Abstract
Some figures are irresistibly seen as complete objects. Others are represented as mere collections of line segments. How does the visual system represent figures in between these extremes — say, when part of a figure strongly suggests a complete object, while another part strongly suggests an open contour? More generally, how do we represent figures that are ambiguous or underdetermined with regard to 'shapehood'? We explored these questions using the "tap-the-shape" task, which reveals the so-called "skeletons" (internal symmetry-defined geometric structures) that underlie human shape representation: when a subject is shown a shape on a touch-sensitive tablet computer and is instructed simply to tap the shape once anywhere they please, the aggregated taps from hundreds of observers collectively form that shape's skeleton (Firestone & Scholl, 2014, Psychological Science). Here we exploit this striking phenomenon to ask a more fundamental question: what is a shape in the first place? Across several experiments, we showed thousands of observers ambiguous or incomplete shapes, and asked them to tap “inside” those figures (with one tap and one shape per subject). Remarkably, ambiguous figures — e.g. a rectangle with one missing edge — were represented in a hybrid manner: touches near the 'inside' edge formed clear skeletal branches, while touches near the open edge devolved into noise. We also tested 'open' figures that were ambiguous between multiple complete interpretations. For example, a trapezoid missing its top edge — which can be seen as either a trapezoid or a triangle — was tapped as though it had both the shape skeleton of a triangle and the shape skeleton of a trapezoid. We discuss the significance of these apparent hybrid representations for theories of shape, and for use of the "tap-the-shape" method to study previously stubborn questions about visual representations.
Meeting abstract presented at VSS 2015