Abstract
While the effect of stimulus similarity has been well-studied for auditory working memory, the impact of object similarity on visual working memory is not well known. Across eight experiments, we compared accuracy in a change detection task with displays of four objects.  The display could consist of four objects  from a single category (one-category condition) or two objects from two separate categories (two-category condition). The initial experiment used complex faces and houses, and results showed better change detection accuracy for two faces and two houses together versus either four faces or four houses alone. Experiments 2–6 tested additional categories (butterflies, bodies, and watches) to examine how generalizable this effect is. Results demonstrated that accuracy was better in the two-category condition only when faces were part of the set. Otherwise, when faces were not part of the set, accuracy in the two-category condition was never greater than in the one-category condition. Experiments 7–8 used inverted faces to examine the face-specific memory effect, and an advantage for inverted faces was not found. Together, these results suggest that upright faces are unique to visual memory (Curby & Gauthier, 2007). However, faces remain subject to hard limits of memory capacity, as remembering four faces was never better than remembering four other objects. For all other object categories that were tested, similarity does not affect visual memory processes.