How does a tiny fish swim upstream without being swept away?
A new study from Duke University School of Medicine and the Swiss Federal Institute of Technology in Lausanne (EPFL) reveals the brain-body teamwork and efficient wiring that makes it possible.
The findings also challenge the way we study the brain.
Using larval or “baby” zebrafish, researchers led by Duke neurobiologist Eva A. Naumann, PhD, tracked brain activity and behavior as the fish reacted to visual cues like those seen while drifting in a river.
Working with Auke Ijspeert, PhD, head of the Biorobotics Laboratory at EPFL, first authors Xiangxiao Liu, PhD, of EPFL, and 2024 Duke neurobiology alum Matthew D. Loring, PhD, collaborated to build and test a virtual fish — a computer model that mimics both brain and body circuits that help the fish stay on course in moving water.
Then the team at EPFL built a physical robot, named ZBot, to further test their findings.
The study was published Oct. 15 in Science Robotics, which features ZBot swimming up Switzerland’s Chamberonne River on its cover.
The work revealed just how closely the brain and body work together to guide movement based on what we see.
A fish, a simulation, and a robot
The research focused on zebrafish, a tropical, freshwater fish whose transparent, 4-millimeter larvae allow scientists to observe brain activity in real time.
At Duke, Naumann, an assistant professor in the Department of Neurobiology, and colleagues use calcium imaging to monitor zebrafish brains as the animals perform the “optomotor response.” This involuntary visual tracking behavior helps them swim in place when water flows past.
Those real-life observations fed into the computer simulation called simZFish, which replicates the biomechanics of swimming, the physics of water, and the zebrafish’s neural network — from retina to motor neurons. Loring helped test and refine the simulation, comparing its predictions to actual zebrafish behavior and neural activity in the live brain.
But the virtual fish was just the beginning. Researchers brought it into the real world by building a robotic counterpart: ZBot, a machine 200 times larger than a real zebrafish, with camera eyes and a motorized tail. Powered by the same neural circuits as simZFish, it swam against the river’s current, relying entirely on visual cues to maintain position.
One of the study’s interesting findings came from testing how different parts of the simulation’s visual field influence behavior.
"What we’re discovering is how whole circuits in the brain work together to guide behavior. We’re also confirming that those neural circuits we are studying in the lab actually work in reality." — Eva Naumann, PhD
The team discovered that certain brain cells, called pretectal neurons, respond preferentially to motion in a specific area: the lower rear of the visual field. That’s where optic flow, or the movement of the riverbed relative to the fish, is most useful for swimming upstream.
“This is evolution solving an engineering problem,” said Naumann. “By connecting only the optimal part of the visual field to movement circuits, the brain maximizes useful information while minimizing unnecessary neural wiring.”
This selective input may be nature’s way of making the brain’s circuits more efficient, said Naumann, who has an appointment in the Duke University Department of Biomedical Engineering and is a member of the Duke Institute for Brain Sciences.
“What we’re discovering is how whole circuits in the brain work together to guide behavior. We’re also confirming that those neural circuits we are studying in the lab actually work in reality,” she said.
New direction for brain research
The three-step process — lab experiments, computer simulation, and robotic testing — showed that lab-studied neurocircuits can be translated into real-world behavior.
The work offers a clear message: understanding the brain requires understanding the body and the world around it.
“Traditional neuroscience studies the brain as if it exists in isolation,” Naumann said. “But brains evolved within bodies, moving through environments. Our work shows that embodiment fundamentally shapes neural architecture.”
The researchers believe this approach can apply beyond sight. They’re planning follow-up studies examining how other sensory systems — including touch and proprioception — integrate with vision to guide behavior.
Additional authors: Luca Zunino, Kaitlyn E. Fouke, François A. Longchamp, and Alexandre Bernardino.
Funding: The European Research Council, the National Institutes of Health, the Whitehall and Alfred P. Sloan Foundation.