPER 940/xx 39% - -

Wright & (VIPER) 134/134 58% 21% 21%

Horry & (VIPER) 1,359/xx 40% 26%

34%

Memon & (VIPER) 1,044/xx 44% 42 15%

30

%Ave [complete-wtd] 16,897/xx 40% 26

%34%

As shown the table, Slater (1994) examined identification attempts by 843 British witnesses who viewed 302 suspects. Slater found that foils were mistakenly identified by 190 witnesses (22.5 percent). Similarly, Wright and McDaid (1996) examined identification attempts of 1,561 British witnesses who viewed 616 suspects in live lineups and found that 611 witnesses (39.1 percent) picked the suspect, 310 witnesses picked a known-innocent foil (19.9 percent), and 640 (41 percent) made no identification. In all, these studies report results on nearly 17,000 identification attempts. Among the studies where complete data are available, nearly 40% of positive identifications were identifications of an innocent foil (26% foils / (26% foils + 40% suspects) = 39.4%). While these studies are informative with respect to the rates at which witnesses appear willing to make choices/guesses from arrays (we know, authoritatively that foil IDs are incorrect) archival studies cannot tell us whether suspect IDs or lineup rejections are correct – for an extensive consideration of the weaknesses inherent in archival studies, see Horry, et al. (2014).

Witness Are Frequently Guessing

Witness error rates are likely to be significantly higher than these numbers immediately indicate due to another problem: it is likely that some of those arrays did not contain the perpetrator. If 75% of those arrays contain the perpetrator and 25% have innocent suspects and innocent and guilty suspects are selected equally often by witnesses, it turns out that less than half of witnesses (45%) who make a selection pick a guilty person (.40 suspect selection rate x .75 guilty rate = 30% guilty picks) and more than half pick an innocent person (.40 suspect selection rate x .25 not-guilty rate = 10% innocent suspects + 26% innocent foils selections = 36% innocent selections (more than one-fourth of whom are innocent suspects who are now in trouble).

But performance is likely affected by another factor that is not obvious in the overall numbers. It is likely that witnesses in these studies were confronted with nine-person arrays, the British standard. This would mean that the average foil drew guesses from 3.3% of witnesses (26/8=3.3)—if these were fair arrays (a dubious proposition in light of evidence of lineup bias) we would expect that another 3.3% of witnesses simply guessed the suspect in the same way that other witnesses guessed innocent foils. If the arrays were biased in a typical manner (see discussion below), it is more likely that between 6% and 10% of witnesses guessed the suspect (some of whom are innocent)—if so, then as few as 40%-10%=30% of witnesses might have selected the suspect based on a sound memory. In sum, a combination of poor memory, willingness to guess, a modest percentage of innocent suspects and small arrays conspire to create a situation in which a substantial portion of suspect identifications are innocent suspects (around one-quarter in the examples above).

Overall, 34% of witnesses did not identify anyone in the array—these non-identifications reflect one of two types of error—either the witnesses “missed” a guilty suspect or the police offered witnesses an array which did not contain the perpetrator. Some reliable witnesses would correctly reject arrays which do not contain the guilty party—other witnesses may (luckily) reject the array because their memory is poor and they are not inclined to make a guess (Penrod, 2003).

Behrman and Davey (2001) examined actual Sacramento, California police records for 271 cases involving 349 crimes (the vast majority armed robberies). They examined 289 photographic lineups and 58 live lineups. They found that approximately 48% of the witnesses who observed a photographic lineup identified

31

the suspect as the perpetrator. Unfortunately, they do not know how many of those were identifications of the actual perpetrator as opposed to mistaken identifications and, because police records were incomplete, Behrman and Davey also could not determine how many additional identifications were made of known-innocent foils—the numbers from the UK studies suggest that 20-25% of witnesses would have identified foils. In live lineups (which did report mistaken identifications of foils—probably due to the presence of defense counsel), Behrman and Davey found that 50% of the witnesses identified the suspect (an unknown percentage of those identifications were erroneous identifications of innocent suspects), the false identification of foils rate was 24% and 26% of the witnesses rejected the lineup. Here, one in three positive identifications was clearly wrong, and, it is likely that between 10% and 15% of the 50% suspect IDS were merely guesses (using the logic applied above with respect to UK arrays—with the US arrays producing higher rates of suspect guesses because the arrays have 6 rather than 9 members) and either the witness or the police made a mistake with respect to the cases in which no ID was made.. In short, errors were rampant.

Probably the best data on witness performance in American lineups come from a recent study by Wells, Steblay & Dysart (2011) of 497 witnesses who attempted identifications from photoarrays. Overall 26% of witnesses picked a suspect and 15% picked a foil – as in the studies from the UK a substantial portion of witnesses making a selection clearly made an error (15/(15+26) = 37%. Because these were 6 person arrays, the 15% who chose a foil were selecting foils at an average rate of 3% (15%/5) – and if these arrays were biased to the extent of arrays studied by others (see below), there is a good chance that 6%-9% of the 26% of witnesses who picked a suspect similarly guessed the suspect (that is, between 23% and 35% of suspect selections might be characterized as “calculated guesses”)!

In short, in both experimental studies and real police cases, one-third or more of witnesses who make a positive identification are wrong and a high proportion of witnesses are guessing—those who guess a foil reveal their errors but someone else—police, prosecutors, judges or jurors--has to detect the errors made by witnesses who guess suspects.

The “Pleading Effect” Elevates the Portion of Trial Defendants Who Have Been Misidentified

Though these analyses imply that a significant portion of identified suspects are innocent, the rates of mistaken identification that appear for trial are likely to much higher than the rates of mistaken identifications. As Wells, Memon and Penrod (2006) note, archival research indicates people charged with crimes plead guilty in a high percentage of cases—(Liptak [2011] reports the rate was 97% in federal courts in 2010 and 94% in state courts in 2006). In contrast Garrett (2008) reports that just 4% of the defendants in the 200 DNA-exonerated cases he studied entered a guilty plea. What impact might this have on the rate of innocent suspects in court? An archival analysis (Kellstrand, 2006) of cases with both DNA and eyewitness identification evidence in the San Diego County District Attorney’s Office, reported that the individuals suspected of crimes were innocent in only 5% of cases. If we assume that just 5% of suspects identified are innocent and 95% are guilty these numbers imply that in state courts, 6% of the 95% (5.7%) of the guilty go to trial, whereas 96% of the 5% (4.8%) of the innocent suspects go to trial. Consequently, at the trial level, nearly half (45.7%) of the defendants (4.8% of the 10.5% going to trial) will be instances of mistaken identification (at the federal level the mix is 2.9% guilty + 4.8% innocent go to trial and 62% are innocent). Charman and Wells (2006) termed this the ‘‘pleading effect’’ and it illustrates how the mistaken-identification rate can be much, much higher at the trial level than at the lineup level.

VIII. Witnessing Factors that Influence the Accuracy of Identifications

32

Witness Intoxication

Alcohol can have a variety of effects on witnesses. It can influence attention as shown in a study by Clifasefi, et al. (2006) who found that at only half the legal driving limit in the US, participants were at a significantly increased risk of failing to notice an unexpected object compared to their sober counterparts.

Yuille and Tollestrup (1990) had 120 participants watch a staged theft. Prior to viewing the staged event, 47 of the participants were intoxicated to a blood-alcohol level of at least .10, 46 received a placebo and 26 acted as controls. Participants were told that they would be viewing a staged event. Following the viewing 58 participants (half intoxicated, half not) where interviewed following standard police procedures and 62 were let go. A week later, all participants returned and were interviewed and participants were asked to select the thief’s photo out of an 8-photo simultaneous array--half of which did not include the thief. The control group recalled 21% more information than the intoxicated groups during the immediate interview. One week later 32% less information was supplied by intoxicated participants. In addition, participants who witnessed the theft under the influence of alcohol made 14% more false identifications..

In a study by Werth and Steinbach, (1991), participants were asked to recognize hand drawn faces. 20 participants were shown 4 faces, each for 1 minute, and asked to remember them for the following day. Half of the participants were exposed to the faces while intoxicated (a BAC of 1 g ethanol per kg) and then asked to identify the faces while sober. The other half of the participants were exposed to the faces while sober and asked to identify the faces while intoxicated. 24 control participants viewed and recognized the face while sober. The participants were shown 12 faces, one at a time—there was no time limit. Participants who attempted to identify the faces when intoxicated (initial exposure while sober) correctly identified the faces at a rate of 51% and falsely identified faces at a rate of 27%. Participants who attempted to identify the faces when sober (following exposure while intoxicated) correctly identified the faces at a rate of 65% and falsely identified the faces at a rate of 20%. All differences between groups were significant (p<.05). Both experimental conditions performed worse than controls and had significantly higher rates of misses (target present) and false positives (false alarms).

Read et al. (1992) tested identification accuracy one week after a staged event using a six-person lineup and found that intoxication while witnessing the event was associated with a lower rate of correct identifications, when the level of arousal (which was manipulated by varying participants’ perceptions of the probability of getting caught while participating in a simulated theft) was low during the event. The alcohol conditionparticipants (31%) made only one half as many correct identifications as placebo no-alcohol participants (68.8%) when arousal was low. When arousal was high, the groups performed equally well (56%). Participants were not tested on target-absent arrays and after one week the participants were no longer intoxicated which begs two questions: what is the effect of alcohol if the target is not present in the array what is the effect of intoxication at viewing and identification?.

Dysart, et al (2002) note that popular belief is that, due to mental impairment, intoxicated witnesses are less accurate than sober witnesses. However, one theory concerning “alcohol myopia” predicts an interaction between blood alcohol level and identification procedures such that intoxicated witnesses will be less accurate only in target-absent conditions. The theory suggests that, relative to intoxicated witnesses, sober

33

witnesses will encode more information/cues about the perpetrator that will facilitate correct rejections in target-absent procedures. Intoxicated witnesses are likely to encode only salient cues from the target and erroneous identifications will result where more subtle cues would have indicated that the suspect was not the target. On the other hand “salient cue matching” will be effective for intoxicated witnesses when the target is present.

Dysart, et al. examined the effect of alcohol consumption on identification accuracy using showups, the identification procedure most likely to be used by police with intoxicated witnesses. As predicted, they found that in the target-present showup condition, blood alcohol level was not significantly related to correct identification rate, but in the target-absent condition, higher blood alcohol levels were associated with a higher likelihood (52%) of a false identification than were lower blood alcohol levels (22%).

Similarly, in a dissertation study by King (2006) participants were shown 6 target photos and were tested for recognition of 2 target faces at 1 hour post exposure, 6-hrs post exposure, and 24-hrs post exposure. The alcohol group produced an increased number of false identifications (64% mistaken IDs versus 34% for those not given alcohol) in target absent lineups while there were no significant differences in performance when the target was present. The alcohol impairment was present at all three time periods—sobering-up did not reduce the alcohol-induced impairment—in fact, overall accuracy declined from 76% at 1 hour to 66% at 6 hours and 51% at 24 hours.

Hilliar et al. (2010) compared the face recognition performance of intoxicated and sober participants in a cross-race context (participants of East/South-East Asian and Western European descent). Half the participants were given alcohol prior to studying the faces on which they were tested. In the alcohol condition, participants’ breath alcohol concentrations just prior to testing averaged .05 which the researchers characterize as a low-to moderate level of intoxication. As shown in the figure below, there was cross-race impairment and overall reduced performance in the alcohol condition.

Accuracy of Witness Reports and Identification Accuracy

In general witness recall of crime events and actors is only moderately good, with people closer to events recalling more than those who are not central to events. Due to conflicting findings, it is not clear that more complete recollections are more correct. Yuille and Cutshall (1986) examined the accuracy of police reports made by 13 witnesses to a gunshop robbery which ended in the shooting death of the robber. The reported information was divided into three categories (a) people details including descriptions of hair color and style, clothing color and style, and age, height, and weight estimations—76% accurate and the source of more than

34

40% of all errors; (b) object details--89% accurate and (c) descriptions of objects -- 85% accurate. “Central witnesses,” who viewed most of the incident from a central vantage point averaged 71.9 details in the police interview versus 24.3 details for non-central witnesses. Accuracy levels for the two groups did not differ.

Similar patterns of results were reported by Christianson & Hubinette (1993) in their study of 110 robbery witnesses—tellers recalled much more accurate information than did customers: e.g. regarding the robbers’ hood (95% vs 77%), robbers’ clothing (90% vs. 69%), robbers’ hair color (50% vs. 31%). Van Koppen & Lochun (1997) studied a very large sample of robberies (n=582) and nearly 2300 descriptions provided by over 1300 witnesses. They found “both accuracy and usefulness decrease if witnesses provide more descriptors of the offender” (p. 676). However, Fahsing, Ask & Granhag (2004) found that recall of more details by their 250 robbery victims was associated with greater accuracy but the reverse was true for general information recalled. Fully accurate recall of clothing information was supplied by a little more than half of witnesses who offered reports: e.g., upper body clothing (recalled by 91% of witnesses with 58% accuracy), trousers (74% with 53% accuracy) and shoes (31% with 55% accuracy) Of course, it is one thing to remember and describe crime events, objects and perpetrators and another matter to recognize such details when attempting to identify a perpetrator. There is little research comparing recognition accuracy of faces versus clothing, but Pryke, et al. (2004) obtained correct facial identifications from 70% of participants in target-present conditions versus the most selected innocent face chosen by 15% in target absent; in voice lineups the respective numbers were 13.3% correct, 35.0% false identifications; forbody lineups: 23.3% correct, 11.6% false identifications, and for clothing lineups, 50.0% and 10.0%, respectively.

In Desmarais & Read's (2010) meta-analysis of 23 surveys assessing 4,669 lay respondents' knowledge of eyewitness issues, they reported that 91% of American respondents understood intoxication could impair identification performance (though as detailed above that does not mean that this understanding would be reflected in their assessments of trial evidence). This nonetheless begs the question of whether respondents understand that intoxication may particularly impact mistaken identifications as shown in the Dysart, et al. study.

Weapon Focus.

Weapon focus refers to the attention witnesses give to a perpetrator’s weapon during a crime. It is expected that the attention the witness focuses on a weapon will reduce their ability to later recognize the perpetrator. Researchers have assessed eyewitness accuracy in an attempt to assess the effects of weapon-focus effects on memory; the relevant studies were reviewed in a meta-analysis by Steblay (1992). The meta-analysis included 19 studies with a total sample of 2,082 participants. The weapon-focus effect on identifications was statistically significant and reflected a modest impairment overall—however, the effect was larger in target-absent lineups and when memory was generally impaired.

A good illustration of the sort of impairment that can be obtained under realistic conditions comes from a study by Maass & Kohnken (1989). In this study 86 non-psychology research participants were approached by an experimenter who was either holding a syringe or a pen and either did or did not threaten to administer an injection. The confederate held, at participants’ eye level, a transparent plastic syringe with a blue label and needle protection cap that was partially filled with a yellow liquid. In the no-threat condition, the confederate (holding either a syringe or a pen) apologized for disturbing the experiment and explained she had to pick up a drug used in a different experiment. She then put the pen or syringe on a tray, took a medicine bottles and left the room. Participants were asked to identify the target from a seven-person, target-absent array. As shown in the table, exposure to the syringe greatly decreased lineup recognition (while

35

enhancing the accuracy of recall for hand cues). The results also show that threat/stress similarly increased error rates.

In a recent dissertation study by an active duty Connecticut police chief, DeCarlo (2010) used a videotaped robbery and found that in a no weapon condition, witnesses were able to correctly identify the target 78% of the time. When a weapon was implied by the perpetrator waving his hands around in his pocket, accuracy dropped to 55% and when a weapon was actually shown, accuracy dropped to 33%. When the perpetrator was absent from the lineup the correct rejection was 89% for the no weapon condition, but when a weapon was implied, accuracy dropped to 76% and when a weapon was actually shown, correct rejections dropped to 65%. A number of studies have demonstrated that a variety of surprising objects (including a whole, raw chicken) can draw attention away from the perpetrator and that novelty, rather than threat, is likely to be the critical ingredient in the effect—indeed, the study by Maass & Kohnken alluded to above, demonstrated either novelty/attention-grabbing or threat can independently produce impairment.

Research by Mitchell, Livosky, and Mather (1998); Pickel (1998, 1999); and Shaw and Skolnick (1999) indicates that any surprising object can draw attention away from the perpetrator and that novelty, rather than threat, may be the critical ingredient in the effect. Researchers have tried to detect weapon-focus effects in field studies, and although the results are conflicting (likely because the methods used are fraught with problems detailed by Pickel (2007)), Tollestrup, Turtle, and Yuille (1994) examined the effect of weapon focus on the rate of suspect identification and obtained data consistent with laboratory findings.

A new meta-analysis (Fawcett, et al., 2013) confirms these conclusions using data from 47 comparisons and further notes that the size of the effects was “unaffected by whether the event occurred in a laboratory, simulation, or real-world environment.”

As emphasized earlier, research shows that jurors overestimate the accuracy of identifications, jurors fail to distinguish accurate from inaccurate eyewitnesses, jurors tend to ignore viewing conditions that are known to predict identification accuracy and instead based their decisions in part on eyewitness memory for peripheral details and witness confidence--both of which tend to be poor predictors of identification accuracy. In Desmarais & Read's (2010) meta-analysis of 23 surveys assessing 4,669 lay respondents' knowledge of eyewitness issues, they reported that just 47% of American respondents understood that weapon focus would impair identification performance (though as detailed above that does not mean that this understanding would be reflected in their assessments of trial evidence). Clearly, there is substantial disagreement among laypersons about the impact that the presence of a weapon can have on identification performance and this is not the kind of disagreement that can really be resolved through logic or argument—as with all the factors discussed below, the nature of the relationship between weapon presence and identification accuracy is an empirical, scientific question—eyewitness expert testimony would provide

36

jurors with scientifically-grounded guidance on this issue and, as detailed above, increase juror sensitivity to weapon focus.

Tracking the Identity of Multiple Individuals

It has been demonstrated in a variety of contexts that witnesses have significant difficulty keeping track of the identity of multiple individuals. Just as a weapon can divide the attention of a witness, multiple perpetrators can divide their attention and identification accuracy and accuracy of descriptions have been shown to decline as the number of perpetrators increases (Clifford and Hollin, 1981; Fahsing, Ask, and Granhag, 2004; Megreya, et al 2006; Shepherd, 1983). This impairment effect was clearly illustrated in the dissertation research by DeCarlo (2010) who showed his witnesses a videotaped robbery involving one or two perpetrators. When the perpetrator was present in the lineup, the single perpetrator condition produced a correct identification rate of 55% versus 33% in the multiple perpetrator condition. In the multiple perpetrator condition, 82% of witnesses correctly rejected perpetrator-absent arrays versus 92% in the single perpetrator condition. More recently, Megreya & Bindeman (2012) compared the performance of witnesses seeing one or two perpetrators. Their data showed that when the perpetrator was present in an array fewer correct identifications (29%) and more incorrect decisions (71%) were made for the double-perpetrator condition compared to the single-perpetrator condition (54% and 46%, respectively). Over a third of witnesses made misidentifications from perpetrator-absent arrays. The authors note: “the double-perpetrator disadvantage is not simply due to a difficulty in separating the facial characteristics of two people that are encountered simultaneously, which therefore cannot be encoded and stored accurately (see Palermo & Rhodes, 2002). Instead, we suggest that this effect might reflect the need to divide attention between two concurrent culprits, which limits the depth or detail of the stored descriptions that can be formed for these persons (see Fahsing et al., 2004).” p.448

Neither Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues nor Benton et al. (2006) provides insight into lay understanding of the effects of multiple perpetrators on identification accuracy. Because both multiple perpetrators and weapon focus appear to involve distracted attention, perhaps the results reported above with respect to understanding of weapon focus (47% believing weapon presence impairs performance) provides some guidance about the likely disagreement among laypersons about the impact of this factor.

