Abstract
Recent evidence suggests that observers show accurate knowledge of the mean size of a group of similar objects, a finding that has been interpreted to suggest that sets of multiple objects are represented in terms of their statistical properties, such as mean size (Ariely, 2001; Chong & Treisman, 2003, Marchant & De Fockert, 2009). A surprising finding is that this effect can be shown across different set sizes (from 4 to 16 members) with little or no detriment to judgements of mean set size (Ariely, 2001; Chong & Treisman, 2005b, exp 1). However, these studies have always held heterogeneity constant whilst manipulating set size. Here, we present data that replicate past findings when heterogeneity is held constant, but show that the accuracy of mean set size estimations decrease when both set size and heterogeneity increase. Our findings suggest that summary representations are not always obtained by averaging together the whole set of items; a feat that often requires a greater capacity than known focused attention processes and therefore has required the proposal of a new perceptual mechanism (Chong & Treisman, 2003, 2005a, 2005b, Treisman, 2006). Instead, summary representations may be based on a sub-sample from the set, within the capacity of focussed attention (Myczek & Simons, 2008). Increased variation in the set leads to more variation in the possible sub-samples and a less accurate approximation of the set summary statistic. These results have implications for current theory and the proposed role this mechanism plays in scene perception (Treisman, 2006; Oliva and Torralba, 2007).
ESRC award PTA-030-2006-00212.