Abstract
Many influences on figure/ground assignment have been identified, but the influence of shape is poorly understood. Here we investigate the influence of axiality - the degree to which a shape has a clearly articulated skeletal structure - by pitting axiality against a known f/g cue, symmetry. We designed displays in which a multi-part axial shape shared a common boundary with a symmetric shape (created by reflecting one side of the shape about an imaginary vertical midline). We used two methods to assess perceived f/g along the common boundary, one local, the other global. Exp. 1 used the motion probe method from Kim & Feldman (2009), in which a small motion signal is placed along the shared boundary, and the subject indicates which region moved, allowing f/g to be measured locally and independently at each point. Exp. 2 used a conventional explicit response in which subjects reported global f/g. We quantified axiality using the Bayesian MAP skeleton framework of Feldman & Singh (2006), measuring the degree to which each shape was well explained (assigned high posterior probability) by its estimated skeleton. In both experiments, we found a systematic influence of axiality on f/g responses; the skeletal posterior ratio - the degree to which one shape's skeleton explained the common boundary better than the other's - systematically influenced the assignment of border ownership. Moreover, the local posterior ratio (computed only with respect to the skeletal point responsible for a given contour point) had a measurable influence on motion probe responses collected at each boundary point in Exp. 1. We conclude that (a) more axial regions tend to own their boundaries, and hence to be perceived as figural, (b) the magnitude of the Bayesian posterior effectively captures “axiality”, and (c) motion probe responses can indeed provide meaningful measurements of local figure/ground assignment.