How does visual working memory solve the binding problem?

Abstract

Location has been highlighted as playing a crucial role in the relationship between memory for features and memory for bindings, with features purportedly bound to one another when they share a location. In three experiments (N = 24, 20, and 24), we show that feature-feature bindings are formed effectively in parallel when stimuli are separated in space but are disrupted when two objects are simultaneously presented in the same location. This pattern holds when conditions are equated with regard to memory for individual features. These findings confirm a prediction from a two-stage model of encoding in visual working memory, in which initial parallel encoding of features in spatial maps is followed by a subsequent sequential binding process that forms object representations that no longer rely on spatial location.