Abstract
The visual system efficiently and accurately extracts summary statistics, such as the mean, from sets of similar objects. Evidence for this phenomenon, known as ensemble perception, has been found across numerous features, such as orientation, size, and faces. However, when perceiving variability, the visual system is biased to overestimate variability in ensembles, particularly when objects are more similar and have less variability. This bias is found when perceiving line orientation, size, and hue, suggesting the visual system fundamentally exaggerates the variability in sets of similar objects. To further test this bias, we manipulated the underlying distribution stimuli were drawn from, exposure time, line length, and line contrast. Again, the bias to overestimate variability was found and was larger in magnitude when stimuli were more similar and had less variability. One possible mechanism for this bias is the boundary effect: objects close to a conceptual boundary, such as the same or not, are repulsed away from the boundary. We tested whether the conceptual boundary of sameness relates only to relevant features (line orientation) or if variation across irrelevant features (line length or contrast) would lessen the bias by pushing the set of stimuli away from the boundary. Participants judged the variability of a set of nine lines presented sequentially by adjusting the variability of nine lines presented simultaneously to match the target stimulus. In all experiments, people were biased to overestimate variability, and the magnitude of the bias was strongest for sets of lines with less variability, regardless of variation in irrelevant factors. The irrelevant feature of line length did not lessen the bias, providing evidence against the idea of a boundary effect related to sameness. If a boundary effect is responsible for the overestimation, the conceptual category is specific to only the relevant feature, not the object as a whole.