Assessing the Probative Value of Identifications, Non-Identifications and Multiple Identifications

Multiple-witness cases are fairly common: archival studies report that multiple-witness cases comprised 29% of 662 cases studied by Yuille & Tollestrup (1992), 38% of 76 cases studied by Tollestrup et al. (1994) and 58% of lineups involving 661 suspects in a study by Wright and McDaid (1996). The Innocence Project (http://www.innocenceproject.org/docs/Eyewitness_ID_Report.pdf) reports that 38% of the DNA exoneration cases involving eyewitness identification had more than one witness making the same misidentification and in 13% of cases there were 3 misidentifications.

In 1974 Robert Buckhout published an article in Scientific American which contained the following photographs of two different individuals who had been mistakenly identified and arrested for crimes actually committed by the person in the middle picture.

37

Below are an example of a case where three witnesses misidentified an innocent person who was convicted and cleared while in prison when the real perpetrator was apprehended after committing similar crimes (see Deckinga picture below) and a case with two misidentifications where the innocent person was cleared while awaiting sentencing (see Stevens picture below). A classic case in the literature involved seven misidentifications of a priest as a purse-snatcher. The real culprit, who wore a hat while committing the crimes, was identified during the priest’s trial—after the identifying witnesses had already testified (see picture below—from Ellison & Buckhout, 1981). Although the actual culprit was 14 years younger than the priest, if one covers their heads, one sees a similarity between the two men.

In these instances of misidentification (and most other I am familiar with) the innocent person and the actual perpetrator bore a resemblance to one another and similarity of appearance can clearly be one of the ingredients contributing to misidentifications.

Work by Clark & Wells (2008) provides us with the most extensive analysis of multiple “identifications” and their potential weaknesses. Clark and Wells assume independence of eyewitnesses—which is to say that one eyewitness’s decision does not influence other eyewitness’s decisions.   They assume equivalence among witnesses—e.g., that witnesses had equally good opportunities to view the perpetrator, received the same instructions when making identification attempts, viewed the same lineup, made their identification attempts at equal time periods after a crime, etc.

Clark & Wells note that they: “consider TP (Target Present) and TA (Target Absent) lineups together to evaluate diagnosticity… diagnosticity refers to the probability that the suspect is guilty given a particular response, whereas accuracy refers to the probability of a particular response given that the suspect is guilty or

38

innocent. From the standpoint of the trier of fact, diagnosticity is the key question, which can only be addressed by considering TP and TA lineups together.”

Clark & Wells use standard Bayesian analysis to examine a number of scenarios observed in the research literature. For these analyses they assume that half of lineups are TP and half are TA ()—this means that the “prior probability” the suspect is guilty is .5--and they illustrate the calculation or the “posterior probability” the suspect is guilty following various witness responses using the examples in their Table 1 below.

Clark and Wells examine a number of scenarios (using observed levels of performance from studies noted in the bottom left of each panel) using various combinations of suspect (S), filler (F) and no-choice (N) conditions. Note that suspects may be guilty or innocent depending on whether they are selected from TP or TA arrays. One can see that under varying conditions the diagnosticity of IDs is reduced or enhanced. These include variations the fairness of photoarrays and identification procedures. Perhaps the most important point, for present purposes, to emerge from these analyses is that multiple suspect identifications (SS and SSS) can, under some conditions, add nothing or very little to the probability the suspect is actually guilty.

Low similarity foils (which make the innocent suspect stand out) yield lower diagnosticity for suspect identifications.

Note that with the low similarity foils used in the example study suspect identifications had no diagnostic value (because the guilty and innocent suspects were equally likely to be chosen)—indeed, only foil

39

identifications had diagnostic value with respect to the suspects guilt (only because foil choices were more frequent in target-present arrays—an outcome that would be useless to police.

The Probative Value of Non-Identifications

Though naïve logic may suggest that multiple eyewitness identifications are stronger evidence of guilt than single identifications, it turns out that logic can and does break down under a number of circumstances (as evidenced by the DNA exonerations and my examples). For the past 30 years the eyewitness research community has recognized that non-identifications can be substantial evidence that a suspect is not guilty. As early as 1980 Wells and Lindsey questioned the criminal justice system's policy of treating eyewitness identifications of suspects as highly probative while treating non-identifications (both in the form of no-choices and foil choices) as non-probative. Wells and Lindsey used a Bayesian model of information to demonstrate that if an identification of a suspect increases the probability the suspect is the perpetrator, then a nonidentification must reduce that probability.

They illustrated these results using data from a study by Loftus (1976) in which (a) when the perpetrator was present in the lineup, 84% of witnesses chose him, 4% of witnesses made nonidentifications, and 12% chose a foil; (b) when the perpetrator was absent, 60% chose an innocent suspect 24% made nonidentifications and 16% chose foils. Wells & Lindsay compared the information gained (increase in the probability the suspect is guilty following a suspect identification or decrease in the probability the suspect is guilty following a nonidentification (both “none of the above” plus foil identifications). As shown in the figure below, nonidentifications were more probative of innocence than suspect identifications were of guilt.

40

Wells & Lindsay then compared the probative value of “none of the above” and foil identification responses and as shown in the figure below, found that “none of the above” responses were much more probative than foil identifications (which were roughly comparable to the probative value of suspect identifications as shown in the prior figure.

Wells and Olson (2002) carried the Bayesian analysis further by examining the information gain (changes in the probability the suspect is the perpetrator) as a function of the prior probability the suspect is guilty (based

41

on evidence prior to the identification results). They combined data from three different studies of sequential and simultaneous lineups and the figure below provides a good summary of their findings: filler identifications, non-identifications and “don’t know” responses all have probative value with respect to the suspect’s innocence and sometimes filler identifications and non-identifications have more probative value (with respect to innocence) than positive identifications of the suspect have with respect to guilt.

Clark, Howell and Davey (2008) carried these types of analyses further in a meta-analysis of 94 comparisons between target-present and target-absent lineups. Their analyses showed that suspect identifications were more diagnostic regarding a suspect’s guilt or innocence than other responses, but that non-identifications were also diagnostic of the suspect’s innocence. They note that their analyses indicate that a suspect identification is less informative if the lineup is biased and that the probative value of a suspect identification is undermined if suggestive lineup instructions increase the willingness of witnesses to make an identification. They note: “Nonidentifications also are straightforward. They are diagnostic of the suspect’s innocence. We reiterate the point made by Wells and Lindsay (1980) that nonidentifications are not merely “failures” to identify the suspect, but rather carry important information whose value should not be overlooked” (p. 211). They further note that foil identifications appear to be diagnostic of innocence when foils are selected based on their match to the description of the perpetrator given by a witness, but not when foils are selected on their match to the suspect

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues did not include non-identifications as a factor, nor did others such as Benton et al. (2006). As noted above, research on other factors certainly suggests there will be disagreement among laypersons about the impact that these factors have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

Stress

Despite the importance of knowledge in this area, one cannot simulate violent crimes or pose a threat to the well-being of naive experimental subjects. Researchers have therefore resorted to a variety of manipulations including the use of violent versus nonviolent videotaped crimes. Increased violence in videotaped reenactments of crimes has been shown to lead to decrements in both identification accuracy and eyewitness recall (Clifford & Hollin, 1981; Clifford & Scott, 1978; Johnson & Scott, 1976; Sanders & Warnick, 1980).

42

Deffenbacher (1983, 1991) pointed to the "Yerkes-Dodson Law" when explaining the effects of arousal on identification. Stress or arousal demonstrates an inverted U-shaped relationship with identification accuracy. Low levels of arousal, such as when waking up, produce low attentiveness; moderate levels of arousal, such as that felt by an athlete preparing to compete, serve to heighten perceptual and attentiveness skills; and, higher levels, such as that felt by an individual under extreme danger or duress, debilitates perceptual skills.

Deffenbacher, et al. (2004) published a meta-analysis of stress effect studies. The meta-analysis was conducted on 27 tests of the effects of heightened stress on identification accuracy. This sample included work published between 1974 and 1997, with a total of 1727 participants involved in relevant tests of the stress-- the mean proportions correct for TP lineups under high and low stress conditions were .39 and .59, respectively (with false alarm rates of 34% and 19% respectively). The effect of stress was larger for target-present than for target-absent lineups—that is, stress particularly reduced correct identification rates. The effect was twice as large for eyewitness-identification studies that simulated eyewitness conditions (e.g., staged crimes) than for studies that induced stress in other ways.

These effects are confirmed and extended in a study by Morgan et al. (2004). Morgan et al. examined the eyewitness capabilities of more than 500 active-duty military personnel enrolled in a survival-school program (see Table 3). After 12 hours of confinement in a mock prisoner-of-war camp, participants experienced both a high stress interrogation with real physical confrontation and a low-stress interrogation without physical confrontation. Both interrogations were 40 minutes long; they were conducted by different persons. A day after release from the camp, and having recovered from food and sleep deprivation, the participants viewed a 15-person live lineup, a 16-person photo spread, or a sequential presentation of photos of up to 16 persons. Regardless of the testing method, as the Table below shows, memory accuracy for the high-stress interrogator was much lower overall than for the low-stress interrogator.

High Stress Low StressHITS [Target-Present] Live line-up method 27% 62% Photo-spread method 36% 76% Sequential photo method 49% 75%

FALSE ALARMS [Target-absent] Live line-up method 45% 50% Photo-spread method 48% 61% Sequential photo method 0% 0%

Another more recent confirmation of findings comes from a study by Valentine & Mesout (2008) conducted in the “Horror Labyrinth” of the London Dungeon tourist attraction. The eyewitness study was conducted using self-report measures of anxiety previously validated against heart-rate changes. Tourists were recruited to participate in the study and they encountered a “scary person” while slowly walking around the labyrinth for approximately 7 minutes. About 45 minutes later, after they completed their tour the purpose of the experiment was explained to them and they were tested to see if they could identify the scary person from a nine-person photo-array. Complete data were obtained from 56 participants who were divided into two groups reflecting the 50% with the highest reported anxiety scores and the 50% with the lowest scores. As shown in the table below, 75% of the witnesses experiencing lower anxiety were able to identify the scary person/culprit but only 18% of those with higher anxiety. Over 50% of the higher-anxiety group incorrectly identified/guessed one of the eight “innocent” foils (which logically suggests that two of the five higher-anxiety witnesses also simply guessed the culprit).

43

Stress-induced retrograde amnesia. The preceding studies focused on the effects of stress on identification accuracy, but other studies have looked at the effects of stress or shock on other aspects of memory. One illustration of this phenomenon comes from a study by Loftus & Burns (1982) in which two groups of participants different versions of a filmed robbery. In one version a young boy was shot in the face and in the other the robbers left without shooting the boy. The participants who viewed the shooting produced more accurate recall of the scene but were less able to recall details shown just before the boy was shot—28% of the viewers of the nonviolent version of the film could recall the number on the boy’s shirt, compared to only 4.3% of participants who swathe shooting. In a more recent illustration of retrograde amnesia, Moreno et al (2006) had participants witness someone fall down some stairs (versus no fall). Compared with the no-fall group, significantly fewer participants in the trauma group recalled information from the presentation (66.6% vs. 21.8%). Of course, these studies do not produce trauma to the witnesses and certainly do not involve trauma to the brain which might well have occurred in the present case and might have produced even greater retrograde amnesia—but evidence on that point would require someone with expertise on physical trauma and memory.

Earwitnesses and Stress. Yarmey (1995) reports the results of a study by Yarmey and Pauley (n.d.) who presented particpants videotaped scenarios involving a masked perpetrator either uttering or not uttering abusive and sexually suggestive language and displaying or not displaying a weapon. The combination of the weapon and violent language produced significantly more false negative responses--guilty suspects in voice lineups were more frequently missed—the same pattern of results reported in the Deffenbacher et al. meta-analysis and in the Morgan et al. and Valentine and Mesout experiments.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues did not include stress as a factor, but Benton et al. (2006) reported that 68% of American respondents believed that stress could impair identification performance. This means there is still substantial disagreement among laypersons about the impact that stress can have on identification performance—eyewitness expert testimony can provide jurors with scientifically-grounded guidance on this issue and, as detailed above, increase juror sensitivity to the effects of stress.

Time Estimation and Exposure Duration

Daniel Yarmey, a Canadian psychologist, and his colleagues have conducted the most extensive research on the accuracy of estimates of duration accuracy. In a fairly elaborate study of time estimation Yarmey (2000) noted in his introduction:

Theoretical and empirical investigation of duration estimations date back to the nineteenth century with Vierordt's (1868) discovery that short intervals tend to be overestimated and longer ones under-estimated. This finding is now commonly referred to as `Vierordt's Law'.1 Consistent with this law, forensically related research has shown that witnesses tend to overestimate the duration of relatively short events lasting a few minutes or less (Loftus et al.,

44

1987; Orchard and Yarmey, 1995; Yarmey, 1993), and underestimate relatively long events lasting 20 minutes and more (Yarmey, 1990). … Yarmey and Yarmey (1997) found that witnesses in field situations (involving)… a 15-second interaction with a `culprit' … overestimated this encounter by a 3 to 1 ratio … (p. 47).

In his 2000 study Yarmey found that witnesses to what he termed “variant events” – non-routinized events which do not occur over and over again such as withdrawing money from an ATM machine – were, on average, overestimated by between 25% (for an event lasting 13 minutes) and more than 100% (events lasting 24 seconds or less). An event lasting 46 seconds was overestimated by 67% and events lasting one minute 20 seconds and two minutes 58 seconds were over-estimated by 40%.

Earwitness time Estimates. Durations of two-person conversations speaker statements have been found to be significantly overestimated. These include durations of 15 s (Yarmey et al., 1994), 30 s (Orchard & Yarmey, 1995), 72 s (Yarmey & Matthys, 1990), 3 min (Yarmey, 1992), and 8 min (Orchard & Yarmey, 1995).

In his review of earwitness research Yarmey (2007) notes that longer opportunities to listen to a speaker yield greater the accuracy of identification (Cook & Wilding, 1997b; Hammersley & Read, 1985; Orchard & Yarmey, 1995; Read & Craik, 1995; Yarmey, 1991b; Yarmey & Matthys, 1992). In Yarmey (1991) longer exposures produced higher correct identification rates (24% accuracy in a 6-voice lineup with an average of 3.2 minutes exposure, 30% with 4.3 minutes and 48% with 7.8 minutes). Mistaken identification rates in target-absent voice lineups were relatively high, averaging 48% across exposure times. Yarmey and Matthys similarly found relatively poor accuracy with speech sample durations of 36 seconds or less (31%) compared to samples of 2 minutes or longer (55%). Performance in these studies is obviously rather poor under the best of circumstances.

The time available for viewing a perpetrator is positively associated with the witness's ability to subsequently identify him. Shapiro and Penrod's (1986) study of more than 100 experiments showed that exposure time was a reliable predictor of accuracy. However, it should be noted that even a long exposure is no guarantee of accuracy—in the Morgan, et al. (2004) study the interrogations of soldiers whose memory was later tested lasted 40 minutes and yet the rate of correct identifications among high-stress soldiers ranged from only 27 to 49% when targets were present and foil identification rates approached 50% using standard lineups and photospreads.

Memon et al. (2003) examined the performance of young adults (ages 17-25) and older adults (ages 59-81) who viewed a simulated crime in which they saw the culprit’s face for 12 seconds or 45 seconds. . The crime was a BBC- produced video re-construction of a bank robbery BBC involving professional actor. No weapons are directly shows a well-dressed man entering the building and joining the queue. He ultimately passes a teller a note stating a robbery is underway and that he has a gun. He instructs the teller to fill a sports bag with money. The teller does so and the man calmly leaves the building accompanied by a man wearing a cycling helmet. Both men are shown getting into the passenger side of a waiting car. The complete recording, used in the long exposure condition, lasted 1 minute 40 seconds and involved a clear exposure to the target of approximately 45 seconds including several profile and front views. For the short duration recording extended views of the target were edited so the recording lasted 1 minute 7 seconds with approximately 12 seconds exposure to the target in full-face and profile view. Participants were then tested, within an hour, with a target present lineup or a target absent. Identification accuracy rates for both young and older adults were significantly higher under the long exposure conditions.

Bornstein

45

Proportions of Hits, False Alarms and Non-choices by Young Adults and Seniors in the Short and Long Exposure Groups with Target-Present and Target-Absent Lineups

Short Exposure Long ExposureHit False Alarm Non-Choice Hit False Alarm Non-Choice

Target Present Young .29 .42 .29 .95 .05 .00 Seniors .35 .45 .20 .85 .10 .05Target Absent Young -- .90 .10 -- .41 .59 Seniors -- .80 .20 -- .50 .50

Note. Correct responses (in italics) were hits in the target present condition and non-choices in the target-absent condition.

Bornstein, Deffenbacher, McGorty & Penrod (2011) conducted a meta-analyses of studies of the effects of exposure time using 33 independent estimates of effect size drawn from 25 articles. This meta-analysis confirms the significant effect of exposure time on identification accuracy—a mean effect size of r = .32.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues indicates 51% of respondents understood the effects of exposure time on identification accuracy. Their analyses did not include time estimation as a factor, nor did Benton et al. (2006). Some indication of the likely levels of agreement can be adduced from the fact that the mean percentage of jurors agreeing with expert conclusions across 16 different factors was 52% in Desmarais a & Read with the lowest rate of agreement = 40% for confidence and accuracy. This means there is widespread general disagreement among laypersons about the impact that these factors have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue and, as detailed above, increase juror sensitivity to the impact of short exposure.

Distance and Illumination

Wagenaar and colleagues have studied the ability of witnesses to recognize faces at varying distances and levels of illumination. The following table (Wagenaar & Van Den Schrier, 1996) illustrates light levels associated with various conditions.

46

Wagenaar & Van Den Schrier’s results from a study of recognition of strangers (1996—shown in the following table) led them to conclude: “The results clearly indicate a systematic increase of recognition performance with decreasing distance and increasing illumination. The end result is a practical rule of thumb, the Rule of Fifteen: Even in ideal conditions the desired diagnostic value of 15 (the ratio of correct to incorrect identifications) is reached at not more than 15 meters, not less than 15 lux.

Loftus and Harley (2005) have also examined performance as a function of distance and their paper gives some idea of how small faces are (in one's field of view) at 5, 43 and 172 feet (their Figure 1—in the published version their topmost picture is about 4” wide) and how much detail can be picked up at varying distances (their Figure 4).  They note (p.63) 25% and 75% identification accuracy for celebrity faces at 77 and 34 ft.  

47

48

Lampinen 2012: Participants viewed eight live human targets, displayed at one of six outdoor distances that varied between 5 and 40 yards. Participants were shown 16 photographs, 8 of the previously viewed targets and 8 of non-viewed foils that matched a verbal description of the target counterpart. Participants rated their confidence of having seen or not having seen each individual on an 8-point scale. Long distances were associated with poor recognition memory and response bias shifts.

Proximity to perpetrators matters. In general witness recall of crime events and actors is only moderately good, with people closer to events recalling more than those who are not central to events. Yuille and Cutshall (1986) examined the accuracy of police reports made by 13 witnesses to a gunshop robbery which ended in the shooting death of the robber. “Central witnesses,” who viewed most of the incident from a central vantage point averaged 71.9 details in the police interview versus 24.3 details for non-central witnesses. Accuracy levels for the two groups did not differ. Similar patterns of results were reported by Christianson & Hubinette (1993) in their study of 110 robbery witnesses—tellers recalled much more accurate information than did customers: e.g. regarding the robbers’ hood (95% vs 77%), robbers’ clothing (90% vs. 69%), robbers’ hair color (50% vs. 31%).

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues did not include illumination and distance as a factor, but Benton et al. (2006) reported that just 33% of their respondents understood that low light levels impair color vision. Some indication of the likely levels of agreement can be adduced from the fact that the mean percentage of jurors agreeing with expert conclusions across 16 different factors was 52% in Desmarais a & Read with the lowest rate of agreement = 40% for confidence and accuracy. This means there is widespread disagreement among laypersons about the impact that these factors have on identification performance—eyewitness expert testimony would provide jurors with

49

scientifically-grounded guidance on this particular issue and, as detailed above, increase juror sensitivity to its effects.

