Abstract
Human observers can recognize actions in point-light biological sequences with relative ease. This skill is believed to be the consequence of motion-based visual analyses, although the exact nature of these computations remains unclear. Typically, point-light animation sequences are depicted as luminance-defined tokens that are easily detected as biological by our first-order motion system. Contrast-defined (second-order) point-light sequences are also readily recognized as biological (Ahlström, Blake, Ahlström, 1997). More recently, some findings have implicated attention-based (third-order) motion as the critical motion analysis in biological motion perception (Thornton, et al., 2000; Garcia & Grossman, 2008). To determine whether third-order motion analyses are sufficient for biological motion perception, we have constructed biological motion displays that are defined by alternating features (e.g. Blaser et al., 2000), and are thus encoded by attention-based motion systems. Target tokens within a larger array of gabors depict human actions by coherently varying on key dimensions (contrast, spatial frequency, gabor orientation, phase or drifting speed). We measured the magnitude of the feature differences (e.g. contrast increments) for threshold discrimination and detection performance in second-order displays using an adaptive staircase. The alternating feature displays were created by varying each gabor dimension, frame-by-frame, of target tokens relative to the background. In these alternating feature displays, the global motion signal is constructed by tracking these salient differences across feature space. We found that for second-order motion, subjects require more feature differences (e.g. higher contrast, larger orientation tilt) for biological motion discrimination compared to detection, as expected. However, we also found that observers can readily detect and discriminate the third-order alternating feature displays. These findings are evidence that attention-based third-order motion analyses may promote biological motion perception through feature tracking.