Abstract
Three studies examined cognitive processing of visually presented dynamic scenes. Garsoffky, Schwan and Hesse (2002) showed that the recognition of dynamic scenes is viewpoint dependent, i.e. visual recognition becomes worse, if the viewpoint used for the test stimulus differs from the viewpoint from which a dynamic scene was initially learned. Based on this finding it can be asked whether this viewpoint deviation effect diminishes if special viewpoints are used, namely canonical viewpoints which are assumed to deliver more information than other viewpoints (Palmer, Rosch & Chase, 1981). As experimental material, short video clips of four moving balls were used. Based on the paradigm of Palmer et al. (1981), experiment 1 examined the existence of canonical viewpoints for dynamic scenes: Participants saw the scenes from various viewpoints and had to decide from which viewpoint in their opinion the scene was presented best. As a main result, participants preferred the viewpoint perpendicular to the axis of main movement of the scene over any other viewpoint. The more the viewpoints deviated from this angle, the worse they were rated. Experiment 2 and 3 used a recognition paradigm and examined if the usage of canonical viewpoints, i.e. the preferred viewpoints in experiment 1, weaken the viewpoint deviation effect in visual recognition due to their information richness, either in the learning phase (experiment 2) or the test phase (experiment 3). Experiment 2 found a general viewpoint deviation effect both for canonical and less canonical viewpoints in the learning phase, although it was weaker for the canonical views. In experiment 3 the use of canonical viewpoints in the test phase had no influence on the viewpoint deviation effect. This means that the influence of canonical viewpoints on the viewpoint deviation effect is more important during storage than during retrieval.