Abstract
Visual search is guided by templates maintained in visual working memory (VWM). Currently, there is debate over the architecture of this guidance, with one theory claiming that only a single VWM item can guide attention at a time (single-item template hypothesis, SIT) and another claiming that multiple items can guide attention simultaneously (multiple-item template hypothesis, MIT). Thus far, relevant evidence has come primarily from costs when switching guidance between different VWM items and from capture when multiple items are maintained in VWM. However, switch costs do not necessarily distinguish between theories, and capture effects do not probe strategic processes. In the present study, we developed a redundancy gains paradigm to test this question for the core construct of interest: strategic attentional guidance by VWM. Participants searched for a target defined by either of two cued features (e.g., color, shape), which varied on a trial-by-trial basis. When present, the target matched the cue on either one feature value or both. Analysis of RT distributions tested for redundant trials that were faster than predicted by two independent guidance processes operating in parallel (i.e., violations of the race model inequality, Miller, 1982). Violations are consistent with a specific architecture in which both cue values guide attention in parallel and sum on the priority map (i.e., a coactive architecture, consistent with the MIT). With evidence in favor of a particular parallel model, all serial models can be eliminated (e.g., SIT). Robust violations were observed in four experiments. In particular, violations were observed when the two cue values came from the same dimension (two colors), were associated with different objects in the cue display, and could not have been easily combined into an integrated template. Together, the results provide strong evidence that features from multiple objects in VWM guide attention in a parallel, coactive manner.