Joseph Schmidt¹, Steven Ford¹, Ashley Ercolino¹, Mark Neider¹, Corey Bohil²; ¹University of Central Florida, ²Lawrence Technological University

Over a decade of research demonstrates that object categories can guide visual search. However, a prominent category learning theory suggests that two systems govern categorization (e.g., Ashby et al., 1998, 2002). The explicit learning system relies on working-memory (WM) and easily verbalizable rules (Rule-Based; RB), whereas the implicit learning system relies on associative learning and difficult to verbalize rules (Information-Integration; II). We tested these systems’ contributions to categorical search by training participants to categorize sinusoidal gratings best characterized by RB or II learning, followed by categorical search which included a concurrent spatial working memory task on half the trials (i.e., search between memory and probe arrays). We found search for II categories relative to RB categories resulted in stronger search guidance as measured by the proportion of trials in which the target was the first object fixated, replicating our prior work (Bohil, et al., 2023). However, guidance was largely unaffected by the addition of the WM task. Taken together with the recent success of computational models of search which utilize inverse reinforcement learning (e.g., Yang et al., 2020; Zelinsky et al., 2021), this suggests search guidance may be governed by implicit systems. Importantly, object categorization times (dwell times on fixated targets and distractors) demonstrated clear effects of separate learning systems. RB classification slowed considerably when paired with the WM task but II classification was largely unaffected by the WM task. Collectively, this suggests that visual search guidance may be governed by implicit systems, whereas, object categorization times show clear impacts of the separate learning systems, with explicit RB categories being strongly affected by concurrent WM demands, whereas implicit II categories remain largely unaffected by a concurrent WM task.

Acknowledgements: This research was supported by the National Eye Institute of the National Institutes of Health under Award Number R15EY029511.