Ulrich Ansorge, University of Vienna, Austria
We investigated the conditions under which repeated visual information attracts attention and eye fixations in dynamically changing images. We hypothesized that human attention is primed for previously fixated image content whenever there is the (implicit or explicit) task to relate two subsequent images to each other. Such situations are ubiquitous in many everyday situations, including viewing series of photographs, edited (or cut) videos, imaging devices, or graphical user interfaces. In the first part, we will present data from a series of image recognition experiments where human observers made explicit decisions about the familiarity of repeated vs. novel images. The collected eye tracking data showed that under recognition conditions, participants showed a pronounced bias for previously presented and fixated image content. Moreover, recognition crucially depended on these repeated fixations. In the second part, we discuss results from our experiments with edited videos. Here, we showed our participants two videos on the same screen that frequently contained cuts at the same time. Crucially, the videos randomly switched or kept their positions at the time of the cuts and participants had to track and keep their gaze on only one of the movies (and ignore the other). Using reaction times and accuracy measures we were able to show that participants were significantly better in this task when the visual similarity of the two images separated by the cut was high than when it was low. In the final part, we summarize our findings and discuss the role of repeated visual information for the guidance of attention and eye movements in dynamically changing images in relation to previous research with more strongly constrained stimuli. We also outline our efforts in developing a flexible and universal model of visual attention and gaze control in videos based on our experimental results.