Abstract
It has been proposed that visual working memory (VWM) is better for real-world objects than for sparse laboratory objects (Brady, Konkle, Oliva, & Alvarez, 2009). One of many gaps between "sparse" and "real world" objects is that real-world objects can have repeating structures that observers might use to encode object features and predict how it will appear after rotation. VWM capacity for mental rotation of sparse objects comprised of multiple oriented bars drops to near 1 (Xu & Franconeri, 2015), but it is possible that adding additional object structure may improve VWM for mentally rotated object features. Here, observers saw novel highly structured objects comprised of 6 connected cubes, with two cubes forming the "head and body" of the object and four "feet". The object then disappeared and underwent a cued rotation in depth (blocked design) that was small or large. The rotated object appeared with the same organization or with swapped colors of two of the cubes. Task performance relies on accurate binding, updating, and comparison of object features. Observers successfully performed large object rotations with no swaps, and were mildly impaired when a swap involved the top two cubes, even though all large rotations involved large changes in the retinal positions of cube colors. But when swaps occurred between two feet, accuracy at these same large rotations was devastated. This particular sensitivity to upper objects and lack of sensitivity (or failure to update) lower objects is consistent with prior findings on low capacity for rotation of simpler features, and suggests that structure alone is insufficient to produce VWM capacity benefits. Further work will continue to map the space between "sparse" and "real-world" objects, and to elucidate what characteristics, if any, promote improved VWM capacity for real-world objects.
Meeting abstract presented at VSS 2018