Courtney G. Nutt¹ (), Tanner L. Lumpkin², Patsy E. Folds³, Erin A. Conway⁴, Ralph G. Hale⁵; ¹University of North Georgia

Human visual memory capacity and limitations of these systems have been highly investigated, often highlighting the impact these factors have on our cognitive abilities and biases. Many visual memory errors relate to a reliance on gist information instead of precise details. These errors can have real-world implications (e.g., eyewitness testimony). Previous research has suggested familiarity and aesthetic value of visual information are likely to affect semantic engagement, strength of encoding, and veridical memory accuracy. In the present study, participants were immersed in a VR environment and asked to study 36 AI-generated paintings. All stimuli consisted of paintings displayed on the wall of an art gallery, with half mimicking famous paintings and half representing novel paintings. After each image, participants rated the item’s familiarity and aesthetic value on an 8-point scale. A VR distractor task followed the study block to prevent items from being held in working memory. The test block consisted of half identical and half changed trials. Painting change options were color changes, inversions, and new images of the same style. Change/no-change data was collected at test. As expected, participants found the paintings that mimicked familiar paintings to be more familiar than the novel paintings. Interestingly, the novel paintings were found to be more aesthetically pleasing overall. However, neither familiarity nor aesthetic value significantly impacted the overall accuracy of detecting changes. Also, no significant differences in accuracy were found between change types. While inconsistent with previous research stating the importance of top-down factors like familiarity, these findings are consistent with models suggesting differences in visual memory ability for, and errors related to, gist versus verbatim information. By testing these factors in a more ecologically valid VR environment, this study improves our understanding of visual LTM and factors that contribute to accurate encoding, particularly related to familiarity and aesthetics.