Global Mean Position Perception of Multiple Spatially-Separated Clusters

The visual system is remarkably efficient at extracting summary statistics from the environment, which drive the overall perception of a scene and inform judgments about individual objects within it. Most ensemble research focuses on the perception of one specific characteristic (e.g., average size) from a single group of stimuli, but natural visual environments usually consist of many groups of objects including outliers that are distributed over space. We evaluated how people perceive the aggregated ensemble mean position (i.e., center of mass) perception in the presence of multiple spatial clusters using a visualization task with scatterplots. Consistent with previous results, when there are distinct two clusters of dots, we find that people reliably overweight the group of smaller cardinality, including if it is a single outlier. For example, for two clusters, made up of 8 vs 16 dots, people report the average position too close to the cluster of 8. However, when the larger cluster is equally partitioned into multiple clusters (e.g., the 16-item cluster is split into two 8-item clusters), the overweighting of the small cluster dampens significantly, despite the total cardinality and center of mass of the large cluster being held constant. Furthermore, the bias towards the partitioned clusters increases at larger cardinality and increasing number of partitions. These results suggest that people are affected by the presence of distinct clusters, and partially rely on the average position of these clusters rather than solely the average position of the individual objects. Thus, Gestalt organization of clusters significantly alters the aggregated mean perception. Overall, we show that people make large systematic errors when judging the aggregated mean of multiple clusters, which is a realistic task that occurs in domains like calculating the average value in a scatterplot or identifying the balancing point of a set of objects.