Abstract
Humans can quickly and efficiently organize point-light stimuli into a compelling global percept of biological motion. Point-light animations have been used extensively to study visual perception of animated motions, in part because of the ease of control over kinematic and form content. In contrast, fully illuminated depictions have received considerably less attention until recent years. Nonetheless, both point-light and fully illuminated depictions clearly contain useful kinematic and structural information, with fully illuminated versions being more ecologically similar to real-world biological motion. Here we use the “Bubbles” technique (Gosselin & Schyns, 2001) to investigate the spatio-temporal dynamics of diagnostic information for discriminating human walkers. In this experiment, we present randomly selected intervals of walkers masked with white noise, but visible through a number of Gaussian apertures. Using the limited stimulus information presented on each trial, observers were instructed to discriminate the walking direction of the target walker. We used reverse correlation to compute dynamic classification images using statistical techniques from random field theory (Chauvin et al., 2004) to illustrate the spatio-temporal pattern of critical information for perceiving point-light and fully illuminated biological walkers. Overall performance varied sinusoidally across the gait cycle. This is similar to previous reports (Thurman & Grossman, 2008) and evidence for diagnostic information at regular intervals in the gait cycle. This diagnostic interval was slightly delayed in the point-light versions as compared to the fully illuminated. Space-time classification images of fully illuminated walkers show that observers tend to rely extensively on the upper body, while in contrast, for point-light walkers the upper body carries little diagnostic information. Instead the most significant information for point-light depictions is found spread around the extremity dots (i.e. hands and feet). These results highlight the different strategies for perceiving biological motion from point-light and fully illuminated representations of human actions.