This is a pre-publication version of the following article: Ingham, M. P., Colloff, M. F., Smith, H. M., & Flowe, H. D. (2021). Innovating eyewitness identification procedures: Bridging the gap between the basic and applied literatures. The Cognitive Psychology Bulletin, Spring(6), pp. 42 - 49.
Introduction
Eyewitness identification from police lineups can have profound consequences. In a lineup, the suspect, who may be guilty or innocent, is presented alongside fillers, who are known by the police to be innocent of the crime in question, and who physically resemble the suspect. The goal of the police in administering the lineup is to test whether the witness can detect the suspect from the fillers. Eyewitness identification evidence plays an integral role in criminal investigations and forms a key component of prosecutions. Eyewitness testimony has been cited as one of the most influential types of evidence that can be presented at trial (Benton et al., 2006). The identification of a guilty suspect can tie a defendant to criminal actions and to forensic evidence, and thereby lead to a successful prosecution. However, mistakes can lead to a guilty suspect going free, or a wrongful conviction.
On the one hand, little research has examined how often eyewitnesses in actual cases make correct identifications. One unpublished master’s thesis compared eyewitness identification outcomes to DNA evidence in cases prosecuted by a District Attorney’s Office in the US. The study found that more than 95% of the time there was a DNA match between the biological evidence left at the crime scene and the DNA of the suspect that the witness had identified (Finklea & Ebbesen, 2007). On the other hand, with respect to inaccurate identification, to date, 365 wrongful convictions have been overturned in the US on the basis of DNA evidence indicating that the person identified by the witnesses did not match the biological evidence left at the crime scene. The Northern Ireland Department of Justice has reported 84 unsafe convictions in the UK between 2007 and 2017, costing the UK government £9 million in compensation costs alone (BBC, 2019). The human and financial costs of these errors have led international organisations (e.g., National Academy of Sciences, National Institute of Justice, American Bar Association) and researchers (e.g., Brewer & Wells, 2011) to call for the development of improved lineup procedures. Most recently, the US National Research Council convened an expert working group to review the state of the science (National Research Council, 2014). The report called on researchers to develop innovative technologies for conducting police lineups to improve eyewitness identification accuracy.
In this paper, we introduce and review research on interactive lineups, an innovative procedure that enables witnesses to actively engage with the lineup members’ faces. The goal of this work is to improve people’s ability to discriminate between innocent and guilty suspects. Procedures that improve discrimination accuracy should be of particular interest to policy makers, because such procedures maximize the likelihood of correct guilty suspect identifications while minimising the likelihood of incorrect innocent suspect identifications (Clark 2012; Gronlund et al., 2015; National Research Council, 2014). Before introducing interactive lineups, we review the literature on lineup procedures, and the theoretical perspectives from basic face recognition research that have informed the development of interactive lineups.
Lineup procedure research
While the procedures police use to administer lineups vary worldwide, there have been few innovations in how lineups are conducted. In some countries (e.g., South Africa), a live lineup procedure is used where the lineup members are shown together to the witness in person. In the UK, witnesses see video parades, wherein a video clip of each lineup member is presented one at a time, with each lineup member turning their head to the left and the right. In other countries (e.g., the US, Germany, Canada, Australia), the dominant procedure is a static frontal pose photographic lineup, wherein witnesses are presented with photographs of each lineup member facing the camera, in frontal pose, from the shoulders up (Fitzgerald, Price, & Valentine, 2018). Live, video, and photo lineups are the primary way in which lineups are conducted, and these procedures have been in use for decades.
