Capacity limitations in the way humans store and process information in working memory have been extensively studied, and several memory systems have been distinguished. In line with previous capacity estimates for verbal memory and memory for spatial information, recent studies suggest that it is possible to retain up to four objects in visual working memory. The objects used have typically been categorically different colors and shapes. Because knowledge about categories is stored in long-term memory, these estimations of working memory capacity have been contaminated by long-term memory support. We show that when using clearly distinguishable intracategorical items, visual working memory has a maximum capacity of only one object. Because attention is closely involved in the working memory process, our results add to other studies demonstrating capacity limitations of human attention such as inattentional blindness and change blindness.capacity ͉ vision ͉ working memory D emonstrations of limitations in the way humans process and store information have generated much controversy among researchers in the cognitive sciences. A major theme concerns the capacity of working memory. Separate verbal, spatial, and visual object working memory systems can be distinguished (1-3), but similar estimates of their capacities have been established. These estimates have sometimes been made in terms of ''magical numbers,'' which have ranged from seven to about four words, numbers, or locations (4). In line with these results, recent studies suggest that it is possible to retain information of up to four objects in visual working memory (5-9).Studies of visual working memory have routinely used a change detection paradigm (5-7, 10) where a number of distinct objects or colors are presented briefly in a sample array. After a short blank interval, a test array is presented that is identical to the sample array, except for one object that may have changed. The task is to indicate whether all objects in the sample and test arrays are identical or whether one of them has changed. The objects and colors typically used in visual working memory studies have been few, repeatedly presented, and without difficulty classified into distinct categories easily given verbal labels such as red, green, blue, square, disk, etc. Whether the categorical boundaries are rule-, prototype-, or exemplar-based, the categorization process needs support by long-term memory storage. Repeated presentations of a few items, initially not separated by category boundaries, may lead people to develop boundaries and consequently use long-term memory support. Also, because items separated by category boundaries are easily assigned verbal labels, verbal working memory may be activated, leading to overestimation of visual working memory capacity (11). Additional potential problems with the traditional changedetection paradigm are that relational coding influences the results (12) and that spatial memory may be used to assist performance (1). A task that is supposed t...
