Elephants may call each other by name. Scientists still disagree.

Who recorded what, and where

Researchers followed families of female elephants and their young in two parts of Kenya: Amboseli National Park (recordings from 1986 to 1990 and from 1997 to 2006) and the Samburu and Buffalo Springs reserves (2019 to 2022). They recorded from a vehicle, and they knew every elephant by the shape of its ears.

From these recordings they kept 469 rumbles in which the receiver was most certain: calls from 101 different elephants to 117 different elephants. There were no adult males among the callers. A rumble is the elephant’s main call, a low sound made in most situations.

Three kinds of call

• Contact callTo an elephant out of sight, more than about 50 metres away. The one it was meant for is the only adult who was separated, or the only one who answered.
• GreetingWhen one elephant walks up to another, to within touching distance.
• CaregivingWhen a mother or another older female comforts, nurses or wakes a calf.

Could a computer tell who each call was for?

The researchers gave a computer measurements of each call’s sound and asked it to find who the call was meant for. Trained on shuffled measurements, which carry no information, it was right for about 8 calls in 100. On the real ones, for the 437 calls whose caller was also certain, it was right for 27 or 28 in 100.

That is far better than chance, and still wrong for most calls. The authors think that a name, if there is one, makes up only a small part of a rumble, and that many calls carry none.

Which calls gave the computer a signal?

Take the 327 calls the computer got consistently right or consistently wrong, after dropping calls with an uncertain caller or context and receivers heard only once. It found the receiver in 42% of contact calls and 47% of caregiving calls, but in only 4% of greetings.

The authors wonder, with a “perhaps”, whether mothers put a label in caregiving rumbles “to help calves learn the labels with which others address them or because hearing the label is comforting for calves.”

Calls by adult females were matched far more often (33% of 274 calls) than calls by juveniles (4% of 53 calls). The authors suggest this may be because the behaviour “takes years to develop”.

Is it just copying a voice?

Dolphins and some parrots address each other by copying the other’s own signature call. If elephants did that, a call to one elephant would sound like that elephant. Human names are not like that: nobody’s name is an imitation of them.

In 236 calls the researchers could compare a call with the voice of the elephant it was meant for. In 141 of them (59.7%) the call was less like the receiver than that caller’s other calls usually were; in 95 it was more like. The computer could find the receiver in both groups, so the authors conclude that copying is probably not how it works, and add that a firm answer needs every caller’s whole repertoire recorded.

One more limit. Different elephants calling the same receiver did not use an obvious common sound: across different callers the computer found the receiver for 1.1% of calls, about what chance gives (1.4%). Yet calls from different callers to the same receiver were slightly more alike than calls to different receivers. The authors call this “mixed evidence”.

Do elephants notice a call meant for them?

In Samburu the team placed a loudspeaker, camouflaged or in the research car, about 49 metres from an elephant (between 40 and 59) and played a recording. Seventeen elephants took part: fifteen adult females, a nine-year-old female and a young male. Most heard two calls from the same caller, at least a week apart: one that had been addressed to that elephant, and one addressed to someone else. (One elephant heard only the first kind, one only the second, and one heard two pairs.)

After the call meant for them, the elephants approached the speaker sooner, called back sooner and called more often. All three differences were statistically significant (P = 0.009, 0.013 and 0.006).

The authors point out that they could not be sure every played call contained a label, which makes the difference more notable. In 2025 they fixed five data errors in this test: some figures changed, and the three differences above stayed statistically significant.

Why scientists disagree

In April 2026 Guha Dharmarajan, of Krea University in India, published a reanalysis of the same data in the journal Animal Behaviour. He concludes that the computer’s success comes from other patterns in the data, such as who was calling and when, and writes: “the use of name-like calls by elephants is inconclusive.” His abstract does not discuss the loudspeaker test.

Inferring “labels” or even “names” is “fanciful and the data do not permit such inferences.”

Robert Barton, anthropologist, Durham University

“I found no evidence that African elephants address one another with individually specific name-like calls.”

Kurt Hammerschmidt, German Primate Center

“I have no doubt that they’re addressing them with these, you know, unique labels.” Are they nicknames? Are they names? “Where do they come from?”

Karl Berg, University of Texas Rio Grande Valley

The study’s own lead author is careful too: “People might assume that elephant names work in exactly the same way as human names, which is not necessarily true.” Nobody has found where in a rumble a name would sit.

Why would an elephant need a name?

Elephant families spread out and regroup through the day, so an elephant is often out of sight of the ones it is closest to. A call that reaches the right listener would help.

Elephant calls travel far. In Etosha, Namibia, elephants responded to recorded calls of several kinds played from 1.2 and 2.0 kilometres away, and the researchers estimated that the loudest calls would be audible to elephants at least 4 kilometres away.

And elephants know many voices. Researchers estimate that an adult female in Amboseli would need to know the contact calls of about 14 families, around 100 adult females.

Somewhere in Kenya this evening an elephant is calling to someone across the grass. We do not yet know what the name sounds like, or whether it is a name.