Eyewitness memory researchers study how different lineup identification procedures affect performance using controlled laboratory experiments (see Flowe, Carline, & Karoğlu, 2018 for a review). Participants in these studies usually witness a staged crime, (e.g., a theft), which is normally presented via video. After a delay, typically 5 minutes, participants are presented with a lineup, with half of participants seeing a lineup in which the culprit is present (i.e., a target-present lineup), and the other half seeing one in which the culprit is absent (i.e., a target-absent lineup). When the target is present, selecting the culprit is a correct decision (i.e., a hit), while selecting a filler (i.e., a false alarm), or rejecting the lineup (i.e., a miss) are incorrect decisions. When the target is absent, rejecting the lineup is a correct decision (i.e., a correct rejection), while selecting a filler (or the innocent suspect if the researcher has designated one) is incorrect (i.e., a false alarm). Looking across experiments, participants make an incorrect identification decision around half of the time, regardless of lineup medium (see Fitzgerald et al., 2018). Moreover, research to date suggests that discrimination accuracy, which measures the ability of the witness to distinguish guilty from innocent suspects, does not consistently vary in relation to the procedure (photo, live, video) used to display the lineup members to participant witnesses, with results mixed across studies (e.g., Clark et al., 2015; Cutler & Fisher, 1990; Rubínová et al., in press; Seale-Carlisle & Mickes, 2016; Seale-Carlisle et al., 2019; Valentine et al., 2007; for reviews, see Cutler et al., 1994; Fitzgerald et al., 2018). However, from a theoretical standpoint, it is surprising that discrimination accuracy does not vary across procedures.
Theoretical Perspectives
According to the diagnostic feature detection theory, discrimination accuracy (i.e., the ability to distinguish guilty from innocent suspects) should be boosted by lineup procedures that help witnesses detect features that are unique to the culprit (i.e., diagnostic features) and rule out features that are shared across all lineup members (i.e., nondiagnostic features) (Wixted & Mickes, 2014; Wixted et al., 2018). Therefore, a procedure that allows a witness to view more facial features—such as a video lineup, wherein the members’ faces are shown moving and from all angles, should increase discrimination accuracy over a procedure in which the witness is able to see fewer features. This hypothesis is consistent with a wealth of research and theory from the basic face recognition literature. In face recognition studies, participants are presented with a list of unfamiliar faces, usually displayed one at a time and, after a short distractor task, view faces and are asked to decide whether or not they have seen each face before. Therefore, a key difference is that participants do not see an episodic event in basic face recognition studies.
From the basic face recognition literature, three key insights about how to improve discrimination accuracy in lineups emerge. First, according to the representation enhancement hypothesis, facial movement facilitates the perception of the three-dimensional structure of a face, and this in turn benefits memory retrieval (O’Toole et al., 2002). Scenes containing people (speaking, talking, and moving) that are learned in motion (i.e., shown dynamically) are better recognized when participants are tested in motion compared to a static format (e.g., Buratto, Matthews, & Lamberts, 2009). Considering that witnesses view culprits in motion when crimes are committed, the presence of motion cues in a lineup, therefore, should improve discrimination accuracy compared to when witnesses can only see the lineup members presented in static frontal pose photographs. However, other basic face recognition research suggests that the availability of motion cues at test does not increase recognition accuracy for faces that are learned in frontal pose and displaying non-rigid motion (i.e., speaking sentences) (Butcher, Lander, Fang, & Coston, 2011). These results suggest that recognition accuracy could depend more on whether a face is learned in motion rather than tested in motion. They also underscore the need to test different types of motion cues (i.e., rigid motion, such as head turning, as well as elastic or non-rigid motion, such as smiling) at learning and at test in the eyewitness memory domain. Second, it has been widely shown that face recognition accuracy is viewpoint dependent. Recognition accuracy is higher when a test face is presented in the same orientation in which it was studied (e.g., Bruce, 1982; Carbon & Leder, 2006). Therefore, eyewitness performance should benefit from the opportunity to view lineup faces in the same pose as the culprit was encoded. Third, procedures that allow for the active exploration of faces at test should improve discrimination accuracy in lineups. Active exploration refers to the test taker being able to interact with the test face on a computer screen, by rotating it, using a mouse, which allows for seeing the different viewpoints of the face for the length of time desired. Object and face recognition research demonstrate that active over passive exploration during recognition enhances memory retrieval (e.g., Harman, Humphrey, & Goodale, 1999; Liu, Ward, & Markall, 2007).
