Abstract
Much recent research has demonstrated how statistical learning biases spatial attention toward targets in the external environment. However, attention can also be directed to internal representations; does statistical learning guide attention within working memory representations? To address this, we had participants encode four items in working memory and then search for one "target" item within the encoded representation. To promote statistical learning, the target appeared more frequently in one quadrant (i.e., rich quadrant) than the other three quadrants (i.e., sparse quadrants). Results showed participants were faster and more accurate to locate targets in the rich compared to sparse quadrants. This suggests that participants learned to prioritize their working memory search toward the rich quadrant. However, an alternative explanation is that participants learned to selectively memorize the item in the rich location during the encoding stage. To rule out this "encoding bias" possibility, we modified the task to include multi-feature items. Specifically, participants had to remember the color and shape for each of the four items. Then, one feature target, either a colored circle or a white shape, appeared in the center. Participants reported the feature's location in the memory display. Unbeknownst to participants, color targets were more frequently located in one "color-rich" quadrant, while shape targets were more frequently located in another "shape-rich" quadrant. Results showed that performance was faster and more accurate for color – but not shape – targets when they appeared in the color-rich quadrant; we found the same pattern for shape – but not color – targets when they appeared in the shape-rich quadrant. These results show that observers encoded each multi-feature object equally but biased their memory search depending on the statistics associated with each feature. We thus confirm that statistical learning can guide search within working memory representations.
Meeting abstract presented at VSS 2016