Detection of a Point-Light Walker within a Mask: Effect of Display Orientation

How, if at all, does display orientation affect the extraction of a known structure in biological motion? A computer-generated 11-point-light walker was simultaneously masked by 66 dots that corresponded to scrambled triads of points on his limbs. At each of 5 randomly presented orientations between upright and inverted (0°, 45°, 90°, 135°, 180°), observers saw a sequence of 210 displays, half of which comprised a walker within a mask, while the rest presented only a 77-dot mask. In a confidence-rating procedure, they had to judge whether a walker was present. A sample of the target was demonstrated before each experimental sequence. ROC analyses showed that detectability decreased with changing orientation: performance began to deteriorate already at 45°. For orientations of 90°, 135°, and 180°, the ROC curves were situated close to one another. Nevertheless, given the complexity of the mask and the brief viewing period (1 s), with inversion (180°) detectability was still surprisingly high (cf Bertenthal and Pinto, 1994 Psychological Science 5 221 – 225). Comparison with data on spontaneous recognition of biological motion (Pavlova, 1996 Perception 25 Supplement, 6) suggests that display orientation affects bottom - up visual event processing more strongly than top - down. We consider the relative power and interrelation of processing constraints (such as axis of symmetry, dynamical constraints) in the perception of invariant structure from biological motion.