Archerfish live in streams, rivers and estuaries where they inhabit mostly the surface shallows. They are renowned for their ability to “shoot down” insect prey by spitting beads of water out of the mouth with incredible accuracy. In this way, they target insects sitting on vegetation on the bank, as well as airborne insects that fly overhead.
There is no parental care in archerfish and so the juveniles have to learn how to aim and shoot their jet-propelled water bullets without adult guidance. They also have to learn to distinguish between what is a potentially juicy insect meal from a worthless piece of vegetation.
“Object recognition is critical for animal survival,” said Ronen Segev from Ben-Gurion University of the Negev, referring to the skill that these fish exhibit in respect of distinguishing dinner. But exactly how do these remarkable fish differentiate a tasty treat from an unpalatable plant? “It’s a complex task,” said Segev, who recently published a research paper on this topic. He and his colleagues were interested in establishing which visual features archerfish depend on when identifying an insect for dinner.
Initially, the researchers wished to find out whether archerfish could recognize individual objects. Fortunately, the fish can be trained to squirt their water jets at images on a computer screen, in return for a food pellet. Segev and Svetlana Volotsky first trained the sniper fish to aim at images of one specific spider while at the same time trying to distract them with an image of a piece of vegetation.
Volotsky then presented the fish with images of the same spider, but taken from unfamiliar angles. If the fish had really learned to recognize the spider, they should be able to identify it from a completely new and unexpected view, even when presented simultaneously with an image of a different spider or a piece of vegetation. Indeed, the fish selected the initial spider for a target, irrespective of the novel view they were presented with. They could clearly recognize individual objects. But would they recognize that an insect they had never before set eyes on was also a potential prey item?
In the next set of experiments, Volotsky trained the fish to spit at images of a variety of insects, ranging from ants and beetles to flies and spiders, while still trying to distract them with images of plants. Once the fish had learned to aim their water jets at the insects, she tried presenting them with images of unfamiliar insects, to see whether the fish could recognize that any old insect is an animal, and not a plant. Astoundingly, the fish still targeted the insect images, even though they had never seen those creatures before.
“Archerfish can generalize from examples to make object recognition of natural object classes,” explained Segev.
The researchers, along with colleagues Ohad Ben-Shahar and Opher Donchin, then tested what visual cues the fish were actually using to enable them to distinguish plants from animals. They broke down images of insects, flowers and leaves into 18 different component features and built a computer program (known as a support-vector machine), that can learn to classify different types of information. Using this program, they mimicked the decision-making process of the fish to find out which features are necessary for archerfish to distinguish between animals and plants.
Amazingly, the fish only needed six essential features to distinguish animals from plants. These included the loose perimeter encircling the object, how jagged or smooth the shape was, and the texture of its surface. The first two factors were the most important in identifying the object correctly as an animal.
These findings, published today in the Journal of Experimental Biology, are even more remarkable because humans use a similar, although more elaborate, strategy to recognize each other’s faces.
The research team hopes to apply the lessons they have learned from archerfish to understanding visual recognition in other animals and to design man-made object recognition systems.
By Alison Bosman, Earth.com Staff Writer