Conceptual size ensembles cannot be predicted by individual item size representations

Abstract

The visual system compresses redundant information into efficient, ensemble representations by averaging features across items. Ensemble perception operates with remarkable flexibility, even integrating object information at a conceptual level. For example, given a sufficiently strong depth cue, the visual system represents the perceived size of a set of triangles as opposed to their physical size (i.e., it accounts for size constancy; Suresh, Thomasson, & Haberman, VSS, 2017). In the current set of experiments, we explored whether conceptual ensemble size representation may be predicted by the size representation of the individual items composing the group. In every trial, observers viewed an individual triangle with and without linear perspective cues and judged whether a subsequently presented test triangle was larger or smaller than the preceding triangle. Whereas observers were biased to perceive the average size of multiple triangles as larger when presented in the context of linear perspective cues (i.e., conceptual size averaging), they did not take those cues into account when estimating the size of a single triangle. That is, observers perceived a single triangle in the context of linear perspective cues as the same size as a single triangle without linear perspective cues. These results suggest the generation of a conceptual size ensemble cannot be predicted by the individual item representations, which points to an emergent calculus that depends on judgments at the group level.