Does Speed Indicate Lineup Identification Accuracy

AMERICAN JOURNAL OF FORENSIC PSYCHOLOGY, VOLUME 32, ISSUE 2, 2014 / 1

DOES SPEED INDICATE LINEUP IDENTIFICATION ACCURACY? EXAMINING CHILDREN’S AND ADULTS’ REACTION TIME Kaila Bruer M.A. and Joanna D. Pozzulo, Ph.D.

The purpose of this study was to assess young children’s lineup identification performance compared to adults and to determine whether developmental variability exists in reaction time when making correct and incorrect identification decisions across target-present and target-absent lineups. Adults (𝑀𝑎𝑔𝑒 = 20.00) and young children (𝑀𝑎𝑔𝑒 = 4.69) were exposed to an unfamiliar target and the time taken for them to make a lineup decision was automatically recorded by touching the picture on a computer screen. Children were found to have fewer hits (i.e., correct identifications and rejections) and more false alarms than adults in both lineup types. In addition, results support the hypothesis that a faster decision was related to improved accuracy for adults, while a slower decision may be suggestive of improved accuracy for children. The developmental variability may suggest that developmental cognitive factors may influence lineup behavior. Implications of the findings and recommendation for future research are discussed. Many crimes, such as abductions, thefts, and sexual assault occur where an investigation and/or a conviction may be dependent upon a child eyewitness (1). Eyewitness testimony, particularly recognition or identification testimony is a frequent form of evidence relied upon during criminal investigation and in court (2). Despite its value and importance, numerous cases have transpired where eyewitnesses have erred in their identification of a perpetrator, resulting in the conviction of innocent people (3). Research has consistently revealed that eyewitnesses, including child witnesses, can make mistakes (4). Although children can be as accurate as adults when presented with a photo array containing the culprit (i.e., target-present lineup), developmental differences emerge when the culprit is not included in the photo array (i.e., target-absent lineup). Specifically, children are more likely to identify an innocent person from the lineup than adults (1, 4, 5). While various research initiatives have tried to understand or explain this age-related difference in identification accuracy (e.g., 4, 6), no clear explanation has emerged. It has been suggested that young children’s false positive responding is, in Copyright 2014 American Journal of Forensic Psychology, Volume 32, Issue 2. The Journal is a publication of the American College of Forensic Psychology, PO Box 130458, Carlsbad, California 92013.

2 / BRUER AND POZZULO: EYEWITNESS IDENTIFCATION AND REACTION TIME

part, the result of social factors (e.g., social pressure to make a selection [4, 6, 7]); however, it is not clear what role “cognition” may have on children’s and adults’ identification decisions. By proxy, one option to examine the role of cognitive development is to examine an eyewitness’s reaction time when making a lineup decision. Reaction time (RT) has been used in other areas of study to indicate cognitive processing development (8, 9). In the eyewitness literature, a considerable amount of research has examined the relationship between reaction time and identification (ID) accuracy for adult populations. This research has generally revealed that adult witnesses tend to make accurate identifications faster than inaccurate decisions (10). However, there has been minimal research examining this relationship with a young child population. Therefore, it is not clear whether identification decisions by child eyewitnesses result in similar patterns to adult eyewitnesses. As a result, the present study examined whether developmental variability exists in both identification accuracy in TP and TA lineups and in reaction time among young children and adults— specifically, between those who make correct versus incorrect identification decisions. AGE AND LINEUP PERFORMANCE

The underperformance of children in lineup identification tasks has been well established (1, 4, 5, 11). What this research has revealed is that children can be as accurate as adults when presented with target-present (TP) lineup; however, children are also more disposed to identifying an innocent person from a target-absent (TA) lineup than adults (4). However, a significant portion of this literature has focused on older children (i.e., eight years and older [1, 5]) given the infrequent appearance of younger children in courtroom settings. Only a handful of studies have examined lineup performance of young children (i.e., 5-year-olds [4]) compared to adults and, thus, the current study is intended to add to this literature. THEORETICAL FOUNDATIONS

