Abstract
Where does one attend when viewing dynamic scenes? Research into the factors influencing gaze allocation during free-viewing of dynamic scenes have reported that motion is highly predictive of gaze location (Mital, Smith, Hill, & Henderson, 2010; Cog.Comp.). However, it is currently unknown whether this is due to exogenous control of attention or due to a correlation of motion with higher-order, endogenously prioritized features such as animate objects. In dynamic real-world scenes filmed from a static viewpoint, most motion is caused by animate objects such as people and animals or the objects they carry or are carried by, e.g. vehicles. This social motion may be prioritized by attention due to its intrinsic relevance to human observers.
Across two experiments we show that a) the gaze of multiple viewers is more coordinated when free-viewing dynamic compared to static versions of the same scene, b) the predictive power of gaze by motion is due to a correlation with people, c) prioritization of moving people can be overridden by a viewing task that requires attention to be allocated to the static background (e.g. Identify The Location), and d) following task completion, gaze returns to a default prioritization of people. These results emphasize the strong endogenous control of gaze during dynamic scene viewing and our tendency to adopt a default viewing mode which prioritizes social elements of a dynamic scene.