Abstract
Since the seminal work by Johansson, it has been widely accepted that mostly motion information is relevant for the perception of point light walkers. Beintema & Lappe (2002) have proposed a novel point light stimulus (sequential position walker, SPW) that strongly degrades local motion information by randomly displacing the dots along the limbs. Despite this degradation, their subjects were still able to judge the direction of walking, suggesting that the perception of these stimuli was based purely on form information, presumably by fitting an internal kinematic model. To test this hypothesis we designed another point light stimulus that is composed of regions containing either purely randomly moving dots, or dots with deterministic opponent motion along the horizontal axes, but random vertical motion. Despite the fact that this stimulus is highly incompatible with the kinematics of a walker subjects perceived it as a walking person. By slight variation of the stimulus outline the percept could be reliably shifted between rightward and leftward walking. Psychophysical experiments show no difference in recognition between the two stimuli (N=5, p>0.7). This result points against a relevance of kinematic models in the recognition of point light stimuli. Additional evidence against the necessity of form information was obtained in a quantitative modeling study using a neurophysiologically plausible model with a motion and a form pathway. The motion pathway of the model that analyzes local motion information achieves reliable recognition of SPW stimuli, whereas the form pathway contributes only weakly. We conclude that, though form information undoubtely can be exploited for the recognition of biological stimuli, it might not be sufficient for the recognition of point light walkers. Instead, our experimental results suggest that their recognition might implicate the use of opponent motion features that can be extracted by relatively simple neural detectors.