Abstract
A Glass pattern consists of randomly distributed dot pairs (dipoles) whose orientations are determined by a geometric transform. To perceive the spatial structure in a Glass pattern, an observer needs to group dipoles to get overall shape. To understand this grouping effect, we investigated how the discrimination threshold between a concentric Glass pattern (target) and a random-dot pattern can be affected by the presence of another Glass pattern (mask). The Glass patterns contained randomly distributed dipoles who orientation were arranged to produce concentric, radial, vertical, and spiral global forms. We used a 2AFC paradigm in which a mask was presented in both intervals while the target was randomly presented in one interval and a random dot pattern with the same number of dots as the target in the other. We measured the target dot density required for an observer to detect the target at 86% correct level (density threshold). For the concentric and the spiral masks, the target threshold first decreased (facilitation) then increased (masking) as mask density increased. The radial and vertical masks only produced masking effect at high mask density. The concentric and the spiral mask also produced a greater threshold increase at high density than the radial and vertical masks. Both the facilitation and masking effects decreased with the curvature of the spirals. The data can be explained by a divisive inhibition model. In this model, the global form detector sums the responses of the local dipole detectors whose preferred orientation conforms that of a concentric pattern. The response of the form detector is this summed input raised to a power and divided by the sum of an additive constant and an inhibition input that is a nonlinear combination of all local responses. Our result suggested that divisive inhibition played an important role in global form perception.
Supported by NSC 96-2413-H-002-006-MY3.