SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Purpose. Showups are common, yet little research has investigated the biasing factors that may influence showup identifications. We investigated the effects of cross-race conditions and clothing bias on showup identification decisions. Additionally, we explored identification decisions made in a subsequent lineup dependent on race, clothing, and showup-target-presence.

Methods. Participants watched a mock crime and were presented with a showup in which suspect race, target-presence, and the clothing worn by the suspect were varied. Following a delay, participants viewed a target-present or -absent lineup and were asked to make a second identification decision.

Results. Presentation of the suspect in the clothing worn by the perpetrator increased choosing rates in both own-race and other-race conditions. Despite this, differential patterns of decision response latencies indicated that eyewitnesses may use clothing information differently when making own-race compared to other-race identification decisions. No evidence for an own-race bias in showup identifications was found; however, other-race lineup identifications were less accurate than own-race lineup identifications. Further, participants in own-race and other-race conditions differed in the extent to which they were affected by multiple identification procedures. Viewing an own-race innocent suspect in a showup increased subsequent false lineup identifications, while choosing the innocent suspect from the showup was necessary to increase false lineup identifications in other-race conditions.

Conclusions. Different situational factors may affect the identification accuracy of eyewitnesses in own-race and other-race conditions for both showup and lineup procedures. Particular caution is advised when showups are clothing-biased and multiple identification procedures are used.


Background

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

A showup is a type of identification procedure in which only one individual is viewed by the witness, who must then make a decision about whether or not he or she is the perpetrator (Dysart & Lindsay, 2007). Showups comprise a considerable percentage of identification procedures used in the United States (Behrman & Davey, 2001; Gonzalez, Ellsworth, & Pembroke, 1993), yet only a fraction of studies on eyewitness accuracy have examined showups. Further, although a considerable percentage of showups in criminal cases may be cross-racial (Behrman & Davey, 2001), no published study has examined accuracy in cross-racial showups. It is also unknown how showup decisions in cross-racial conditions might be affected by clothing cues, despite the fact that the temporal proximity of a showup to the crime (Wells, 2001) makes it particularly likely that a suspect in a showup will be presented in clothing similar to that described by an eyewitness (Dysart & Lindsay, 2007). The current study investigated the impact of cross-racial conditions and clothing cues on showup identification decisions. We also explored how these factors affect identifications from a subsequent lineup identification task.

The showup is an inherently suggestive procedure because only one person, the police suspect, is viewed (Dysart & Lindsay, 2007; State v. Dubose, 2005). Although the courts have stopped short of suppressing all identifications made using the procedure, they have expressed concern about its use and set limits on the admissibility of showup identifications (e.g., Simmons v. United States, 1968; State v. Dubose, 2005; Stovall v. Denno, 1967). Despite this concern, it is unlikely that police will discontinue the use of showups for two reasons: they may prevent guilty suspects from going free (e.g., by providing a basis for an immediate arrest), and they may allow innocent suspects to be cleared without undue delay (Wells, 2001). However, police may conduct a second identification procedure to ‘confirm’ a showup identification – both to assess witness reliability and/or alleviate concerns regarding the admissibility of the showup identification – thus increasing the chances that multiple identification procedures will be used with showups (a second procedure may even be required in the United Kingdom; Valentine, Davis, Memon, & Roberts, 2012).

Laboratory research paints a mixed picture in terms of the degree to which apprehensiveness about showups is warranted. A meta-analysis comparing showups to lineups (Steblay, Dysart, Fulero, & Lindsay, 2003) found that the correct decision rate was higher in showups, largely due to lower showup choosing rates. When filler identifications were taken into account, however, lineups and showups had similar false identification rates. Further, in studies that designated an innocent suspect a priori (as in actual criminal investigations), showups yielded a higher false identification rate than lineups.

Within this limited body of research, many factors that have been shown to influence identification accuracy in lineups have not been investigated with showups. For example, although one showup study used Black and White suspects (Gonzalez et al., 1993), participant race was not reported, making a cross-race analysis impossible. There is reason to believe that cross-racial conditions may affect showup accuracy: research over the last several decades has shown that people are better able to recognize own-race than other-race faces, and more likely to falsely identify other-race faces, a phenomenon known as the own-race bias (ORB; Brigham, Bennett, Meissner, & Mitchell, 2007). Meta-analyses have consistently found evidence for the effect in facial recognition and lineup studies (Anthony, Copper, & Mullen, 1992; Bothwell, Brigham, & Malpass, 1989; Meissner & Brigham, 2001; Shapiro & Penrod, 1986). Whether the ORB has the same effect in the showup has not yet been empirically investigated, nor have potential moderators of the effect with showups, such as clothing bias. In the current study, we included a direct cross-race comparison, and combined this with an examination of the increased suggestibility resulting from presentation of the suspect in similar clothing.

