Abstract
Although the idea of visual averaging of multiple objects is widely recognized, there is still no consensus among researchers about its mechanism. Some authors suggest that size averaging is carried out in parallel over all items (Ariely, 2001; Chong & Treisman, 2003; Chong et al., 2008), while other consider it to be based on limited-capacity sampling (Myszek & Simons, 2008). Advocating the latter viewpoint Marchant, Simons, and De Fockert (2013) have recently reported a size averaging study where they manipulated set size and set heterogeneity independently. Their observers were good at averaging when a set included only two sizes distributed among all items (regular sets), but got poorer when the number of sizes increased with the set size (irregular sets). However, in their irregular condition the range of size variation was increasing systematically with the set size, while in the regular condition they used always the same two sizes that were very close to the mean. We suggest, therefore, that Marchant et al. could falsely recognize the effect of size variation as the failure of parallel averaging. We tested this in two experiments. In Experiment 1, we repeated Marchant et al.'s study and replicated their effects. In Experiment 2, we used the same range of variation for all conditions and changed the distribution within that range. Here, regular sets consisted of only smallest and largest items (instead of mean-similar ones in Experiment 1), and in irregular sets the intermediate sizes were added between these extremes. We found in the result that averaging accuracy remained almost the same for all regularity conditions and all set sizes, with even slight advantage of the regular condition. This indicates that size averaging still appears to be parallel but the accuracy of such averaging depends on the total range of variation of the target feature.
Meeting abstract presented at VSS 2014