The purpose of the proposed research is to link a longstanding framework for understanding how recognition memory decisions are made to the forensically relevant question of how witnesses make a recognition memory decision when faced with a lineup. In the field of experimental psychology, our understanding of how recognition memory decisions are made has been effectively guided by signal-detection theory since Egan's (1958) seminal report was published more than 50 years ago. By contrast, in the applied literature, signal-detection-based efforts to understand decision-making on recognition memory tasks are virtually nonexistent. The wide chasm separating experimental and applied investigations of recognition memory is surprising because issues that may be informed by signal-detection theory (e.g., the relationship between confidence and accuracy) are of considerable interest in both fields. In the applied literature, it has been repeatedly noted that signal-detection theory could inform eyewitness memory but, to date, no signal-detection model of lineup-based recognition memory has been seriously pursued.
The goal of the proposed research is to produce a simple signal-detection framework that will be useful for helping to understand a variety of empirical phenomena that have been investigated in the applied literature, including the relationship between confidence and accuracy. Our strategy will be to test the viability of a signal-detection-based model of eyewitness memory using tasks that are in some ways similar to standard list-memory tasks (e.g., each subject studies a list of items) but that have been modified to be more forensically relevant (e.g., the stimuli will consist of faces, and the recognition tests will involve lineups). After testing and developing a simple signal-detection model using those relatively convenient laboratory procedures, we plan to then test the model using more forensically relevant methods that are commonly used in the applied literature (e.g., where each subject is drawn from a diverse population, views one simulated crime, and provides one recognition decision). The results should help scientists to better inform policymakers about how to improve police lineup procedures.
The purpose of this grant was to develop a signal-detection-based model of eyewitness identification and to investigate the utility of a technique known as receiver operating characteristic (ROC) analysis. This method is widely used in the field of medicine for testing the diagnostic accuracy of different tests, but it had not been used in the field of eyewitness identification. As part of this research project, we investigated the diagnostic accuracy of simultaneous vs. sequential lineups. Recent experiments from our lab using ROC analysis have consistently favored the simultaneous lineup procedure (Mickes et al., 2012), a suprising result that was almost immediately independently replicated by several other labs. We developed a signal-detection based theory of why simultaneous lineups are diagnostically superior to sequential lineups. In terms of societally relevant outcomes, the broader impacts of our NSF-supported research consist of its potential to influence the assessment of eyewitness memory in real-world criminal investigations and in subsequent judicial proceedings. The work we have produced thus far offered the first detailed description of how to assess different lineup procedures using ROC analysis. A National Academy of Sciences report recently confirmed that ROC analysis is a significant improvement in the empirical assessment of eyewitness identification procedures. The societal impact of this work is far reaching. Up to 30% of law enforcement agencies that use photo arrays have already switched (perhaps prematurely) to using the diagnostically sequential procedure (Police Executive Research Forum, 2013). The switch was made because prior work used non-ROC analyses that mistanly suggested that sequential lineups were superior to simultaneous lineups.