Cross-Race Identification

Research on cross-race identification impairment began forty years ago and has included various mixes of Caucasian, Asian, Hispanic, Black, and middle-eastern witnesses. For example, Platz and Hosch (1988) conducted a field study examining the identification performance of Mexican American, Black, and White convenience store workers in identifying customers who had interacted with them earlier in the day. A Mexican-American, Black, or White “customer” went into the store and made a fairly involved purchase or asked for directions from the clerk. Two hours later a pair of students posing as law office interns asked the clerk for help in identifying the individual using a series of lineup photos. Platz and Hosch found a significant cross-race impairment for all three of the racial groups: clerks of each group were better recognizing customers of their own race than customers from either of the other two races.

Luce (1974) tested the face recognition ability (for White, Black, Chinese and Japanese faces) of 65 Chinese and 60 Japanese S’s who attended a California state university. As emphasized by Luce, all had been raised exclusively in California, lived in the multi-ethnic city of San Francisco, attended a multi-ethnic university, attended school with all ethnic groups, and reported specific friendships among these ethnic groups. Both groups displayed some impairment with Black faces as compared to their own ethnic group (Figures 3 and 4 below).

Luce concluded (based also on impaired cross-race performance by Whites and Blacks): “With respect to other-race recognition, the current study reveals that despite lifelong association with a given ethnic group, there is no assurance that its members will be readily recognized. The study, therefore, offers tentative support for the “they all look alike” notion on the part of some ethnic groups towards other ethnic groups. Finally, implications of these findings for witness identification are clear. Reliable recognition performance cannot be assured whenever a member of one ethnic group is called upon to attempt to recognize a member of a different ethnic group. Perhaps the clain (sic) “Yes that’s the man who held me up”, must be looked upon with considerable skepticism when the alleged culprit is of a different ethnic group than is his identifier.” (p. 41)

50

The performance of several racial groups (including Whites, Blacks, Latinos, and Asians) in recognizing both White and Black faces was examined by Teitelbaum and Geiselman (1997).  Asian participants showed significantly lower hit scores on the Black faces than Black participants--as shown in this table (d' is a measure of "discriminability"--how effectively participants can identify faces)

Anderson (1999) tested thirty White, Black, Hispanic, and Asian observers (presumably from New York as the research was conducted at Brooklyn College--N=120), with uniracial parentage (both parents of the same race). All observers were required to have lived in a uniracial home for the first three years of life. Each group displayed fairly similar impairment on cross-race faces (as reflected in a measure of recognition performance (d’).

Participants with mixed parental backgrounds (N=107) were also tested and displayed impairment for groups not represented by one or the other of their parents. In particular, among Asian participants, the following figure shows that only participants with Asian/Black parent combinations did not show impairment on Black faces.

51

Anderson also tested 227 participants (with varied parental racial mixes) ability to recognize the perpetrator (White, Black, Asian or Hispanic) in a videotaped robbery (in which the presence of a weapon was implied). The video lasted three minutes and 15 seconds during which the perpetrator was visible of two minutes and thirty seconds. Participants were tested with a six-person perpetrator-present photoarray. As the following figure illustrates, the same pattern of results emerged with respect to Asian participants and Black perpetrators: only participants with one Black parent outperformed other combinations.

Gross (2009) also tested college-age participants from four major ethnic groups (White, Black, East Asian, and Hispanic) who were asked to view, remember, and later recognize faces from each of those groups. On average all groups had lived in California 11 years or more. As the following table demonstrates, there was significantly better performance (indexed by d’—with higher numbers indicating better performance) in all

52

groups on faces from their own ethnic backgrounds—with particularly poor Asian on Black performance. As the second table shows, Asians were 50% more likely to misidentify a new Black face than a new Asian face.

Meissner & Brigham (2001) have written what I consider to be the best review of research on the problems of what have sometimes been called other-race or cross-race identifications or own-race bias (ORB). Meissner and Brigham analyzed data from 39 research articles, with 91 independent samples involving nearly 5,000 witness participants. Measures of correct identifications and false alarm rates, and aggregate measures of discrimination accuracy and response criterion were examined. Overall, they reported that when the perpetrator is present the ratio of correct to incorrect identifications was 40% higher for same-race identifications. The ratio of mistaken identifications to correct rejections in target-absent arrays was 56% greater for other-race identifications. Overall, the ratio of correct to incorrect identifications was more than 2.2 greater for own-race faces as compared with performance on other-race faces.

Meissner and Brigham reported that cross-race impairment was greater with short exposure times (especially with respect to false identifications), longer periods of time between viewing and identification were associated with larger ORB (especially with respect to false identifications), Whites showed more ORB than other groups (especially with respect to false identifications of Blacks).

53

Meissner and Brigham also explored the question of whether cross-race contact might reduce these effects and found that overall, contact played a small role in the ORB, accounting for just 2% of the variability across participants. They also found that the amount of viewing time available to witnesses significantly influenced accuracy in the ORB--particularly through an increase in false alarms to other-race faces when study time is limited. That is to say, there is little reason to think other-race contact would play an important role in reducing own-race bias and some reason to think the brief exposure to the perpetrator would aggravate the problem.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that 51% of their respondents understood that cross-race identifications are less reliable than same-race identifications As with the other factors discussed above, this means there is significant disagreement among laypersons about the impact that these factors have on witness confidence—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

Changes in Appearance/Disguise/Obscured Views

Full face masks stockings, hats and hoods can be quite effective in diminishing the facial feature cues that are necessary for recognition. Cutler (Cutler, Penrod, & Martens, 1987a, 1987b; Cutler et al., 1986; O'Rourke et al., 1989) examined the effects of masking a target's hair and hairline cues on subsequent identification accuracy. In these experiments participants viewed a videotaped liquor store robbery and later attempted an identification from a videotaped lineup. In half of the robberies the robber wore a knit pullover cap that covered his hair and hairline. In the other half the robber did not wear a hat. The robber was less accurately identified when he was disguised. For example, in one of the experiments (Cutler et al., 1987a) 45% of the participants gave correct judgments on the lineup test if the robber wore no hat during the robbery, but only 27% gave a correct judgment if the robber wore the hat during the robbery. Pozzulo & Marciniak (2006) found that correct identifications fell by half when a target’s hairstyle changed (short versus long hair).

There is research (Mansour, et al., 2012) on the effects of stockings pulled over the head (and the effects of a variety of other disguises--as shown in the figure below which is taken from Figure 2 of the Mansour, et al., 2012 article). As the figure shows, the stocking they used blurred the facial characteristics of the target individual and compared to no-stocking views, correct identifications dropped from 80% to 55% and false identifications went from 21% to 36%.

54

More extreme disguises can produce even more substantial impairment. This is illustrated in a study Fisher and Cox (1995) where celebrities, known to the research participants, were identified by only 12% of participants when shown only the celebrities’ eyes (mouths alone produced recognition by 3% of participants). Eyes+noses+mouths yielded 57% recognition (versus 100% when full faces were shown). The figures below provide a sense of the increasing levels of exposure which produce increasing levels of accuracy. Of course it is important to emphasize that the participants were endeavoring to identify persons known to them through multiple exposures over extended periods of time—not the faces of strangers seen briefly.

In Shapiro and Penrod's (1986) meta-analysis, experiments were coded for whether or not the facial stimuli had undergone changes in facial features between the encoding and recognition phases. Facial transformations included changes in facial hair and deliberate disguises such as those used in the experiment just described. Nontransformed faces were more accurately recognized (d = 1.05; 75% non-transformed vs. 54% for transformed) and less often falsely identified (d = .40; 22% vs. 30%) than transformed faces.

Neither Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues nor Benton et al. (2006) provides insight into lay understanding of the effects of disguise on identification accuracy. This does beg the question of whether laypersons would have any clue about the abominable performance shown in the Fisher and Cox study of celebrity recognition from eyes and mouths!

Police vs Civilian Witnesses

Several researchers have compared the performance of police versus civilians on memory for criminal events and identification accuracy. Although there indications that police are better at remembering critical details in crime situations, the evidence (when examined across studies) also indicates police are slightly better than

55

civilians when making person identifications (even though, in individual studies, the differences are not statistically significant)..

However, if you average performance across those four studies, there is a slight advantage for police in making correct identification when the perpetrator is present (police 8% better) and a slight advantage for police when the perpetrator is absent (police 5% better).

Christianson, et al. (1997) tested Swedish police officers and civilians performance in a witness situation in which two versions of a film with a simulated violent robbery were shown. The perpetrator was either an immigrant or a native Swede… “a test of the proportion of subjects in each group who identified the correct person revealed no statistically significant differences. There were no significant differences between the two groups with regard to mistaken identification. Both groups were significantly more likely to make an incorrect rejection in the line-up when they had seen the immigrant perpetrator than when the perpetrator was a Swede… A chi-square test where the observed frequencies of mistaken identification of immigrant and of Swedish foils were compared to the expected frequencies, showed that innocent immigrants were mistakenly identified more often and innocent Swedes less often than what could be expected by chance” The results are shown below.

Christianson, et al. (1998) described their methods this way:.” Sixty-one university students, 31 teachers, 60 police recruits and 59 police officers with at least three years' professional experience were presented with a series of slides depicting a simulated crime (a woman going for a walk in a park and eventually being assaulted by a man with a knife). After the presentation of the slides, the four groups were tested on what they remembered of different aspects of the crime and their ability to recognize the perpetrator. ,,… Overall, the performance level is somewhat low. Although the police officers were more correct in their identifications as compared to the other groups, a test of the proportion of participants in each group who correctly identified the perpetrator revealed no statistically significant differences. Furthermore, no statistical differences were found between groups with regard to the proportions who had mistakenly identified someone as the perpetrator (false alarms).” Their results are shown below.

56

DeCarlo (2010) conducted two studies comparing police and civilian identification performance. The first experiment included 323 participants (165 police and 158 citizens). The mean age of police officers was 37.7 years and citizens was 40.7 years. All participants were from the same Northeastern state. Police came fr.om 6 agencies which ranged from 49 to 611 sworn officers (M = 173). The mean length of service for officers was 10.6 years. The Hit rates in this study were 61.6% for police versus 46.1% for civilians and the false alarm rates were 20.2% for police and 25.6% for civilians—neither difference reached statistical significance.

The second DeCarlo (2010) study with similar research participants found Hit rates of 44% for police versus 49.3% for civilians and the false alarm rates were 10.4% for police and 15.6% for civilians—neither difference reached statistical significance.

Identification from Faces versus other Features

 Pryke, et al. (2004) had witnesses attempt identifications based on face, body, or voice. The target was presented to students by a course instructor. The target spoke for approximately 1 minute. Following this participants first responded to a simultaneous lineup of faces, then attempted to identify the voice of the target by listening to 1-munute recordings of men’s voices one at a time. Finally, participants were shown a simultaneous lineup of bodies from the shoulder down. Correct face identifications were obtained from 74% of participants in the target present condition and the most-selected innocent face was selected by 20% of participants in the absent condition (using both those rates, 78% of choices were guilty). Bodies and voices proved less useful. For the body lineup, 53% of participants correctly identified the target and 81% falsely identified the most-selected innocent person (39% of choices guilty). For the voice lineup, 40% of participants correctly identified the target and 35% falsely identified the most-selected innocent person (53% guilty). Pryke, et al. also note that although nearly 75% of witnesses successfully chose the target from the facial lineup, only 55% of witnesses selected the target from the facial lineup and either or both of the voice and body lineups.

A second study added a test for clothing recognition with a similar pattern of results. In addition, the researchers provide information about the percentage guilty choices associated with various combinations of features. Faces prove to be an extremely important contributor to correct identifications. If a witness picked only a face they were correct 63% of time versus 25% for body only, and only combinations of selections including the face produced identifications of the guilty person—other selection combinations did not. Combinations of choices were not extremely common: in target-present conditions, 30% of witnesses made no selections, 20% made one selection, 33% selected two features, 10% three features and 7% made a selection on all four features.

Retention Interval57

Common sense tells us that memory declines over time. But common sense does not tell us about the pattern of memory decline. Basic research in memory had long-ago documented the loss of memory over time (with a typical pattern being one of rapid early loss of information followed by a more gradual decline). Can we expect eyewitness identification accuracy to decline as the time between the crime and the identification test increases? Shapiro and Penrod included retention interval in their 1986 meta-analysis. When studies that manipulated retention interval were grouped into long versus short time delays, longer delays led to 10% fewer correct identifications and 8% more false identifications. Across all experimental cells in all the studies examined in the meta-analysis (including those that did not directly manipulate retention interval) retention interval also proved an important determinant of correct identifications (r=-.11, p<.05).

Studies which test recognition accuracy at varying time intervals can tell us more than the simple comparisons in the meta-analysis or the field study (which lump observations into just two categories). Classic research by Egan, Pittner, and Goldstein (1977) found that the percentage of mock witnesses making false identifications increased significantly over time: 48% at 2 days, 62% at 21 days, and 93% at 56 days. The percentage of subjects making no errors declined from 45% (2 days) to 29% (21 days) to 7% (56 days).

As illustrated in the following table, memory loss is most rapid in the first hours and days following an event. Tollestrup, Turtle & Yuille (1994) studied suspect identification rates (suspects may, of course, be innocent) in actual robbery cases—the sharpest drop in suspect identification rates was reflected in the comparison of .5 days delay versus 3.6 days (an average drop of 8% per day, versus an average drop of less than .3% per day over the next 116 days).

Another study of suspect identifications from police files (Behrman & Davey (2001), showed that 55% of the witnesses picked the suspect from a photo lineup if the identification attempt was within a 0–7-day delay period (a time period in which substantial loss has already occurred) and 45% did so after a delay of greater than 7-days.

Deffenbacher, Bornstein, McGorty and Penrod (2008) demonstrate that loss of eyewitness memory is most rapid early on…in the first minutes and hours. The forgetting curves/loss of memory which emerges from the studies in that meta-analysis shows that memory has typically declined by 15%-20% within 2 hours, with a further 4-5% drop in the next 10 hours. From that point forward the loss is less dramatic…though it continues—as is evident in the studies cited above. The following figure plots the loss of memory found in studies (from the meta-analysis) with varying retention intervals.

58

The rapid early loss of memory is illustrated by the results from a study by Yarmey, Yarmey & Yarmey (1996) shown in the table below.

Earwwitnesses and Retention Interval. In some of the earliest earwitness research McGehee (1937) tested voice recognition at time intervals from 1 day to 5 months. She found over 80% accuracy up to 7 days, but at

59

2 weeks, accuracy dropped to 69% and after 3 weeks to 51%. At 3 and 5 months correct identification rates were 35% and 13% respectively. McGehee (1944), reported similar results: at 2-days, recognition accuracy was 85%. After 2 weeks, it dropped to 48%, to 47% after 4 weeks and 45% after 8 weeks. Clifford (1980) found significant changes in performance from 41% to 19% accuracy over a range of retention intervals of 10 minutes, 24 hours, 7 days, and 14 days. Clifford, Rathbom. and Bull (1981) examined accuracy at intervals of 10 minutes (55%), 24 hours (32%), 7 days (30%), and 14 days (37%). Clifford (1983) found voice identification rates declined over 1-week, 2-week, and 3-week delays --from 50% to 43% to 9% accuracy, respectively. Clifford and Denot (1982) had tested voice recognition among witnesses to a live scene--correct identification rates were 50% after 1 week, 43% after 2 weeks, and 9% after 3 weeks.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that just 53% of American respondents understood the “forgetting curve” reflecting rapid early loss of information from memory. As with other factors this means there is widespread disagreement among laypersons about the pattern of memory loss and the effect of the passage of time on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

(Mis-)Recognition of Familiar Faces

Research shows that, counter-intuitively, identifications of familiar faces - as with identifications of unfamiliar faces - are influenced by a host of variables, including interaction time, contextual information, expectation, postevent information, and own-race bias. ... these variables simultaneously increase the probability that an individual will identify a given face as familiar and inflate the individual's confidence in the identification, regardless of whether the face in fact is familiar or the identification accurate.

From the outset it is important to understand that witnesses in cases where the witness claims to know the perpetrator regularly make identification errors. In a 2003 study of 640 actual eyewitness identification attempts—56 of which involved suspects putatively known to the witness--Valentine et al. found that in 21% of cases where the suspect was supposedly known, the eyewitness failed to identify the suspect and in another 5% of cases the witness identified an innocent filler. In a larger study with 632 witnesses who said they knew the perpetrator (Memon et al, 2011) a comparable percentage--6.1%--of those witnesses misidentified an innocent filler. 

Factors known to influence stranger identifications can similarly influence non-stranger identifications. Kerstholt et al. (1992) report:  "For commonplace situational factors like poor viewing conditions and facial typicality, what is surprising is the magnitude of the deleterious effects. In one study, bad lighting and partially concealing the subject's face and hair reduced correct identifications of known subjects by 18 percent."

These results beg the question: why do witnesses misidentify people who are supposedly known to them?

A 1985 study by Young et al. (1985) included 922 records of instances in which people had difficulty recognizing other people. These included 314 misidentifications--272 instances in which an unfamiliar person was misidentified as a familiar person and 42 mistaken identifications of one familiar person for another. Young et al.’s summary (p. 505) of these cases is informative both with respect to the circumstances that gave rise to the errors (akin to the conditions in the present case) and the circumstances which gave rise to corrections of errors (not akin to the present case) : "…the diarists often tended to think that the unfamiliar person they misidentified was someone who was highly familiar (54 per cent) and often

60

seen (more than once a week in 45 per cent of the records). ... In short, then, the most salient features of misidentifications of an unfamiliar person as a familiar person are that they tended to be brief, and were often associated with poor conditions; they usually ended when a better view was obtained."

Factors other than actual familiarity can influence the perception of familiarity and (mis)-recognition. For example, Vokey & Read (1992) demonstrated that typical-looking faces were associated with stronger perceptions of general familiarity and led to false recognitions. They report: "for every unit increase in general familiarity, the false alarm rate increased by 5% or 6% " (familiarity was measured on 4-point scales, so the effect could be quite large).

It turns out that prior familiarity can actually reduce the reliability of an eyewitness identification. A 1995 study by Read demonstrates that greater familiarity can sometimes lead to greater error. Read’s study was an "experiment that examined the effect of increased interaction time on store clerks' ability to accurately recognize and identify a female target. Half of the clerks spoke with the female target for 30-60 seconds ("short-interaction" clerks); the other half were interviewed by the female target for between four and 12 minutes ("long-interaction" clerks). Two days later, all clerks were asked to pick out the targets' photographs in various photographic lineups; some lineups contained the targets' photos (target-present), while others did not (target-absent). On the whole, the most reliable effect of increased exposure to the female target was the clerks' greater willingness to identify someone from the lineups, but their choices were wholly unrelated to increased ability to choose the correct photograph."For both target-present and target-absent lineups, the long-interaction clerks more often erroneously picked the photograph of a  stranger as the interviewer. Moreover, the long-interaction clerks' performance deteriorated most severely – and most disturbingly for eyewitness identification purposes - in target-absent lineups: the percentage of clerks who incorrectly said the target was present more than doubled as their interaction time with the female target increased."

The misidentification error rates in the Read study were not trivial—63% for the short interactions but 88% for the long interactions.

The latest relevant study is by Pezdek & Stolzenberg (2014). Their results confirm the earlier findings and are particularly relevant to the present case--here are excerpts from the abstract--- sophomores (N 139) in two small private high schools viewed yearbook pictures of (a) graduated students from their school who were seniors (fourth year) when participants were freshmen (first year) (familiar) and (b) unfamiliar individuals, and responded whether each was ‘familiar’. The design was completely crossed; familiar faces at each school served as unfamiliar faces at the other school. Although individuals’ familiarity judgments were diagnostic of prior contact, accuracy was low (mean hit rate 0.42; mean false alarm rate 0.23), rendering an eyewitness’s report of having seen a perpetrator casually in the past of limited forensic value.

Of course, the identifications made in the Pezdek & Stolzenberg study were not made while the “witnesses” were viewing a shooting which occurred quickly and was committed by someone wearing a hat and wielding a weapon!

What the jury needs to hear in the present case is how witnesses can be wrong in what are asserted to be non-stranger cases, so they can judge the reliability of the identification in the present case instead of assuming that the witness is correct because the witness asserts they are familiar with the defendant. Part oft he point of expert testimony in the present case would be to explain factors which have been shown to influence stranger identification accuracy and which, on logical grounds—like lighting and concealment—could influence non-stranger identifications.

