Eyewitness Identifications and Assessments of Them

Abstract

Eyewitness memory plays a critical role in criminal justice, yet misidentifications remain a leading cause of wrongful convictions. Decades of research show that eyewitnesses often provide observable indicators of identification accuracy, such as confidence. The effective evaluation of these cues is therefore essential for safeguarding justice. This thesis is guided by two core aims: first, to establish whether these behavioural indicators serve as valid signals of eyewitness identification accuracy; and second, to examine whether and how external evaluators, such as jurors and legal professionals, can assess accuracy on the basis of these cues.
Chapter 2 presents a mega-analysis of more than 58,000 identifications, based on the re-analysis of raw data from published papers. This synthesis demonstrates that high-confidence and rapid identifications consistently exceed 90% accuracy across varied contexts, providing robust evidence for their validity as indicators of accuracy. Chapter 3 reports an identification experiment with over 500 participants, showing that eyewitnesses can distinguish guilty from innocent suspects, and providing the first evidence that retrieval effort cues, alongside confidence and response time, predict identification accuracy. This chapter also generated real eyewitness identification videos for subsequent research. Chapter 4 builds directly on this work by testing whether external evaluators can perceive and use these indicators. Although evaluators relied on perceived eyewitness confidence and response speed, they misjudged retrieval effort and were unable to distinguish correct from incorrect identifications, performing at chance level.
Together, these studies progressed from re-analysing large-scale published data, to conducting a controlled identification experiment, and finally to examining evaluators’ judgments using authentic video materials. The findings reveal a paradox: eyewitness identifications contain valid indicators of accuracy, yet these indicators are not adequately perceived or effectively used by evaluators. Overall, this thesis advances scientific understanding of eyewitness accuracy indicators and points to future research on how such cues can be better recognised and applied.

Date of Award	9 Dec 2025
Original language	English
Awarding Institution	University of Bristol
Sponsors	China Scholarship Council
Supervisor	Laura B Mickes (Supervisor) & Chris Kent (Supervisor)