Fast mapping, the ability to rapidly learn an association between two things after very little exposure, is a key tool responsible for the vast repertoire of human language. It’s the reason we can recognize voices from another room and why newborns prefer to listen to their mothers read them The Cat in the Hat over other women. While fast mapping is generally thought of as a human ability, zebra finches can also distinguish dozens of other finches’ vocalizations, doing so with very little exposure and retaining those memories for at least a month, researchers report today (November 13) in Science Advances.
“The reason that this study is groundbreaking is because zebra finches are the very first vocally learning species aside from ourselves with any evidence that fast mapping takes place,” Samantha Carouso-Peck, a behavioral neuroscientist at Cornell University who researches social influences on vocal learning in zebra finches but was not involved in the current work, tells The Scientist.
Prior to the new study, evidence of fast mapping in nonhumans had been suggested for only a single nonhuman species: dogs. Rico the border collie is capable of distinguishing the name of more than 200 individual objects, while another dog named Chaser could discriminate between more than 1,000. Whether this is truly evidence of fast mapping, however, has been debated among scientists who believe that only species capable of language, specifically humans, can be said to truly fast map.
To test zebra finches’ ability to recognize individual calls, and the number of exposures they need to do so, researchers at the University of California, Berkeley, designed a five-day “learning ladder.” Birds were initially trained on a small set of call data that became increasingly larger over several days.
On the first day of testing, each bird in the training arena went through a series of trials during which they heard brief clips of either a song or a distance call used by birds to locate their group when it is out of sight. These two distinct vocalizations were chosen because they are thought to have the most variation between individuals. Nineteen birds were trained on songs, while another 19 heard only distance calls. The vocalizations were parsed, so that one set led to a food reward, while the other set didn’t.
In that first test, each finch only needed to discriminate between two individuals’ vocalizations, one that provided a food reward and one that did not. By pecking a key, the finch would start the sound file. It could then either peck the key again to start a new trial or listen to the entire clip to receive a food reward. Repeating this process over several trials, the finches learned to associate vocalizations that led to a reward from the ones that did not and to quickly skip vocalizations that weren’t rewarding.
Each successive day, sounds from additional individuals were added to the training set. While finches were originally trained to discriminate between one pair of individuals, by day two it jumped to four pairs, and then eight on days three through five, for a total of 16 different individuals’ vocalizations. The researchers began evaluating the finches’ performance only on days four and five, when each bird had been exposed at least once to each possible vocalizer.
To be certain the finches weren’t simply memorizing the sound files, each bird providing the vocalization was recorded 10 different times. “We were pulling randomly from a large library of calls that we have, so the calls were all different every time,” Frederic Theunissen, an auditory scientist at the University of California, Berkeley, tells The Scientist. “Not only are they different vocalizers, but they heard different renditions from the same vocalizer.”
Even with this added layer of complexity, Theunissen says he was “really impressed” by how well the finches performed. All 19 finches trained using song files were able to discriminate between the vocalizers they recognized based on their individual signatures and those they did not, and the same was true for 18 of the finches trained using distance calls.
Beyond their ability to recognize other birds based on their calls, the team was also struck by how quickly the finches were able make the distinctions. The majority of birds could hear as few as 10 exposures to an individual and memorize their signature, while some required fewer than five, strong evidence for fast mapping in the species.
“We didn’t expect the birds to be so good at this,” Theunissen tells The Scientist. “We thought this was going to be interesting, but the fact that the number of vocalizers is so high, and that they can do this so quickly, was really quite astonishing.”
In the study, the team decided to create what Theunissen calls “an impossible task” to test the limits of the finches’ performance. He created a new set of stimuli that included both songs and distance calls from 56 distinct vocalizers. Only four randomly-selected birds went through this mega-task, but on average, they were able to remember 42 individual vocalizers.
A month later, he tested the recall of two of the four birds by running them through the same, 56-vocalizer test and found that they were still able to distinguish between individuals—they had retained their memories of the individual calls. Both the speed of their learning and their retention over time, Theunissen says, “point towards these really remarkable abilities that these birds have in terms of making auditory memories.”
Because these birds are so social, living gregariously in colonies of more than 100 individuals, it isn’t unexpected that zebra finches could discriminate between dozens of individuals, says Sarah Woolley, a neuroethologist at McGill University who studies the neurological basis for social behavior in songbirds. “We’ve always been pretty sure that birds could do that, zebra finches included,” Woolley says. “But no one had really demonstrated it, certainly not to the degree that they do in this paper.”
While the results are “very interesting evidence” that zebra finches are capable of fast mapping, Carouso-Peck says that more work on a larger sampling of birds is needed “before we can make weightier claims about what they’re capable of.” For example, one of the criticisms of the previous work with Rico the border collie is that fast mapping is a tool for learning language, and therefore only species capable of language truly do it. While Carouso-Peck says she doesn’t necessarily agree, as a next step she would like to see if finches are also capable of fast mapping verbally. For example, how many times must young zebra finches hear their father’s song before they can reproduce it for themselves?
“Language is something which we humans tend to hold up as our great crowning achievement, something that we are capable of which absolutely no [other] species is capable of,” Caruso-Peck tells The Scientist. “Over the last few decades, we’ve had to change the goalposts for what is language in order to maintain that high and mighty position.”
K. Yu et al., “High-capacity auditory memory for vocal communication in a social songbird,” Science Advances, doi:10.1126/sciadv.abe0440, 2020.
Correction (November 17): An earlier version of this story stated that fast mapping had only been shown in one dog, but two border collies have been studied by researchers. The Scientist regrets this error.