During a lab meeting about two years ago, Logan Grosenick presented research describing how fish determine the pecking order within their group — a process that depends not only on the physical attributes of individuals but also on their social interactions. The presentation inspired a lengthy discussion between Grosenick and Tricia Clement, another member of his lab at Stanford University in California. The two, along with their supervisor Russell Fernald, devised an experiment to test whether fish can figure out the social rank of other fish simply by watching how they interact. “We sat in front of a whiteboard and sketched the essentials of the project right then and there,” recalls Grosenick.

The experiment used a large, square tank partitioned into several compartments and the extremely territorial native African fish Astatotilapia burtoni (see page 429). The central unit of the tank contained the designated 'bystander' fish, whose role was to observe. Five surrounding units each contained size-matched male fish — fish A to E — designated as 'rivals'. The researchers could lift an opaque barrier between any two compartments, so that the fish, while still physically separated, could see one another.

With the opaque barriers down, Grosenick moved a fish from one unit into another fish's unit. Because the fish are so territorial, the resident fish in the unit, fish A, would attack the intruder, fish B. Grosenick then removed the barrier, making it possible for the bystander fish to watch the fight. The resident fish was always victorious. After seven minutes, the barrier would be replaced, and Grosenick would remove the defeated fish, returning him to his unit, and then stage a second fight.

This 'training' went on for 11 days, with the bystander fish always seeing fish A defeat fish B, B defeat C, C defeat D and D defeat E. “We were able to create an artificial dominance hierarchy for the bystander fish,” says Grosenick. The observations were time-consuming and demanded a lot of patience. “If anything startled the fish I would have to wait half an hour to let things settle down and then start again,” adds Grosenick.

At the end of the training period, the researchers tested whether the bystander fish could infer from what he had witnessed that B was dominant over D, even though the two had never actually fought one another. To do so, Grosenick lifted the opaque dividers between B, D and the bystander. He then timed how long the bystander spent in the vicinity of B compared with D. In most cases, the bystander spent more time near the 'weaker' fish, indicating that he knew to keep his distance from the dominant one. “We had expected a difference, but the magnitude of the difference was surprising,” says Grosenick. The researchers were also surprised that the fish could recognize each other on the basis of appearance alone, because they are not very distinctive-looking.

The ability to make this sort of logical connection is known as transitive inference, and is described as a developmental milestone in children by renowned psychologist Jean Piaget. Transitive inference had previously been demonstrated in primates, rats and birds, but never in fish. “There has been a long debate about whether fish can reason in this way,” says Grosenick, who is completing a master's in statistics. “This is compelling evidence that they can.” He plans to extend the results using a more 'natural' environment. “I have been talking to an electrical engineer about tagging fish so that we can follow their behaviour while they move around a tank unconstrained,” he says.