Abstract
Recent studies suggest that visual working memory (VWM) is best described by a model that enables the storage of a discrete number of items with limited precision. Motivated by known similarities between neural mechanisms of visual selection and working memory, we asked whether performance on an attention-demanding selection task could be described by a similar model. Observers were cued to monitor a variable number of locations in a masked visual display and discriminate the orientation of a single target. Performance on this task was well-described by a model assuming that observers may select a fixed number of spatial locations with limited precision, while encoding no information from other locations (R2 = .94). In contrast, a shared resource model that assumes no fixed selection limit and an inverse relationship between the number of selected locations and the precision of information that can be extracted from any one location provided a relatively poor fit to the observed data (R2 = .46). Furthermore, selection capacity estimates obtained in this task were strongly predictive of VWM capacity estimates obtained in a memory-limited task that employed the same stimulus set. Finally, a cue-evoked N2pc (an ERP component thought to reflect the selection and individuation of objects) was strongly predictive of the number of locations that observers could successfully monitor. This predictive relationship suggests that behavior in this task was limited by the selection of multiple positions rather than the subsequent encoding or storage of information at these locations. Together, these findings suggest that visual selection and VWM storage depend on a common fixed capacity system that enables the selection or storage of a discrete set of positions or items.