Working- and long-term memory target templates converge to produce coactive guidance of attention during visual search

Templates that guide attention during visual search can be maintained in visual working memory (VWM) or in long-term memory (LTM). A central challenge for theories of attention is explaining how these sources of bias combine to determine selection. Here, we tested whether they converge on a common priority signal. Specifically, we contrasted a coactive model of guidance, assuming convergence, with three non-coactive alternatives, each holding that guidance from VWM and LTM operate independently. Participants searched for the presence of a target defined by either of two cued colors. One cued color changed on each trial (dynamic, engaging VWM), whereas the other remained constant (static, engaging LTM). The target could match either one of the cued colors (single-target trials) or both (redundant-target trials). Through distributional analysis, we tested for the presence of redundant-target RTs that were faster than could have been generated by the two individual guidance processes operating independently in parallel. Such "violations" of the race model inequality (RMI) are observed only within a coactive system. For single-target trials, searches were substantially more efficient for dynamic-match targets than for static-match targets. Nevertheless, we observed robust violations of the RMI on redundant-target trials. Together, these results demonstrate that target templates can be maintained in qualitatively different states yet ultimately converge to jointly influence a common attentional priority signal.