61

Voice Identification

In his review of research on voice identification, Dan Yarmey (2007) observes that due to everyday experiences in listening to and distinguishing among the voices friends and acquaintances, people often believe that speaker identification is highly accurate (Regina v. Morin, 1995; Yarmey et al., 2001). There is some evidence voice identifications of highly familiar voices can be fairly accurate. Yarmey, et al. (2001) compared recognition performance on voices of people who varied in degree of familiarity. High-familiar speakers included immediate family members and best friends, moderate-familiar voices included co-workers, team-mates and general friends, low-familiar speakers included casual acquaintances and neighbors with whom the listener would have spoken for a few minutes on occasion in any week and a fourth group included unfamiliar voices listened to for varying lengths of time. Yarmey, et al. found 79-85% correct and 5-13% incorrect identifications for familiar voices versus 49%-55% correct for low-familiar and stranger voices and 23% to 45% mistaken identifications). Yarmey et al. (2001) reported significant confidence-accuracy correlations for witnesses who were highly familiar with a speaker’s normal tone of voice (family members, best friends)--in contrast to non-significant correlations for speakers who were only moderately familiar (co-worker, teammate) or unfamiliar (stranger). Another group of participants (potential jurors) was asked to predict the percentage correct for these groups and predicted that between 99% and 84% of listeners would be correct across conditions in which the actual accuracy rate ranged from 85% to 49%).

With regard to strangers, accuracy of identification is much worse for voices as compared to faces (Hollien, Bennett, & Gelfer, 1983; McAllister, Dale, & Keay, 1993; Pryke, Lindsay, Dysart, & Dupuis, 2004; Yarmey et al., 1994). For example, in the Yarmey et al. (1994) study, witnesses interacted with target individuals for fifteen seconds in public settings and five minutes later were presented with a target-present or target-absent six-person photo or six-person voice lineup or a one-person voice or photo show-up (the voice presentations consisted of a 15-word sentence). The mean percentage correct identifications for the same targets presented in photo showups and photo lineups were 57% and 46%, respectively. However, the mean percentage correct identifications for these targets in voice showups and lineups were much lower: 28% and 9%, respectively. The mistaken identification rates in target-absent presentations were, respectively, 12% and 28% for photos and 30% and 98% for voices. There were no significant correlations between confidence and accuracy in the voice recognition conditions.

In Pryke, et al. the witnesses viewed and listened to the target for one minute and a few minutes later were tested using target-present and target-absent six-person photo and voice arrays (with voice recordings of about 1 minute each). Across target-present and absent presentations, 58% and 63% of identifications from faces-only were correct in two studies versus 17% and 0% from voices-only.

Yarmey (2007) notes that longer opportunities to listen to a speaker yield greater the accuracy of identification (Cook & Wilding, 1997b; Hammersley & Read, 1985; Orchard & Yarmey, 1995; Read & Craik, 1995; Yarmey, 1991b; Yarmey & Matthys, 1992). In Yarmey (1991) longer exposures produced higher correct identification rates (24% accuracy in a 6-voice lineup with an average of 3.2 minutes exposure, 30% with 4.3 minutes and 48% with 7.8 minutes). Mistaken identification rates in target-absent voice lineups were relatively high, averaging 48% across exposure times. Yarmey and Matthys similarly found relatively poor accuracy with speech sample durations of 36 seconds or less (31%) compared to samples of 2 minutes or longer (55%).

Faces Interfere with Voice Recognition. There is good reason to think that seeing and hearing someone actually interferes with voice recognition. Cook and Wilding (1997), for example, used a lineup procedure in which participants were presented with either an audio or audiovisual sequence involving both a female and a male speaker saying one sentence each. After 1 week, participants were presented with a target-present

62

audio-lineup. They found that the presence of the face at initial study significantly impaired subsequent voice recognition. McAllister, Dale, Bregman, McCabe, and Cotton (1993) confirmed this interference effect. They tested some participants in a face recognition lineup having either seen, or seen and heard the target. Other participants in were tested in a voice recognition lineup having either heard or seen and heard the target. Performance indicated significant impairment in the face+voice condition compared to the face or voice alone conditions. The impairment was not symmetrical; voice recognition was significantly impaired by presentation of the face at study, whereas face recognition was only partly impaired by presentation of the voice at study. A similar pattern of results has been reported by Stevenage et al. (2010) who found that face+voice increased false alarms in voice recognition compared to voice alone.

Changes in Tone and Voice Recognition. Saslove and Yarmey (1980) altered the tone of voice that was originally heard when it was presented later for identification. This had a significant effect on correct identification rates with an average of 86% correct identifications for the unchanged tone conditions and 39% correct identifications when the tone changed. In this study it was speculated that the correct identifications were higher than would be normally expected after a tone change because of the highly distinctive nature of the voices.

A study by Read & Craik (1995) particularly resonates with the present case (their study was motivated by an interest in a widely-publicized murder case in Canada). They said about their study:

Our participants listened to brief utterances spoken by different speakers. One of the utterances was the emotional shout "Help me, help me. Oh God, help me!" and others were nonemotional phrases such as, "I went to a movie last Saturday night." Some time later, participants were reminded of the target phrase and asked to identify the voice speaking that phrase from voice samples that varied in the degree to which they matched the content and tone of voice of the original utterance of that phrase. The questions addressed by the three experiments were as follows: First, what is the probability of recognizing a voice 2 weeks later when the listeners were unprepared for a later recognition test, when the initial sample is brief, and the tone and content of the voice have changed markedly? Second, is that probability changed when both content and tone of voice more closely approximate that which was originally experienced? Third, does mild prior familiarity with the speaking voice increase the probability of recognition? Fourth, is recognition influenced by the emotionality of the original utterance? (p. 8)

The test tape had two sets of six statements. The first set (conversational voices) was composed of each of six speakers reading the same 20-s excerpt from a play. The second set contained the original recordings of the six speakers giving their version of the emotional statement. There was a 10-s pause between eachvoice sample within the two tests. 17 days after hearing the original six statements participants attempted tochoose the voice that originally uttered the emotional statement after hearing the samples played twice . Recognition of conversational voices (20% correct) was no better than chance but performance rose to 66% in the identical voices condition.

Read & Craik (1995) conclude:

The probability of recognizing a voice some weeks after hearing it when the initial sample isbrief and the tone and content have changed is apparently at chance level. Participants did not exceed a chance level of recognition when trying to recognize the target voice in a 20-s conversational fragment (Experiments 1 and 2). Indeed, even when nominal content (the words) is held constant but tone of voice is changed, we found voice recognition to be at

63

chance level (Experiment 3). Voice recognition increases as the study and test voice samples become more similar in tone and content….

Experiment 2 also showed that prior familiarity with the speaking voice had no effect on voice recognition; as long as the voice was not recognized during study, there was no enhancement of recognition, even for moderate degrees of voice familiarity (having been heard for more than 1 hr). This is a surprising result perhaps, but Table 3 shows clearly that in none of the three test conditions did recognition levels associated with relatively familiar voices (3 and 4) exceed those associated with less familiar voices (1 and 2). Finally, Experiment 3 revealed no influence of the emotionality of the original utterance on its recognizability. (p. 15)

Decision Speed

Research indicates that correct identifications are made more rapidly than incorrect identifications (e.g. Brewer, Caon, Todd, & Weber, 2006; Brewer, Gordon, & Bond, 2000; Dunning & Stern, 1994; Sporer, 1993, 1994; Weber, Brewer, Wells, Semmler, & Keast, 2004). In addition, self-reports by witnesses that they have made an automatic, rather than a deliberative decision (Dunning & Stern, 1994), and that they have used an absolute, rather than relative or comparative judgment strategy (Lindsay & Bellinger, 1999), are associated with accurate identification decisions. Dunning and Perretta (2002) argued that a 10-12 second rule did a good job of classifying identification decisions as fast (and, therefore, likely to be accurate) versus slow (and, therefore, not likely to be accurate). However, in subsequent research the optimum time boundary was found to vary (a) across different targets and lineups (Weber et al., 2004), and (b) across retention intervals and lineup size (Brewer et al., 2006).

As an alternative method for exploiting decision times to differentiate accurate and inaccurate decisions, Brewer, et al. (2008) tested a procedure in which witnesses were shown an array for 8 seconds and either asked if they could make an identification or reject the array (optional condition) or were forced to make a decision (forced condition). If optional condition witnesses could not, they were given an additional 8 seconds to view the array, asked again about a decision and if they were not prepared to decide, were given unlimited time to view the array. Across two targets, the optional witnesses who decided at 8 seconds were correct in 69% of decisions, those forced to decide at 8 seconds were accurate in 44% of decisions, those who decided with 8 more seconds were correct in 44% of decisions and those with unlimited time were correct in 46% of decisions. Clearly, the fast decisions were the most reliable.

IX. Aspects of Identification Procedures that Affect Identification Accuracy

Composites

A considerable body of research on composite production systems indicates that the likeness of composites to their respective targets is low (Brace, Pike, Allen & Kemp, 2006; Davies & Valentine, 2006; Davies et al., 2000; Ellis, Davies, & McMurran, 1979; Frowd et al., 2005; Frowd et al., 2007; Kovera, Penrod, Pappas, & Thill, 1997; Paine, Pike, Brace, & Westcott, 2008; Wells & Hryciw, 1984). The Kovera et al. (1997) illustratrates the low resemblance of composites to target faces. In this study the researhers instructed 10 college students (who had attended 5 different high schools--2 students from each school) to create composites of 10 randomly selected faculty members (5) and class mates (5) from their own school using a computer-based composite system. This resulted in a total of 100 composites. A sample of 10 composites from each school was taken (5 composites of faculty and 5 of student) and those 50 composites were then shown to 50 students from each school who were told that these were composites of former high-school

64

classmates; they were asked to identify them, and make familiarity and confidence ratings. Only 3 identifications out of a total of 167 positive identifications were correct. Neither the witnesses nor the operators were effective in estimating when a likeness was of good or poor quality. The researchers concluded that “the findings…raise doubts about the likelihood that composites prepared under field conditions will yield a pinpointed identification of a perpetrator by individuals who know the perpetrator” (Kovera et al., p. 245).

More recently Brace, Pike, Allen & Kemp (2006) tested the identification rates of composites of well-known celebrities created by operators working alone from memory (23% identified) or from a photograph (22%). Of course, in crime situations the operator does not know what the target person looks like. The identification rates of composites created by the operators working with describers (the operator did not know who the describer was describing) was just under 13% when the composites were constructed from describer memory (the best analog to crime situations). The rates was just over 18% when the describer could work from a photograph.

Of course, both the Kovera et al. and Brace et al. studies were testing the usefulness of composites prepared by individuals describing people generally known to them (high school classmates and teachers and celebrities). In situations where composites are based on short views of strangers the usefulness of composites is similarly poor. For example, Koehn and Fisher (1997) had witnesses observed a target person for several minutes and two days later either constructed a facial composite of the target. The resulting composites were of low quality and of little value to individuals who saw the composites and tried to select the target from a group of similar-looking people—only a 7% correct identification rate.

Multiple Suspect Arrays

It is noteworthy that the first instruction in the NIJ Guide is to have only one suspect per array. To do otherwise—as was apparently done in this case—completely defies the logic of having a multi-person array. Multi-person arrays are intended to protect innocent suspects from over-eager and poor-memory witnesses. If a witness makes a “guess” in a one-suspect six-person array where the suspect is innocent, there is only a one in six chance they will pick the suspect. If everyone in the array is a suspect (and in the worst-case scenario, they are all innocent) an innocent person will be selected every time a choice is made. The protections afforded to innocent suspects are dramatically diluted in an all-suspect array—it is akin to showing six showups all at once. The foulness of this practice is underscored by the research cited above which shows that nearly 40% of witnesses who make identifications from single-suspect arrays are indisputably mis-identifying innocent people – fortunately those innocent people are known to be innocent because they were selected to serve as innocent foils. What do we imagine those 40% who pick innocent fillers are going to do if presented with all-suspect arrays!?!?

Mugshot Exposure Effects

Deffenbacher, Bornstein & Penrod wrote in 2006:

The phenomenon of the photobiased lineup has existed for as long as police have, in the furtherance of their criminal investigations, engaged in the practice of exposing eyewitnesses to mugshots. If a witness is exposed to mugshots subsequent to viewing a perpetrator and prior to an additional test of recognition memory, there is the possibility that exposure to the mugshot photos may bias the witness’s decision at that test, usually a photo or live lineup. Defense attorneys have labeled a photobiased lineup as one in which a possibly innocent person has been arrested because his/her photo was initially included in a set of mugshots but not identified by the witness;

65

subsequently, however, he/she was identified by that witness at a lineup. Any bias thus engendered would be due to the previous exposure of the witness to the mugshot photo of the suspect identified at the subsequent lineup.

In such cases, the witness may well have suffered a failure of memory for the circumstances of previous encounter of the face identified at the lineup. This sort of failure of memory for facial source or context is all the more problematic when viewing of the perpetrator has occurred under less than optimal viewing conditions (Brown, Deffenbacher, & Sturgill, 1977). Indeed, the United States Supreme Court had opined that this was the case, even before there was empirical support for their view (Simmons v. United States, 1968).

The first empirical examination of the possibility of bias imposed by mugshot exposure more generally as well as the more specific sort of bias that might be engendered by a photobiased lineup was that of Brown et al. (1977). In all three of their experiments, these investigators noted that recognition memory for a once-seen face is much greater than is the capacity to recall the circumstances under which that previously unfamiliar face had been encountered. In Experiments 2 and 3, they employed what we will henceforth refer to as a transference design to test for specific biasing effects of mugshot presentation between exposure to live targets and a subsequent lineup. …

At the lineup, eyewitnesses in Experiments 2 and 3 of Brown et al.’s (1977) study did indeed identify persons seen previously only in mugshots at significantly higher rates than they did previously unfamiliar foils. Interestingly, under Experiment 3’s viewing conditions of only a brief glimpse of target persons, witnesses were as likely to indict a person on the basis of a single mugshot encounter as they were a target person encountered only once, that is a target person not also presented in mugshots.

…Deffenbacher, Carr, and Leu (1981) tested whether a transference design would produce significant loss of memory for context of facial encounter at retention intervals of 2 min and 2 weeks. At both retention intervals, they found significantly higher false alarm rates to faces seen once previously in mugshots than to previously unfamiliar foils. …

In addition to the passive effect of familiarity gained by mere mugshot exposure to a previously unfamiliar face, Gorenstein and Ellsworth (1980) sought to determine whether there was a further increase in interference with witness memory for the target face engendered by the witness actively choosing the mugshot of an innocent person. Would witness commitment to a prior choice of a particular mugshot as the target person cause the mugshot image to be retained as strongly or even more strongly than that of the target? In brief, then, a mugshot commitment design involves a control group and an experimental group both being exposed to a target person, followed by mugshot exposure for the latter group. Witnesses in the experimental condition may be urged to pick an innocent person from the mugshots, despite the target’s face not being included. At the lineup the target face may or may not be included, but the previously chosen mugshot image definitely is. Gorenstein and Ellsworth (1980) found that whereas control witnesses identified the target from a six-person photo array at a significantly above chance rate (.39), mugshot commitment witnesses failed to identify her at an above chance rate (.22) but did identify their mugshot choice’s image at an above chance level (.44), a level somewhat greater than the rate at which control witnesses identified the target person. (pp. 288-290).

In their (2006) meta-analysis of 15 mugshot studies Deffenbacher, Bornstein & Penrod found the overall proportion of correct identifications from target-present and target-absent arrays following mugshot exposure was .43 whereas the comparable figure for the no-mugshot control conditions was .50 correct--indicating that mugshot exposure inserted between exposure to the target and a subsequent lineup test has a negative effect on eyewitness memory. In 11 of those 15 studies only innocent people were shown in the mugshots and the negative effects of exposure were even greater: the proportion of correct identifications from target-present and target-absent arrays following mugshot exposure was .48 versus .61 for the no-mugshot control

66

condition. If the witness actually made an identification of a mugshot image as that of the target, the subsequent appearance of that image or person at an identification procedure had an even larger negative impact upon lineup accuracy, with mean proportion correct of .32 for the mugshot commitment condition versus .55 for the control condition. Identification of an innocent mugshot interferes with a witness’s memory of the actual target so as to reduce both the hit rate and correct rejection rate for target present and target absent lineups, respectively.

We also specifically examined witness performance when they are shown mugshots and a subsequent presentation that does not contain the target/perpetrator. The overall false alarm rate at a lineup for mugshot exposure conditions was .37, and the comparable figure for the no-mugshot control conditions was .15, less than half the former rate. More alarmingly, if the witness selected an innocent person in a mugshot and that person was included in a subsequent target-absent array, the mean false alarm rate for the mugshot commitment condition was .53 versus just .20 for the control condition.

Goodsell, et al. (2015) further illustrates the problem—in their study 85% of participants made an erroneous choice from a 50-photo mugbook following a 25 second view of a perpetrator and the question was what those witnesses would do when shown a photoarray containing there erroneous choice or a replacement, the perpetrator or another look-alike, a non-selected filler from the mugbook and several new faces: “A commitment error occurs when participants search through mug shots for the perpetrator, make a selection from the mug book, and then select that same individual from a subsequent lineup, even when faced with the actual perpetrator. The present study replicated a performance decrement arising from the commitment effect. …Commitment resulted in participants not recognizing the perpetrator, which supported a memory-blending/replacement hypothesis.“ (p. 219)

As the following table shows—among witnesses shown an array including their erroneous choice (the CI column), nearly 70% stayed with their erroneous choice (vs 8% who shifted to the perpetrator and 11% who picked a different foil). When the perpetrator was not in the array, 81% stayed with their erroneous prior choice and 14% picked another innocent foil. When the erroneous choice was not included in the array (CNI column) there were more perpetrator choices (18%) but even more foil choices (the previously vied but nont-selcted foil jumped to 28% and other foils drew 15% of choices). When neither the previously chosen foil nor the perpetrator was present, 38% picked a foil they had seen in the mugshots. Goodsell, et al. conclude: “This study provides additional evidence that once an eyewitness has made an identification, no further identifications should be attempted because access to memory for the perpetrator is diminished.” (p. 219) and “Our results are consistent with prior research on mug shot exposure and suggest that witnesses who are exposed to a mug shot search should not participate in a subsequent lineup identification task. Witnesses who commit to a prior choice are unreliable because their memory for the perpetrator is affected. Those who made a commitment error failed to recognize the actual perpetrator..." (p. 231)

67

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that 61% of American respondents understood mugshot exposure can increase the risk of misidentifications. This means there is still substantial disagreement among laypersons about the negative impact that mugshot exposure can have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue and, as detailed above, increase juror sensitivity to its effect.

Unconscious Transference

Problems induced by seeing multiple people can go beyond recognition impairment and include what has been referred to as “unconscious transference” –a term applied when there is a “transfer of one person’s identity to that of another person from a different setting, time, or context” (Read et al., 1990, p. 3). Williams (1963) originally coined the phrase in reference an English murder case, in which an innocent person may have been executed. Williams indicates that one of the eyewitnesses who identified the defendant had briefly seen him before the crime and may have unconsciously transferred the identity of the actual perpetrator to the defendant.

Bystander and Prior Viewing Errors. Studies analogous to the case Williams (1963) described have been conducted by a number of researchers. A classic study of the confusion of roles was conducted by Buckhout, et al. (1974) who staged a mock assault before a group of 141 students. When the students were later asked to pick the assailant from a group of six photographs including the assailant, an innocent bystander to the assault and four fillers, 40% were able to select the assailant and 25% selected the person who had been at the scene as an innocent. Loftus (1976) demonstrated a substantial transference effect in which a bystander was selected at greater than chance rates by witnesses from a target-absent array. In that study college students listened to an audiotape of a crime and photographs from a high school yearbook

68

illustrated the characters in the story. After three days participants attempted to identify the assailant from one of two lineups which contained the assailant plus four unfamiliar faces, the face of an innocent bystander who was seen in the story plus four unfamiliar faces. When the assailant was in the lineup, 84 percent of the participants correctly identified him. However when the bystander was in the lineup, 76 percent of the participants misidentified someone, and 60 percent of these participants misidentified the familiar bystander.

