Memory detection in virtual reality: The effect of scene modifications

Can your eyes give away your deepest secrets? In two preregistered studies, we examined whether eye movements toward modified areas of a crime scene enable to detect crime knowledge. After committing either a mock crime or an unrelated task, guilty (Study 1 n = 12; Study 2 n = 63) and innocent (Study 1 n = 21; Study 2 n = 74) participants viewed the crime scene in virtual reality to which several salient modifications were made. Guilty participants looked more (Study 1 d = 0.88; Study 2 d = 0.65) and earlier (Study 1 d = 0.66; Study 2 d = 0.52) at these modifications as compared to innocent participants. It was possible to significantly distinguish guilty participants from innocent ones based on their eye movements toward the modified areas (Study 1 AUC = 0.75; Study 2 AUC = 0.68). Furthermore, in Study 2, we added a third group of participants who were guilty but revisited the crime scene in virtual reality without modifications, to test the contribution of these modifications. We found that the guilty-modifications group looked significantly more at the modified areas than the guilty-no modifications group (d = 0.35), although from a Bayesian perspective this evidence was only sufficiently strong (BF10 = 2.06). Despite the modest effect sizes, we argue that this approach has several important advantages in comparison to traditional memory detection tests