Abstract
Viewpoint changes caused by an observer's locomotion lead to better scene recognition performance than that following equivalent movement of the scene. In studies testing change detection of a tabletop scene, the benefit of observer locomotion has been attributed to spatial updating through body-based information (Simons & Wang, 1998), or knowledge of a change of the reference direction gained through locomotion (Mou et al, 2009). In the current study, six experiments were performed using a similar paradigm but measuring both accuracy and reaction time of the response. Experiment 1-4 examined the effect of an external visual indicator introduced during the testing phase signalling the learning direction (as in Mou et al, 2009). Experiments 1 and 2 compared performance in the locomotion condition with table rotation condition, Experiments 3 and 4 compared performance in a short walking condition with conditions where body-based information was not reliable (disorientation or walking a long, curved path). Experiments 5 and 6 examined the effect of intrinsic reference direction information by aligning the orientation of the dominant axis of all the objects in the scene. Experiments 1-4 show that even with the visual indicator, performance in conditions that lacked normal locomotion was still significantly worse than that in the observer locomotion condition and that the visual indicator did not improve performance at all. However, in Experiments 5 and 6, performance for a scene composed of objects with consistent orientations was: (1) better than that for a scene composed to objects with random orientation and (2) comparable to that in observer locomotion condition. Overall we show that the body-based information in observer locomotion provides the most prominent information, while knowledge of a reference direction is useful but might only be effective in limited scenarios, such as scenes with an obvious and dominant orientation.
Meeting abstract presented at VSS 2014