Back

From Exploration to Structured Navigation: Learning Dynamics of Freely Moving Marmosets during Foraging

El Mahmoudi, N.; Lanzarini, F.; Ziaie Nezhad, F.; Surendran, D.; Laurens, J.

2025-03-24 animal behavior and cognition
10.1101/2025.03.22.644580 bioRxiv
Show abstract

How do structured, memory-guided behaviors emerge in freely moving primates? We addressed this question by training common marmosets to perform a foraging task in a semi-naturalistic environment, where they retrieved food from eight dispensers over multiple sessions. Animals received no instruction or shaping and were free to visit dispensers in any order. We found that animals engaged with the task in most trials and, within engaged trials, transitioned from exploratory to efficient foraging behavior. Learning was marked by performance gains and the emergence of reproducible movement patterns between specific dispensers. Using trajectory analyses and probabilistic modeling, we found that animals formed stable route segments linking specific dispensers, reflecting the emergence of local navigation motifs. These segments became increasingly regular and predictable with experience. Yet rather than being rigidly replayed, they were flexibly recombined into variable global sequences. This indicates that animals adopted a hybrid navigation strategy, in which reusable route segments are embedded within a topological structure. These findings demonstrate how efficient, adaptive navigation can emerge through self-guided experience in complex environments. Our approach provides a naturalistic and longitudinal framework for studying the formation of structured spatial strategies in non-human primates, bridging ecological behavior with theoretical models of learning and memory.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.