Increasing the Number of Context Elements in Early Vision: More is Better

Abstract

Studies in visual search have shown that in feature search conditions in which the target differs from the context in a single feature (like orientation), search time increases only slightly (or sometimes not at all) as a function of increasing number of context elements. This is a typical result for early vision conditions, and is taken as an indicator of parallel processing of the whole stimulus.

In visual search experiments the manipulation of increasing the number of context elements is confounded with the fact of increasing their spatial density. And as is known from other typical early vision parameters, eg texture segmentation, increasing density improves performance. So, parallel visual search may be a byproduct of the (implicit) density manipulation.

In two texture-segmentation experiments, the number of context elements was increased while their spatial density was kept constant. This manipulation leads to an enlargement of the whole stimulus. The results show that under these conditions increasing the number of context elements even improved performance. This improvement was observed particularly at a retinal eccentricity of more than 3 deg. The results are explained within a framework of a spatial pooling mechanism.