Clothing bias refers to the presentation of only the suspect in an identification procedure in clothing resembling the eyewitness’ description (Dysart, Lindsay, & Dupuis, 2006). It has been found to increase false identification rates while leaving correct identification rates unchanged (e.g., Dysart et al., 2006; Lindsay, Wallbridge, & Drennan, 1987; Yarmey, Yarmey, & Yarmey, 1996). We hypothesized that clothing bias may also interact with cross-racial identification situations in its effect on identification performance. Specifically, eyewitnesses may have greater difficulty accurately recognizing an other-race than an own-race perpetrator's face; thus, they may be more likely to incorporate non-facial cues into their decision-making process. If clothing is a visible non-facial cue, eyewitnesses in other-race conditions may be more likely to consider clothing and thus may be more susceptible to the effects of clothing bias.

Our proposed clothing/race interaction is consistent with recent automatic recognition hypotheses (Charman & Cahill, 2011; Wilford, Smalarz, & Wells, 2011) proposing that a witness presented with a lineup will either automatically recognize the perpetrator (i.e., he will ‘pop out’) or automatic recognition will fail, in which case the witness will divert attention to the lineup fillers. In the absence of fillers – that is, in a showup – attention may be diverted to other cues (e.g., clothing). These hypotheses have implications for both accuracy and response latency, and are consistent with the relationship found between response latency and accuracy in research with lineups (Brewer, Caon, Todd, & Weber, 2006; Dunning & Perretta, 2002) and with showup-non-choosers (Sauerland, Sagana, & Sporer, 2011). Automatic recognition would be associated with shorter latencies and greater accuracy, while failure of automatic recognition would be associated with longer latencies and reduced accuracy.

Cross-racial identifications may also be less reliable when multiple identification procedures are used. Viewing and/or selecting a particular face in one identification procedure can influence the results of a subsequent procedure (e.g., Deffenbacher, Bornstein, & Penrod, 2006; Godfrey & Clark, 2010; Hinz & Pezdek, 2001; Valentine et al., 2012). This influence may occur because of source confusion (see Johnson, Hashtroudi, & Lindsay, 1993): the eyewitness recognizes the suspect for some innocuous reason (e.g., because he was a bystander to the crime) and then misattributes the source of his familiarity to having seen him commit the crime (Loftus, 1976). Alternatively, it may be commitment to a specific previous choice that decreases accuracy (Dysart, Lindsay, Hammond, & Dupuis, 2001; Gorenstein & Ellsworth, 1980; Haw, Dickinson, & Meissner, 2007). Memory for the context in which faces are viewed has been shown to be impaired for other-race compared to own-race faces (Horry & Wright, 2008); thus, we expected greater source confusion to occur in other-race than in own-race identifications. Although the exact mechanisms behind commitment have not yet been elucidated, eyewitnesses may be more likely to commit to other-race than own-race identification decisions for two reasons. The act of choosing a person may make source confusion more likely (e.g., because it may increase the eyewitness's ability to envision the suspect committing the crime). It is also possible that greater difficulty in making identification decisions at all may make eyewitnesses confronted with other-race suspects more likely to simply rely on their previous decision rather than making an independent evaluation of the suspect's match to their memory for the perpetrator.

In the current two-part study, we investigated the effects of clothing and race on showup identification procedures, and then explored subsequent own-race and other-race lineup identification decisions. Our hypotheses were as follows: (1) identification accuracy would be reduced in other-race procedures, but this effect would be greater when the suspect wore the same clothing worn by the perpetrator; (2) longer showup decision response latencies would be associated with reduced accuracy, other-race conditions, and ‘different-clothing’ conditions; (3) lineup accuracy would be reduced for participants who had previously viewed a target-absent showup rather than a target-present showup and for participants who had previously made an inaccurate rather than an accurate showup decision; (4) other-race lineup identification accuracy would be poorer than own-race accuracy; and (5) eyewitnesses would exhibit greater source confusion in other-race than in own-race conditions.

Method

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Participants

Participants were 304 undergraduates at two large urban universities. Seven participants did not complete the second part of the study and were dropped from all analyses leaving a final sample of 297 participants. The sample included 214 females (72.1%) and 83 males (27.9%) with a mean age of 20.6 (SD = 4.2). Of these, 118 (39.7%) were Hispanic, 74 (24.9%) were White, 57 (19.2%) were Black, and 48 (16.2%) were from other racial groups (all of the latter participants were included in other-race conditions).