Eyewitness and cognitive paradigms, alike, have comprehensively examined the relationship between decision time and accuracy when recognizing faces with adult populations. Differences in reaction time can be understood at both a theoretical and empirical level. Dual-processing models of recogni-


tion are often used to examine differences in RT. These dual-processing theories of recognition (e.g., remember/know [12]; relative/absolute [13]), as well as its earlier predecessors (14-16) propose that recognition is based on two levels of processing—a fast level that is based on familiarity, as well as a slow one that is based on learning and conscious effort. For example, Wells (13) proposed that an increased RT is related to inaccurate decisions because witnesses use relative judgments (i.e., comparing different members of a lineup), rather than using an automatic judgment (i.e., immediate decision that does not require comparisons). In other words, these dual processing theories generally propose that when a witness recognizes a face, their reaction time should be based on familiarity and, therefore, be much quicker than when they do not recognize a face. Research has revealed support for dual-processing theories in that witnesses who take longer when making identification decisions have been found to be less accurate than witnesses using short cognitive processes (17). For example, Schooler and Engstler-Schooler (18) examined the relationship between time taken to verbally describe faces and the time taken to identify faces. Using an adult sample, they found when witnesses were asked to make recognition judgments, those who made judgments within five seconds were more accurate than those who took longer to make a decision. Thus, this suggests that when witnesses make recognition judgments quickly they are more accurate than those making judgments slowly. Similarly, Dunning and Stern (19) proposed that automatic (i.e., fast) cognitive processes are more strongly associated with accurate decisions than witnesses who explicitly think about their decision (i.e., take longer to think about decision). In a meta-analysis, Dunning and Stern (19) examined four separate studies that involved participants viewing a staged crime and then provided a selfassessment of their decision process when recognizing the culprit. Three of the four studies found that accurate witnesses were more likely to report fast (M = 38.4 seconds), “automatic” (e.g., face just stood out) recognition decisions, while inaccurate witnesses were more likely to report longer (M = 63.9 seconds), “process of elimination” decisions. Despite the limitations with relying on participants’ self-reporting (e.g., inaccurate reporting, social desirability), this study demonstrates support for cognitive influences on lineup decision-making.


REACTION TIME AND ACCURACY: LINEUP TYPE

Although the majority of research findings support the notion that increased speed of reaction time is related to accuracy in lineup identifications, some research suggests that this relationship is more complex (17). It has been suggested that the impact of reaction time on accuracy differs depending on whether the eyewitnesses make an identification decision, or reject all members of the lineup. Sporer (20), for example, investigated the relationship between reaction time and accuracy using both simultaneous and sequential lineups. Using an adult sample, participants were shown a film depicting a crime and were asked to make a decision in a lineup task. In targetpresent lineups (TP; i.e., the guilty person is among the photo array), the results indicate that witnesses who made an accurate identification (i.e., choosers) did so significantly faster (M = 3.61 seconds) than witnesses making an inaccurate decision (M = 8.06 seconds). However, in target-absent lineups (TA; i.e., the guilty person is not among the photo array), witnesses who correctly rejected all members of the lineup (i.e., non-choosers) took slightly longer to make their decisions than the inaccurate witnesses. Similar results were found in other studies by Hosch et al. (21), Smith et al. (22) and Sporer (23). Specifically, these studies found that in target-present lineups, as decision time increased, the decision accuracy decreased; however, in targetabsent lineups, an increase in decision time may be related to increased accuracy. These findings may be suggestive of differing cognitive processes involved in making a decision in a TA and TP lineup. As a result, it was hypothesized that accurate identifications in TA lineups would result in slower RT for adult witnesses than in TP lineups. REACTION TIME AND ACCURACY: AGE Adults

