Abstract
In “Hybrid Search”, observers search for any member of a set of targets held in memory. Current hybrid search models propose working memory (WM) serves to pass representations of attended items one at a time from early visual processes into long-term memory (LTM) for comparison to target representations (Drew, et al., 2015). Here, we sought to better understand the role of WM in hybrid search using an ERP component associated with WM load, the Contralateral Delay Activity (CDA). Participants (N=20) memorized setsizes (MSS) of 2 or 16 categorical targets (dogs, cats, tables, or dressers), counterbalanced across participants. Next, they passed a recognition memory test twice with >=80% accuracy. Then, for 700 trials (per MSS), participants were cued to attend to one lateralized search item and indicate if a target was present. The cued object was either a target (e.g., target cat), a non-target from the target category (e.g., non-target cat), a non-target from a similar category (e.g., dog), or a non-target from a dissimilar category (e.g., table). ERP waveforms were time-locked to search onset. Overall, CDA amplitude was larger for objects from the target category, which provides support for the idea that processing is terminated for irrelevant objects following initial object identification (Cunningham & Wolfe, 2014). For objects within the target category, CDA amplitude increased with MSS, suggesting LTM load-dependent WM involvement in hybrid search. Furthermore, N2pc amplitude increased as the non-targets became more similar to the object category, but only for MSS16. This suggests more non-targets with similar features were selected for further processing in memory at larger MSS. Together, these findings demonstrate WM plays a larger role in hybrid search than previously thought. Specifically, to identify objects within the target category, WM resources may be deployed to compare the target set held in LTM to the potential target.