Design

The study was a 2 (showup-target-presence: target-present, target-absent) × 2 (race-condition: own-race, other-race) × 2 (clothing-condition: same-clothing, different-clothing) × 2 (lineup-target-presence: target-present, target-absent) between-subjects design conducted in two parts separated by approximately 4 days. Part 1 included the showup procedure, while Part 2 included the lineup procedure.

Materials

We created 30-s videos showing a male confederate stealing money and a laptop from an unoccupied office. Due to concerns about stimulus generalizability (see Wells & Windschitl, 1999), three Black, three White, and three Hispanic males were used, for a total of nine videos.1 Confederates were approximately 20 years of age and clean-shaven with short hair. All confederates wore the same black hooded sweatshirt; the hood was always worn down. More uniquely recognizable clothing may have a greater impact than non-distinct clothing (Dysart et al., 2006; Valentine et al., 2012) and therefore we made the sweatshirt more recognizable by placing two 4 × 2 inch patches on it, one on each side of the chest. One patch was red and white while the other was blue and white; both were quite noticeable.

An independent group of participants (N = 36) searched an online database of offender photographs for individuals similar to the general description for each confederate. A second group of participants (N = 23) rank-ordered and rated the resulting 20 photographs per confederate in terms of their similarity to the confederate. The photograph with the highest score was designated the innocent suspect, while the photographs with the second through sixth highest ratings were selected as lineup fillers. To evaluate lineup fairness, a third group of participants (N = 47) served as mock witnesses. Using Tredoux's E′ (Tredoux, 1998), the resulting mean effective size was 4.34 (SD = 0.79) for target-present lineups, and 4.20 (SD = 0.75) for target-absent lineups; effective sizes for confederates from different racial groups were comparable.

Showup photographs2 depicted the head and torso of the suspect and measured approximately 550 × 720 pixels. Suspects wore the same hooded sweatshirt worn in the video in same-clothing conditions, and a range of other types of clothing in different-clothing conditions. Photographs for the lineups depicted roughly the clavicle up and very little clothing was visible; however, no lineup members wore a black-hooded sweatshirt. Lineup photographs were presented simultaneously and the position of the suspect was rotated throughout all six positions of the lineup.

Procedure

To minimize experimenter expectancy effects (see Clark, Marshall, & Rosenthal, 2009; Greathouse & Kovera, 2009; Rosenthal, 1966, 1967), both parts of the study were computerized. Part 1 was conducted in the laboratory using MediaLab (Jarvis, 2006). Participants were randomly assigned to condition and viewed a video, after which they were asked to describe the perpetrator and his clothing. Following a 5-min delay, participants read unbiased instructions (i.e., the perpetrator may or may not be present) and viewed the showup. MediaLab automatically recorded response latencies for identification decisions.

Two to three days after they completed Part 1, participants were e-mailed the link to Part 2, which was conducted online via http://www.SurveyMonkey.com. After entering the site, participants were reminded of what had transpired during Part 1 (i.e., that they had watched a video and viewed a photograph) and asked to report their showup decision to ensure they remembered their prior decision.3 They were then given unbiased instructions and presented with the lineup. Not all participants completed the survey within three days; therefore, delay in responding was coded as a separate variable.

Results

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Part 1: Showups

Accuracy and choosing

Overall, 57.9% of participants made a correct showup decision and the choosing rate was 33.3%. In target-present showups, 46.5% of participants made correct identifications, while 41.3% of participants incorrectly rejected the procedure and 12.3% participants said they did not know if the suspect was the perpetrator. In target-absent showups, 70.4% of participants correctly rejected the procedure, 19% made a false identification and 10.6% responded ‘don't know’. ‘Don't know’ rates were similar across conditions and preliminary analyses indicated that they were not affected by our independent variables; therefore, they were subsumed into the incorrect category for all analyses not pertaining specifically to false alarms.4

We first looked at the effects of showup-target-presence, race-condition, clothing-condition, and their interactions on showup identification accuracy (see Table 1). Our logistic regression model was significant, χ2(7, N = 297) = 25.67, p = .001, inline image. Significant main effects of showup-target-presence, Wald = 11.51, p = .001, OR = 5.36 (95% CI: 2.03, 14.13), and clothing-condition, Wald = 5.44, p = .02, OR = 2.82 (95% CI: 1.18, 6.74), were qualified by a significant interaction between the two, Wald = 4.45, p = .04, OR = 0.25 (95% CI: 0.07, 0.91). In target-present conditions, correct identifications increased by 14.6% when the suspect wore the same (54.1%) compared to different clothing (39.5%). In target-absent conditions, correct rejections decreased by 8.4% when the suspect wore the same (66.2%) compared to different clothing (74.6%). No main effect of race was found, Wald = 2.35, p = .13, OR = 2.03 (95% CI: 0.82, 5.02), nor were any other interactions significant (all ps > .1).

