Order Constrained Analyses of Eyewitness Memory Accuracy

In psychological research, there are several commonly reported patterns in the relationships between decision accuracy, confidence, and response times. Firstly, there is a negative correlation between the similarity of the target and lures and response accuracy, as noted in studies like Huang et al. (2021). Secondly, studies report a positive relationship between confidence and accuracy, as mentioned in research by Brewer & Wells (2006). Thirdly, response time and accuracy are inversely related, as reported in De Boeck & Jeon’s (2019) work. We submit these regularities to a novel test using data from Horry and Brewer (2016)’s Experiments 1 and 2. In these experiments, participants first studied a target face, then tried to identify it from two items (Experiment 1) or four items (Experiment 2), and rated their confidence. We employed order-constrained inference using the QTEST software (Regenwetter et al., 2014; Zwilling et al., 2019). This approach translates verbal hypotheses into testable statistical ones without introducing unnecessary assumptions. At the individual participant level, we quantified evidence in favor or against each of the three hypotheses using Bayes factors (against an unconstrained baseline). We also assessed each of the three hypotheses jointly across participants using group Bayes factors. The Bayes factors obtained from the analyses indicated “decisive” evidence in support of the three hypotheses in Experiment 1 and “strong” to “very strong” evidence against the three hypotheses in Experiment 2, with notable individual variations. The lack of support in Experiment 2 might not be a flaw of the hypotheses themselves but could stem from the stimulus design or display settings not effectively reflecting the intended manipulation. This highlights the need for careful testing of stimulus design as well as display settings in visual attention research.