Could precise and replicable manipulations of suspect-filler similarity optimize eyewitness identification performance?

Abstract

The optimal level of suspect-filler similarity in police lineups remains undefined. Difficulties inherent in pinpointing precise and replicable variations in face similarity create challenges for examining the effects of suspect-filler similarity on identification outcomes and providing decisive lineup construction recommendations. We tested the relationship between suspect-filler similarity and identification outcomes using stimuli developed with a combination of face matching and morphing software that could potentially be used by police investigators to standardize lineup composition. A group of fillers selected to be as low in similarity to the suspect as possible, while still matching a basic perpetrator description, were morphed to the suspects. Participants viewed lineups in which the fillers were either unmorphed, a 33% morph, or a 50% morph to the suspect. Results showed a tendency for identification performance -including accuracy, discriminability between guilty and innocent suspects and the confidence-accuracy relationship-to suffer as similarity levels increased beyond matching a basic perpetrator description. Thus, despite the attraction of exploiting modern technology to standardize suspect-filler similarity relations, our findings across multiple sets of encoding and test materials broadly cohere with Wells and colleague's argument (Luus & Wells, 1991;Wells et al., 1993) that identification performance may be optimized when fillers are selected to be as low as possible in similarity to the suspect while still matching description. However, before unequivocally endorsing such a lineup construction approach, it is necessary to conduct a systematic examination of precisely what constitutes an adequate perpetrator description and how this might be consistently obtained.

A stronger memory for the perpetrator may attenuate effects of the identification procedure on eyewitness accuracy

“…If all lineup members resemble each other to a high degree, an eyewitness would not be able to differentiate the faces from each other or from the memory of the perpetrator's face. Empirical evidence suggests an optimal level of filler‐similarity, such that there may be an upper bound to filler‐similarity in which the fillers become too similar to each other and hinder the eyewitness's ability to discriminate between guilty and innocent suspects (Carlson et al, 2019; Colloff et al, 2021; Lucas & Brewer, 2021), and our data support this idea. However, we again reiterate that our findings are based on match‐to‐target rather than match‐to‐suspect due to our lack of a priori innocent suspects, and therefore more research is still needed to test this idea in an ecologically‐valid manner (i.e., matching TP fillers to a target while matching TA fillers to a designated innocent suspect).…”

“…The third main effect, that faces of low match to target yield higher empirical discriminability than faces of high match to target (given that they all match the description well), is a relatively recent finding in the literature. To our knowledge, only two papers have supported this pattern (Colloff et al, 2021; Lucas & Brewer, 2021), which gives more weight to our additional support for this important finding. Low‐similarity fillers may allow eyewitnesses to notice differences between the faces in the lineup, which they can then compare to their memory for the perpetrator's face.…”

“…The majority of the aforementioned lineup size studies held filler similarity constant whereas we will test description‐matched fillers of both low and high similarity with our targets. Several recent papers have shown that filler similarity affects eyewitness ID performance (Carlson et al, 2019; Colloff et al, 2021; Fitzgerald et al, 2013, 2015; Lucas & Brewer, 2021). Wells et al (1993) proposed that lineups should contain a degree of propitious heterogeneity, such that fillers and suspect should not be too similar to each other in physical appearance.…”

“…Carlson et al (2019) recently supported this concept by testing the upper bounds of filler similarity. Moreover, when fillers match the description of the target, low similarity fillers tend to increase eyewitness discriminability compared to high similarity fillers (Colloff et al, 2021; Lucas & Brewer, 2021).…”

“…We preregistered five hypotheses on the Open Science Framework (https://osf.io/yz8p7/?view_only=db1781ee6ced49b89e3a6cf687613b30). The first three are main effects we intend to replicate: greater empirical discriminability for: (a) longer compared to shorter encoding time for a target's face (e.g., Mickes, 2015; Palmer et al, 2013), (b) lineups over showups (regardless of lineup size; Akan et al, 2021; Wooten et al, 2020), and (c) lineups with low‐similarity over high‐similarity fillers (Colloff et al, 2021; Lucas & Brewer, 2021). The first two predictions are robustly supported in the literature, but the prediction involving filler similarity is much more recent and therefore in need of further investigation.…”