Abstract
Drawings of real-world scenes made from visual memory can be highly detailed and spatially accurate, and often contain little information not found in the observed stimuli (Bainbridge et al., 2018). However, little is known about the specific visual detail recalled when episodes are highly similar, and the nature of what object information is likely to intrude on the memory of a competing scene. The current study probes this question by asking participants (n=30) to study and recall 8 complex-real world scene images (4 from a repeated category, 4 from unique categories) using a drawing task. We recorded participants’ eye movements during the study period (10s each) to assess what percentage of detail studied in the scene was later successfully recalled, and to compare the order in which objects were studied versus recalled. After a distractor task, participants completed two memory tasks. For the first, they drew as many scenes as they could remember on a digital tablet that recorded timing and stroke order. Second, they completed an old / new recognition task for the original images intermixed with 8 matched foils. Participants recalled 83.8% of the images across the study, and recognized category-unique images significantly better than repeated category items. Online scorers of the drawings (n=964) were better able to match category-unique drawings to their original images than repeated category drawings, showing that recalled visual memory content was more accurate when there weren’t competing exemplars in memory. Indeed, when a separate group of online workers (n=3,010) were asked to label intrusions (or non-original objects) within the drawings, repeated category drawings were found to contain significantly more intrusions than category-unique drawings. Overall, these results reveal the nature and extent of intrusions that can occur in competing visual memories for complex real-world scenes.