Abstract
When presented with a single brief exposure of 12 dots colored with the same but unknown identical color interspersed among 14 dots of 7 different colors, subjects can accurately report the centroid of either the subset of homogeneous dots or the subset of heterogeneous dots, or the centroid of all the dots. When asked to locate the centroids of both heterogeneous and homogeneous subsets simultaneously (two responses), performances are little affected compared to single response controls. We interpret these remarkable data in terms of subjects' ability to simultaneously create two groups of stimulus items according to higher-order attributes (e.g. homogeneity or heterogeneity), and to compute summary statistics, e.g., centroids, on the items within a group. Whereas perceptual grouping by similarity is a cornerstone of Gestalt theory, the ability of subjects to also concurrently group by dissimilarity makes similarity tautological. Prior experimental demonstrations of grouping have been largely qualitative, the present experiments reveal the relative weight every color of dot among targets and distracters to the judged centroid in both the homogeneous and heterogeneous sets, and thereby demonstrate a high degree of selectivity in the contribution of dots to a particular centroid. The paradigm also provides a lower-bound estimate of the number of dots that contribute to a centroid judgment--over eight from each subset--an extraordinary number. In a partial report procedure, when reporting whether a cued dot was present or absent in the stimulus display, subjects have conscious access to less than 4 stimulus dots, enormously fewer than the >16 available to the pre-conscious centroid computation. The direct quantitative measurements enabled by the centroid paradigm reveal a great selectivity and an extraordinary information capacity of the pre-conscious perceptual grouping process.
Meeting abstract presented at VSS 2018