You’ve heard of elephants such as Dumbo and Lumpy before, but these are just names given by people. However, recent research has made the astonishing discovery that these highly intelligent gentle giants call each other by name. These specific and unique vocalizations are, of course, not as articulate as human speech but rather distinct low rumbling sounds. Still, they’re names nonetheless.
These remarkable findings position elephants as the first non-human animals to use a form of address that doesn’t mimic the receiver’s call, a trait previously observed in dolphins and parrots.
The names of giants, revealed by AI
In northern Kenya’s Samburu ecosystem and southern Kenya’s Amboseli National Park, researchers led by behavioral ecologist Michael Pardo from Colorado State recorded over 600 elephant calls. You may be familiar with their iconic trumpet-like calls, but elephants also produce low-frequency noises between 1 to 20 Hertz, too low for the human ear to hear. However, these so-called infrasounds can travel over vast distances as large as 10 kilometers (6 miles).
The researchers then implemented a machine learning algorithm, which identified specific rumbles for 119 individual elephants, or close to 20% of cases. They separated rumbles from certain individuals by observing what elephants were separated from or approached the herd during these vocalizations.
Some of these rumbles were played back to 17 wild elephants. When they heard their name, they were more likely to move quickly toward the sound source and vocalize faster in response. These rumbles were remarkably consistent with the receiving elephant.
Elephants make different types of rumbles when they greet each other, encounter a predator, or want to play. But the ‘name’ calls are distinct. They’re not generic sounds either, such as those a mother might use to solicit the attention of their offspring.
The researchers also noted that calls to the same elephant by different callers were similar, hinting at a multi-layered communication system. According to the new study, “receivers of calls could be correctly identified from call structure statistically significantly better than chance.”
In a lecture available on YouTube, Pardo says his discoveries “blur the line” between “what we think is unique to human language versus what is found in other animal communication systems.” The new research was published in pre-print on bioRxiv and is waiting to clear peer review.