Mice strategically move to gather hidden visual information, according to research published on June 30. A team led by Mackenzie Weygandt Mathis, professor at the Bertarelli Foundation Chair of Integrative Neuroscience at EPFL, found that mice perform a specific version of active sensing known as infotaxis when navigating virtual environments.
The study used a custom-built virtual reality system where mice were placed in an arena with a screen displaying a 3D scene rendered from their perspective. The animals' positions and movements were tracked using an overhead camera and DeepLabCut-Live, a marker-less tracking platform developed by Mathis's group. Mice were trained to identify the location of a target object among distractors and indicate their choice by walking to one side of the arena.
In key experiments, researchers introduced virtual walls that occluded most of both objects except for a narrow central gap. When only 10% of each object was visible from the starting area, mice walked closer to the screen before making decisions, slowed down during their approach, and took more winding paths. They sometimes reversed direction mid-trial when new visual evidence became available.
The team tested five levels of occlusion and observed that infotaxic behavior scaled continuously: as less of the target was visible, mice moved closer before choosing. Those who approached more closely also made more correct choices under difficult conditions. The behavior appeared immediately upon encountering occluded objects after learning the task, suggesting an internal understanding of environmental structure.
Despite having lower visual acuity than humans and lacking foveas for sharp central vision or fine detail perception, mice actively move to improve what they see rather than passively reacting. The researchers have made their VR platform open source for future studies combining brain recordings with active visual behavior.
The experiment was conducted under Swiss animal welfare legislation, with approval from relevant veterinary authorities.
