The research team, led by postdoc Roy Harpaz, used VR (virtual reality) technology to project blobs of light that move like zebrafish into the water alongside real zebrafish. These VR “neighbors” can be precisely targeted to address questions about how zebrafish process visual information.
Because zebrafish are an extensively studied model organism, scientists have proposed several different hypotheses for how zebrafish process sensory information and make decisions about where to swim within a school.
“We used to assume that the fish know a lot about their neighbors—where they are and where they’re swimming to—and we saw that it’s probably a lot simpler,” says Harpaz. The most important visual cue was simply the total space that swimming neighbors took up in the images on the fish’s retina, which simplifies the calculations the fish brain must make to decide where to go next.
Harpaz says that these streamlined computations are more like the way we drive our cars and unlike the way computers navigate traffic. “Self-driving cars have computers, and the computers actually follow everything around you,” he explains. “The cars have cameras that are able to analyze every single pedestrian, every single car, every single thing that’s happening around you. When we drive, we probably don’t do that.” And neither do zebrafish.
Instead, the fish rely on cues such as how much space other swimmers take up in the visual field. “Instead of saying, I know where Nemo and Dory and Oscar are, I have the sum of the images they cast on my eyes,” Harpaz says.
The researchers showed the fish VR neighbors that varied in height, width, movement, and spacing, and they were surprised to find that changes along the blob’s vertical axis were more likely to change the fish’s behavior. “We could stretch the stimulus on the horizontal axis so big, and the fish would not even care about it,” Harpaz says. “But you start increasing it on the vertical dimension, and suddenly [the fish] will completely move away from it.”
He adds that this behavior might represent the way zebrafish estimate distances, because a neighbor’s vertical size (or height) is much less variable and does not depend on the neighbor’s orientation relative to the fish looking at the neighbor.
Based on these data from virtual reality experiments, the Engert Lab team created virtual simulations of zebrafish schools that closely mimicked observed collective behavior in live fish. The accuracy of the models in capturing real fish behavior led the researchers to think that zebrafish brains primarily use these simple visual cues to guide schooling behavior.
These observed responses to the retinal image of neighbors also provide important clues about where the neural circuits governing this cognition would be in zebrafish brains. Harpaz says that finding these neural circuits is the next step for his research.
The Nature Communications study also ties in with another recent Engert Lab paper that was published in Science Advances. In the Science Advances study, the Engert Lab collaborated with Mark Fishman of HSCRB to investigate how mutations that affect social behavior in humans affect schooling behavior in zebrafish. In these results, “visual occupancy”—or the amount of space taken up by neighbors—was also an important cue. along with motion cues, and it was affected by the specific single gene mutations.
“When you put these mutant fish together, some of these mutations make them more social so they huddle together,” Harpaz says. “And some mutations make them less social so they swim farther apart.”
He adds that these studies were only possible through strong collaborations with other Engert Lab members, such as Minh Nguyet Nguyen, who was the second author on the Nature Communications paper. He also thanks then MCB postdoc and current professor at the University of Konstanz Armin Bahl and then HSCRB postdoc Ariel Aspiras who were key authors on the Science Advances study.
Harpaz says, “The long term goal of these projects is to understand how brains in general solve complex problems and find the heuristics they utilize to do so. The heuristics or ‘hacks’ we found in zebrafish are probably also utilized by other species that need to solve similar complex problems.”
“This story is a beautiful example of a general strategy in the Engert lab, where we try to explain seemingly complex and high level behaviors, such as social interactions and schooling, as an emerging phenotype that is fully explained by sequential execution of simple behavioral motifs or ‘primitives,” says MCB faculty and senior co-author on both papers Florian Engert. “The reason why we believe that this approach works is that it is in line with evolutionary theory, where simple behaviors have evolved first; and the emergence of more complex traits usually happens based on the successful implementations of simple reflexes that work well.”