These retrieval cues—motion, pose reinstatement, active exploration—which have been shown to facilitate accurate memory retrieval in basic face recognition paradigms, have not been adequately considered in the eyewitness context. While there have a been a number of eyewitness studies comparing lineup procedures (video, photo, live), a major shortcoming of these studies is that the pose in which the culprit was encoded was not controlled or manipulated, with many studies not even reporting crucial information about the extent to which participants could see the culprit moving or could see different angles of the culprit’s face. Further, only one study (Bailenson, Davies, Beall, Blascovich, Guadagno, & McCall, 2008) has investigated active exploration in an eyewitness context. Bailenson et al. (2008) employed an interactive virtual reality eyewitness memory paradigm with avatars, not real faces. The results indicated that multiple face viewpoints increased accuracy, but only when the perpetrator was present in the lineup (Bailenson et al., 2008). However, the study did not measure discrimination accuracy (i.e., the target present and target absent trials were analysed separately). Therefore, an outstanding matter is whether basic face recognition theory regarding motion, pose reinstatement, and active exploration, generalizes to the eyewitness identification context.
There are a number of differences between basic face recognition and eyewitness memory paradigms that could limit the generalizability of the findings across research contexts. First, participants in eyewitness memory studies are each given just one study-test trial, whereas those in face recognition studies are each given numerous study-test trials. In basic face recognition studies, participants may attend to cues, such as pose and motion, more so than they would in a more naturalistic context. This is because in a face recognition paradigm, the cue being varied (i.e., motion, or pose) at encoding and test across faces may be particularly salient to participants because everything else is held constant across trials (see Grayson, 1998). Second, subjects in standard recognition paradigms make an identification decision about a single face, whereas in a lineup task, witnesses have to detect whether the perpetrator is present among a number of similar-looking faces. This requires subjects to reject the filler faces and identify the culprit’s face. The presence of multiple faces during the test may fundamentally alter memory retrieval processes compared to when there is only one face (see Colloff & Wixted, 2019). In a similar vein, one basic recognition study found that pose is less influential as a retrieval cue when the target is very similar to one or more distractors (Logie et al., 1987). Together, it is not clear whether eyewitnesses can use motion, pose, or active exploration to boost discrimination accuracy and given the important differences between the literatures, this requires testing in an eyewitness paradigm.
Interactive Lineups
We have been investigating whether cues (i.e., rigid motion, pose reinstatement and active exploration) that have been found to improve accuracy in the basic face recognition literature can improve discrimination accuracy in lineups (Colloff et al. 2020a; Colloff et al. 2020b) and in face matching (Smith et al., 2020), as well as exploring children’s ability to make identifications using the system (Windsor et al., 2020). In all of our studies, we have employed an eyewitness memory paradigm and large samples of participants and have controlled the manner in which participants encoded the culprit. Further, we have analysed the data using ROC analysis, which has been employed for decades in the basic scientific literature on memory. This means that our results are tethered to signal detection theory, a longstanding model of decision-making. ROC analysis was applied in the context of eyewitness lineups for the first time in 2012 (Mickes et al., 2012). Before this, the applied field of eyewitness identification did not rely on a principled mechanistic understanding of memory, perception, and decision-making (Albright & Rakoff, 2020). ROC analysis is recommended as a replacement for the intuitive, yet potentially misleading, measurement methods previously used in the eyewitness field to compare the performance of lineup procedures (see Mickes et al., 2012). Although ROC analysis is now a primary way in which eyewitness memory performance is measured, it should be noted that some researchers argue for using their own ROC-like alternatives instead of traditional ROC analysis (e.g., Smith, Yang, & Wells, 2020). The novel alternatives being advocated are not tied to any formal model of decision-making, and therefore are flexible and may result in misleading conclusions.
To investigate the effects of motion, pose reinstatement and active exploration in an eyewitness identification context, we have developed an interactive lineup procedure, wherein participant witnesses can rotate and hold the lineup members’ faces in any position desired along the vertical axis, from left-profile to right-profile. This allows witnesses to dynamically view each face at different orientations (see https://tinyurl.com/t4nc9gp). The system we have developed also records mouse movements; therefore, we know what angles the participant explored and for what length of time.
