SYM 17: Exploring the Boundary Conditions of the Confidence-accuracy Relationship in Eyewitness Identification Decisions
| Friday, June 13, 2025 |
| 3:30 PM - 4:50 PM |
| Fountain Suite |
Overview
Symposium organiser: Mitchell Eisen
Details
New research will be presented exploring the boundary conditions of the confidence-accuracy relationship in eyewitness decision-making. First, Ryan Fitzgerald will describe how suspect similarity in photo-arrays can influence confidence in suspect-identifications. Next, Ruth Horry will present new work examining how the accuracy of highly confident suspect-identifications are affected by conditions that impair discriminability. Then, Mitch Eisen will describe new field-simulation research showing how confidence and response-time postdict accuracy quite differently when using lineups versus showups for both choosers and rejectors. Finally, Amy Bradfield-Douglass will then present new co-witness research comparing Japanese to American participants, and Steve Lindsay discuss these findings.
Speaker
Dr Ryan Fitzgerald
Associate Professor
Simon Fraser University
Reduced accuracy for high confidence identifications in cases with appearance-based reason of suspicion
Symposium Presentation
Eyewitness confidence can reflect accuracy, but it depends on the reason of suspicion in the case. If a suspect is put in a lineup because of their appearance (e.g., looks similar to CCTV evidence), it increases the risk that an innocent suspect would resemble the actual culprit. Other reasons of suspicion would not have this effect (e.g., history of committing similar crimes). We conducted a lab experiment and manipulated how much an innocent suspect resembled the culprit. Accuracy of high confidence suspect identifications reduced from 98% for low similarity innocent suspects to 82% for high similarity innocent suspects.
Paper Number
449
Dr Ruth Horry
Senior Lecturer
Swansea University
Do highly confident suspect identifications imply high accuracy under conditions that impair discriminability?
Symposium Presentation
The extent to which the accuracy of highly confident suspect identifications (HCIDs) is affected by conditions that impair discriminability is contentious. This is a difficult question to test, as conditions that impair discriminability yield few HCIDs, necessitating very large samples. As a consortium, we collected 12,532 lineup decisions from 3,133 participants, who were pseudo-randomly allocated to an immediate (n = 350) or 1-week delay condition (n = 2,783). This approach yielded 192 immediate HCIDs and 286 delayed HCIDs. Despite a large difference in discriminability, the accuracy of HCIDs was similarly high in both conditions (Immediate = 95.8%; Delayed = 97.2%).
Paper Number
485
Dr Mitchell Eisen
Professor
California State University, Los Angeles
The Suggestive Pressure of Showups Conducted in the Field Yields Overconfident Choosers But Highly Accurate Rejectors
Symposium Presentation
Participants (N = 979) witnessed a staged-crime and were immersed in what they were led to believe was an actual police investigation using the field-simulation paradigm in which they attempted to make identifications from photographic-lineups or live-showups. Confidence-accuracy characteristic analyses (CAC) for confidence and response-time data yielded findings that diverge form lab research. Consistent with previous lab research, for lineups, confidence and response-time post-dicted accuracy for choosers, but not rejectors. However, for showups, the opposite pattern was found, as confidence and response-time post-dicted accuracy for rejectors, but not choosers. As expected, choosers at showups were overconfident regardless of accuracy.
Paper Number
530
Dr Amy Douglass
Whitehouse Professor Of Psychology
Bates College
Post-identification feedback and co-witness confidence
Symposium Presentation
The current research tests whether confirming post-identification feedback affects witness behavior in two samples of witnesses: generally collectivist participants in Japan versus generally individualist participants in the United States. In both sets of participants, we collected retrospective confidence coded conversations for non-verbal behavior and verbal content. The primary dependent variable is co-witness conformity which is measured after pairs of witnesses watch a stimulus crime together, complete independent memory reports, and then engage in a conversation about their memory reports. Predicted results include increased conformity among witnesses whose partners received confirming feedback (vs. no feedback) and among Japanese (vs. American) participant-witnesses.
Paper Number
566
Prof Steve Lindsay
Professor
University Of Victoria
Discussant: Reflections on Confidence and Accuracy in Eyewitness Identification
Symposium Presentation
Psychological science exploring eyewitness identification has greatly influenced police practice, a shining example of applied cognitive psychology. But there remain many points of debate regarding underlying mechanisms and their implications for practice. I will comment on Ryan Fitzgerald’s work on selecting suspects based on their resemblance to CCTV images of the culprit, Ruth Horry’s evidence that high confidence IDs are likely to be accurate regardless of delay; Mitch Eisen’s field-simulation studies indicating that confidence and RT posdict accuracy differently for show-ups versus lineups; and Amy Bradfield Douglas’s studies with Yui Fukushima comparing co-witness conformity in Japan versus the US.
Paper Number
461
