ABSTRACT
Abstract The primary aim of this experiment was to discover why one set of experiments (visual search) which tap early perception typically reports that features are independent and a second set (texture segregation) suggests that they are not. One explanation (Callaghan, 1989) is that lowering the similarity of elements within target regions in segregation displays is the crucial factor in interference. This hypothesis was tested in an experiment that attempted to bridge the gap between visual search for single elements and texture segregation for groups of elements by using a single visual search task, where the size of the target searched for was varied. The stimuli were rectangles varying in orientation and luminance. Three stimulus factors were varied. The first was the number of target elements (1,2, 6). The second was whether the background elements were constant or varied on a second dimension irrelevant to the segregation ference (constant, variback). The third factor, varied only for the multi-element target arrays, involved having either identical elements within the target region (high similarity) or variation on the irrelevant dimension of target elements (low similarity). It is apparent from the results that within-region similarity is an important factor in visual search, but it is not as simple an effect as it first seemed. First, it appears as though more than two elements are needed for the grouping process to have a significant impact; here we see effects of withinregion similarity only for six-element target displays. Second, lowering the similarity of the elements within the target region sometimes impairs the ability to find the target, and sometimes helps it. The particular effect depends on the property that is varied, as well as on whether one or both of the regions in the display contains variation. These results are discussed as they relate to existing views of early perception.