Using this interactive procedure to investigate whether pose reinstatement improves discrimination accuracy over static frontal pose lineups, we found that discrimination accuracy is significantly higher when the viewing angle of the lineup members is the same rather than different from the perpetrator at encoding (Colloff et al., 2020a). Participants were also found to reinstate the angle of the interactive faces to match the angle in which they encoded the perpetrator. These findings highlight the importance of encoding specificity for eyewitness identification (e.g., Tulving & Thomson, 1973), and demonstrate that people actively seek out cues in the testing environment that match the study environment, and this aids memory retrieval.
We have also studied whether interactive lineups can attenuate own race bias, a phenomenon whereby discrimination accuracy is higher for own compared to other race faces both in basic face recognition and eyewitness memory paradigms (Meissner & Brigham, 2001). We found that interactive lineups enhanced witnesses’ discrimination accuracy compared to static frontal pose lineups for both own-race and other-race identifications (Colloff et al., 2020b). In this work we also explored the effects of presenting the members of the interactive lineup simultaneously (all at once on the computer screen) compared to sequentially (one at a time). Previous research has found higher discrimination accuracy for simultaneous compared to sequential lineups (see Seale-Carlisle, Wetmore, Flowe, & Mickes, 2019 for a meta-analysis). Similarly, we found that interactive simultaneous compared to sequential lineups lead to higher discrimination accuracy. We also explored whether discrimination accuracy is higher in joint movement (i.e., all of the faces moving at once) compared to independent movement (i.e., one face can be moved at a time), and found that while joint movement leads to better performance the difference was not statistically significant (Colloff et al., 2020b).
We have also extended the interactive testing procedures to forensic face matching (Smith et al., 2020), with the results underlining the wide-ranging utility of this procedure. Participants either compared two frontal images, or one frontal and an interactive image, and were asked to decide whether both images featured the same person. When they were able to use the interactive procedure to explore the face and control the angle from which it was shown, overall matching accuracy was higher for both typical and superior face recognisers. The procedure also improves performance when one of the images is low quality (i.e. pixelated).
Conclusion
Eyewitness identification plays a key role in the justice system, but eyewitnesses can make errors, often with devastating consequences. Our findings show that procedures that enable witnesses to reinstate pose and interact with the lineup member can enable witnesses to make more decisions in lineups. In our future work, we will be exploring how to optimize an interactive lineup procedure, comparing its efficacy against other types of lineup procedures, such as video and simultaneous photo lineups. Our programme of work will continue to exemplify how basic science can be brought to bear on important policy issues regarding how best to conduct a police lineup and reduce eyewitness errors.
References
Albright, T. D., & Rakoff, J. S. (2020). A Clearer View: The Impact of the National Academy of Sciences Report on Eyewitness Identification. Judicature, 104(1), 21-29.
Bailenson, J. N., Davies, A., Blascovich, J., Beall, A. C., McCall, C., & Guadagno, R. E. (2008). The effects of witness viewpoint distance, angle, and choice on eyewitness accuracy in police lineups conducted in immersive virtual environments. Presence (Cambridge, Mass.), 17(3), 242-255. https://doi.org/10.1162/pres.17.3.242