As previously indicated, a significant amount of research has examined the relationship between RT and accuracy with adult populations. Dunning and Perretta (10), for example, conducted a meta-analysis to examine the relationship between RT and accuracy for adult witnesses by grouping the RT into specific ranges based on accuracy. The authors found that identifications made within 10 seconds were accurate 87% of the time, identifications made between 11-20 seconds were accurate 53% of the time, and identifica-


tions made between 21 and 30 seconds were accurate 51% of the time. In a similar study, Smith et al. (22) examined RT using ranges and found that those who made a decision (i.e., choosers) in 15 seconds or less were accurate 70% of the time, whereas participants who took more than 30 seconds to make their decisions were accurate only 18% of the time. For the participants who were non-choosers, decision time and accuracy were not significantly correlated. Thus, these studies suggest that a loose cut off of 10-15 seconds can be used to estimate ID accuracy. Sauer et al. (24) tested the validity of the 10-12 second rule identified by Dunning and Perretta (10) for discriminating accurate from inaccurate identifications. This study tested participants’ memory of a crime video using two different sequential lineups. In one of the lineups, accurate identifications were found to be faster (i.e., approximately 4 seconds) than inaccurate identifications. More recently, Sauerland and Sporer (25) further investigated the relationship between reaction time and accuracy and found that, for TP lineups, accurate participants made faster decisions (M = 8.6 seconds) than inaccurate participants (M = 12.1 seconds), while no differences were found in accuracy for TA lineups. These results were replicated in research by Kneller et al. (26), Sporer (23), and Weber and Brewer (27). Given the extensive support for a relationship between increased RT and recognition accuracy for adults (10), it was predicted that reaction time would be related to accuracy, such that faster reaction times will be more accurate than slower reaction times. While much research has examined the relationship between RT and accuracy in adult populations, a common limitation of the existing research is the lack of precision in measuring the decision times. For example, in the study conducted by Sporer (20), reaction times for the participants were measured by the researcher using a stopwatch, and they were recorded as the time elapsed between the uncovering of the lineup display in front of the participant, and the participant’s completion of the sentence that either indicated a positive choice or a rejection of the whole lineup. Several of these studies relied upon a third party to measure the decision time of the witnesses, raising questions about the accuracy and precision of these results (25). The present study provides additional information for the existing knowledge base on adults’ RT and identification accuracy through using current technology (i.e., touch screen computers and EPrime software). The present study also


adds to the existing literature given that it provides a more efficient method of measuring precise reaction times. Children

Despite the support for use of dual-processing models to understand the relationship between RT and identification accuracy for adults, these frameworks (e.g., remember/know) have not been applied to a young child population. There is some evidence from related areas of research that suggest that young children (under six years old) undergo different cognitive processes than adults when recognizing faces (28-31), thus, speed may not be directly related to accuracy. The present study examined the relationship between RT and ID accuracy in young children. The only study we are aware of that has examined developmental differences in reaction time in an eyewitness context is that by Brewer and Day (32). This study found that children were marginally faster than adolescents and that children were faster when they made correct versus incorrect decisions. However, this study examined an older group of children (8- to 10year-olds) than the present study, and thus the findings may not be representative of younger children. To the best of our knowledge, there has been no research that has examined children’s reaction time (i.e., decision time) with young children in an eyewitness context (i.e., making a lineup decision). There has, however, been some related research on the development of facial recognition that sheds some light on how decision time may be related to accuracy for young children. For example, Carey and Diamond (29) examined children’s (6- and 10-year-olds) RT when recognizing different facial compositions and found a developmental decrease in RT, such that 6-yearolds were slower (1339 ms) than 10-year-olds (1023 ms), who, in turn, were slower than adults (980 ms). Similarly, Ellis and colleagues (33) conducted a study that examined the influence of priming in children’s recognition of faces. Again, developmental differences in RT emerged, such that 5-yearolds were slower (2478 ms) than 8-year-olds (1795 ms), who, in turn, were slower than 11-year-old children (1457 ms). While these studies do not examine ID accuracy in relation to RT, these data do suggest that there may be an age-related decrease in the time needed to process and recognize faces, such that young children may require more time to make an ID decision than older children and adults. This is consistent with research suggesting that


young children may process (and recognize) faces differently than older children and adults (31). Therefore, it was hypothesized that young children’s increased RT would be related to accuracy in recognizing faces (for TP and TA lineups). PRESENT STUDY