Read et al. (1990) conducted five field experiments involving the prior viewing of witnessing of bystanders and targets by retail store clerks or by students in university classrooms going about their normal situation-appropriate activities. Tests of the difference between transference and control conditions in regard to the proportion of false positive identifications of the bystander produced evidence of transference effects. In one of their experiments control and transference participants viewed an “assailant” (an electrician who interrupted class to fix an electrical problem) and transference participants were also exposed to a bystander who was seen distributing exam materials in another classroom); control participants did not see the bystander. Two weeks later participants attempted to identify the assailant from a photoarray containing the bystander and four unfamiliar foils. Twenty-five percent of the transference witnesses misidentified the bystander, as compared with just 12 percent of the control subjects. In addition, 13 percent of the transference participants who misidentified him thought the assailant and the bystander were the same person and they had seen him in two different places.

Ross, Ceci, Dunning and Toglia (1994) students watched a film with a series of short segments showing teachers interacting with children. At the conclusion of the film a female teacher walks into a cafeteria and is robbed. Everyone saw the same film but transference witnesses also saw a male bystander to a group of children several minutes before the robbery. The assailant and bystander were both seen for about thirty-four seconds. Between three to five minutes later participants were asked to identify the assailant from an array that contained the bystander and four unfamiliar foils. Transference witnesses were nearly three times more likely to misidentify the bystander than control participants, (61 percent versus 22 percent) and were less likely to respond correctly that the assailant was not in the lineup (34 percent versus 64 percent). Sixty-six percent of transference participants mistakenly believed they had seen the assailant outside the cafeteria -- 95 percent mistakenly said the assailant was reading to the children. Few of the control participants (4 percent) made this error.

In Phillips, et al. (1997) 650 undergraduates viewed a videotaped robbery. Participants were significantly more likely to select a bystander from a photoarray than the actual perpetrator and were more confident in their misidentifications of the bystander. Participants shown a photoarray without the bystander were over six times more likely to select the perpetrator than participants shown an array that included the bystander. When participants saw a similar film showing both the bystander and the perpetrator for a few seconds in the same frames of the video the majority of participants correctly selected the perpetrator rather than the bystander (by a ratio of 3: 1)--but among those who made a misidentification, the bystander was selected more often than chance (.34 vs .20),

In their 2006 meta-analysis Deffenbacher, Bornstein & Penrod reviewed eight studies using these types of designs and found that the effect is a reliable one with an overall transference misidentification rate of 27% for previously-seen bystanders versus 17% for the same individuals when they had not been viewed as bystanders.

Role Confusion. In related research Geiselman, McArthur and Meerovitch (1993), participants viewed a videotape of a staged robbery and were shown a photoarray containing either the assailant, an accomplice, or neither perpetrator. The results revealed that half of the witnesses who recognized the accomplice mis-identified him as the assailant, and overall, witnesses displayed a three-to-one bias toward labeling any

69

person they identified as being the assailant. After a one-week delay the participants were more confident in mis-identifying the accomplice as the assailant than correctly identifying him as the accomplice. The authors note that this result: “is important because it suggests that with a time delay, a witness's high level of confidence cannot be used to alert investigators or jurors that the witness is incorrect about the role of the identified person. Instead, it appears that the witness becomes confident in relying on his or her bias to label any recognized person as the central figure in the crime.” [p. 12]

Geiselman, et al 1996 used the same driver/assailant tape in a second experiment: in about three-fourths of the instances where a participant made a identification in an assailant-present array, the participant correctly labeled the identified person as the assailant. In contrast, less than half of the participants correctly identified the accomplice as the accomplice. Participants making false-identifications labeled the person as the assailant two-thirds of the time. A similar pattern was observed for the target-absent array as well: many participants assumed that different pictures of the assailant appeared in each array. Overall, 69 percent of the participants who made false identifications on both arrays labeled the two persons with the same role and 84 percent of the time that role was the assailant.

Change Blindness. Another related body of research involves work on “change blindness.” In one illustrative study, Simons and Levin (1998) devised a scenario in which confederate asked pedestrians for directions around campus. During that interaction, one confederate swapped places with a third confederate while the pedestrian’s view was blocked and the third confederate continued the conversation. Surprisingly, half of the pedestrians did not notice the swap.

In a paper entitled “‘Unconscious Transference’ Can Be an Instance of ‘Change Blindness” Davis, Loftus, Vanous & Cucciare (2008) reported experiments investigating the role of change blindness in mistaken eyewitness identifications of innocent bystanders. Two innocent people appeared in a 4-minute scenario filmed in a supermarket. The ‘continuous innocent’ (CI) walked down the liquor aisle and went behind a stack of boxes from which the perpetrator emerged and shoplifted a bottle of liquor. The scenario promoted the illusion of continuity between the perpetrator and innocent bystander. A second, ‘discontinuous innocent’ (DI) was shown immediately afterward in a produce aisle. All three individual were shown for 13 seconds. In one study (see Table below—where OI=Other Innocent fillers) 74% of participants made a mistaken identification, 60% failed to notice the change between the CI and the perpetrator, and among those who did not notice the change, more misidentified the ‘CI’ than the ‘DI’, a pattern that did not hold for those who did notice the change. Another study demonstrated that participants were less likely to notice the change when they were distracted while watching the video.

70

Change blindness has also been investigated in direct interactions which are closer to robbery scenarios where a robber and a clerk interact across a counter (Levin, Simons, Angelone, & Chabris, 2002). In that study, participants approached a counter where they briefly interacted with the experimenter, signed a consent form and handed it to the experimenter. At that the point the experimenter ducked down behind the counter—apparently to file the form---and a different experimenter rose to finish the interaction. 75% failed to notice the change in experimenters. Indeed, almost half had heard about change blindness research—some within the last hour. Nonetheless, two-thirds of these participants failed to detect the change.

It should be noted that the transference effect is not a uniquely forensic phenomenon. In addition to the previously cited studies of eyewitness memory, memory researchers have for the past quarter of a century obtained strong support for a variety of transference-related phenomena. These phenomena include “familiarity without awareness” Mandler (1980), “retention without awareness” (Roediger, 1990), the “false fame effect” (Jacoby, Kelley, Brown, & Jasechko, 1989), and failures in reality monitoring (Johnson & Raye, 1981) and source monitoring (Johnson, Hashtroudi, & Lindsay, 1993).

Second Identification Attempts

In a study of the suggestibility of showups, Dickinson (2006) found—in line with the mugshot exposure effects noted above, that when an innocent suspect appears in both a showup and a lineup, the chances the suspect be identified increases substantially. Indeed, merely telling a witness that he/she will have a chance for a second viewing may be sufficient to affect the accuracy of the identification (Duckworth & Kreiner, 2009). In that study the researchers manipulated whether participants were informed that they would have a second chance to view a sequential lineup via a randomly-ordered lineup (administered on a compute) versus a control condition in which participants viewed the lineup only once. Participants who were not told they would have a second viewing made more correct identifications of the perpetrator and made fewer incorrect rejections of the lineup.

In related study, Godfrey & Clark (2010) conducted two experiments measuring the effects of multiple identification procedures when the time between the first and second attempted identification varied

71

(Experiment 1: one week delay; Experiment 2: 30 minute delay). Following the viewing of a robbery video, witnesses were asked to write down anything they could remember about the crime and the perpetrator together with their confidence in heir ability to identify the perpetrator. Witnesses were then shown either a photo showup or participated in a filler task. In the first experiment, witnesses returned to the laboratory the following week and were shown either a guilty-suspect or innocent-suspect lineup. In the second experiment, participants saw the showup, were given a second filler task and then were presented with the lineup only thirty minutes after the showup. When the interval was 30 minutes few participants changed from a showup nonidentification to a suspect-identification in the lineup, whereas nearly half of the participants made this change when the interval was one week. When the interval was 30 minutes, there was an increase in foil identifications from showup to lineup, but no change in correct or false identifications. When the interval was long, there was an increase in both correct and false identifications from showup to lineup, but no change in foil identifications. Suspects (both guilty and innocent) were more likely to be identified in multiple identification procedures but the identifications had little probative value. Witness confidence was not a reliable indicator of eyewitness accuracy.

In a more recent study (Steblay, Tix, & Benson, 2013) participants watched a crime video crime and were asked to identify the perpetrator from perpetrator-present and –absent lineups of the same format (sequential or simultaneous), with a two-week time period between the two lineups. Participants who made choosing errors in the first lineup continued with, rather than corrected, their identification error at the second lineup. There was no difference in confidence between those who made correct and false identification—though choosers were significantly more confident than non-choosers.

A related issue is whether witnesses are given an opportunity to see all the photos in a sequential lineup more than once. Horry, Brewer, Weber and Palmer (2015) tested the effects of two “laps” (either optional or required—the latter is the practice in the UK) in a sequential lineup on witness’ identification decisions. Choosing rates for both fillers and perpetrators increased in the second lap (40% changed their original response--mostly from nonidentification to positive identifications-- but identification accuracy did not differ as a function of number of laps. However, Abraham (2012) found overall accuracy was higher when witnesses viewed each mugshot once (47.6%), compared to twice (19.7%) and one viewing was better than two laps accompanied by a warning about the potential negative impact of multiple viewings (31.1%).

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues indicates that 68% of laypersons thought unconscious transference impacted identification performance, but Benton et al. (2006) reported that 30% of American respondents held that belief. In either event these results mean there is still substantial disagreement among laypersons about the impact that unconscious transference can have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue and, as detailed above, increase juror sensitivity to the effects of unconscious transference.

The Effects of Suggestive Identification Procedures

Eyewitness identifications take place in a social context in which the eyewitness's performance can be influenced by his or her expectations and inferences, which in turn can be influenced by the verbal and nonverbal behaviors of investigators, the structure of the identification test and the environment in which the identification test is conducted. At the outset, the fact that the investigator has taken the time to put together a photoarray, made an appointment with the eyewitness and driven across town to meet her may suggest to the eyewitness that the police think they have the perpetrator. This is likely to be so even if the police are unsure about whether the suspect is the perpetrator. For example, the police might be conducting a

72

photoarray identification test as a "shot in the dark" or in order to eliminate a suspect. But there would be no reason for an eyewitness to know this and he or she may be inclined to think that the police are reasonably certain that they know the identity of the perpetrator. It is reasonable to expect that the eyewitness will be motivated to act on this inference and make a positive identification--whether correct or incorrect.

The tendency to make a positive identification may be further strengthened by a number of factors. One such factor is the degree to which the investigator pressures or encourages the eyewitness into participating in an identification test. An eyewitness who is told that it is very important for her to view a photoarray or lineup immediately is more likely to infer that the investigators have identified the perpetrator than is an eyewitness who is told that she could drop by the station whenever it is convenient for her. Another factor might be the zealousness of the investigator. The more zealous the investigator, the more confident the eyewitness might be that the investigator knows the perpetrator's identity. A cooperative eyewitness might therefore "do her part" by making a positive identification. While there is limited research testing hypotheses concerning the effects of effort exertion and zealousness on the part of investigators [Phillips, et al., 1999], if these factors operate as described, we would deem them suggestive procedures for the reasons described below.

Suggestive procedures are aspects of the identification test that are under the control of police investigators and that enhance the likelihood that an eyewitness will make a positive identification--whether it is correct or not. The criminal justice system acknowledges that mistaken identifications occur and that suggestive identification procedures used by police or prosecutors enhance the likelihood of mistaken identifications. Said Supreme Court Justice William Brennan, writing for the majority in U.S. v. Wade (1967), "The vagaries of eyewitness identification are well-known; the annals of criminal law are rife with instances of mistaken identification . . . A major factor contributing to the high incidence of miscarriage of justice from mistaken identification has been the degree of suggestion inherent in the manner in which the prosecution presents the suspect to witnesses for pretrial identification. ... Suggestion can be created intentionally or unintentionally in many subtle ways."

Showups

As evidenced by a Kassin et al. (2001) survey of eyewitness experts, the majority of psychologists are critical of the showup procedure. Criticisms tend to fall into two categories (Malpass & Devine, 1981; Wells, Lindsay & Ferguson, 1979). The first category concerns eyewitness reliability (Ellison & Buckhout, 1981). In a one-person identification task the witness is asked to choose between two alternatives: ‘Yes’ this is the person that I saw commit the crime or ‘No’ this is not the person I saw. Logically, this task does not provide innocent suspects with the same protections that are associated with standard lineups. The greater number of plausible choices in the standard lineup (6), versus the showup (1), reduces the likelihood that a suspect will be selected purely on the basis of chance (Malpass & Devine, 1981). Suspects in lineups are somewhat protected from witnesses who are merely guessing, from those who think they are helping police by giving a positive identification and those that are motivated to have someone held responsible for the crime.

Lineups also provide the added benefit of identifying unreliable witnesses. Because lineups consist of a suspect and known innocent fillers, when a witness identifies a known filler, it reflects their poor memory for the perpetrator. Filler identifications serve the purpose of weeding out those “witnesses who do not possess very good information about the identity of the offender but who feel they must attempt identification” (Malpass & Devine, 1984, p. 87).

73

A second category of concern is the unique environmental cues associated with the one-person showup. It is not uncommon in showups for a witness to view the suspect in custody (e.g., handcuffed, and/or in or near a police vehicle) or at the scene of the crime. Custody is highly suggestive of guilt and may increase the likelihood that the witness will make a positive identification. To date relatively little of the research on the effects of the number of alternatives presented to witnesses has studied the effects of custodial suggestiveness—rather, the research has focused on the effects of one versus multiple-member arrays.

A meta-analysis of 12 studies involving 3013 participants compared witness performance in one-person presentations to witnesses [typically one-photo presentations in which custodial biases do not play a role) versus [typically] six-person presentations indicates that mistaken identifications are significantly more likely with show-ups. In that meta-analysis, Steblay, Dysart, Fulero, & Lindsay (2003) found that in target-present presentations, the showup and lineup will produce approximately the same results (46% vs. 45% correct identifications), but that dangerous false identifications of innocent suspects are much higher in showups than lineups, 23% versus 10%. A major reason for the difference is that when there are no fillers/foils, witnesses who are making guesses have only one person to guess—the suspect.

A more recent experiment (Wetmore, et al. 2014) compared showups and lineups on immediate tests and tests conducted after 48 hours. The authors report: “Participants (N = 1486) viewed a mock-crime video and then were presented with a showup or a simultaneous lineup, either immediately or a 48 h delay. Receiver operating characteristic (ROC) analyses revealed that a showup never resulted in better identification accuracy than a lineup. Their results are illustrated in the following figures…in which the desirable outcome is the represented by lines which are closest to the upper left corner of the figure (higher correct ID rates coupled with low mis-ID rates).

Figure: ROC curves comparing showups (SU) and lineups (LU; top panel), showups and lineups at the different retention intervals (middle panel), and the lineups separated by fairness (bottom panel). The thick diagonal line in each panel indicates chance performance.

Wetmore, et al. conclude: “lineups were more diagnostic than showups. This was true when showups were compared to lineups immediately and after a 48 h delay. These results replicate previous work comparing these identification procedures (Gronlund et al., 2012)… It also was true that both fair and biased lineups resulted in better diagnostic accuracy than showups…”

More recently, similar results have been reported by Key, et al. (2015) as shown in the following figure:

74

Key, et al. note: “In contrast to …similar correct identification rates, the showup had more than double the percentage of false identifications obtained in lineups.” p. 14. In addition: “Calibration curves revealed that showups yielded a poorer confidence–accuracy relationship than fair lineups. Not only do witnesses viewing a showup exhibit poorer discriminability, but they also express overconfidence when doing so. Specifically, these witnesses are overconfident at high levels of confidence, which is problematic for the legal system because confident witnesses are perceived as more believable by jurors (Cutler et al., 1990).” p. 16.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that 59% of their respondents understood that XXX. As with the other factors discussed above, this means there is significant disagreement among laypersons about the impact that these factors have on witness confidence—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

Clothing Bias

Clothing bias refers to bias induced by presenting a suspect in clothing matching the description of clothing worn by the perpetrator. It has been found to have an effect on identification accuracy in lineups, in mugshots and in showups (Dysart & Lindsay, 2007; Yarmey et al., 1996). The rate of correct identifications in these procedures have not been found to be affected by clothing, but false identifications have been found to be more likely when clothing is similar (Dysart et al., 2006). Accuracy of eyewitness identifications has been shown to be reduced in situations that prompt witnesses to make more choices (Clark, Howell, & Davey, 2008), and presentation of the suspect in the clothing similar to that worn by the perpetrator may have that effect. Dysart, Dupuis, and Lindsay (2001) found strong evidence of clothing bias in showups, and further found that the type of clothing worn by the perpetrator may interact with other factors, such similarity of the innocent suspect and target. When their target wore distinct clothing (a black Harley-Davidson t-shirt with a motorcycle, blue eagle wings and the Harley-Davidson logo on the front), the rate of false identifications increased significantly for both high and low similarity innocent suspects when they wore identical or similar distinct clothing (a black Harley-Davidson T-shirt with a motorcycle, howling wolf, and the Harley-Davidson logo on the front). When high similarity innocent suspects wore identical or similar distinct clothing false identification rates roughly doubled (from 23% to 50%). For innocent suspects rated low in similarity to the perpetrator the effects were even more pronounced--there were no false identifications were when different clothing was worn, but 37% of witnesses misidentified him when identical distinct clothing was worn and 14% when similar distinct clothing was worn.

Lineup Instruction Bias

75

Instructions given to an eyewitness prior to an identification test can vary in their degree of suggestiveness. Suggestive instructions strongly convey to the eyewitness the impression that the suspect is in fact in the photoarray or lineup, thereby increasing the likelihood that the eyewitness will make a positive--though not necessarily correct--identification. How can instructions convey this message? The following experiments examine this question empirically.

Buckhout, Figueroa & Hoff (1975) examined the influence of suggestive instructions combined with suggestive presentations of a photoarray. Subjects who received "high-biased" instructions and viewed a "leading photoarray," identified the assailant at a significantly higher rate10 (61.3%) than subjects in the remaining three conditions (which averaged about 40%). In short, this experiment demonstrates that suggestive instructions combined with a leading photoarray can lead to higher rates of identification.

1GENERAL REFERENCES

Behrman, Bruce W. & Davey , Sherrie L. Eyewitness Identification in Actual Criminal Cases: An Archival Analysis, Law and Human Behavior 2001 25, 475-491.

Bothwell, R. K., Deffenbacher, K. A., & Brigham, J. C. (1987). Correlation of eyewitness accuracy and confidence: Optimality hypothesis revisited. Journal of Applied Psychology, 72, 691-695.. Journal of Applied Psychology, 1987 Nov Vol 72(4) 691-695.

Brewer, N., Keast, A. & Rishworth, A. (2002). The confidence-accuracy relationship in eyewitness identification: The effects of reflection and disconfirmation on correlation and calibration. Journal of Experimental Psychology: Applied, 8, 44-56.

Brigham JC. 1990. Target person distinctiveness and attractiveness as moderator variables in the confidence–accuracy relationship in eyewitness identification. Basic and Applied Social Psychology 11: 101–115.

Brigham, J. C., & Bothwell, R. K. (1983). The ability of prospective jurors to estimate the accuracy of eyewitness identifications. Law and Human Behavior, 7, 19-30.

Brigham, J. C., Maas, A., Snyder, L. D., & Spaulding, K. (1982). Accuracy of eyewitness identifications in a field setting. Journal of Personality and Social Psychology, 42, 673-681.

Brigham, J. C., & Wolfskeil, M. P. (1983). Opinions of attorneys and law enforcement personnel on the accuracy of eyewitness identification. Law and Human Behavior, 7, 337-349.

Brown, E., Deffenbacher, K., & Sturgill, W. (1977). Memory for faces and the circumstances of encounter. Journal

Buckhout, R., Alper, A., Chern, S., Silverberg, G., & Slomovits, M. (1974). Determinants of eyewitness performance on a line-up. Bulletin of the Psychonomic Society, 4, 191-192.