Benton, T. R., Ross, D. F., Bradshaw, E., Thomas, N.W., & Bradshaw, G. S. (2006). Eyewitness memory is still not common sense: Comparing jurors, judges and law enforcement to eyewitness experts. Applired Cognitive Psychology, 20(1), 115-129. https://doi.org/10.1002/acp.1171
Brewer, N., & Wells, G. L. (2011). Eyewitness identification. Current Directions in Psychological Science, 20(1), 24-27. https://doi.org/10.1177/0963721410389169
Bruce, V. (1982). Changing faces: Visual and non‐visual coding processes in face recognition. British Journal of Psychology, 73(1), 105-116. https://doi.org/10.1111/j.2044-8295.1982.tb01795.x
Buratto, L. G., Matthews, W. J., & Lamberts, K. (2009). When are moving images remembered better? Study-test congruence and the dynamic superiority effect. The Quarterly Journal of Experimental Psychology, 62(10), 1896-1903. https://doi.org/10.1080/17470210902883263
Butcher, N., Lander, K., Fang, H., & Costen, N. (2011). The effect of motion at encoding and retrieval for same-and other‐race face recognition. British Journal of Psychology, 102, 931-942. https://doi.org/10.1111/j.2044-8295.2011.02060.x
Carbon, C. C., & Leder, H. (2006). The Mona Lisa effect: Is ‘our’ Lisa fame or fake? Perception, 35, 411-414. https://doi.org/10.1068/p5452
Clark, S. E. (2012). Cost and benefits of eyewitness identification reform: Psychological science and public policy. Perspectives on Psychological Science, 7(3), 238-259. https://doi.org/10.1177/1745691612439584
Clark, S. E., Benjamin, A. S., Wixted, J. T., Mickes, L., & Gronlund, S. D. (2015). Eyewitness identification and the accuracy of the criminal justice system. Policy Insights from the Behavioral and Brain Sciences, 2(1), 175-186. https://doi.org/10.1177/2372732215602267
Colloff, M. F., Flowe, H. D., Smith, H. M. J., Seale-Carlisle, T. M., Meissner, C. A., Rockey, J. C., Pande, B., Kujur, P., Parveen, N., Chandel, P., Singh, M. M., Pradhan, S., & Parganiha, A. (under review). Active exploration of faces in police lineups increases discrimination accuracy. [Preprint]. Available at: https://psyarxiv.com/tvga4/
Colloff, M. F., Seale-Carlisle, T. M., Karoğlu, N., Rockey, J. C., Smith, H. M. J., Smith, L., Maltby, J., & Flowe, H. D. (under review). Criminal identification viewed from a different angle: Enabling witnesses to reinstate perpetrator pose during a lineup test increases accuracy. [Preprint]. Available at: https://psyarxiv.com/2rwgh/
Colloff, M. F., & Wixted, J. T. (2019). Why are lineups better than showups? A test of the filler siphoning and enhanced discriminability accounts. Journal of Experimental Psychology: Applied, 26(1), 124-143. https://doi.org/10.1037/xap0000218
Cutler, B. L., Berman, G. L., Penrod, S. D., & Fisher, R. P. (1994). Conceptual, practical, and empirical issues associated with eyewitness identification test media. In D. F. Ross, J. D. Read, & M. P. Toglia (Eds), Adult eyewitness testimony: Current trends and developments (pp. 163-181). New York, NY: Cambridge University Press. http://dx.doi.org/10.1017/CBO9780511759192.009
Cutler, B. L., & Fisher, R. P. (1990). Live lineups, videotaped lineups, and photoarrays. Forensic Reports, 3(4), 439-448.
Finklea, K. M., & Ebbesen, E. B. (2007). Eyewitness accuracy in the real world: DNA evidence as "ground truth" for eyewitness accuracy rates.
Fitzgerald, R. J., Price, H. L., & Valentine, T. (2018). Eyewitness identification: Live, photo, and video lineups.Psychology, Public Policy, and Law, 24(3), 307-325. http://dx.doi.org/10.1037/law0000164
Flowe, H. D., Carline, A., & Karoğlu, N. (2018). Testing the reflection assumption: A comparison of eyewitness ecology in the laboratory and criminal cases. International Journal of Evidence & Proof, 22, 239-261. https://doi.org/10.1177/1365712718782996
Grayson, C. E. (1998). When a 5 isn’t a 5: Sex, prototypes, stereotypes and standards of comparison in emotion judgments of self and others. Dissertation Abstracts International: Section B: The Sciences and Engineering, 59, 0912.
