Applying Signal Detection Models to Investigate the Effect of Sequential Item Presentation on the Police Lineup Task

Abstract

Much research investigating the police lineup has argued that presenting items sequentially is superior to presenting them simultaneously, because sequential presentation reduces rates of innocent suspect identification, minimising the chance of false conviction. However, the research program that arrived at this conclusion was resolutely applied in focus, directing less attention to developing theories that might explain how sequential item presentation achieves this outcome. Recent research has addressed this issue by applying signal detection theory to understanding the lineup task. This mathematical modelling framework characterises observed performance on a recognition memory test, such as the police lineup, as resulting from two latent variables; discriminability, the ability to distinguish target items (guilty suspects) from non-target items, and response bias, conceptualised as willingness to choose. Research employing this framework suggests that sequential item presentation achieve its reduction in innocent suspect identifications by encouraging witnesses to choose less readily than simultaneous presentation, rather than by increasing discriminability. Some studies also find that discriminability is greater for simultaneous presentation and that, on this basis, it should be preferred. However, this body of recent research has employed analysis techniques that fail to capture the unique constraints of sequentially presented lineup tasks. This may compromise the measurement of discriminability (and response bias), leading to incorrect conclusions when comparing sequentially presented lineup tasks to the simultaneous lineup. This thesis addresses this limitation by developing signal detection models that capture the structural constraints of sequentially presented lineup tasks. These models are used in studies one and two to compare simultaneous lineup presentation to two sequentially presented lineup tasks, one on which identification of the current item terminates the task (stopping rule) and another on which two full laps of the items is completed before an identification decision is made, as used in the United Kingdom (UK). Study three develops a model for examining changes in discriminability and response bias by serial position in the sequential stopping rule lineup. Each study involved the collection of new experimental data and studies one and three also analysed previously published datasets. The results of studies one and two imply that sequential item presentation may have a small negative effect on discriminability compared to simultaneous presentation, but this effect was not consistently observed. Effects on response bias were larger and more reliable; the sequential stopping rule lineup was associated with the most conservative overall choosing, followed by the simultaneous lineup, then the UK sequential lineup. In study three, discriminability increased from serial position one to position two in the sequential stopping rule lineup, but not beyond. Changes in response bias by serial position differed depending on whether an identification was made before or after the presentation of the guilty suspect. Taken together, these results imply that there is no compelling reason for policymakers to prefer sequentially presented lineups to the simultaneous lineup. The insights generated from this thesis demonstrate the value of a formal modelling and approach and the need to consider carefully the match between model, task, and research question.