Abstract
Representing summary statistics of visual ensembles is an efficient tool to rapidly encode the features of multiple objects at a glance. However, even when "seeing" average, we still perceive numerous distinct objects and are aware of their variation to some degree. What information is still stored about the individuals during ensemble perception? Brady and Alvarez (2011) showed that the stored size of individuals is typically biased to an ensemble average. We attempted to test how ensemble characteristics affect rapid representation of individuals when attention is focused on that individual or distributed across the entire ensembles. Observers were presented with sets of 1 to 16 circles of variable size (4 distinct sizes could be involved in a trial). On the one half trials, the location of an arbitrary chosen circle was precued (focused attention), on the other half trials, the location was postcued (distributed attention). Two test items were presented after the set forcing observers to choose one looking like the cued circle. On a half test pairs, a larger member was correct choice. Correct responses were measured as accuracy index and the probability of choosing a larger member was measured as bias index. We found in the result, that average accuracy was predictably lower in distributed attention condition for all set sizes including 1 item. Accuracy tended to fall down with set size for both focused and distributed attention, however, for distributed attention the fall was more dramatic indicating increasing noisy interactions between individuals. The general bias trend was underestimating the size of individuals. In focused attention condition, this underestimate was slight. In distributed attention, this underestimate trend remarkably increased with an item size. This indicates strong tendency to compress estimated size around smaller values. In this sense, representations of individuals indeed appear to be more averaged in distributed attention.
Meeting abstract presented at VSS 2013