Gronlund, S. D., Wixted, J. T., Mickes, L., & Clarke, S.E. (2015). Conducting an eyewitness lineup: How the research got it wrong. Psychology of Learning and Motivation, 63, 1-43. https://doi.org/10.1016/bs.plm.2015.03.003
Harman, K. L., Humphrey, G. K., & Goodale, M. A. (1999). Active manual control of object views facilitates visual recognition. Current Biology, 9(22), 1315-1318. https://doi.org/10.1016/S0960-9822(00)80053-6
Innocence Project (2015). The Causes of Wrongful Conviction. Available at https://innocenceproject.org/causes-wrongful-conviction./
Liu, C. H., Ward, J., & Markall, H. (2007). The role of active exploration of 3D face stimuli on recognition memory of facial information. Journal of Experimental Psychology: Human Perception and Performance, 33(4), 895-904. https://doi.org/10.1037/0096-1523.33.4.895
Logie, R. H., Baddeley, A. D., & Woodhead, M. M. (1987). Face recognition, pose and ecological validity. Applied Cognitive Psychology, 1(1), 53-69. https://doi.org/10.1002/acp.2350010108
Meissner, C. A., & Brigham, J. C. (2001). A meta-analysis of the verbal overshadowing effect in face identification. Applied Cognitive Psychology, 15(6), 603-616. https://doi.org/10.1002/acp.728
Mickes, L., Flowe, H. D., & Wixted, J. T. (2012). Receiver operating characteristic analysis of eyewitness memory: Comparing the diagnostic accuracy of simultaneous versus sequential lineups. Journal of Experimental Psychology: Applied, 18(4), 361-376. https://doi.org/10.1037/a0030609
National Research Council. (2014). Identifying the culprit: Assessing eye-witness identification. Washington DC: The National Academies Press.
News, B. B. C. (2019). Millions spent compensating wrongful convictions. Retreived from
https://www.bbc.co.uk/news/uk-northern-ireland-47973826
O'Toole, A. J., Roark, D. A., & Hervé, A. (2002). Recognizing moving faces: A psychological and neural synthesis. Trends in Cognitive Sciences, 6(6), 261-266. https://doi.org/10.1016/S1364-6613(02)01908-3
Rubínová, E., Fitzgerald, R, J., Juncu, S., Ribbers, E., Hope, L., & Sauer, J. D. (in press). Live presentation for eyewitness identification is not superior to photo or video presentation. Journal of Applied Research in Memory and Cognition. https://doi.org/10.1016/j.jarmac.2020.08.009
Seale-Carlisle, T. M., & Mickes, L. (2016). US line-ups outperform UK line-ups. Royal Society Open Science, 3(9). 160300. https://doi.org/10.1098/rsos.160300
Seale-Carlisle, T. M., Wetmore, S. A., Flowe, H. D., & Mickes, L. (2019). Designing police lineups to maximize memory performance. Journal of Experimental Psychology: Applied, 25(3), 410-430. https://doi.org/10.1037/xap0000222
Smith, A. M., Yang, Y., & Wells, G. L. (2020). Distinguishing between investigator discriminability and eyewitness discriminability: A method for creating full receiver operating characteristic curves of lineup identification performance. Perspectives on Psychological Science, 15(3) 1-19. https://doi.org/10.1177/1745691620902426
Smith, H. M. J., Andrews, S., Baguley, T., Colloff, M. F., Davis, J. P., White, D., & Flowe, H. D. (under review). Performance of typical and superior face recognisers on a novel interactive face matching procedure. [Preprint]. Available at: https://psyarxiv.com/kbm2g/
Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80(5), 352-373. https://doi.org/10.1037/h0020071
Valentine, T., Darling, S., & Memon, A. (2007). Do strict rules and moving images increase the reliability of sequential identification procedures? Applied Cognitive Psychology, 21(7), 933-949. https://doi.org/10.1002/acp.1306
Winsor, A.,, Flowe, H. D., Seale-Carlisle, T. M., Kileen, I. M., Hett, D., Jores, T., Ingham, M. P., Lee, B., Stevens, L., & Colloff, M. F. (under review). Child witness expressions of certainty are informative. [Preprint]. Available at: https://psyarxiv.com/8gj46/
Wixted, J. T., Mickes, L., & Fisher, R. P. (2018). Rethinking the reliability of eyewitness memory. Perspectives on Psychological Science, 13(3), 324-335. https://doi.org/10.1177/1745691617734878.
Wixted, J. T., & Mickes, L. (2014). A signal-detection-based diagnostic-feature-detection model of eyewitness identification. Psychological Review, 121(2), 262-276. https://doi.org/10.1037/a0035940