Mouse lockbox: a sequential mechanical decision-making task to investigate complex mouse behavior

Recent advances in automated tracking tools have sparked a growing interest in studying naturalistic behavior. Yet, traditional decision-making tasks remain the norm for assessing learning behavior in neuroscience. Here, we present an alternative sequential decision-making task to study complex mouse behavior. We developed two different 3D-printed mechanical puzzles, so-called lockboxes, that require a sequence of four steps to be solved in a specific order. During the task, the mice move around freely, enabling the emergence of complex behavioral patterns. We observed that mice exhibit a high level of motivation, willingly engage in the task, and learn to solve it in only a few dozen trials. To analyze the strategy the mice use to solve the task, we used three cameras to capture different perspectives and developed a custom data analysis pipeline. The pipeline allows the automated detection of interactions of the mice with the different lockbox parts for a large corpus of video material (>300h, 12 mice). We find that an increasing number of mice are capable of solving the lockbox task across
trials. The mice are significantly more engaged with the lockbox for trials in which the task is solved and they furthermore express a higher-than-random preference towards the state-advancing lockbox parts. Preliminary analyses suggest that the increased solving capability is not due to an increased interaction time with the task, but potentially due to low-level motor learning and/or due to learning of a high-level solution strategy. Although our data analysis is preliminary, we find that freely moving mice can rapidly learn to solve complex, multi-step mechanical puzzles that are more challenging than most standard tasks. We believe that this task provides a promising balance between natural behavior and a well-defined task that provides anchor points for the analysis of both the behavior and – in future – neural recordings.