Table 1. Percentage (and standard error) of correct decisions followed by response latencies in target-present and target-absent showups as a function of race and clothing conditions
  Own-raceOther-race
  1. aNote that frequencies represent the number making a correct decision, not the total number of participants in the cell.

Target-present
 Same-clothing50.0 (9.1)56.8 (7.5)
   n = 15a n = 25
 Different-clothing48.6 (8.2)31.8 (7.0)
   n = 19 n = 14
Target-absent
 Same-clothing70.0 (8.4)63.4 (7.5)
   n = 21 n = 26
 Different-clothing77.8 (6.9)71.4 (7.6)
   n = 28 n = 25
Response latency
 Same-clothing M = 7.65 M = 10.14
  95% CI = 5.75, 9.1895% CI: 8.70, 11.58
 Different-clothing M = 9.42 M = 8.38
  95% CI: 7.87, 10.9795% CI: 6.88, 9.88

It is likely that the clothing results were driven by differences in choosing.5 Our model for showup choosing, again including showup-target-presence, race-condition, clothing-condition, and their interactions, was significant, χ2(7, N = 297) = 33.39, p < .001, inline image, but we found only a main effect of clothing type, Wald = 5.44, p = .02, OR = 2.82 (95% CI: 1.18, 6.74). Participants were more likely to choose the suspect when he wore the same clothing worn by the perpetrator (37.9%) than when he wore different clothing (28.9%). No other main effects and no interactions were significant (all ps > .1).

Response latency

Accurate participants made significantly quicker decisions (M = 7.74s, SD = 6.10) than inaccurate participants (M = 10.70s, SD = 6.10), t (295) = 3.81, p < .001, Mdiff = 2.96 (95% CI: 1.23, 4.49). An ANOVA examining the effects of showup-target-presence, race-condition, and clothing type on response latency, revealed no main effects of showup-target-presence, F(1, 289) = 0.12, p = .74, inline image, race-condition, F(1, 289) = 1.07, p = .30, inline image, or clothing-condition, F(1, 289) = 0.02, p = .90, inline image; however, we found a significant interaction between race and clothing, F(1, 289) = 5.53, p = .02, inline image, suggesting that differential decision-making processes might be employed in own-race and other-race identifications dependent on clothing. In own-race conditions, participants spent more time making their decision when the suspect's clothing was different than when it was the same (see Table 1). In other-race conditions, however, participants spent more time making their decision when the suspect's clothing was the same than when it was different. No other two-way interactions or the three-way interaction were significant (all ps > .1).

Part 2: Lineups

Overall

The period of delay between Parts 1 and 2 was not significantly related to correct lineup identification decisions, t(294) =−0.14, p = .89, d = .02; therefore we collapsed across delay in all subsequent analyses. Correct lineup decisions were made by 44.8% of participants, while the choosing rate was 54.2%. In target-present conditions, 51.0% of participants made a correct identification, while 13.5% identified a filler and 24.5% incorrectly rejected the procedure. In target-absent conditions, 38.0% correctly rejected the procedure, 21.0% identified a filler and 21.8% made a false identification. Correct decisions were significantly more likely in target-present than in target-absent conditions, χ2(1, N = 297) = 5.02, p = .03, Φ = 0.13. For the full breakdown of lineup decisions see Table 2.

