Abstract
Theories of biological motion perception differ in their assumptions about the form of the internal representation. We used ideal observer analysis to quantify the information provided by different display types. Our working assumption is that displays that are more similar to the internal representation will be processed more efficiently. It follows that human efficiency with different types of displays may provide valuable clues as to the nature of the internal representation.
A walker was presented in four display types. A point-light displays points representing joints of the walker. A skeleton connects appropriately the joints of a point-light display. A silhouette displays the side view of the walker. A contour displays the outline of the silhouette. Observers either detected the presence of a walker, or discriminated walking direction (left or right). A walker stimulus was embedded in dynamic white noise. Contrast threshold was measured for each display type in separate blocks. The position of the walker was randomly jittered from trial to trial.
For both detection and discrimination, the lowest contrast thresholds for humans and the ideal observer were obtained with the silhouette display, which was expected since the total area of the walker was the largest in a silhouette display. However, efficiency in the detection task was higher with the skeleton (0.86%) and point-light (0.82%) displays than with contour (0.46%) and silhouette (0.51%) displays. In the discrimination task, efficiency was higher with skeleton (0.94%) and silhouette (1.01%) displays than with contour (0.49%) and point-light (0.45%) displays. Our findings suggest that the effective internal representation of biological motion may be closer to that of a skeleton, utilizing both the points of articulation and the structural links between them.