Efficient behavior requires the rapid attentional selection of task-relevant objects. Previous research has shown that target-selective neurons in visual cortex increase their baseline firing rate when participants are cued to search for an object. Such preparatory activity represents a key finding for theories of visual search, as it may reflect a top-down bias that guides spatial attention, favoring processing of target-matching input for subsequent report. However, in daily life, visual search is often guided by non-target objects that are neither externally cued nor reported. For instance, when looking for a pen, we direct our attention to the office desk where we expect the pen to be. So-called anchor objects (e.g., the desk) thereby guide search for associated objects (e.g., the pen) in scenes.
Here, we used fMRI and eye tracking to test whether preparatory activity represents the externally cued target (the pen) or the internally generated anchor object (the desk). In an anchor-guided search task, participants (N=34) learned associations between targets and anchors and searched for these targets in 3D rendered scenes. To fully dissociate target from anchor processing, target-anchor associations reversed across scene contexts. Participants’ first fixations were reliably guided towards the target-associated anchor. Importantly, preparatory fMRI activity patterns in object-selective cortex represented the target-associated anchor rather than the target. Whole-brain analyses additionally identified a region in the right intraparietal sulcus that similarly represented the anchor. Our results show preparatory activity represents a self-generated guiding template, not necessarily the cued target, contributing to the efficiency of visual search.