Table 2. Lineup identification decision rates (and standard error) as a function of race condition, showup-target-presence, and showup accuracy
  Lineup condition and response
Target-presentTarget-absent
Correct IDFiller IDRejectionDon't knowCorrect rejectionFiller IDFalse IDDon't know
Overall 51.0 (4.0)13.5 (3.3)24.5 (3.5)11.0 (2.5)38.0 (4.1)21.1 (3.4)21.8 (3.5)19.0 (3.3)
   n = 79 n = 21 n = 38 n = 17 n = 54 n = 30 n = 31 n = 27
Own-race 52.2 (6.0)11.6 (3.9)29.0 (5.5)7.2 (3.1)53.1 (6.2)15.6 (4.5)17.2 (4.7)14.1 (4.4)
   n = 36 n = 8 n = 20 n = 5 n = 34 n = 10 n = 11 n = 9
Showup condition
 Target-present 59.0 (7.9)12.8 (5.3)20.5 (6.5)7.7 (4.3)78.6 (7.8)7.1 (4.9)0.0 (0)14.3 (6.6)
   n = 23 n = 5 n = 8 n = 3 n = 22 n = 2 n = 0 n = 4
 Correct decision90.0 (6.7)0 (0)10.0 (6.7)0 (0)84.6 (10.0)0.0 (0)0.0 (0)15.4 (10.0)
   n = 18 n = 0 n = 2 n = 0 n = 11 n = 0 n = 0 n = 2
 Incorrect decision26.3 (10.1)26.3 (10.1)31.6 (10.7)15.8 (8.4)73.3 (11.4)13.3 (8.7)0.0 (0)13.3 (8.7)
   n = 5 n = 5 n = 6 n = 3 n = 11 n = 2 n = 0 n = 2
 Target-absent 43.3 (9.0)10.0 (5.5)40.0 (8.9)6.7 (4.6)33.3 (7.9)22.2 (6.9)30.6 (7.7)13.9 (5.8)
   n = 13 n = 3 n = 12 n = 2 n = 12 n = 8 n = 11 n = 5
 Correct decision39.1 (10.2)13.0 (7.0)39.1 (10.2)8.7 (5.9)38.5 (9.5)23.1 (8.3)23.1 (8.3)15.4 (7.1)
   n = 9 n = 3 n = 9 n = 2 n = 10 n = 6 n = 6 n = 4
 Incorrect decision57.1 (18.7)0.0 (0)42.9 (18.7)0.0 (0)20.0 (12.6)20.0 (12.6)50.0 (15.8)10.0 (9.5)
   n = 4 n = 0 n = 3 n = 0 n = 2 n = 2 n = 5 n = 1
Other-race 50.0 (5.4)15.1 (3.8)20.9 (4.4)14.0 (3.7)25.6 (4.9)25.6 (4.9)25.6 (4.9)23.1 (4.8)
   n = 43 n = 13 n = 18 n = 12 n = 20 n = 20 n = 20 n = 18
Showup condition
 Target-present 61.5 (6.7)5.8 (3.2)15.4 (5.0)17.3 (5.2)36.1 (8.0)30.6 (7.7)16.7 (6.2)16.7 (6.2)
   n = 32 n = 3 n = 8 n = 9 n = 13 n = 11 n = 6 n = 6
 Correct decision95.5 (4.4)4.5 (4.4)0.0 (0)0.0 (0)47.1 (12.1)35.3 (11.6)11.8 (7.8)5.9 (5.7)
   n = 21 n = 1 n = 0 n = 0 n = 8 n = 6 n = 2 n = 1
 Incorrect decision36.7 (11)6.7 (2)26.7 (8)30.0 (9)26.3 (5)26.3 (5)21.1 (4)26.3 (5)
   n = 11 n = 2 n = 8 n = 9 n = 5 n = 5 n = 4 n = 5
 Target-absent 32.4 (8.0)29.4 (7.8)29.4 (7.8)8.8 (4.9)16.7 (5.8)21.4 (6.3)33.3 (7.3)28.6 (7.0)
   n = 11 n = 10 n = 10 n = 3 n = 7 n = 9 n = 14 n = 12
 Correct decision37.5 (9.9)25.0 (8.8)33.3 (9.6)4.2 (4.1)22.2 (8.0)22.2 (8.0)18.5 (7.3)37.0 (9.3)
   n = 9 n = 6 n = 8 n = 1 n = 6 n = 6 n = 5 n = 10
 Incorrect decision20.0 (12.6)40.0 (15.5)20.0 (12.6)20.0 (12.6)6.7 (6.5)20.0 (10.3)60.0 (12.6)13.3 (8.8)
   n = 2 n = 4 n = 2 n = 2 n = 1 n = 3 n = 9 n = 2

We initially looked at lineups as a whole and then examined target-present and target-absent lineups separately. We first examined the effects of showup-target-presence, race-condition, clothing-condition, and showup accuracy on overall lineup accuracy. Our model was significant, χ2(4, N = 297) = 56.87, p < .001, inline image, and we found significant main effects of showup-target-presence, Wald = 33.65, p < .001, OR = 0.19 (95% CI: 0.11, 0.33), race-condition, Wald = 6.23, p = .01, OR = 1.92 (95% CI: 1.15, 3.19), and showup accuracy, Wald = 23.04, p < .001, OR = 0.24 (95% CI: 0.14, 0.44). Participants were less likely to make a correct lineup decision when they had viewed a target-absent (30.3%) than a target-present showup (58.1%), and when they had made an incorrect (32.8%) than a correct showup decision (53.5%). They were more likely to make a correct decision in own-race lineups (52.6%) than in other-race lineups (38.4%). Relatedly, the likelihood of responding ‘don't know’ was somewhat higher in other-race (17.7%) than in own-race lineups (10.5%), but this difference did not reach significance, χ2(1, N = 297) = 3.04, p = .08, Φ = 0.10. There was no effect of the clothing worn in the showup on subsequent lineup accuracy, Wald = 0.06, p = .81, OR = 0.94 (95% CI: 0.57, 1.56).

