Animals do not experience the world passively. Hawks track prey by tilting their heads. A person leans forward and reads a sign. Scientists call this “active sensing.” In other words, move your body to gather better information.
A specific version of active sensing is infotaxis. This describes how animals move strategically to maximize the information available from their surroundings. Despite the central role of mice in neuroscience research, whether mice use this strategy remains an open question.
Mice have poor eyesight, approximately seven to eight times worse than humans. It also lacks a fovea (fovea fovea). The fovea is a small, specialized area in the eye’s retina that allows us to have sharp, clear central vision, color perception, and fine detail.
Because of these apparent deficiencies, researchers believe that mice rely more on smell, whiskers, and hearing than vision. At the same time, we know that mice use vision for a variety of tasks, from detecting predators to capturing prey to navigating space.
A team of scientists led by Mackenzie Weigand-Mattis, Bertarelli Foundation Professor of Integrative Neuroscience at EPFL, has shown that mice do indeed perform visuotaxis. Using a custom-built virtual reality (VR) system, they showed that mice move strategically to seek out more informative views of partially occluded objects, and that this behavior is precisely tuned to the amount of visual information available.
The works are published in current biology.
black and white tears
The researchers built a free-roaming VR “arena” that displayed a 3D scene rendered on screen in real time from the perspective of a mouse. They tracked the animals’ positions and movements using a 100 Hz overhead camera and DeepLabCut-Live, a marker-free tracking platform developed in 2020 by Mathis’ group.
Mice were trained to discriminate the location of a target object, a white teardrop, from a distractor, a black teardrop, and indicated their choice by walking to the corresponding side of the arena.
Now comes the important action. The screen places a virtual wall in front of both the target and distractor objects, leaving only a narrow gap in the middle. In the most restrictive condition of the first experiment, only 10% of each object was visible from the starting area. However, as the mouse moved closer to the screen, the viewing angle widened and more hidden objects became visible.
When the teardrop was mostly hidden, the mouse moved much closer to the screen before making a selection, slowing down and taking a more winding path on the way. Sometimes they changed direction during the trial as new visual evidence became available.
Completely open source platform
The research team tested five levels of occlusion and found that the mice’s chemotactic behavior changed continuously. Once the target is out of sight, the mouse will move closer before you can select it. Mice that moved closer also tended to make more correct choices under the most difficult conditions, suggesting that this strategy helped them solve the task.
The mice displayed this behavior immediately when they first encountered an occluded object, after already learning the task. This suggests that they used their internal understanding of the environment to cope with new visual challenges.
This study shows that even mice, despite their relatively poor visual acuity, actively move to gather better visual information, rather than simply reacting to what they see. The research team has made the platform completely open source and proposes that it is ideal for future studies that combine brain recordings with active visual behavior, helping researchers understand how looking and moving are coordinated in the brain.
The experiment was carried out under the conditions defined by the Swiss Animal Welfare Act and approved by the relevant veterinary authorities.
sauce:
Lausanne Federal Institute of Technology
Reference magazines:
Benkett, C. others. (2026). Visual uncertainty and task demands shape active sensing strategies in mice. Current Biology. DOI: 10.1016/j.cub.2026.06.011. https://www.cell.com/current-biology/fulltext/S0960-9822(26)00722-0
