Abstract
The allocation of gaze is commonly thought to be determined by features of visual stimuli. A number of studies have analyzed the statistical properties of features such as luminance, contrast, and spatial frequency content at locations selected by gaze versus randomly chosen locations. Significant differences were found between these two classes of regions but it remains controversial whether such effects are correlational or causal. However compelling evidence exists for the task dependence of visual attention. Here we report results from an experiment that demonstrates that in a navigational task in which subjects avoid obstacles and pickup objects the feature statistics at fixation location are dependent on the ongoing task and can be explained by the different control laws used in navigating towards or around objects.
Ten human subjects walked along a path in a virtual environment and were instructed to stay on the path and either avoid or pickup rectangular objects. As previously reported, luminance contrast was elevated at fixation locations. However, mean contrast was significantly different when subjects looked at objects when approaching them to pick them up (−5%) and when they looked at the same objects in the context of walking around them (+3%). The explanation for this is that the spatial distribution of fixations with respect to the objects depended on the type of interaction. When approaching objects for pickup, subjects tended to look at their center while when avoiding obstacles they tended to look at the object's edges.
This suggests that the local luminance contrast differences are not causal but correlational. Depending on the ongoing task, subjects selected different gaze targets such as edges versus centers so as to obtain relevant information for the control of movement. Thus different control laws in navigating around or towards objects determine which information is selected by the visual system.