The present study observed the relationship between RT and accuracy for child and adult witnesses across both target-present and target-absent lineups. Given that Pozzulo and Lindsay’s meta-analysis (4) revealed that children under five years were found to produce a lower correct identification and rejection rate compared to adults, it was predicted that young children would have a lower correct ID rate compared to adults. In addition, this research was expected to reveal differences in RT between children and adults who make correct lineup identifications and those who make incorrect identifications. If differences in RT are found between correct/incorrect decisions and age categories, it could suggest a more cognitive development explanation. Although we do not measure cognitive development specifically, we are using decision time as a proxy for impulse control (i.e., underdeveloped executive functioning in children). METHOD Participants

Participants included 64 children ranging in age from 4 to 6 years (Mage = 4.69, SD = .71; males = 38; females = 26) recruited from daycare centers in Eastern, Ontario. Sixty-two adults, ranging in age from 18 to 24 years (Mage = 19.66, SD = 3.14; males = 26; females = 35), were recruited from the introductory psychology participant pool from a school in eastern Ontario. Design

A 2 (age: child versus adult) x 2 (lineup: TA versus TP) between-group experimental design was used. Materials

Demographic form. Each participant (or parental guardian) was provided with a response form requesting demographics. Three questions were asked


requesting the participant’s age, gender, and primary language. Adult participants completed the form themselves. For child participants, the form was completed by the parent/guardian. Target video. One male Caucasian university student (22 years old) was used as a target. The target was filmed completing an everyday task for sixseconds (e.g., video clip displayed a male putting on his coat and exiting his home). The video provided a 3-second close-up of the individual’s face. The target video was filmed in color. Foils. The target was photographed in a different outfit than what was worn during the video clip. The foils were selected from a pool of 90 male faces. The foil photographs were selected based on similar appearance to the target. Similarity was measured in terms of general facial structure, hair length, and color. Three raters selected 4 foils for the target. Target and foils were closely cropped such that their face, neck and the tops of their shoulders were photographed. Target-present lineups contained the target and three foils. Target-absent lineups contained four foils. All photos were presented in color. Lineup presentation. A lineup was presented using the computer program E-Prime (Version 2). A simultaneous procedure was used to present the lineup. That is, for each lineup, all pictures were shown at once. In addition to the lineup members, each lineup also included a silhouette to represent the possibility of an absent target. Each participant saw the video and was shown one photoarray. The position of the target/replacement was stationary across photoarrays. The videos and photoarrays were displayed using a 13-inch touch screen laptop using the program E-Prime. Instructions for lineup identification. The following instructions were provided prior to the display of each lineup: “I am going to show you some pictures on the computer screen. They will look something like this [indicating on a piece of paper]. There will be four pictures at the top and one at the bottom. Please look at the pictures. The person you saw in the video might be one of these pictures, or they might not be one of these pictures. If you see the person from the video, please touch his face. If you do not see the person, please touch the box at the bottom [indicating].” Free recall descriptions. All participants were asked an open-ended question to describe everything they could remember about each video clip.


