Visual working memory (VWM) is key to many daily tasks, such as remembering visual information about traffic when crossing a busy street. Despite extensive research, the extent to which VWM abstracts out sensory properties not relevant to identification, such as object size, remains unclear. In three experiments, we examined how object size affects VWM, with size defined in two ways: retinal size, referring to the image’s size on the screen (small or large photos), and canonical size, referring to the typical size of objects in the real world, from big (e.g., a tower) to small (e.g., an egg). Experiments 1 and 2 tested memory for real-world objects, classified into four types based on their photo size and canonical size. VWM was better for large rather than small photos—a retinal-size effect—and for canonically small than big objects—a canonical-size effect. These effects were stronger when participants remembered a mix of different-sized objects than when all objects in a display were of the same size. Experiment 3 tested memory for colored squares that were either big or small on the screen. Size had no effect when displays consisted of colored squares of the same size, but big squares were remembered better when mixed with small squares. These results suggest that seemingly irrelevant sensory properties affect VWM, favoring objects that stimulate more neurons. The effect is stronger when size conditions are mixed, indicating that retinally larger objects are better attended.