Next we examined the effects of showup-target-presence, race-condition, clothing-condition, showup choosing, and lineup-target-presence on lineup choosing. Our choosing model was significant, χ2(5, N = 297) = 30.83, p < .001, inline image. We found significant main effects of showup-target-presence, Wald = 4.26, p = .04, OR = 1.72 (95% CI: 1.03, 2.89), showup choosing, Wald = 12.07, p = .001, OR = 2.67 (95% CI: 1.53, 4.63), and lineup-target-presence, Wald = 15.84, p < .001, OR = 0.37 (95% CI: 0.23, 0.60). Participants were more likely to choose from the lineup if they had previously chosen from the showup (66.7%) than if they had not (48.0%), if they viewed a target-present lineup (64.5%) than a target-absent lineup (43.0%), and if they had viewed a target-absent showup (55.6%) than a target-present showup (52.9%; although the absolute magnitude of the differences based on showup-target-presence was quite small). No other predictors were significant (all ps ≥ .09).

Target-present lineups

We next examined target-present lineups using showup-target-presence, race-condition, and showup accuracy as predictors (clothing had no effect in the overall lineup analyses and was removed from the target-present and target-absent lineup models). Our model was significant, χ2(3, N = 155) = 33.42, p < .001, inline image, and we found significant main effects of showup-target-presence, Wald = 15.81, p < .001, OR = 0.18 (95% CI: 0.08, 0.42), and showup accuracy, Wald = 20.59, p < .001, OR = 0.14 (95% CI: 0.06, 0.33). Correct identifications were less likely when the target had been absent (37.5%) than present (60.4%) from the showup, and when participants had made an incorrect (33.3%) than a correct (64.0%) showup decision. There was no effect of race on accuracy in target-present lineups, Wald = 0.00, p = 1.0, OR = 1.0 (95% CI: 0.49, 2.03).

Target-absent lineups

For target-absent lineup conditions, we again included showup-target-presence, race-condition, and showup accuracy. Our model was significant, χ2(3, N = 142) = 31.98, p < .001, inline image, with significant main effects of showup-target-presence, Wald = 16.88, p < .001, OR = 0.17 (95% CI: 0.07, 0.40), showup accuracy, Wald = 4.40, p = .04, OR = 0.41 (95% CI: 0.18, 0.94), and race-condition, Wald = 12.30, p < .001, OR = 4.11 (95% CI: 1.87, 9.07). Participants were less likely to make a correct decision when they had viewed a target-absent (24.4%) than a target-present (54.7%) showup and when they had made an incorrect (32.3%) than a correct (42.2%) showup decision. For target-absent lineups, own-race accuracy rates (53.1%) were more than double other-race accuracy rates (25.6%).

The overall pattern of results suggests that when race is not accounted for, commitment drives decision making in multiple identification procedures. Participants were significantly more likely to falsely identify the innocent suspect if they had viewed him in the showup (32.1%) than if they had not (9.4%), χ2(1, N = 142) = 10.59, p = .001, Φ =−.27. However, when choosers and non-choosers at the showup were examined separately, the relationship was significant only for choosers, χ2(1, N = 47) = 12.96, p < .001, Φ =−.53 [for non-choosers, χ2(1, N = 95) = 2.75, p = .10, Φ =−.17]. More than half of participants who had chosen the innocent suspect from the showup identified him a second time at the lineup (52.9%), which was twice the percentage of participants who identified him at the lineup after viewing him in the showup without choosing him (26.2%).

Different patterns of results emerged when own-race and other-race identifications were examined separately: viewing the innocent suspect in a showup was sufficient to increase false own-race lineup identifications, but choosing the innocent suspect from the showup was necessary to increase false other-race lineup identifications (for clarity, the showup-to-lineup data pertaining to commitment and source confusion extracted from Table 2 can be seen in Table 3). No participants (0%) identified an own-race innocent suspect at the lineup if they had not viewed him at the showup. For participants in own-race conditions who viewed the innocent suspect at the showup, the percentage identifying him increased significantly regardless of whether they had previously chosen him (40.0%), χ2(1, N = 18) = 5.85, p = .02, Φ =−.57) or had viewed him without choosing him (29.0%), χ2(1, N = 46) = 5.41, p = .02, Φ =−.34). The difference between own-race-showup choosers and non-choosers was not significant, χ2(1, N = 36) = 0.24, p = .62, Φ = .08; however, the small number of own-race showup choosers precludes drawing conclusions based on this result.

