My favorite story I wrote in 2019 was about a research study that taught rats to drive, an activity that the rats appeared to enjoy. Today, we have another tale of lab animals getting behind the wheel, but this time, the motorists in question weren’t mammals—they were goldfish that learned how to drive a fish-operated vehicle, or “FOV,” in a terrestrial environment.
What’s the point of this experiment? In the driving-rat study from 2019, the researchers were trying to look at environmental stress, and driving is an activity that turned out to reduce stress levels in the rats. This study, conducted by Shachar Givon and colleagues at Ben-Gurion University of the Negev in Israel and published in Behavioral Brain Research, aimed to discover something a little different.
Specifically, the idea was to see if the fishes’ navigation skills are universal and work in extremely unfamiliar environments, a concept known as domain transfer methodology. And you have to admit that driving a tank inside an enclosure in a research lab is a pretty unfamiliar environment for a goldfish.
Nice fish tank—how fast does it go?
The FOV is, in essence, a fish tank on wheels. Unlike the rat-mobile, there are no physical controls for the fish to learn to use. Instead, a downward-looking camera tracks the fish’s position in the tank. If the fish is near one of the tank walls and is facing outward, the fish-control algorithm (which runs on an onboard Raspberry Pi 3B+) will move the FOV in that direction. A lidar sensor on the same mast overrides the fish control algorithm if the FOV comes within 20 cm of the walls of the terrestrial environment, a 4×3 meter enclosure.
Six fish took part in the experiment and were taught to drive in 30-minute sessions, conducted three times a week, every two days. To begin with, the fish were given a reward if they navigated to a target (a pink corrugated board) from the center of the enclosure.
The fish did indeed learn to drive the FOV, becoming more proficient in its operation over time both in terms of achieving tasks (navigating to a target) and in the time taken to complete that task. The authors wanted to check that the fish weren’t just memorizing a set of movements to earn a reward, so they repeated some tests with a new starting position inside the enclosure or added decoy targets that were a different color.
Not only were the goldfish able to cope with those changes, but they were also able to approach targets from different angles, suggesting that the animals have an internal representation of the world around them. And that’s despite the fact that the interface between the air, the tank’s plexiglass walls, and the water would create nonlinear refractions that presumably look quite different in fish-vision than the environment in which those fish eyes evolved.
Perhaps goldfish aren’t quite as forgetful as the jokes would have us believe.