Abstract
Visual working memory enables the online maintenance of a limited quantity of information in an “online” or readily accessible state. Various paradigms have confirmed that this system is subject to relatively severe capacity limits. One influential paradigm for assessing these capacity limits is the change detection procedure, in which observers are asked to remember the content of a sample array for a brief period and then indicate whether any of the remembered items have changed in a subsequent test array. Multiple investigators have used this procedure and concluded that even with very simple objects, the number of items that can be maintained in working memory is limited to about four (e.g., Luck and Vogel, 1997; Pashler, 1988). Further work has shown that when more complex objects are presented, change detection performance declines even more (e.g., Alvarez and Cavanagh, 2004; Eng et al., 2005). We recently observed, however, that the influence of object complexity may be better explained by limitations in the resolution of representations in working memory rather than the number of items that are maintained (Awh et al., in press). This conclusion was supported by the observation that capacity estimates from a change detection procedure were equivalent for complex and simple objects when comparison errors were minimized by reductions in sample/test similarity. Here we extend these findings by describing a new analytic procedure for estimating the probability of comparison errors (i.e., the functional resolution of memory representations) while taking into account errors that result from limitations in the number of items that an individual can represent. This analysis coupled with new data demonstrates that the resolution of representations in working memory declines monotonically as the number of items increases, thereby clarifying how number and resolution interact during the maintenance of information in visual working memory.