Abstract
Previous studies have shown that the visual system can identify objects much faster than they can be stored in memory. The speed of this identification system coupled with the small capacity of visual working memory (4 items), suggests the need for an attentional mechanism to selectively store only a subset of the identified objects in a scene into visual working memory. In this study we attempted to establish the spatial properties of this mechanism by comparing it with an earlier, perceptual-level attention mechanism. We used a visual working memory task in which a sample array of eight colored squares was presented briefly in a configural circle. After a short blank interval, the subject's memory for half of these items was tested. Spatial precues were presented at the locations of four objects in the sample array, indicating to the subject which items would be subsequently tested. These cues directed attention to objects at either four contiguous locations, four noncontiguous locations, or all eight locations (neutral trials). Performance was found to be equally accurate for both the contiguous and noncontiguous cues, indicating that working memory-level attention can be divided among four noncontiguous locations with no cost. In a second condition, we examined perceptual-level attention using the same task by partially occluding a single item in the sample array with small colored masks. In this condition, accuracy was significantly lower for noncontiguous cues than for contiguous cues, indicating that perceptual-level attention cannot be easily divided among noncontiguous locations. These results indicate that the functional properties of attention are different for perceptual tasks versus working memory tasks, consistent with the idea that different mechanisms of attention operate within different cognitive subsystems.