Impact of spatial grouping on mean size estimation.

People represent summary statistics of visual scenes, but it is not fully clear whether such summary statistics are extracted automatically. To determine whether different levels of summary representation (i.e., at the perceptual-group or the entire-display level) may be formed differently, in two experiments we investigated how people extracted summary statistics for displays consisting of spatially segregated groups. Participants were asked to report the mean sizes of either entire sets or perceptual groups in precue and postcue conditions. There was no precueing advantage in the mean size estimations of entire sets. However, when these precues identified target perceptual groups, participants reported the perceptual-group means more accurately than when postcues were used. In the postcue condition, participants were biased toward the entire-set mean even when they were probed to report the perceptual-group mean. There was also greater bias toward the entire-set mean for more erroneous perceptual-group summaries. These findings suggest that ensemble representations are extracted more efficiently for the whole than for the perceptual parts and that ensemble perception is not a uniform process across perceptual groups and entire sets.