Amy Poole¹ (), Eline R Kupers², Adrian Wong¹, Clayton Curtis³, Kendrick Kay¹; ¹University of Minnesota, ²York University, ³New York University

Several studies have shown that gaze patterns during natural scene viewing are different across individuals (Castelhano & Henderson, 2008), directed towards salient objects (Anderson et al., 2015), and task-dependent (Yarbus, 1967). To better characterize these effects, we conducted an eye tracking experiment in which we independently manipulated the natural scene viewed as well as the task performed. Participants (n=11) engaged in 6 different tasks on 30 scenes that were balanced with respect to content and spatial layout (a total of 180 trials). Each scene was shown for 2 seconds. The tasks were to (1) detect a contrast change of the entire scene, (2) categorize the scene as indoor or outdoor, (3) classify the dominant object, (4) determine the location of the dominant object, (5) decide how one would act in the scene, and (6) hold the scene in working memory (and after an 8-s delay, decide whether content was added or removed). Eye tracking data were recorded with an Eyelink 1000, and a density map indicating fixated locations was created for each trial. We find that, after collapsing scenes and tasks, individuals’ density maps differ significantly from one another. Moreover, when collapsing participants and tasks, density maps also differ significantly across scenes. Finally, collapsing participants and scenes, we see that most tasks have distinct density maps. In particular, the working memory task provoked the most widespread gaze, while the contrast change task provoked the least widespread gaze. We conclude that all three factors—individual, scene, and task—affect where people look. Future analyses will evaluate how well existing models of eye movements can account for these data.

Acknowledgements: NIH R01 EY034118 to KK and CC