What does visual suffix interference tell us about spatial location in working memory?

A visual object can be conceived of as comprising a number of features bound together by their joint spatial location. We investigate the question of whether the spatial location is automatically bound to the features or whether the two are separable, using a previously developed paradigm whereby memory is disrupted by a visual suffix. Participants were shown a sample array of four colored shapes, followed by a postcue indicating the target for recall. On randomly intermixed trials, a to-be-ignored suffix array consisting of two different colored shapes was presented between the sample and the postcue. In a random half of suffix trials, one of the suffix items overlaid the location of the target. If location was automatically encoded, one might expect the colocation of target and suffix to differentially impair performance. We carried out three experiments, cuing for recall by spatial location (Experiment 1), color or shape (Experiment 2), or both randomly intermixed (Experiment 3). All three studies showed clear suffix effects, but the colocation of target and suffix was differentially disruptive only when a spatial cue was used. The results suggest that purely visual shape-color binding can be retained and accessed without requiring information about spatial location, even when task demands encourage the encoding of location, consistent with the idea of an abstract and flexible visual working memory system