The researchers recorded each child participant’s responses, while the adult participants recorded their own responses. This task was used as a brief filler between exposure of the video and presentation of the lineup. Approximately two minutes elapsed between each video exposure and lineup presentation. Lineup Administrators. Two or three female experimenters showed children the video clips and photoarrays. Experimenters were sure to always dress in professional-casual attire (e.g., sweater/blouse and dress-pants) in order to reduce external visual cues of authority (see Lowenstein et al. [34]). More specifically, experimenters were “neat” in appearance but not overly formal (e.g., no uniforms or lab coats). Procedure

Adults. Upon entering the laboratory, each participant was given a consent form that explained that they would be participating in a study examining videos. Participants who gave consent were then shown the target video. The participants were asked to pay attention because following the video they would be asked some questions. After the video, the participants were provided with a sheet asking a free recall question, “What did the person in the video look like?” The participant then wrote down all they could remember about what they saw on the video. This question was followed up with, “Do you remember anything else about the person in the video?” Once that was completed, the experimenter presented the participant with the touch screen laptop, using E-Prime, randomly displaying either a target present (TP) or target-absent (TA) lineup. The experimenter asked the participant to identify (by touching) the person they saw in the video if he was present. The experimenter informed the participant that the person they saw may not be there and demonstrated that, in this case, the participants should select the option that corresponds to the silhouetted photograph. Time elapsed between viewing the target and lineup identification was approximately five minutes. Following completion of the video and lineup, the participants were given a demographic questionnaire and were debriefed and thanked for their participation. Children. Parents/guardians of the children attending daycares in the community were supplied with a written consent form, as well as a demographics sheet. Upon receiving written consent forms and completed de-


mographic forms, two to three female experimenters and one female facilitator arrived at each daycare. Only children with consent were invited to participate. The researchers were introduced to the students as a group from the university doing a memory game with computers. During the introduction and invitation to participate, the researchers made it clear to the children that they could change their minds at any time and not get into trouble. In order to create a level of comfort with the children, the researchers worked with the children to make some crafts prior to engaging the children in the experimental task. Experimenters tested children individually. The procedure for data collection was comparable to that used with adults. Once the child was comfortable, the experimenter played the video. Each child was told that he/she would be watching a video and was told to pay special attention because, following the video, they would be asked some questions about what they saw. Following the video clip, the experimenter asked the child one free recall question about what they remembered about what the person in the video looked like. Following the child’s response, the experimenter asked a non-specific, probing question, i.e., “Do you remember anything else about what the person looked like?” After recording (in writing) the information provided by the child, the experimenter displayed the lineup on the touch screen laptop to the child. The experimenter asked the child to identify the person they saw in the video by touching the person’s face. The experimenter instructed the child that the person they saw may or may not be there and demonstrated that, if the correct person was not there, they should touch the silhouetted box at the bottom. Time elapsed between viewing the target and lineup identification was approximately five minutes. Following the end of the study, the children were thanked and given a small token (i.e., crayons and coloring book). RESULTS Statistical Assumptions

Prior to analysis, the accuracy of the data and the assumptions of an ANOVA (i.e., independence, normality, and homogeneity of variance) were examined. Four cases with extremely high z scores (i.e., < 3.29 SD) on mean reaction time were identified as a univariate outlier. Because the outlier was determined to be part of the population of interest (just an extreme case), the


scores were adjusted to fall within 3.29 SD of the mean. In addition, variables met the assumptions of independence and normality given that there were more than twelve participants per cell (35). Therefore, the analysis proceeded without correction. Age Differences

In order to ensure there were no age effects, a series of ANOVAs were conducted between ages within each age group. No significant differences were found in the reaction time between the children, F (1, 61) = 2.51, p =.09, nor the adults, F (1, 54) = .85, p =.55. Descriptive Statistics

Means and standard deviations on key dependent variables are presented in Table 1. Correct identification and rejection rates are presented in Table 2. ID Accuracy

Data were divided into target-present versus target-absent lineups given the identification decision differs for each and children and adults differ in identification as a function of age (4). Target Present

As predicted, young children compared to adults produced a significantly lower rate of correct identification (i.e., hits) in the target-present condition (.26 vs. .74). X 2 (1, N= 65) = 8.68, p =.003.

Target Absent

As predicted, young children compared to adults produced a significantly lower rate of correct rejection in the target-absent condition (.26 vs. .73). X 2 (1, N = 62) = 16.67, p