Looking at an archerfish, it’s possible to deduce a few key features about its biology. Its pale, mottled color mimics the dappling of sunlight on water, suggestive of disguise, and its eyes and mouth both point skyward, divulging a predatory hunting strategy. Sure enough, just as their name implies, archerfish are skilled hunters. In addition to targeting aquatic prey in murky water, archerfish can also shape their mouth just so to shoot a jet of liquid capable of dislodging insects and even small lizards from overhanging leaves meters above the water. Because archerfish are able to accurately estimate a prey’s distance from below the water’s surface, the jet hits its peak velocity as it strikes its target—delivering a blow six times more powerful than one levied by an average vertebrate muscle.
Fortunately for researchers, archerfish (family Toxotidae) readily spit at targets in a lab in exchange for snacks, says University of Oxford visual ecologist Cait Newport, who adds that the fish are “very hard workers . . . and generally not too shy.” These features make them a powerful model for studying visual cognition in fish, a group that diverged from its most recent common ancestor with humans roughly 420 million years ago. Even though they lack a cortex, the part of the mammalian brain that processes visual stimuli, archerfish can still perform many of the same behavioral tasks as mammals, making direct comparisons between the two groups possible. “Anything that you can place in the form of a multiple-choice question, you can actually study with the archerfish,” says Ronen Segev, a neuroscientist at Ben-Gurion University of the Negev in Israel.
In 2016, Newport demonstrated that the archerfish Toxotes chatareus can discriminate between human faces. It was ostensibly an odd thing to study, but faces, Newport says, were a good way to test the fishes’ ability to detect novelty in objects that “the animal has not evolved to learn.” Newport and her colleagues used a set of images designed by researchers at the Max Planck Institute for Biological Cybernetics to train fish to spit at a single face displayed among a series of novel ones; correct choices elicited a food reward. Then, the team began adding more faces and standardizing their shape, color, and brightness. Even still, the fish were able to identify the familiar face from as many as 44 distinct pictures. In a follow-up study, the researchers expanded their work to show that archerfish can also recognize the same face from several different angles.
Fortunately for researchers, archerfish readily spit at targets in a lab in exchange for snacks.
These findings, Newport notes, argue against a simplistic image-matching process where fish compare new images with an internal catalog. Instead, they suggest that archerfish may be using a complex mechanism more akin to mammalian visual processes. Object recognition is “a very complicated task, and humans have a portion of our brain that’s dedicated just to recognizing human faces,” Newport says. Many fish, including archerfish, have brains that are relatively small for their body size compared to mammals, she adds, “but we’re seeing them do these amazing processing tasks.”
Segev and Svetlana Volotsky, a PhD student working in his lab, recently took this idea one step further, opting for ecologically relevant pictures of ants, spiders, and plants. They found that T. chatareus could pick out individual insects they had been trained to spit at in return for food. The study also showed that the fish can recognize novel insects as prey without additional training, even from different angles. To do this, Segev says, the archerfish likely has an internal set of rules it uses to classify new things. “You don’t memorize the fact that my face is in the middle of the screen; you learn the rules that actually describe my face.”
Researchers trained archerfish to shoot at targets on a screen in return for food.
To determine just what those rules entail, Volotsky broke down each image into its parts—its shape, texture, curvature, and symmetry, among others. Using a machine learning algorithm, the group identified five features, mostly related to the object’s shape, that appeared to be important for a fish to recognize an object as food. Volotsky next tested these features individually with the fish in pared-down pictures. The fish saw just the rough shape of a leaf and a spider, for example, or just their visual texture. Their choices about whether or not to spit reinforced the model’s conclusions: “You can remove all texture and the fish can still identify and classify objects, but they cannot do that if you keep [only] the texture,” Volotsky says.
Several researchers now want to pick apart the visual system in freely swimming archerfish to determine how the brain processes visual input in real time. The challenge is a technological one—developing tools that are small enough to be worn by the fish and that are able to function in water. Segev and his colleagues have made a device capable of measuring neuronal activity in freely swimming goldfish, and Volotsky is developing a separate technique that she says can record individual brain cells in tethered archerfish to determine which neurons are activated when the team shows the fish images of various shapes. Trevor Hamilton, a behavioral neuroscientist at MacEwan University in Canada who studies visual cognition in zebrafish, says there may be tools from his field worth borrowing, including optogenetic approaches and portable electroencephalographs designed using fish species that are much smaller than archerfish.
Object discrimination isn’t the only surprising thing that archerfish seem to be able to do without a cortex, and archerfish as a group remain a fruitful source of research ideas, says Shai Gabay, an evolutionary neuroscientist at the University of Haifa in Israel. Gabay’s lab uses archerfish to complement his research on the human brain and to challenge what he calls a “cortico-centric bias” in the neuroscience literature. His team has argued that archerfish respond to stimuli in a volitional manner—that is, by processing and acting on information in a way that goes beyond purely reflexive reactions—and preliminary work suggests that the animals engage in prosocial behaviors such as food sharing. In findings recently presented at a conference by a member of Gabay’s lab, the team found that an archerfish will consistently choose a target that rewards both itself and a passive tankmate with food, so long as the chooser receives at least as much as its neighbor. While it was thought that a highly evolved brain that includes a cortex was needed for social behaviors, “now we can see very complex social behaviors even in fish,” Gabay tells The Scientist, suggesting that alternative mechanisms can lead to sociality.