Schneider, A. M.; McGregor, J. N.; Song, M.; Amme, J. L.; Zheng, S.; Wu, D.; Tu, J.; Yao, G.; Eslinger, E.; Chitalia, J.; Powers, J.; Sinha, V.; Dyer, E. L.; Levenstein, D.; Hengen, K. B.

Ethological tasks promise to engage the integrated perception, memory, and decision-making that define natural behavior, yet laboratory implementations are often so sparsified that they may fail to recruit the very cognitive processes of interest. We tested whether increasing environmental complexity in a standard task could expose this hidden cognition. Mice that were already expert hunters in a bare arena were challenged to capture live insect prey in arenas filled with objects that obstruct movement, occlude vision, and offer the prey places to hide. Despite their prior mastery, the added complexity revealed an entire layer of learning that the simple task failed to engage: rather than refining the sensorimotor details of pursuit, mice reorganized how they searched the environment. Across trajectory, kinematic, and object-referenced analyses, learning was expressed predominantly within the search state. To analyze behavior in explicit relation to environmental structure, we developed an open-source framework-a compact ethogram with hierarchical, pose- and object-based classification-that links each action to its environmental context. Unsupervised analyses revealed structured search dynamics across multiple timescales, and a minimal, interpretable agent-based model showed that short-term spatial memory and object-specific value are together sufficient to reproduce the non-random structure of search, including a learned, non-backtracking bias that emerged within the first days of object exposure. Classifiers further showed that mice selectively acquired the object interactions most likely to expose hidden prey. Reproducible with inexpensive materials, the paradigm and its analysis tools offer a sensitive behavioral readout of search, memory, and strategy for studies that conventional low-dimensional assays leave unresolved.