Seeing Both the Forest and the Trees: Distinct Resolution of Hierarchical Representations in Visual Working Memory

The capacity limit in visual working memory (VWM) has been well established, but the way in which it is limited remains under debate. For example, an item limit was proposed by the discrete slot models whereas a flexible allocation of information was suggested by the flexible resource models. Despite the discrepancies on the nature of VWM capacity limit, both models assume that items are processed independently. Recent evidence, however, revealed that VWM is hierarchically structured. Given that an item can be represented both at the ensemble and the item level, existing measures of VWM capacity cannot distinguish between them. To separately measure an item at both ensemble and item levels, we designed a novel paradigm by combining the continuous report paradigm with a categorical judgement task. In the present series of four experiments, we found that participants’ performance at the ensemble and the item level was differentially affected by the layout and the number of objects in the display. Similarly, hierarchical Bayesian models fit to the empirical data suggest distinct effects of layout and set size on ensemble and item representations, respectively. Moreover, separating ensemble from item representations further allowed us to address their relationship and we uncovered a surprising possibility of independence between these two levels of representations. Taken together, the present dissociation and the demonstration of independence suggest distinct mechanisms of item and ensemble processing, which further sheds light on the nature and structure of representations in VWM.