Table 3. Source confusion and commitment: percentage (and standard error) of eyewitnesses who did or did not view the innocent suspect in the showup who subsequently made a false identification of the innocent suspect versus correct rejection at the lineup*
  Lineup decision
False identificationRejected procedure
  1. *Only participants making one of these two lineup decisions are included. For information on participants making other types of decisions, see Table 2.

Own-race
Showup condition and/or choiceDid not view innocent suspect0.0 (0)78.6 (7.8)
   n= 0 n= 22
 Viewed innocent suspect30.6 (7.7)33.3 (7.9)
   n= 11 n= 12
  Showup chooser 40.0 (21.9)40.0 (21.9)
   n= 2 n= 2
  Showup non-chooser 29.0 (8.1)32.3 (8.4)
   n= 9 n= 10
Other-race
Showup condition and/or choiceDid not view innocent suspect16.7 (6.2)36.1 (8.0)
   n= 6 n= 13
 Viewed innocent suspect33.3 (7.3)16.7 (5.8)
   n= 14 n= 7
  Showup chooser 58.3 (14.2)8.3 (8.0)
   n= 7 n= 1
  Showup non-chooser 23.3 (7.7)20.0 (7.3)
   n= 7 n= 6

Participants in other-race conditions were more likely to misidentify the innocent suspect at the lineup if they had previously identified him at the showup (58.3%) than if they had not (11.8%), χ2(1, N = 29) = 7.13, p = .008, Φ =−.17). In other-race conditions, however, participants were equally likely to identify the innocent suspect at the lineup when they had viewed him at the showup without choosing him (23.3%) and when they had not viewed him at all (16.7%), χ2(1, N = 66) = 0.46, p = .55, Φ =−.08). In contrast to own-race conditions, participants in other-race conditions were more likely to misidentify the innocent suspect at the lineup if they had previously chosen him (58.3%) than if they had viewed him without choosing him (23.3%), χ2(1, N = 42) = 4.73, p = .03, Φ = .34.

Discussion

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study investigated the effects of race and clothing on showup identification decisions. We additionally explored whether there were differences in the manner in which eyewitnesses responded to multiple identification procedures when the suspect was from their own versus another racial group. We found that presenting the suspect in the same clothing as that worn by the perpetrator increased choosing rates. Race did not affect showup choosing or accuracy, although we found an interaction between race and clothing on response latencies at the showup. We also found evidence that decisions made in own-race versus other-race multiple identification procedures may be driven by different mechanisms, in addition to a lineup ORB.

Regarding the showup, our results were both consistent and inconsistent with prior research. In contrast to prior studies (e.g., Dysart et al., 2006; Lindsay et al., 1987; Yarmey et al., 1996), we found that clothing affected both correct and false identification rates via its effect on choosing: presenting the suspect in the same clothing worn by the perpetrator made participants more likely to choose the suspect, regardless of guilt or innocence. No interaction was found between clothing and race on choosing or accuracy; however, the response latency data indicated that participants in own-race and other-race conditions may have used clothing cues differently. In own-race conditions, participants spent more time making their decision when the suspect's clothing was different than when it was the same, while those in other-race conditions spent more time making their decision when the suspect's clothing was the same than when it was different. This suggests that, contrary to our expectations, clothing may be seen as more diagnostic of guilt in own-race than in other-race identifications. An alternative explanation, however, is that participants were more likely to reject an other-race showup out of hand if the clothing was different (which would make them correct in target-absent conditions but incorrect in target-present conditions); as seen in Table 1, other-race participants who viewed a guilty suspect in different clothing had the lowest accuracy rates. A more in-depth exploration of this effect is warranted, particularly as we did not find significant race interactions in our accuracy analysis. These results contradict our speculations regarding race and clothing on the basis of automatic recognition hypotheses (Charman & Cahill, 2011; Wilford et al., 2011). However, these hypotheses were supported in the overall response latency pattern: quicker responses were associated with greater accuracy, also consistent with previous research (e.g., Brewer et al., 2006; Dunning & Perretta, 2002; Sauerland et al., 2011).

