Abstract
People can rapidly extract average size information from a set of objects (Ariely, 2001; Chong & Treisman, 2003). Here we explored whether this average is computed using a late representation of the display that has undergone iterative processing, or an initial representation. We asked participants to judge the average size of a set of six circles presented for 30 ms, along with two additional circles that were either smaller or larger than the average. These additional circles were surrounded by four dots that either disappeared simultaneously with the circles, or remained for an additional 320 ms. When the dots disappeared simultaneously, the additional circles were visible. When the dots remained longer, however, the additional circles were invisible due to object substitution masking (Enns & Di Lollo, 1997). If the averaging process uses a late representation of the display, the masked circles should not influence average size judgments. In contrast, we found that masked circles strongly influenced average size judgments, suggesting that the averaging process does not require conscious perception. Average size was also overestimated when the circles were invisible relative to when they were visible, but this difference was not observed in the control condition where the mask did not contain a circle. We speculate that the size of the mask, which always subtended a larger area than the biggest circle in a display, participated in the global average. That is, the mask size was counted into the average calculation when it made the inside circles invisible. In summary, we found that masked circles strongly influenced size averages even though they were never consciously perceived, and that the four-dot mask itself could also contribute to the size average, but only when it blocks the conscious access to the masked objects