Abstract
As we interact with the visual world, we encounter substantial redundancy. A field of grass, for example, is comprised of many similar blades of grass. Coding every one of these elements is an inefficient representation scheme, and thus the visual system engages in ensemble coding, a heuristic that represents large amounts of information through summary statistics. Ensemble coding occurs across several dimensions-observers can extract a precise estimation of the mean size of an array of dots, the mean orientation of a set of gabors, and even the mean emotion of a set of faces. Although it is clear the visual system favors a statistical representation of sets, the mechanism of ensemble coding remains unresolved. Some evidence suggests that parallel processes mediate ensemble coding, while other research shows that serial processes offer a more parsimonious explanation. Using a modified visual search task, we present evidence that supports ensemble coding as a parallel process. On every trial, observers saw a set of 1–6 faces that varied in expression from neutral to disgusted, simultaneously presented with a single test face. In one condition, observers had to indicate whether the test face was a member of the set (regular search). In the other condition, observers had to indicate whether the test face was more neutral or more disgusted than the mean emotion of the set (mean search). Reaction times for the regular search condition showed a significant increase as a function of set size (a positive slope), the hallmark of serial search processes. The reaction times for the mean search, however, showed no such dependence on set size, suggesting that observers were able to extract a statistical representation regardless of the number of items used to establish the mean emotion. This result supports the hypothesis that ensemble coding is mediated by parallel processes.