Abstract
More often than not, objects that we see are partially hidden by other objects. As such, the images that fall upon our retinae are typically composed of object fragments separated by regions of occlusion rather than complete figures. Nevertheless, we perceive the world as being populated by cohesive forms rather than collections of independent parts. These image fragments must therefore be bound together in order to be perceived as wholes, a process known as ‘visual completion’. But how well do we integrate the parts of objects when forming perceptual wholes? Conventional wisdom suggests that our visual system should be exceptionally efficient at such a process--that a perceptually organized whole should be processed in a manner that is superior to what one would simply predict from performance with each individual part shown in isolation. We tested this idea by using a psychophysical summation-at-threshold technique (Nandy & Tjan, 2008). Specifically, we measured an observer’s contrast sensitivity S for discriminating perceptually complete figures as well as each of their constituent parts shown in isolation. From this, we computed an integration index Φ, where Φ = S2whole / Σni=1 S2part i and n equals the number of individual parts that make up a whole figure. Φ = 1 indicates the whole is processed in a manner that is predicted by performance with its individual parts, Φ > 1 indicates the whole is processed more efficiently than would be predicted by performance with its individual parts, and Φ <1 indicates the whole is processed less efficiently than would be predicted by performance with its individual parts. We find Φ to be significantly less than 1 across a variety of tasks involving the discrimination of perceptually complete figures. That is, we find the process of combining parts together to form perceptually complete wholes to be surprisingly inefficient.
Meeting abstract presented at VSS 2012