Abstract
The visual system's ability to extract ensemble representations from cluttered scenes has been demonstrated for low-level features as well as high-level object properties such as facial expression (Ariely, 2001; Chong & Treisman, 2003; Haberman & Whitney, 2007; Parkes et al., 2001). Our previous work suggests that ensemble information influences visual search efficiency: participants were faster to detect a target face when its expression deviated substantially from the mean of the set, but only within sets containing low rather than high variance in expression (Puri et al., 2009). Thus, the relatively precise summary representations known to arise under low-variance conditions (Dziuk et al., 2009) may provide a basis for deviance detection. Here we tested whether an individual's ability to extract summary information from low- compared to high-variance sets predicts the degree of benefit when searching for deviant targets under the two variance conditions. Participants estimated the mean expression of face sets with either low or high variance. In a separate task, the same participants searched for a particular identity within low- and high-variance sets; the expression of the target face could be either near or far from the mean expression of the set. Across participants, the difference in mean estimation performance for low- versus high-variance sets was correlated with the relative benefit for detection of deviant targets within low- versus high-variance sets. In addition, within individuals, search times were more positively correlated across two separate presentations of the same display when the variance of the set was low. These results suggest that readily extracted ensemble representations enhance deviance detection. Furthermore, the availability of ensemble information may contribute to consistencies in search behavior within an individual.