Abstract
The sharp limit of working memory necessitates the selection of a subset of items from large memory displays to keep capacity from being overloaded. Because each item in the display is equally likely to be tested, this process is assumed to select items randomly from the entire array. However, the presence of grouping cues substantially impacts which items will be selected for storage, even when they are no more likely to be tested than ungrouped items (Woodman et al., 2003). Furthermore, performance for specific configurations of arrays is consistent across participants (Brady & Tenenbaum, 2013), suggesting that similar information tends to be encoded from each display. Here, we tested the hypothesis that participants have a preexisting spatial bias that prioritizes the selection of items in certain portions of the visual field. Analyzing a large dataset of change detection performance (N=271) revealed that changes in the upper left quadrant were detected much more frequently than changes in the bottom right, suggesting that items in the upper left were preferentially encoded into memory. We examined the impact of these spatial biases in a new experiment in which we manipulated the probability of changes in each quadrant, which has been shown previously to bias selection towards those items (Umemoto et al., 2010). Participants (N=28) were unknowingly assigned a dominant quadrant where 80% of all changes occurred. Accuracy dramatically improved for detecting changes in the dominant quadrant over the nondominant quadrants, with one exception: participants assigned to the bottom right condition showed no improvement in that quadrant and continued to show superior performance in all other quadrants. These results suggest that individuals have strong preexisting spatial biases that affect the encoding of information into working memory and that the lower right quadrant may be a "blind spot" for selection within crowded displays.
Meeting abstract presented at VSS 2018