Fish Learn To Use Tools, So Let's Rethink the Definition of Tool Use

Illustration for article titled Fish Learn To Use Tools, So Lets Rethink the Definition of Tool Use

When you think of Atlantic cod, you probably think of a strip of fish, battered and fried. But new research suggests that the fish might be able to use tools, and that might cause us to rethink how we evaluate tool use in other animals.


In a recent study, a group of researchers led by Sandie Millot designed a feeding machine that could be operated with a pull string. If a fish was hungry, it would swim up, pull the string, and the food would be dispensed. Since fish – like all other animals – are capable of associative learning through basic classical and operant conditioning, the fact that fish could learn to operate this sort of basic feeder is not surprising. Indeed, 48 of 56 fish figured out how to get the food. Each of the fish was marked with a small tag with a colored bead on it just in front of the dorsal fin. The tags helped the researchers to identify each individual fish.

Piscine Innovation

The surprising part is what happened next. Three of the fish figured out that they could use the artificial tags, rather than their mouths, to operate the feeder. They learned to swim past the string and hook it onto their tags so that the food would be released that way. Since the researchers caught everything on video, they could carefully analyze the surprising behavior.


For each of the fish, the moment of insight began by accident. The fish appeared to accidentally catch their tags on the feeder's pull-string. As soon as they felt the pull of the string, they showed what's called a startle reaction: they immediately responded with a fast burst of swimming until the tag became unhooked. Eventually, after a bit of trial and error and fine-tuning, all three fish were performing the action with apparent intention, using their dorsal tags alone rather than their mouths to operate the feeder.

Illustration for article titled Fish Learn To Use Tools, So Lets Rethink the Definition of Tool Use

"The cod first adjusted its position by slow swimming movements, then with great precision attached the bead of the tag to the trigger pulley and finally swam forwards before turning to release itself," the researchers write. "At the end of the experiment, the three focal fish showed standardized swimming patterns to catch the trigger with their tag, activate the feeder and reach quickly the feeding zone." And in general, they began to use the "tag technique" exclusively instead of the "mouth technique." One of the first, for example, used its tag to operate the feeder 422 times over the course of the month of testing, but used its mouth only 51 times.

The researchers say it "represents a novel and amazing behavior."

One common definition for innovation comes from philosopher Grant Ramsay: "Innovation is the process that generates in an individual a novel learned behavior that is not simply a consequence of social learning or environmental induction." The researchers argue that the behavior fits those criteria.

The cod did not just swim to a particular location or use another normal behaviour but carefully attached the tag by fine-scale coordinated movements and repeated this behaviour a high number of times. Innovations are not necessarily cognitively complex, and chance events can lead to innovations if the individual learns from the experience, and when it is rewarded, the innovative act is likely to be repeated with increased efficiency. What is striking with the present observation is that the "behavioural accident", i.e. getting the tag attached to the feeder pulley, was presumably initially aversive and totally artificial since cod are not known to use their back or fins to manipulate objects and that cod were yet able to fine-tune this action to a standardized technique to obtain a goal. That the fish had already activated the feeder repeatedly with the mouth before switching to the tag technique, and thus had a functional and non-aversive alternative to obtain food, makes the findings even more surprising.


Why was the "tag technique" preferred to the "mouth technique"? The researchers think that it allowed the fish to visually monitor the area where the food would be dispensed while simultaneously operating the feeder, something impossible using the mouth technique. That way they could be more likely to gobble up the food themselves, rather than allowing another fish to get a free snack. Clever fish indeed.

But is it tool use?

One commonly accepted definition for tool use is "the employment of an environmental object to alter the form, position, or condition of another object, another organism, of the user itself, when the user holds or carries the tool during or just prior to use."


The problem is fish can't hold tools. They don't have limbs to grasp or to orient tools, and being surrounded by water constrains them even further.

Still, Atlantic cod do not have any natural anatomical structures that function as hooks. Therefore, using an object that has been attached to their body ought be considered as significantly outside the range of their normal behavioral repertoire as it is when a dolphin uses a sponge to find food on the seafloor.


If this argument doesn't convince you, then it's worth considering that the most widely used definition of tool use may inherently preclude fish. Which would reflect a shortcoming of the definition, not a failure of the fish.

To that end, the researchers argue that whether the action qualifies as tool use should hinge on cognition rather than behavior. "The question is not how an animal uses a tool or an extension of the body but why." The why, in this case, is to obtain a desired goal.


"If we accept that the cod were aware of the morphological extension (the tag) on the back," they reason – and how could we not, given how their behavior become more fine-tuned over time? – then "using this to activate a feeder could be regarded as a form of tool use." The researchers even go so far as to suggest that the tag could be thought of as a sort of artificial limb!

The surprisingly sophisticated development in fish of this goal-directed behavior suggests that the usual definition may be insufficient to capture the full range of tool use in all animals. Instead, Millot and colleagues offer a more nuanced characterization of tool use, which relies upon understanding the mental machinery that underlies the innovative behavior, rather than simply describing the innovation itself.


It is said that after Jane Goodall phoned Louis Leakey more than fifty years ago with news of chimpanzees' tool use, he wrote, "We must now redefine man, redefine tool, or accept chimpanzees as human." It seems as if we are still redefining what can be considered a tool and in so doing, rethinking our uniqueness on the tree of life.

[Animal Cognition]

Header image: Hans Hillewaert/Wikimedia Commons/CC-BY-SA-3.0


Share This Story

Get our newsletter


My issue with calling this tool use is that the fish made no concious decision to attach the tag. Think of this hypothetical example: say a cod had a spontaneous mutation that grew a long spine on it's back (unusual for cod because the dorsal fins have soft rays, but many other fish already have hard spines in their dorsal fins, so maybe imagine a rockfish instead). It then, with this mutation that is clearly outside the normal range, figures out the same behavior. Is that tool use? Most people would say no, and I would agree. Because humans implanted the tag and the fish has no control over whether the tag is there or not, this is just learning a novel behavior, not tool use. I am impressed anyways because fish are, by and large, not very bright (fish ecologist here), so even learning the behavior is impressive. But I think it's a far cry from tool use.

The dolphin example you stated is much less ambiguous. The dolphins are actively picking something up from their environment and using it for a specific task, then discarding it.