Abstract
Humans are able to track several independently moving objects among identically looking distractor objects. However, the question of whether this ability relies on allocentric or retinocentric coordinates is part of a controversial debate. Liu et al. (2005; J. Exp. Psychol. Human, 31:235–247) showed that multiple object tracking (MOT) in a 3D-scene is robust against smooth movements of the whole scene and concluded that MOT relies on allocentric coordinates. On the other hand, Seiffert (2005; J. Vision, 5:643) showed that retinocentric coordinates are important for MOT in 2D-displays. In her experiments, MOT was impaired if a 2D-display was shifted smoothly across the retina either as a result of eye movements around a stationary display or as a result of display movements while eyes were fixated. MOT was also impaired if a 2D-display was translated abruptly without overlap between the two views. We studied MOT in dynamic 3D-scenes and introduced abrupt and unpredictable viewpoint changes, but gradually varied the change by rotating the entire scene either by 10°, 20°, or 30° during a filmic cut. In the reported experiment, observers tracked targets that moved independently among identically looking distractors on a rectangular floor plane. The tracking interval was 11s. Abrupt rotational viewpoint changes of 10°, 20°, or 30° occurred after 8 s. Accuracy of tracking targets across a 10° viewpoint change was comparable to accuracy in a continuous control condition, whereas viewpoint changes of 20° and 30° impaired tracking performance considerably. This result suggests that tracking is mainly dependant on a low-level process whose performance is saved against small disturbances by the visual system's to compensate for small changes of retinocentric coordinates. Tracking across large viewpoint changes succeeds only if allocentric coordinates are remembered to relocate targets after displacements.