Item

Abstract

To know how an object will behave, it is useful to be able to visually infer its physical
properties, such as its weight, hardness, or elasticity. One important source of information
about an object is the way it moves and interacts with other objects in the scene. Here, we
present two experiments in which subjects had to judge the elasticity of a ball from the way
that it bounces. Previous research on this topic generally used simple 2D graphics to produce
finely controllable, albeit not very realistic, simulations. With modern VR techniques,
however, we can maintain the tight control over the parameters while creating highly realistic
stimuli. Thus, we designed these psychophysical experiments using real-time 3D simulations
of both simple and complex object interactions. Subjects viewed two of these simulations side
by side and had to adjust the elasticity of the ball on the right to match the ball on the left.
The entire simulation was shown either right-side-up or upside-down. Our results suggest that
subjects generally rely on simple low-level image cues to match the elasticity: (1) For the
simple bouncing events, subjects confounded elasticity with the height from which the ball
was dropped; (2) Subjects were generally poor at matching elasticity for the complex
bouncing events, in which the reliability of the low-level cues was minimized; (3)
Performance was similar for right-side-up and for upside-down, suggesting a weak role of
prior knowledge about gravity. We have found that using real-time VR simulations presents a
number of unique challenges, and we will briefly review the successes and failures we
encountered in the design and implementation of these experiments.