Beyond temporal order: Spatial contextual cueing in sequential visual search

Statistical learning (SL) of co-occurring events in space and time shapes visual selection, yet most research using the visual-search paradigm has examined these dimensions in isolation. Work with static displays has revealed SL of spatial contexts, a phenomenon known as contextual cueing, in which repeated distractor configurations form associative links with target locations, facilitating search. By contrast, sequence-presentation approaches that utilize temporally evolving displays have emphasized order-based mechanisms, such as inter-item chaining or position-list coding, as accounts of SL. Here, we developed a novel search task in which items appeared sequentially at constant (vs. variable) positions across space and time, allowing us to track how SL unfolds in dynamic search environments. Across five experiments (N = 125 total), we systematically varied spatial, temporal, and featural regularities. We found that any combination of two constant dimensions (space+time, space+identity, time+identity) was sufficient for SL, but identity alone was not. Crucially, robust contextual cueing emerged even when the temporal order of the distractors was randomized (Experiment 4), demonstrating that sequentially presented items can be learned as spatial configurations rather than through temporal order mechanisms (e.g., chaining or serial position codes). This reveals that SL with sequential viewing can operate through configural spatial representations, extending classical contextual cueing frameworks to dynamic event streams. This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), grant numbers ZI1950/2-1, GE1889/7-2, and WO2552/2-1. Additional funding was provided by the Mentoring Program of Fakultät 11 of LMU, München. notReviewed other