Charman, S.D., & Wells, G.L. (2006). Applied lineup theory. In R.C.L. Lindsay, D.F. Ross, J.D. Read & M.P. Toglia (Eds.), Handbook of eyewitness psychology (Vol. 2, pp. 219–254). Mahwah, NJ: Erlbaum.

Clifford, B. R., & Hollin, C. (1981). Effects of the type of incident and the number of perpetrators on eyewitness memory. Journal of Applied Psychology, 66, 364-370.

Clifford, B. R., & Scott, J. (1978). Individual and situational findings in eyewitness testimony. Journal of Applied Psychology, 63, 352-359.

Cutler, B. L., & Penrod, S. D. (1989). Forensically relevant moderators of the relation between eyewitness identification accuracy and confidence. Journal of Applied Psychology, 74, 650-652.

Cutler, B.L., & Penrod, S.D. (1989a). Moderators of the confidence-accuracy correlation in eyewitness identifications: The role of information processing and base-rates. Applied Cognitive Psychology, 3, 95-107.

76

Malpass & Devine (1981a) staged an act of vandalism during a lecture to about 350 undergraduate students, 100 of whom were asked to identify the vandal from one of two live lineups within the next three days. Half of the eyewitnesses were given instructions suggesting the perpetrator was in the array and the remaining eyewitnesses were given an "unbiased" instruction: "The person . . . may be one of the five individuals in the lineup. It is also possible that he is not in the lineup.” Among eyewitnesses who viewed a vandal-absent lineup, 78% of those who received biased instructions made a positive identification. Of course, all of whom were incorrect. In contrast, only 33% of those who received unbiased instructions made an incorrect positive identification from the vandal-absent lineup. Thus, significantly more false identifications were obtained with biased instructions than with neutral instructions.

Cutler, Penrod & Martens, (1987a) tested instructions that are more subtly suggestive. As in Malpass & Devine (1981a), when the thief was absent from the lineup, eyewitnesses who received biased instructions

Cutler, B. L., Penrod, S. D., & Dexter, H. R. (1989). The eyewitness, the expert psychologist, and the jury. Law and Human Behavior, 13, 311-332.

Cutler, B. L., Penrod, S. D., & Martens, T. K. (1987a). Improving the reliability of eyewitness identifications: Putting context into context. Journal of Applied Psychology, 72, 629-637.

Cutler, B. L., Penrod, S., O'Rourke, T. E., & Martens, T. K. (1986). Unconfounding the effects of contextual cues on eyewitness identification accuracy. Social Behaviour, 1, 113-134.

Dahl, M., Allwood, C. M., Scimone, B. & Rennemark, M. (2015) Old and very old adults as witnesses: event memory and metamemory, Psychology, Crime & Law, 21:8, 764-775, DOI: 10.1080/1068316X.2015.1038266

DeCarlo, J. (2010). A Study Comparing the Eyewitness Accuracy of Police Officers and Citizens. Dissertation completed at City University of New York.

Deffenbacher, K. (1980). Eyewitness accuracy and confidence: Can we infer anything about their relationship? Law and Human Behavior, 4, 243-260.

Deffenbacher, K. A. (1991). A maturing of research on the behaviour of eyewitnesses. Applied Cognitive Psychology, 5(5), 377-402.

Deffenbacher, K. A.; Bornstein, B. H.; & Penrod, S. D. (2006). Mugshot exposure effects: retroactive interference, mugshot commitment, source confusion, and unconscious transference. Law and Human Behavior, 30, 287-307. http://dx.doi.org/10.1007/s10979-006-9008-1

Deffenbacher, K. A.; Bornstein, B. H.; Penrod, S. D. & McGorty, K. (2004). A Meta-Analytic Review of the Effects of High Stress on Eyewitness Memory. Law and Human Behavior, 28, 687-706.

Deffenbacher, K. A., Carr, T. H., & Leu, J. R. (1981). Memory for words, pictures, and faces: Retroactive interference, forgetting and reminiscence. Journal of Experimental Psychology: Human Learning & Memory, 7, 299-305.

Deffenbacher, K. A. & Loftus, E. F. (1982). Do jurors share a common understanding concerning eyewitness behavior? Law and Human Behavior, 6, 15-30.

Douglass, A. B., Neuschatz, J. S., Imrich, J. F., & Wilkinson, M. (2010). Does post-identification feedback affect evaluations of eyewitness testimony and identification procedures? Law and Human Behavior, 34(4), 282–294. doi:10.1007/s10979-009-9189-5

Douglass, Amy Bradfield & Steblay, Nancy. (2006). Memory Distortion in Eyewitnesses: A Meta-Analysis of the Post-Identification Feedback Effect, 20 Applied Cognitive Psychol. 859.

Egan, D. Pittner, M. Goldstein, A.G. (1977) Eyewitness identification: Photographs vs. live models. Law and Human Behavior, 1, 199-206.

Fahsing, I.A., Ask, K. & Granhag, P.A. (2004). The Man Behind the Mask: Accuracy and Predictors of Eyewitness Offender Descriptions, Journal of Applied Psychology, 89, 722–729

Fawcett, Jonathan M.; Russell, Emily J.; Peace, Kristine A.; Christie, John (2013).  "Of guns and geese: A meta-analytic review of the ‘weapon focus’ literature". Psychology, Crime & Law 19 (1): 35–66.

Fisher, G. H., & Cox, R. L. (1975). Recognizing human faces. Applied Ergonomics, 6, 104-109.77

were significantly more likely to make a false identification (90%) than were eyewitnesses who received unbiased instructions (45%). The effects of subtly biased instructions were examined in three additional experiments (Cutler, Penrod & Martens, 1987b; Cutler, Penrod, O'Rourke & Martens, 1986; O'Rourke, Penrod, Cutler & Stuve, 1989). Overall, there was strong evidence for the influence of suggestive instructions on false identifications in data from 895 participants in crime simulation experiments.

Instruction bias research was reviewed by Steblay in a 1997 meta-analysis in which she cumulated the results of 22 different experimental studies [in 18 different papers] of the effects of biased instructions involving nearly 2600 witness-participants. She found that biased instructions were particularly harmful in target-absent lineups in which witness accuracy declined from 60% (unbiased lineups) to 35% (biased lineups). Strikingly, the magnitude of the biasing effect was just as large when witnesses were simply not given a

Fleet, M. L., Brigham, J. C., & Bothwell, R. K. (1987). The confidence-accuracy relationship: The effects of confidence assessment and choosing. Journal of Applied Social Psychology, 17, 171-187.

Garrett, B. L. (2008). Judging Innocence. Columbia Law Review, 108. 55-142.Gorenstein, G. W., & Ellsworth, P. C. (1980). Effect of choosing an incorrect photograph on a later

identification by an eyewitness. Journal of Applied Psychology, 65, 616–622.Gurney, D. J., Vekaria, K., & Howlett, N. (2013).A nod in the wrong direction: Does nonverbal feedback

affect eyewitness confidence?Psychiatry, Psychology & Law.Haber, R.N. & Haber. L. A Meta-Analysis of Research on Eyewitness Lineup Identification Accuracy. Paper

presented at the Annual convention of the Psychonomics Society, Orlando, Florida, November 16, 2001.

Hastie, R., Landsman, R. & Loftus, E. F. (1978). Eyewitness testimony: The dangers of guessing . Jurimetrics Journal, 19, 1-8.

Horry, R., Memon, A., Milne, B. & Wright, D. (June, 2010). System and estimator variables in real eyewitness identifications: a field study of video lineups in England. 20th conference of the European Association of Psychology and Law, Gothenburg.

Hosch, H. M., & Platz, S. J. (1984). Self-monitoring and eyewitness accuracy. Personality and Social Psychology Bulletin, 10, 289- 292.

Jenkins R, White D, Montfort X, Burton AM. (2011). Variability in photos of the same face. Cognition 121:313-323.

Johnson, C., & Scott, B. (1976). Eyewitness testimony and subject identification as a function of arousal, sex of witness, and scheduling of interrogation. Paper presented at American Psychological Association Convention, Washington, DC.

Juslin, P., Olsson, N., & Winman, A. (1996). Calibration and diagnosticity of confidence in eyewitness identification: Comments on what can be inferred from the low confidence–accuracy correlation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1304–1316.

Kellstrand, E. B. (2006). Eyewitness identification accuracy in cases accepted and rejected for prosecution: An archival analysis of criminal case files, Unpublished Manuscript. San Diego: University of California.

Key, K. N., Cash, D. K., Neuschatz, J., Price,J., Wetmore, S. A. & Gronlund, S. D. (2015): Age differences (or lack thereof) in discriminability for lineups and showups, Psychology, Crime & Law, DOI: 10.1080/1068316X.2015.1054387

Krafka, C., & Penrod, S. (1985). Reinstatement of context in a field experiment on eyewitness identification. Journal of Personality and Social Psychology, 49(1), 58

Leippe, M.R. (1980). Effect of integrative memorial and cognitive processes on the correspondence of eyewitness accuracy and confidence. Law and Human Behavior, 4, 261-274.

Leippe, M. R., Manion, A. P., & Romanczyk, A. (1992). Eyewitness persuasion: How and how well do factfinders judge the accuracy of adults’ and children’s memory reports? Journal of Personality and

78

“don’t know” or “not present” option as it was when instructions also included some pressure to make a selection.

Quinlivan et al. (2012) manipulated pre-admonition suggestions and cautionary instructions. That is, in addition to manipulating whether witnesses received biased or unbiased instructions, Quinlivan et al. manipulated whether, before the instructions, the lineup administrator suggested the perpetrator was likely to be in the lineup (i.e., “I could really tell you were paying a lot of attention; surely you are going to be able to pick the person out of the lineup”). When instructions were unbiased, there was an increase in false identifications and choosing rates among witnesses who received the pre-admonition suggestions compared to no pre-admonition suggestions (Quinlivan et al.). The study demonstrate that unbiased instructions are more effective than biased instructions, but only if the administrator refrains from suggesting that the

Social Psychology, 63, 181–197.Leippe, M. R., & Romanczyk, A. (1987). Children on the witness stand: A communicationpersuasion

analysis of jurors’ reactions to child witnesses. In S. J. Ceci, M. P. Toglia, & D. F. Ross (Eds.), Children’s eyewitness memory (pp. 155–177). New York: Springer-Verlag.

Lindsay, R. C., & Harvie, V. L. (1988). Hits, false alarms, correct and mistaken identifications: The effect of methods of data collection on facial memory. Gruneberg, M. M.(Ed), Morris, P. E.(Ed), & et al. Practical aspects of memory: Current research and issues, Vol. 1: Memory in everyday life. (pp. 47-52). Oxford, England: John Wiley & Sons

Lindsay, R. C. L., Wells, G. L., & O'Connor, F. J. (1989). Mock juror belief of accurate and inaccurate eyewitnesses: A replication and extension. Law and Human Behavior, 13, 333-339.

Lindsay, R. C. L., Wells, G. L., & Rumpel, C. M. (1981). Can people detect eyewitness identification accuracy within and across situations? Journal of Applied Psychology, 66, 79-89.

Liptak, A. (October 30, 2011). Supreme Court to Weigh Effects of Bad Plea Advice. The New York Times.Loftus, E. F. (1976). Unconscious transference. Law and Psychology Review, 2, 93-98.Loftus, E. F., Schooler, J. W., Boone, S. M., & Kline, D. (1987). Time went by so slowly: Overestimation of

event duration by males and females. Applied Cognitive Psychology, 1, 3-13.Luus, C.A.E., & Wells, G.L. (1994). The malleability of eyewitness confidence: Co-witness and

perseverance effects. Journal of Applied Psychology, 79, 714-723.Maas, A. & Kohnken, G. (1989). Eyewitness identification: Simulating the "weapon effect." Law and

Behavior, 13, 397-408.Mansour, J. K. , Beaudry, J. L., Bertrand, M. I., Kalmet, N., Melsom, E., & Lindsay, R. C. L. (2012). Impact

of disguise on identification decision and confidence with simultaneous and sequential lineups. Law and Human Behavior, 36, 513-526. doi: 10.1037/h0093937

Megreya, A. M., & Burton, A. M. (2006). Recognising faces seen alone or with others: When two heads are worse than one. Applied Cognitive Psychology, 20, 957–972.

Megreya, A. M., & Burton, A. M. (2008). Matching faces to photographs: Poor performance in eyewitness memory (without the memory). Journal of Experimental Psychology: Applied, 14, 364-372.

Megreya A M, Bindemann M (2012) ``Identification accuracy for single- and doubleperpetrator crimes: Does accomplice gender matter?'' British Journal of Psychology, 103, 439-453.

Memon, A., Havard, C., Clifford, B., Gabbert, F., & Watt, M. (2011). A field evaluation of the VIPER system: A new technique for eliciting eyewitness identification evidence. Psychology, Crime, & Law.17, 711-729.

Mitchell, K., Livosky, M., & Mather, M. (1998). The weapon focus effect revisited: The role of novelty. Legal and Criminological Psychology, 3, 287–303.

Morgan III, C.A., Hazlett, G. A., Doran, A., Garrett, S., Hoyt, G., Thomas, P., et al. (2004). Accuracy of eyewitness memory for persons encountered during exposure to highly intense stress. International Journal of Law and Psychiatry,

79

perpetrator is in the lineup--unbiased lineup instructions may only work when the administrator does not make statements contradicting the instructions.

In conclusion there is convincing evidence that suggestive identification instructions influence eyewitness performance. The research shows that suggestive instructions substantially increase the likelihood of false identifications.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that 71% of their respondents understood that instructions influence performance (41% in Benton, et al.). As with the other factors discussed above, this means there is significant disagreement among laypersons about the impact that this factor has on witness confidence—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

27, 265−279.Olsson, N. (2000). A comparison of correlation, calibration, and diagnosti-city as measures of the

confidence–accuracy relationship in witness identification. .Journal of Applied Psychology, 85, 504–511.

Orchard, T. L., & Yarmey, A. D. (1995). The effects of whispers, voice-sample duration, and voice sample distinctiveness on criminal speaker identification. Applied Cognitive Psychology, 9, 249–260.

O'Rourke, T., Penrod, S.D., Cutler, B.L., & Stuve, T.E. (1989). The external validity of eyewitness identification research: Generalizing across age groups. Law and Human Behavior, 13, 385-396.

Palermo, R., & Rhodes, G. (2002). The influence of divided attention on holistic face perception. Cognition, 82(3), 225–257.

Penrod, S. D. & Bornstein, B. H. (2007). Generalizing Eyewitness Reliability Research. In R.C.L. Lindsay, D. Ross, D Read & M. Toglia, (Eds.), Handbook of eyewitness psychology (Vol. II): Memory for people. Mahwah, NJ: Erlbaum.

Pezdek, K., & Stolzenberg, S. (2014). Are individuals’ familiarity judgments diagnostic of prior contact? Psychology, Crime, & Law. 20(4), 302-314. 

Pickel, K. (1998). Unusual and threat as possible causes of “weapon focus.” Memory, 6, 277–296. Pickel, K. L. (1999). The influence of context on the "weapon focus" effect. Law and Human Behavior,

23(3), 299-311.Pickel, K. (2007). Remembering and identifying menacing perpetrators: Exposure to violence and the

weapon focus effect. L. In Lindsay, Rod C. L. (Ed); Ross, David F. (Ed); Read, J. Don (Ed); Toglia, Michael P. (Ed). The handbook of eyewitness psychology, Vol II: Memory for people. (pp. 339-360). Mahwah, NJ, US: Lawrence Erlbaum Associates.

Pigott, M. A., Brigham, J. C., & Bothwell, R. K. (1990). A field study of the relationship between quality of eyewitnesses' descriptions and identification accuracy. Journal of Police Science and Administration, 17, 84-88.

Pike, G., Brace, N., & Kynan, S. (2002). The visual identification of suspects: Procedures and practice (Briefing Note 202). London: Home Office.

Pryke, S., Lindsay, R. C. L., Dysart, J. E., & Dupuis, P. (2004). Multiple independent identification decisions: A method of calibrating eyewitness identifications. Journal of Applied Psychology, 89, 73-84. doi:10.1037/0021-9010.89.1.73

Rahaim, G. L., & Brodsky, S. L. (1982). Empirical evidence versus common sense: Juror and lawyer knowledge of eyewitness accuracy. Law and Psychology Review, 7, 1-15.

Read, J. D. (1995). The availability heuristic in person identification - the sometimes misleading consequences of enhanced contextual information. Applied Cognitive Psychology, 9, 91-121.

Read, J. D., Tollestrup, P., Hammersley, R., McFadzen, E., & Christensen, A. (1990). The unconscious transference effect: Are innocent bystanders ever misidentified? Applied Cognitive Psychology, 4, 3-31.

80

Foil Bias

The term "functional size" (Lindsay & Wells, 1980) refers to the number of viable lineup members, or the number of lineup members who plausibly match the eyewitness's description of the crime perpetrator. Having other lineup members who resemble the perpetrator in physical appearance affects suggestion by protecting the suspect from the eyewitness's tendency to make a positive identification. For example, if an eyewitness had a poor memory for the crime perpetrator but remembered some general characteristics, such as the perpetrator's long blond hair, then having other lineup members with long blond hair safeguards the suspect from identification by deduction. The quality and the number of foils in an array clearly influence

Sauer, J. D., Brewer, N., & Wells, G. L. (2008). Is there a magical time boundary for diagnosing eyewitness identification accuracy in sequential line-ups?. Legal and Criminological Psychology, 13(1), 123-135.

Shapiro, P. N., & Penrod, S. D. (1986). Meta-analysis of facial identification studies. Psychological Bulletin, 100, 139-156.

Shaw, J. S. (1996). Increases in eyewitness confidence resulting from postevent questioning. Journal of Experimental Psychology: Applied, 2.126-146.

Shaw, J. S., & McClure, K. A. (1996). Repeated postevent questioning can lead to elevated levels of eyewitness confidence. Law and Human Behavior, 20.629-654.

Shaw, J. I., & Skolnick, P. (1999). Weapon focus and gender differences in eyewitness accuracy: Arousal versus salience. Journal of Applied Social Psychology, 29(11), 2328-2341.

Shepherd, J. W. (1983). Identification after long delays. In S. M. A. Lloyd-Bostock & B. R. Clifford (Eds.), Evaluating witness evidence (pp. 173–187). Chichester: Wiley.

Slater, A. (1994). Identification Parades: A Scientific Evaluation. (Police Research Award Scheme, Police Research Group, Home Office (reported in Tim Valentine & Pamela Heaton. (1999) An Evaluation of the Fairness of Police Line-Ups and Video Identifications, Applied Cognitive Psychology, 13, S59-72).

Sporer, S., Penrod, S.D., Read, D. & Cutler, B.L. (1995). Gaining confidence in confidence: A new meta-analysis on the confidence- accuracy relationship in eyewitness identification studies. Psychological Bulletin, 118, 315-327

Steblay, N. M. (1992). A meta-analytic review of the weapon focus effect. Law & Human Behavior, 16(4).Steblay, N., Dysart, J., Fulero, S. & Lindsay, R.C.L. (2003). Eyewitness accuracy rates in police showup and

lineup presentations: a meta-analytic comparison.  Law and Human Behavior, 27, 523-540.Steblay, N. K., Dysart, J. E., Fulero, S., & Lindsay, R. C. L. (2001). Eyewitness accuracy rates in sequential

and simultaneous lineup presentations: A meta-analytic comparison. Law and Human Behavior, 25, 459-473. doi:10.1023/A:1012888715007

Sučić, I., Tokić, D. & Ivešić, M. (2015) Field study of response accuracy and decision confidence with regard to lineup composition and lineup presentation, Psychology, Crime & Law, 21:8, 798-819, DOI: 10.1080/1068316X.2015.1054383

Tollestrup, P., Turtle, J., & Yuille, J. (1994). Actual victims and witnesses to robbery and fraud: An archival analysis. In D. F. Ross, J. D. Read, & M. P. Toglia (Eds.), Adult eyewitness testimony: Current trends and developments (pp. 144–160). New York, NY: Cambridge University Press.

