Author contributions were as follows: LS led study conceptualisation. Study methodology was designed by LS, JD, EJD, NT and PM. Extant face memory data collection was led by JD. Data collection for this study was conducted by LS, EJD and NT. Software for personality and self-selected photograph collection was designed by NT and PM. Data analysis was conducted by LS. First draft was written by LS. Substantial editing was contributed from JD, NT and EJD.

A considerable proportion of crimes involve multiple perpetrators. Yet, little is known about how police officers construct, administer, and record eyewitness identification procedures for multiple suspects. An online survey of law enforcement agents in Sweden, Belgium, and the Netherlands (N = 51) was conducted to obtain an initial understanding of police perceptions of prevalence and characteristics of multiple perpetrator crimes, and to examine identification procedure practices given the little to no guidance provided for police. Practice converged when it came to the use of sequential, photographic lineups, but diverged between and within countries on whether or not suspects of multiple perpetrator crimes should be placed in separate lineups. Results specifically highlight contextual cuing as one critical area for future research in identification for multiple perpetrator crimes (i.e. placing multiple suspects in the same lineup or asking eyewitnesses to look for a specific suspect).

Research in perception and recognition demonstrates that a current decision (i) can be influenced by previous ones (i–j), meaning that subsequent responses are not always independent. Experiments 1 and 2 tested whether initial showup identification decisions impact choosing behavior for subsequent showup identification responses. Participants watched a mock crime film involving three perpetrators and later made three showup identification decisions, one showup for each perpetrator. Across both experiments, evidence for sequential dependencies for choosing behavior was not consistently predictable. In Experiment 1, responses on the third, target-present showup assimilated towards previous choosing. In Experiment 2, responses on the second showup contrasted previous choosing regardless of target-presence. Experiment 3 examined whether differences in number of test trials in the eyewitness (vs. basic recognition) paradigm could account for the absence of hypothesized ability to predict patterns of sequential dependencies in Experiments 1 and 2. Sequential dependencies were detected in recognition decisions over many trials, including recognition for faces: the probability of a yes response on the current trial increased if the previous response was also yes (vs. no). However, choosing behavior on previous trials did not predict individual recognition decisions on the current trial. Thus, while sequential dependencies did arise to some extent, results suggest that the integrity of identification and recognition decisions are not likely to be impacted by making multiple decisions in a row.