Surprisingly, we found no evidence of an ORB on showup identifications, which is consistent with the results of a prior archival study (Behrman & Davey, 2001), but inconsistent with the literature on facial recognition and lineup paradigms (e.g., Brigham et al., 2007; Meissner & Brigham, 2001). Coupled with the findings from the archival study, our results suggest that the presence or absence of an ORB may be specific to the identification procedure used; however, given that the archival study was based on a non-random sample, additional replication of this result is needed. It is also possible that the low rates of choosing in target-absent showups (a result commonly found; Steblay et al., 2003) suppressed the effects of race. One further possibility is that our results were affected by location. The study was conducted in a large metropolitan area, where contact with people from other racial groups is the rule rather than the exception. Increased contact has been hypothesized to reduce the ORB (see Brigham et al., 2007); thus, it is conceivable that a showup ORB might be found with participants from other locations. This possibility is contradicted by our observed ORB with lineups. Participants in own-race and other-race conditions had comparable accuracy in target-present lineups, but those making other-race identification decisions were considerably less accurate in target-absent conditions. It is also contradicted by our observed differential responses to own-race and other-race multiple procedures.

There were differences in how the showup affected own-race and other-race lineup decisions. When participants as a whole were examined, evidence was found only for commitment; however, we found evidence of non-commitment-related source confusion in own-race conditions. To the extent that some prior research has found evidence only for commitment (e.g., Dysart et al., 2001), it is possible that a closer examination of race is warranted. Own-race-showup choosers and non-choosers were more likely to make a false lineup identification if they had viewed the innocent suspect at the showup; in fact, not a single own-race participant who did not view the innocent suspect in the showup falsely identified him at the lineup. Participants in other-race conditions, however, were more likely to falsely identify an innocent suspect at the lineup only if they had falsely identified him at the showup. For other-race-showup non-choosers, the percentage of participants making false lineup identifications was almost identical whether they had viewed a target-absent or target-present showup. It is possible that in other-race procedures, choosing the suspect is necessary for the level of encoding that would make future recognition possible. It is also possible that participants viewing other-race suspects relied on their earlier decision to a greater extent than participants viewing own-race suspects. Dual processing theories suggest that ability plays a role in the level of processing used (see Evans, 2008). It is possible that reduced expertise in other-race compared to own-race identifications makes eyewitnesses viewing suspects from another racial group more likely to rely on shortcuts such as prior decisions, rather than more carefully scrutinizing the lineup members; further examination of this possibility is warranted.

The results of the current study provide information about how eyewitnesses evaluate and make decisions from own-race and other-race showups and lineups and suggest ways in which they might approach them differently. Own-race and other-race showup accuracy was comparable, and for both, presentation of the suspect in the same clothing worn by the eyewitness increased choosing rates. However, the differences in response latency suggest differential evaluation of clothing cues. Additionally, our results indicate that somewhat different factors may affect eyewitnesses participating in own-race and other-race multiple identification procedures. Merely viewing an innocent suspect in a showup was sufficient to increase subsequent false lineup identifications when the suspect was from the same racial group; when the suspect's race differed, it was necessary for eyewitnesses to have previously identified him in order to increase false lineup identifications. Eyewitnesses viewing other-race suspects appear overall to be at a deficit when multiple procedures are used as evidenced by significantly lower accuracy in target-absent lineups.

The risks of using showups may lie not only in their inherent suggestiveness, but in that they may commonly be used in conjunction with another identification procedure. If an eyewitness identifies the suspect from a showup, a lineup may follow in order to provide ‘confirmation’ of the identification for admissibility purposes or to address concerns regarding reliability. Even if an identification is not made, police may present the same suspect to an eyewitness a second time if he remains under suspicion or is drawn to their attention for another reason (e.g., the case of Michael Anthony Green, http://www.innocenceproject.org). Prior research clearly demonstrates that either scenario presents a real danger to an innocent suspect (e.g., Behrman & Vayder, 1994; Godfrey & Clark, 2010; Haw et al., 2007); our results suggest that the implications may differ dependent on whether the procedures were cross-racial.

Footnotes
  • 1

    Patterns of identification decisions with different confederates were largely the same. There were no significant differences in accuracy between the confederates from each racial group at the showup or the lineup. Further, there were no significant differences in accuracy based on different cross-racial combinations.

  • 2

    Although field showups may be more commonly used by law enforcement, photographic showups are not uncommon in the United States (Valentine et al., 2012), and an archival study found similar suspect identification rates for field and photographic showups (Behrman & Davey, 2001).

  • 3

    The vast majority of participants (95.6%) remembered their earlier decision. When we examined results for only those participants who remembered their decisions separately, the same pattern of results was found; therefore, we included all participants in our analyses.

  • 4

    The same pattern of results was found when participants who responded ‘don't know’ were excluded.

  • 5

    Because ‘don't know’ responses were considered incorrect, we performed a separate choosing analysis.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The authors would like to thank Dr. H. Philip Zeigler for his assistance in recruitment of participants and Drs. Gary Wells and Steven Penrod for their comments on an earlier version of the manuscript.

References

  1. Top of page
  2. Abstract
  3. Background
  4. Method
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References