Turtle, J. W., & Wells, G. L. (1988). Children versus adults as eyewitnesses: Whose testimony holds up under cross examination? In M. W. Gruneberg et al. Eds.), Practical aspects of memory (pp. 27-33). New York: Wiley.

Turtle, J. W., & Yuille, J, C. (1994). Lost but not forgotten details: Repeated eyewitness recall leads to reminiscence but not hypermnesia. Journal of Applied Psychology, 79, 260-271.

81

the fairness of the array--as reflected in the tendency for witnesses to make identifications, particularly false identifications.

The 1999 National Institute of Justice “Eyewitness Evidence A Guide for Law Enforcement” (http://www.ncjrs.gov/pdffiles1/nij/178240.pdf) makes similar points in its guideline for the preparation of photoarrays (p. 29): “Select fillers who generally fit the witness’ description of the perpetrator. When there is a limited/inadequate description of the perpetrator provided by the witness, or when the description of the perpetrator differs significantly from the appearance of the suspect, fillers should resemble the suspect in significant features.” The same language is used with respect to lineups.

In a compelling demonstration of foil bias, Lindsay & Wells (1980) staged a theft in view of 96 undergraduates. Shortly after the theft subjects were asked to identify the thief from thief-present or thief absent photoarrays containing six photographs. Both the thief and the innocent suspect who replaced him in

Valentine, T. (2008). Indentifying perpetrators. In Forensic Psychology, edited by Graham Davies, Clive Hollin, Ray Bull, Chichester, England ; Hoboken, NJ : John Wiley & Sons.

Valentine, T., & Mesout, J. (2009). Eyewitness identification under stress in the London Dungeon.Applied Cognitive Psychology, 23(2), 151-161.

Valentine, T., Pickering, A, & Darling, S. (2003). Characteristics of eyewitness identification that predict the outcome of real lineups Applied Cognitive Psychology, 17, 4969-994.

Vierordt, K. von (1868). Der Zeitsinn nach Versuchen [The sense of time according to research]. Tübingen: H. Laupp.

Vokey, J. R., & Read, J. D. (1992). Familiarity, .memorability, and the effect of typicality on the recognition of faces. Memory & Cognition, 20, 291-302

Warnick, Dell H.; Sanders, Glenn S. (1980). The effects of group discussion on eyewitness accuracy. Journal of Applied Social Psychology. 10(3). 249-259.

Wells, G. L., & Bradfield, A. L. (1998). “Good. You identified the suspect”: Feedback to eyewitnesses distorts their reports of the witnessing experience. Journal of Applied Psychology, 83, 360–376.

Wells, G. L., Ferguson, T. J., & Lindsay, R. C. L. (1981).The tractability of eyewitness confidence and its implications for triers of fact. Journal of Applied Psychology, 66, 688–696.

Wells, G. L., & Lindsay, R. C. L. (1980). On estimating the diagnosticity of eyewitness nonidentifications. Psychological Bulletin, 88, 776–784.

Wells, G. L., & Murray, D. M. (1984). Eyewitness confidence. In G. L. Wells, & E. F. Loftus, (Eds.), Eyewitness testimony: Psychological perspectives. New York: Cambridge University Press.

Wells, G. L., Lindsay, R. C. L., & Ferguson, T. J. (1979). Accuracy, confidence, and juror percep- tions in eyewitness identification. Journal of Applied Psychology, 64, 440-448.

Wells, G. L., Memon, A. & Penrod, S. D. (2006). Eyewitness Evidence: Improving its Probative Value . Psychological Science in the Public Interest, 7, 45-75.

Wells, G. L., Small, M., Penrod, S. D., Malpass, R. S., Fulero, S. M., & Brimacombe, C. A. E. (1998). Eyewitness identification procedures: Recommendations for lineups and photospreads. Law and Human Behavior, 23, 603–648.

Wells, G.L., Steblay, N.K. & Dysart, J.E. (2011). An Initial Report of the AJS National Eyewitness Identification Field Studies. American Judicature Society. http://www.ajs.org/wc/pdfs/EWID_PrintFriendly.pdf

Wetmore, S.A. , J.S. Neuschatz, S.D. Gronlund, A. Wooten, C.A. Goodsell, and C.A. Carlson (2014). Effect of Retention Interval on Showup and Lineup Performance. Journal of Applied Research in Memory and Cognition.4, 8-14.

Williams, G. (1963). The proof of guilt. Springfield, IL.: Charles C. Thomas.Wright, D. B., & McDaid, A. T. (1996). Comparing system and estimator variables using data from real line-

ups. Applied Cognitive Psychology, 10(1), 75-84.82

the thief-absent conditions were white males in their 20's with light brown hair and moustaches. Half of the subjects viewed photoarrays in which all of the foils were white males in their 20's with brown to blond hair and moustaches (high similarity condition). The other half viewed photoarrays in which the foils were two oriental and three white males in their late 20's with full black beards and black hair (low similarity condition). Although high similarity photoarrays produced lower correct and false identification rates than low similarity lineups, the effect was significantly greater on false identifications than on correct identification rates. Among subjects shown thief-present photoarrays, 71% of subjects in the low similarity and 58% of subjects in the high similarity conditions made correct identifications. Among subjects shown thief-absent photoarrays, 70% in the low similarity and 31% in the high similarity conditions made false identifications.

What these results underscore is that the selection of foils can dramatically influence identification performance and can tilt identifications towards potentially innocent suspects. How can this be prevented? While psychologists, police and appellate courts have historically advocated maximizing the similarity of appearance between the lineup members and the suspect/perpetrator, Luus & Wells (1991) disagree. They argue that the ideal, given this advice, would be a lineup of clones. The suspect, in a lineup of clones, is protected from mistaken identification, but there is little chance of a correct identification because the witness cannot discriminate among lineup members.

Luus and Wells (see also Navon, 1992) propose, as an alternative, that lineup members should match the descriptions given by the witness at the time of the crime on all features mentioned but should vary on features not mentioned in the witness's description. For example, if the witness, at the time of the crime, describes the perpetrator as a white male, about 6' tall, 180 lbs, broad shoulders, blond hair, moustache and no beard, all lineup members should fit this description. But they should vary on features not mentioned by the witness, such as hair length, eye color, etc. These criteria, argues Wells, should protect the suspect from the witness's tendency to make a positive identification while not making the identification task overly difficult.

Research by Wells, et al. (1993) demonstrated that foils selected with the match-to-description procedure could produce higher rates of correct identifications of targets in target-present arrays than was true in lineups composed of highly similar foils selected by match-to-suspect and Lindsay, et al. (1994) obtained similar results. Although more recent research by Tunnicliff and Clark did not yield an advantage for

Wright, D. B., & Skagerberg, E. M. (2007). Postidentification Feedback Affects Real Eyewitnesses.Psychological Science, 18(2), 172-178.

Yarmey, A. D. (1990). Accuracy and confidence of duration estimates following questions containing marked and unmarked modifiers. Journal of Applied Social Psychology, 20, 1139–1149.

Yarmey, A. D. (1993). Stereotypes and recognition memory for faces and voices of good guys and bad guys. Applied Cognitive Psychology, 7, 419–431.

Yarmey, A. D. (2000). Retrospective duration estimations for variant and invariant events in field situations. Applied Cognitive Psychology, 14, 4557.

Yarmey, A. D., & Jones, H. P.T. (1983). Accuracy of memory of male and female eyewitnesses to criminal assault and rape. Bulletin of the Psychonomic Society, 2, 89-92.

Yarmey, A. D. and Yarmey, M. J. (1997). Eyewitness recall and duration estimates in field settings. Journal of Applied Social Psychology, 27, 330-344.

Yarmey, A. D., Yarmey, A. L., & Yarmey, M. J. (1994). Face and voice identifications in showups and lineups. Applied Cognitive Psychology, 8, 453–464.

83

match-to-description, the authors do note that performance in match-to-description line-ups was more consistent than performance in match-to-suspect line-ups, perhaps because the former are less subjective.

Fitzgerald, et al. (2013) in a meta-analysis of 17 independent studies with 6,650 participants which compared performance on arrays with high, moderate and low similarities among suspects and foils concluded:

In the report developed by the Technical Working Group for Eyewitness Evidence (2003), police investigators are advised that the suspected culprit should not stand out from the lineup members who are known to be innocent. The meta-analysis results provide support for this recommendation. Compared with lineups that had fillers of moderate or high suspect-filler similarity, the rate of innocent suspect misidentifications nearly doubled when lineups contained fillers of low suspect-filler similarity. The group’s report further advises police investigators to ensure that fillers and the suspect are not too similar. The concern is that using extremely similar fillers will essentially result in a lineup of “clones” that would greatly diminish the likelihood that a culprit will be correctly identified. However, our synthesis of the existing literature did not support this assertion. Although the not-too-similar rule has a solid theoretical foundation (Luus & Wells, 1991) that was soon after supported by empirical research (Wells et al., 1993), we found no reliable difference in correct identifications between lineups within the categories of high and moderate suspect-filler similarity.

Wells (1993) reported concern among some eyewitness researchers that choosing fillers with features that vary from those of the suspect could result in lineups with an unintended bias toward innocent suspects. The present research suggests their concern may have been justified. Innocent suspects were significantly more likely to be misidentified from lineups of moderate suspect-filler similarity compared with lineups of high suspect-filler similarity. This increase in innocent suspect misidentifications, taken together with the null effect in culprit identifications, suggests that either (a) the rule of ensuring lineup members are not too similar to the suspect does not improve performance on culprit-present lineups and may actually contribute to wrongful convictions or (b) the inability to obtain fillers who are truly of high resemblance to the suspect has led to an incongruity between theory and practice. In other words, although the rule to avoid highly similar fillers may be theoretically sound, finding such fillers in practice may be more difficult than had been anticipated. (p. 160)

On balance, both the analyses by Luus and Wells and Navon and the results from the empirical research underscore the desirability of minimizing the impact of subjective (and possibly biased) selection of foils for lineups. Probably the most effective method of eliminating police bias is to have an officer who has not seen the suspect select foils based on match-to-description. An argument can be made that these foils might be complemented by foils selected by defense counsel on a match-to-suspect basis.

It is clear from research on lineup and photoarray fairness that police presentations frequently fall short of being fair. One technique for making such assessments uses the “mock witness” method in which the witness’s original description of the perpetrator is presented to 30 or more “mock witnesses” together with the photoarray or picture of the lineup in question. Each witness is asked make an identification based on the description and one can divide the number of mock witnesses by the number of suspect identifications to gauge the effective size of the array (e.g. in a fair 6-person array 5 of 30 witnesses would be expected to select the suspect: 30/5 = 6…the effective size of the array. Studies of actual arrays show that suspects are picked 2 to 3 times the rates one would expect from a fair array (Brigham et al. 1990; Brigham et al. 1999; Wells & Bradfield, 1999). The problem is not confined to American arrays—see Valentine & Heaton (1999) and Py (2003) for similar findings involving British and French arrays.

84

Benton, et al. (2006) reported that 30% of their respondents appreciated the effect of biased lineups on witness performance. As with the other factors discussed above, this means there is significant disagreement among laypersons about the impact that these factors have on witness confidence—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

Simultaneous Presentation Bias

Nearly twenty years of research indicates that sequential lineups can cut the rate of false identifications in half.

Traditionally, in both live and photographic lineup procedures, both the suspect and foils are presented simultaneously and the eyewitness identifies which (if any) is the suspect. A growing body of scientific research demonstrates that sequential lineups, in which a witness views foils and a suspect one at a time and make a judgment about each face as it is presented, is generally superior to simultaneous lineups, in which a suspect and foil are presented to witnesses all at once. The research comparing sequential and simultaneous procedures was launched by Rod Lindsay and Gary Wells in 1985. Lindsay and Wells (1985) found that of those who experienced simultaneous presentation, 43% made a false identification. Among those who experienced sequential presentation, only 17% made a false identification. Sequential presentation substantially reduced false identification rate.

Cutler & Penrod (1988) replicated the results of Lindsay and Wells’ (1985) experiment and found that witnesses who experienced simultaneous presentation were twice as likely to make a false identification (39%) as were subjects who experienced sequential presentation (19%).

In 2001, Steblay, Dysart, Fulero, & Lindsay published a large-scale meta-analysis, examining 23 papers, containing 30 comparisons of sequential and simultaneous procedures, based on 4,145 research participants. These experiments were conducted in the United States, Canada, the United Kingdom, South Africa, Germany, and Australia and employed a wide array of research participants, formats (photo, video, both), types of crime (robbery, theft, other, no crime), and event stimuli (video, live, slides, transparencies). This meta-analysis is the most comprehensive study of sequential and simultaneous lineups to-date, and represents the current culmination of nearly two decades of research.

Steblay, et al. reported that, across 25 studies, witnesses were nearly half as likely to make a false identification from a target-absent array (28% mistaken identifications) than from simultaneous arrays (51% mistaken identifications). This advantage holds up even in studies in which research teams have further tested the perpetrator-absent scenario by placing a particular suspect in the lineup who closely matches the description of the true target. As with the more general situation, sequential lineups foster significantly fewer false identifications of the innocent suspect compared to simultaneous lineups (9% vs. 27%).

Steblay, Dysart & Wells (2011) have updated the 2001 meta-analysis with 72 tests of simultaneous and sequential lineups involving over 13,000 witnesses and find the same pattern of results. In 27 studies that directly compare sequential/simultaneous using both target-present and target-absent arrays, witnesses were again substantially less likely to make a false identification from a target-absent array (32% mistaken identifications) than from simultaneous arrays (54% mistaken identifications). Sequential lineups also yielded significantly fewer false identifications of an innocent suspect compared to simultaneous lineups (15% vs. 28%).

85

Steblay, et al. (2001) concluded, “[u]nder the most realistic simulations of crimes and police procedures (live staged events, cautionary instructions, single perpetrators, adult witnesses asked to describe the perpetrator), the differences between the correct identification rates for simultaneous and sequential lineups are likely to be small or nonexistent. On the other hand, correct rejection rates are significantly higher for sequential than simultaneous lineups and this difference is maintained or increased by greater approximation to real world conditions.”

Steblay, et al. note that the simplicity of the sequential technique, together with the large number of positive research out-comes, has “made it one of the most important of all the practical contributions of eyewitness research to actual eyewitness evidence collection procedures.”

Steblay, et al also note:

The more realistic the stimuli used in the research, the smaller the difference in correct identification rate produced by the simultaneous and sequential lineup procedures. As the experimental conditions become more realistic, the results increasingly approach the pattern of results frequently attributed to lineup procedures: sequential lineups result in approximately the same rate of correct identification and significantly lower rates of false identification than simultaneous lineups. Generalizing to real world identification situations, the tradeoff between correct and false identifications would appear to be not a serious problem. (p. 470).

A meta-analysis by Ebbesen and Flowe (not yet published), available at http://www-psy.ucsd.edu/%7eeebbesen/SimSeq.htm also examined studies of correct and false identifications in studies using simultaneous and sequenial procedures. The Ebbesen and Flowe sample of studies consisted of 113 experiments from 82 papers, with a total of 152 lineups. The sample of 136 lineups with adults included 11 sequential-only lineups, 17 lineups using both procedures, and 108 simultaneous-only lineups.

In 114 simultaneous, target-present lineups, the target was identified 48.8% of the time. In the 23 sequential target-present lineups, a correct identification was made 44.6% of the time. The difference in correct identifications between simultaneous and sequential lineups was not statistically significant – that is, the analysis does not support the conclusion that the difference between the procedures was a result of anything but chance. In target-absent lineups, witnesses made false identifications in simultaneous lineups 48.6% of the time, but only 29.3% of the time in sequential lineups. Sequential lineups were associated with a marked reduction in mistaken identifications.

86

More recently a sequential versus simultaneous field experiment was conducted with the cooperation the Charlotte-Mecklenburg (NC) Police Department, the Tucson (AZ) Police Department, the San Diego (CA) Police Department and the Austin (TX) Police Department (wells, Steblay & Dysart, 2011). Cases in which the eyewitness had prior familiarity with the suspect were excluded as were instances in which the lineup was not conducted double-blind The resulting 497 lineups ranged in seriousness from simple assault to murder. Results demonstrated (as shown in the figure below) comparable rates of suspect identification (25.5% simultaneous versus 27.3% sequential), but higher rates of foil identifications in simultaneous lineups (18.1%) versus sequential (12.2). Further analyses showed that 80.8% of non-identifications for the simultaneous procedure were clear rejections (“no” to all photos) and 19.2% were “not sure” responses. In contrast, 53.5% of the non-identifications for the sequential procedure were clear rejections and 46.5% were “not sure” responses.

Overall, these studies, spanning 25+ years of research on thousands of participants, indicate that an innocent suspect is between 2 and 3 times as likely to mistakenly identified from a simultaneous array as compared to a sequentially presented array.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that just 46% of American lay respondents understood that the risk of misidentification was higher for simultaneous as opposed to sequential presentations. This means there is widespread disagreement among laypersons about the impact that presentation format have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

Non-Blind Presentation

Concern has been expressed about the fact that identification procedures are conducted “non-blind”—that is, the administering officers know who the suspect is. Psychological and medial researchers have completely abandoned non-blind research out of fear that research participants will be unwittingly influenced by researchers who unwittingly communicate their expectations (e.g., about the effectiveness of a new test medicine). Can witnesses be influenced by administrators who know the identify of the suspect? The first

87

study to examine the influence of administrator knowledge paired mock line-up administrators with mock witnesses who had previously viewed a live, staged crime involving two perpetrators (Phillips et al., 1999). Phillips and colleagues manipulated whether the administrator knew the identity of the suspect, the type of lineup presented (simultaneous vs. sequential), as well as the presence of an observer during the line-up task. When the line-up was presented sequentially and an observer was present, administrator knowledge influenced witnesses to choose an innocent suspect.

Additional support for the effects of line-up administrator knowledge was obtained in research manipulating the level of contact between administrators and witnesses (Haw & Fisher, 2004). For some witnesses, administrators had direct contact with them while administering the line-up. For others, administrators sat behind the witnesses out of their direct view, limiting their ability to communicate cues to the suspect’s identity to the witnesses. When the administrator was permitted high contact with the witness and presented a target-absent simultaneous line-up, witnesses were more likely to identify the innocent suspect. In contrast, other studies have failed to find any influence of administrator knowledge on witnesses’ identification decisions (Haw et al., 2003; Russano et al., 2006).

Greathouse & Kovera (2009) manipulated whether an administrator had knowledge of the suspect’s identity, the type of line-up (simultaneous vs. sequential), the presence of the actual perpetrator in the line-up and the type of line-up instructions (biased vs. unbiased). When the witnesses received biased instructions and simultaneous line-ups, they were more likely to make suspect identifications in non-blind than in blind lineups. The pattern of filler and suspect identifications suggested that the increase in mistaken identifications was the result of non-blind administrators influencing those who would have, under blind conditions, made filler identifications to make suspect identifications instead. Line-up rejections did not significantly increase or decrease as a function of line-up administrator knowledge. Suspect identifications were twice as diagnostic for blind administrations as they were for non-blind administrations.

In a more pronounced demonstration of the effects that a non-blind administrator can have, Alberts, Duncan, Wallace and Penrod (2008) trained administrators to ”steer” witnesses to make positive identifications of suspects (both guilty and innocent suspects) and avoid identifications of non-suspects (foils) and to accomplish this steering without arousing the suspicion of witnesses. Steering was highly successful: selection rates for “steered” witnesses were: 57% guilty suspects, 32% innocent suspects and 19% filler identifications versus 18% guilty suspects, 10% innocent suspects and 39% filler identifications. Witnesses were essentially unaware that they had been influenced. Clark, et al. (2013) used similar procedures and obtained similar results.

Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues did not include blind presentation as a factor, nor did Benton et al. (2006). As noted above, research on other factors certainly suggests there will be disagreement among laypersons about the impact that non-blind presentation have on identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

X. Witness Confidence and Witness Accuracy

The problems with respect to witnesses’ confidence about the accuracy of their identifications and the actual accuracy of those identifications are manifold. Some of these problems relate to juror use of witness confidence and some relate to the weakness of the association between confidence and accuracy (CA). Witness confidence can be influenced by post-identification factors such as repeated questioning, briefings in anticipation of cross examination, and feedback about the behavior of other witnesses. Jurors over-believe

88

eyewitnesses, have difficulty reliably differentiating accurate from inaccurate eyewitnesses, and (as detailed below) are not adequately sensitive to aspects of witnessing and identification conditions. A major source of juror unreliability is reliance on witness confidence. Juror reliance on witness confidence appears to be unaffected by traditional safeguards such as cross-examination and judges' instructions in eyewitness cases.

Although there is a presumption among many actors in the legal system that there is a positive confidence—accuracy relation, serious questions about whether that is true at the trial stage have been raised by researchers within the law-psychology community. These issues were discussed in some detail in a 1998 Whitepaper (Wells, et al, 1998) and I have excerpted portions of that paper here:

Surveys Show that Most People Believe Confidence is Diagnostic of Accuracy. Based on survey techniques, it is clear that people believe that there is a strong relation between eyewitness identification confidence and eyewitness identification accuracy (Brigham & Bothwell, 1983; Brigham & Wolfskiel, 1983; Deffenbacher & Loftus, 1982; Rahaim & Brodsky, 1982). Similar findings have been reported in Canada (Yarmey & Jones, 1983), Germany (Sporer, 1983), Australia (McConkey & Roche, 1989) and England (Noon & Hollin, 1987).

Jurors Infer Accuracy from Confidence. There is consistent evidence to indicate that the confidence that an eyewitness expresses in his or her identification during testimony is the most powerful single determinant of whether or not observers of that testimony will believe that the eyewitness made an accurate identification (e.g., see Cutler, Penrod & Dexter, 1990; Leippe & Romanczyk, 1987, 1989; Leippe, Manion, & Romanczyk, 1991; Lindsay, Wells, & O'Connor, 1989; Lindsay, Wells, & Rumpel, 1981; Turtle & Wells, 1988; Wells, Ferguson, & Lindsay, 1981; Wells, Lindsay, & Ferguson, 1979; Wells & Murray, 1984).

Overall, it is clear that jurors do rely on witness confidence as an indicator of witness accuracy—even when, circumstances do not support such reliance.

To what extent does witness confidence predict identification accuracy? Should jurors (and attorneys and judges) rely so heavily on witness confidence as a guide to witness accuracy?

Pre-identification Confidence Reveals Little about Identification Accuracy. Cutler & Penrod (1989a) examined nine studies testing the relation between pre-identification confidence and identification accuracy. Across nine studies the pre-identification confidence-accuracy correlation ranged from .00 to .20--which indicates that pre-identification is a poor predictor of identification performance.

Post-Identification Confidence, if Measured Properly, is Modestly Correlated with Accuracy. Over the past 20 years researchers have examined the results from the growing numbers of studies that measure both witness accuracy and witness post-identification confidence in an effort to arrive at a reliable estimate of the magnitude of their relation. Deffenbacher (1980) concluded that there was little support for a strong reliance on witness confidence as a guide to witness accuracy. In a review of 31 studies Wells & Murray (1984) reported an average r=.07. Bothwell, Deffenbacher, & Brigham (1987) meta-analyzed 35 studies involving staged incidents that yielded an average confidence and accuracy correlation of r=.25. Perhaps the most informative study is by Sporer, Penrod, Read & Cutler (1995) who found a confidence-accuracy correlation of .41 among (forensically-relevant) choosers. These finding suggests that witnesses who are highly confident in their identifications are somewhat more likely to be correct as compared to witnesses who display little confidence.

Earwitnesses. With respect to earwitness identifications, the pattern of results is similar. In a very recent study Stevenage et al. (2010) report confidence and accuracy correlations of r=.26 and .35 for faces

89

versus .15 and .13 for voices. Similarly low correlations from voice identifications emerge from Procter (2002)-- r= .26; Saslove &Yarmey (1980)—r= .26); Yarmey and Matthys (1992) found only non-significant correlations; Orchard & Yarmey (1995) found significant negative correlations; Yarmey (2003) found non-significant correlations (with one exception). Read & Craik (1995) found an average correlation of .21 in their study of changes in speaker voice/emotionality.

But, Witnesses are Over-Confident. Though confidence-accuracy correlations are sometimes relatively high, most research yields relatively low correlations. Some have argued that despite the generally weak confidence-accuracy correlation, accuracy may be very high among the most confident witnesses. One analytic method that addresses this question uses calibration methods which measure peoples' confidence on a percentage scale (0, 10, 20, 30% and so on--not that this is not what the police do), then clumps people together at different levels of confidence to assess their accuracy. In what is probably the first published eyewitness study using calibration methods, Cutler and Penrod (1989) found witness overconfidence of 10-20% (that is, witnesses were making 10-20% more errors than their confidence levels indicated). 

In a study by Juslin, Olsson and Winman (1996) overconfidence ranged from 10-25%. Similarly, though the published numbers are slightly ambiguous, it looks as though the top 21% most confident witnesses in Brigham, Maass, Snyder, & Spaulding (1982) were 85% correct. Brewer, Keast, and Rishworth (2002) found that eyewitnesses who were very confident of the accuracy of their identifications (95% certain) were only about 70%-75% correct. In a 1987 study by Fleet, Brigham, & Bothwell 75% of Ss who rated themselves as extremely confident were accurate. Brigham (1990) found a 74% accuracy rate for the top 27% most-confident of his witnesses.

Sauer, Brewer & Wells (2008) report some of the most informative results: a 40% error rate among witnesses who make identifications and are 90%-100% confident (only 18% of their witnesses expressed this high level of confidence), a 51% error rate among those who are 70%-80% confident and a 59% error rate among those who were 50-60% confident. In short, although there is some evidence of greater accuracy among those who are more confident, error rates can be high among even the small percentage of most-confident witnesses.  Among non-identifying witnesses Sauer, Brewer & Wells (2008) report a similar pattern of improved performance with higher levels of confidence plus overconfidence – they report a 30% error rate among witnesses who make a nonidentification and are 90%-100% confident (21% of their nonidentifying witnesses), a 35% error rate among those who are 70%-80% confident and 42% errors among those who are 50-60% confident.

More recently, Sauer et al. (2010) found that 13% of their witnesses, when tested immediately, made a positive identification with 90-100% confidence—their error rate was 20%. 10% of witnesses to the same events--tested after one week--made identification with 90-100% confidence—their error rate was 25%. Even greater overconfidence was evident in a study by Sucic and colleagues (2015) as shown in the following figure. If the witnesses had been performing perfectly their calibration lines would have overlaid the diagonal reference line. As is evident from the figure neither pre-identification (prospective) nor post-identification (retrospective) confidence was well calibrated and highly confident witnesses were quite substantially over-confident (e.g. a better than 50% error rate for witnesses who were 95% confident about their selections. The authors conclude: “The calibration curve, as well as the values of additional parameters, confirmed correlational findings, which indicated that confidence only weakly postdicts accuracy and that confidence not at all predicts accuracy.” (p. 808)

90

Overconfidence has similarly been demonstrated in memory for events as shown in this figure from Dahl, et al. (2015)

Optimality and Confidence-Accuracy. As early as 1980 Deffenbacher advanced the idea that the confidence-accuracy relationship should be weaker under conditions which promote poor witness performance. Deffenbacher illustrated the effect by comparing the number of studies producing significant and non-significant confidence-accuracy relationships as a function of witness performance. Studies which yielded identification accuracy levels of 70% or higher were deemed high in optimality. As shown in the following table, high optimal condition were much more likely to produce a significant confidence-accuracy correlation.

91

Subsequent research has supported Deffenbacher’s conclusions. Sporer, Penrod, Read & Cutler (1995) in their meta-analysis reported that: “As the percentages of correct identification decisions in these studies increased, the CA relation also strengthened among choosers (simple r = .59).… Thus, when circumstances of encounter permit high quality of encoding, correct identifications will become more likely, and these will more likely be associated with high levels of confidence among choosers.” (p. 323).

Olsson (2000) observed: “Earlier studies (e.g., Cutler & Penrod, 1989) found that optimality of encoding conditions affects the CA relation, in particular the calibration score. For instance, in the study by Cutler and Penrod, low optimality conditions (disguised target persons) produced significantly poorer calibration and over-/underconfidence scores. This result is in line with Deffenbacher's (1980) optimality hypothesis, which posits that the CA relation should be lower under poor information-processing conditions. … the strong covariation observed in this study between task difficulty and calibration and over-/underconfidence, for both earwitnesses and eyewitnesses, is further evidence that the CA relation can vary considerably as a function of encoding and retrieval conditions.” (p. 509).

The general point is that these results are consistent with other measures of the confidence-accuracy relation--even the calibration approach does NOT support the notion that confidence is a highly reliable indicator of accuracy.  Error rates can be high among even the most confident witnesses. 

Earwitnesses. Research with earwitnesses reveals similar patterns of poor calibration—only worse than is true for eyewitnesses. For example the following figures are from studies of earwitness calibration by Olsson & Juslin (1998). In the first study witnesses listened to a tape on which four persons discussed what could be interpreted as criminal activities—voices were heard for 10 seconds each. Later the participants were required to identify the four voices in four separate 8-voice voice lineups, one for each culprit (there were two variants of the lineups—one in which foils were chosen to match descriptions of the target’s voice and another in which they were chosen to match the suspect’s voice). Suspect identification rates were barely better than chance (18%) and foil identification rates were 53% in target-present arrays and 59% in target-absent arrays. As shown in the figure below, calibration was abominable—perfectly calibrated witnesses would have produced confidence judgments that matched their accuracy levels (the “Ideal” line). In fact, witnesses were vastly overconfident—error rates were between 60% and 70% for witnesses who were 95% confident. Similarly atrocious calibration emerges from a study by Zetterholm et al. (n.d.) in which witnesses hear a 2 minute dialogue between two males who simulated the planning of a burglary. Two kinds of voice lineups were used. A “text” voice lineup consisted of the target speaker and five foils reading the same text. A “dialogue” lineup consisted of the same six male speakers having a spontaneous dialogue with another male speaker who was not heard. Even though tested only fifteen minutes later, only 21% of listeners identified the target. Their poor calibration is shown in the second figure.

92

The Problem Grows Worse--Confidence at Trial may be Useless as an Indicator of Accuracy. . Imagine that prosecutors are skimming only the most confident witnesses; there is no artificial confidence-boosting among the witnesses; and we have reliable measures of confidence, not the vague verbal reports currently obtained by police. Among these highly confident witnesses, the results above indicate that 20 to 30% could be in error. But even if the error rate is only 10% for these highly selected and most confident witnesses, they will all appear highly confident to jurors—so confidence cannot help the jurors figure out which witnesses have made errors. Indeed, the simple correlation between confidence and accuracy for these witnesses will be much worse than among all witnesses, because there is very little variability in confidence and maybe no useful variance. Though it is tempting to conclude that jurors might be entitled to assume a fairly high base rate of accuracy among these highly confident witnesses (even if confidence cannot aid them in differentiating accurate and inaccurate witnesses), if guilty (and accurately identified) defendants are pleading guilty at higher rates than innocent defendants it would not be safe to conclude that the accuracy rate is fairly high; indeed, the accuracy rate could be fairly low, because the guilty defendants facing confident witnesses have already pleaded guilty.

In short, the research results and logic call into question the notion that witness confidence can be of significant assistance to jurors.

Survey studies clearly that people believe there is a strong relation between eyewitness identification confidence and eyewitness identification accuracy (Brigham & Bothwell, 1983; Brigham & Wolfskiel, 1983; Deffenbacher & Loftus, 1982; Rahaim & Brodsky, 1982). Similar findings have been reported in Canada (Yarmey & Jones, 1983), Germany (Sporer, 1983), Australia (McConkey & Roche, 1989) and England (Noon & Hollin, 1987). Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues include confidence-accuracy and they reported that just 40% of their respondents understood that confidence and accuracy are modestly related-- the lowest rate of expert-lay agreement across all the factors included in their analyses. This means there is widespread disagreement among laypersons about the relationship between witness confidence and identification performance—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue

Confidence Malleability

93

Even if the research showed that eyewitness-identification confidence and accuracy are related at a level that could have practical utility, this conclusion would come with another huge caveat. Confidence malleability refers to the tendency for an eyewitness to become more (or less) confident in his or her identification as a function of events that occur after the identification. The confidence malleability problem is particularly important because actors in the legal system can contaminate the confidence of an eyewitness in ways that can make an eyewitness’s in-court expression of confidence a meaningless indicator of the eyewitness’s memory. Unfortunately, an eyewitness’s belief that the identified person is the culprit can also arise for reasons other than the eyewitness’s memory (Leippe, 1980; Wells, Ferguson, & Lindsay, 1981; Luus & Wells, 1994; Wells & Bradfield, 1998). For example Hastie, Landsman, & Loftus (1978) found that witnesses who were questioned repeatedly grew more confident about the accuracy of details in their reports (see also Shaw, 1996; Shaw & McClure, 1996; Turtle & Yuille, 1994).

Wells, Ferguson, & Lindsay (1981) demonstrated they could increase witness confidence simply by briefing witnesses about the types of questions they might encounter in an upcoming cross-examination. Unfortunately, the briefing effect occurred among inaccurate eyewitnesses, whose levels of confidence rose dramatically, whereas confidence levels among accurate witnesses were unchanged.

In a dramatic illustration of confidence malleability, Luus & Wells (1994) used a staged-crime to secure false identifications from 136 eyewitnesses. These eyewitnesses viewed a theft in pairs and were separated shortly after the theft. After being separated false identifications were obtained from the witnesses using a photospread (the eyewitnesses were unaware they had made a false identification). After making their identifications eyewitnesses were either told nothing about the identification decision of their co-witness or were given information that their co-witness ostensibly identified the same person, identified someone else, or indicated that the culprit was not in the lineup. They were then interviewed by an assistant to the experimenter (posing as a campus police officer) who solicited the witness' confidence levels (on 10-point scales) in the accuracy of their identifications. Each eyewitness was videotaped while giving statements to the police officer.

There were dramatic increases in the confidence that eyewitnesses expressed in their false identifications in the condition in which they were told that their co-witness identified the same person (the average confidence on a 10 pt. scale was 8.8 versus 6.9 in the no-information control condition). The lowest confidence levels were found among witnesses who were told that the co-witness had indicated that the perpetrator was not in the array (3.6). Even broader effects have been shown to emerge when eyewitnesses are told after their identification that they identified the suspect (versus being told nothing). Wells & Bradfield (1998) also found that eyewitnesses who received confirming feedback were much more confident than the witnesses with no feedback and witnesses with disconfirming feedback. In addition, the confirming feedback witnesses distorted their reports of their witnessing conditions by exaggerating how good their view was of the culprit, how much attention they paid to the culprit’s face while observing the event, and so on. Furthermore, Douglas et al. (2010) have shown that witnesses receiving confirming post-identification feedback are viewed by others as more accurate and confident than witnesses who have not received feedback.

A 2006 meta-analysis by Douglass and Steblay of 20 studies with 2,400 identifications found that witnesses receiving feedback “expressed significantly more . . . confidence in their decision compared with participants who received no feedback. This meta-analysis demonstrated that the effect of confirming feedback was consistent and reliable, with large effect sizes obtained for dependent measures including retrospective certainty, opportunity to view, and attention paid. The broad impact of confirming feedback is well illustrated in the following table from Douglas and Steblay—where d indexes the size of the effect (d is roughly twice as large as r)

94

Gurney and colleagues (2013) took the malleability research one step farther and tested whether positive or negative non-verbal feedback would affect witness confidence. Participants watched CCTV footage of a crime scene and answered a series of questions. While answering questions the participants given either affirming feedback (a head nod), disconfirming feedback (a head shake) or no feedback. Those presented with affirming non-verbal feedback produced inflated confidence judgments compared with those presented disconfirming non-verbal feedback and this was true regardless of their accuracy.

Smalarz and Wells (2014a) investigated other ways in which confirming feedback can affect identification performance. After participants received either no feedback or confirming feedback the researcher/administrator pretended to have given the witness the “wrong” lineup, and then gave the witness the “correct” lineup. Results indicated that feedback impaired witnesses’ ability to identify the culprit in a final memory test (Smalarz & Wells, 2014a). In a second study, Smalarz and Wells (2014b) found that evaluators/jurors could distinguish the accuracy and credibility of witnesses’ videotaped testimony when witnesses were not given feedback, but when witnesses received feedback, there were no differences in evaluators’ ratings of accuracy/credibility (Smalarz & Wells, 2014b).

Dysart, Lawson, and Rainey (2012) tested whether feedback from administrators who were blind to the identity of the suspect had the same influence on eyewitness confidence as non-blind administrators. Participants in their study watched a video of a crime and were asked to identify the suspect from either a target-present or target-absent array. The researchers manipulated whether the participant witnesses did or did not receive feedback from an administrator who was blind or non-blind. Post-identification feedback had an effect on witness ratings of confidence and other judgments only when witnesses received feedback from an administrator who was in the room during the video (and presumably knew the identity of the perpetrator). There was no effect of feedback on witness ratings when the administrator was not in the room during the video and presumably blind to the perpetrator’s identify. Desmarais & Read's (2010) meta-analysis of lay respondents' knowledge of eyewitness issues reported that 59% of their respondents understood that witness confidence can be influenced by factors unrelated to witness accuracy. As with the other factors discussed above, this means there is significant disagreement among laypersons about the impact that these factors have on witness confidence—eyewitness expert testimony would provide jurors with scientifically-grounded guidance on this issue.

XI. Cumulative Impact of Threats to Reliability (pp. 46-47)

95

I have enumerated a large number of factors which have been linked to eyewitness reliability in research and are involved in this case. A strong case can be made that these factors can cumulatively increase identification errors. There is also evidence that the studies reported here typically underestimate the effects of these factors on performance? Are the effects these factors cumulative or does the presence of one factor make the others irrelevant?

Generalizability of Effects

Shapiro & Penrod (1986) noted that in the studies in their meta-analysis overall performance levels were worse in more realistic eyewitness studies higher in the laboratory studies and that those differences were readily accounted for by systematic differences between more and less realistic studies. Lindsay and Harvie (1988) reported a similar pattern: less-realistic studies appear to systematically over-state levels of witness performance. There is other evidence (see Penrod & Bornstein, 2007 for an overview) that less realistic studies understate the impact of studied variables on witness performance. This evidence emerges from meta-analyses in which researchers have tested whether the effect of the factor that is the main focus of the meta-analysis is larger or smaller under more realistic testing conditions. There are many such demonstrations.

Steblay’s (1992) meta-analysis of the weapon focus effect similarly showed that relatively artificial simulations underestimated the magnitude of the effect. Deffenbacher, Bornstein, Penrod, and McGorty’s (2004) found that the effect of stress was twice as large for staged-crime studies than for studies manipulating stress by other means.

Cumulation of Effects

In studies were several factors are manipulated at one time it is regularly observed that they simultaneously influence witness performance. For example, Cutler, Penrod, O’Rourke and Martens (1986) found, across two studies, that disguise, biased instructions, weapon visibility, retention interval, and lineup size all influenced identification accuracy. Cutler, Penrod & Marten (1987a) found simultaneous effects of disguise, retention interval and mugshot viewing. Cutler, Penrod & Marten (1987b) found simultaneous effects for disguise, weapon focus, retention interval and instructions to witnesses.

Shapiro and Penrod (1986) analyzed performance across the hundreds of experimental conditions and found that even when statistically controlling for the effects of other variables most variables still explained significant portions of variability in performance. With regard to correct identification rates, attention variables and the duration of exposure per face were positively related to hit-rate performance. The number of targets studied and retention interval accounted for significant variance. With respect to identification errors, greater attention led to fewer false alarms, the number of targets studied was positively associated with false-alarm rates, there were many more false identifications when participants were confronted with live targets as opposed to still photographs and there was a negative correlation between the false-alarm rate and number of foils.

In short, there is substantial evidence that the effects of the many factors discussed above can and do have a cumulative effect on identification performance.

96