Abstract
Each frame of a Glass pattern consists of a random placement of dots and a spatially shifted copy of this pattern. Thus, each dot has a partner, forming dot-pair dipoles. When shown in succession, motion is perceived along the axis of the spatial shift. The perception of motion in dynamic Glass patterns is believed to be a two-stage process: first, local orientation detectors respond to the orientation signal in the dot-pair dipole; and second, global detectors integrate local orientation signals. We examined the ability to detect rotation in dynamic Glass patterns whose dipoles contained a) the same polarity, b) opposite polarity within a dipole, and c) opposite polarity between dipoles. We also manipulated the extent of spatial shift between the dots in a dipole. A two-interval forced procedure was used, and the proportion of dipoles spatially shifted to be consistent with rotation was varied. The task was to determine which interval (noise or signal + noise) contained rotation. When both dots in a dipole were the same polarity, rotation detection thresholds increased with increasing dot separation. The same pattern was found when dot polarity differed between dipoles. Thresholds were significantly elevated, however, when dot polarity differed within a dipole. Unlike other reports (Wilson, Switkes & DeValois, 2004), we find that observers are capable of discriminating between Glass patterns and random noise even when dot-pair dipoles are of opposite polarity. This outcome suggests an additional role for global mechanisms in the perception of motion from dynamic Glass patterns.