Abstract
At VSS 2008, we presented a series of experiments that sought out common mental representation strategies for three-dimensional shape across the modalities of vision and touch. One of these experiments required subjects to physically sculpt replicas of visually and haptically presented objects. While investigating strategies for comparing the depicted shapes to their respective ground-truth we developed a metric based on spherical harmonic decomposition. An unexpected and surprising artifact of this procedure is that it is also highly predictive of performance in our various discrimination tasks. Here, we present the details of this model as well as a reanalysis of results from other haptic and visual discrimination experiments (Norman et al. 2004, 2006) that also show close agreement with our model. Finally, we present a series of experiments intended to test the limits of our model. We use a spherical harmonic decomposition that shares characteristics with traditional Fourier methods. Subjects performed a paired-comparison discrimination task using objects that varied in frequency (complexity) and phase (relative location of features). It is well known that, in the case of two-dimensional visual images, the phase component contains an overwhelming amount of the information needed for identification and discrimination. Is this true for the visual discrimination of three-dimensional objects as well? Our results show that, for particular ranges of 3D spatial frequency, the phase components dominate, while at other frequencies the amplitude carries the information used for discrimination.