Bernstein member involved: Jonathan Gant

Seeing and moving are directly linked in the brain. When we stand still, the visual scene barely shifts. When we run, by contrast, the environment rushes past, so that the signals reaching the eye become more variable and change faster. Because animals generate this movement themselves, the change is predictable and the brain can adjust at the moment movement starts rather than waiting for new visual input.

However, the strength of this neuronal adjustment varies considerably between animal species. When a mouse runs, neuronal activity in its visual system undergoes significant changes. In marmoset monkeys, these effects are much weaker or absent. “This raises the question of whether rodent and primate vision follow different rules,” says Professor Wiktor F. Młynarski, Professor of Computational Neuroscience at LMU’s Faculty of Biology.

Together with Jonathan M. Gant – a doctoral researcher in his group, the neuroscientist recently published a study in the journal Science Advances, in which they propose an answer to this question. The visual systems of the two animal groups do in fact follow the same basic evolutionary principle of information processing. The differences arise because signals that enter the brain through mouse and primate eyes are not the same.

The brain adjusts to motion

The starting point is the efficient coding hypothesis. This mathematical approach assumes that neurons have adapted to typical patterns of their natural environment over the course of evolution. This allows them to carry the most information for the least energy, such as when processing visual signals.

“We wanted to know how these neuronal patterns change when the animal moves,” says Gant. The two researchers thus extended the efficient coding framework to moving observers. To this end, they filmed natural scenes akin to the visual environment of a mouse while standing still and walking, and analyzed footage from cameras mounted on freely moving mice in the laboratory of Cristopher Niell at the University of Oregon.

To analyze these videos they simulated neuronal processing in the visual cortex of mice and primates. “Based on the theory, we developed computational models of neurons that represent those inputs most efficiently while stationary and moving.”, says Gant.

The theory explains several effects that are known from experimental studies. “During movement, neural responses are stronger, their temporal dynamics change, there are weaker interactions between neighboring neurons, and neural code becomes more accurate,” says Młynarski.

“Starting from the same theoretical principle our approach can explain all of these effects ” adds Gant. The visual system adapts its processing to how sensory impressions change through the animal’s movement.

Why mice have stronger responses

The theory also explains why mice and primates differ. Mouse visual neurons respond to large, coarse patches of a scene, which movement affects strongly. By contrast, neurons in the primate fovea, the high-resolution center of the visual field, handle fine details. Such high-resolution visual signal already fluctuates rapidly at rest, and movement barely changes it.

“We thus predicted that foveal neurons should not be modulated by movement, while peripheral ones, which resemble mouse neurons, should be,” says Gant. Recently published experiments confirmed this prediction. “Consequently, the neurons of mice and monkeys do not behave differently because their brains work differently, but because they process different inputs.”

According to Wiktor Młynarski, the study also demonstrates the contribution theoretical neurobiology can make by not only describing but explaining the function of neurons: “Theories of neural computation can unify seemingly contradictory observations under common principles which, I believe, is one of the key roles of theory in natural sciences”.