Skip to content

ScienceIndependently confirmed2 publishers2 min readPublished

Amboseli elephants make their rumbles more alike as a chorus unfolds

Earth Species Project and Elephant Voices researchers used machine learning on 902 Amboseli recordings and found rumbles converging within choruses. The matching showed up in particular social situations, a hint that the calls track what a family is doing.

The Scientist · Science desk

How we use AISend a correction

Illustration accompanying Amboseli elephants make their rumbles more alike as a chorus unfolds
Generated illustration

What happened

  • The recordings span four decades and come from wild female African savannah elephants in dozens of family groups in Kenya's Amboseli National Park.
  • Calls grew more alike within a chorus when elephants were keeping in contact while separated, coordinating what to do next, or greeting after a separation.
  • Picking out one caller inside an overlapping chorus, which researchers call the cocktail party problem, had made such analysis very hard until recent machine-learning advances.
  • Poole's group reported in 2005 that elephants can learn to imitate one another's voices, and the new work extends that imitation to rumbles.

Why it matters

  • capability Software that can follow one elephant through an overlapping chorus opens a common and previously hard-to-study kind of recording to analysis at the level of individual animals.
  • constraint The team's proposed conservation use, tracking behaviour and movements remotely, needs the context signal to classify calls reliably in the field. The published account does not yet show it is that strong.
  • precedent Old field recordings of other species become worth reanalysing with the same tools, since this result came out of a long-running archive once overlapping callers could be separated.

The study follows individual voices through a single chorus, and that is a good design for this question. The same animals, in the same situation, are compared as the chorus goes on. A drift toward sounding alike is then hard to blame on differences between recordings made on different days [1]. The report does not say how many of the recordings were choruses, how they split across the six situations, how large the convergence was, or what baseline it was tested against [3].

The result gets interesting in how the situations split, and that part needs care. Preparing to move and comforting calves are not among the converging situations the report names [16]. If convergence simply went with group organisation, preparing to move would be an obvious place to find it. Coordinating what to do next is on the list and preparing to move is not [16]. The count is also uncertain. "Greeting after a separation" could match one or both of the two greeting categories, greeting close family members and joining large reunions, so convergence turned up in three or four of the six [15].

Sara Keen of the Earth Species Project [13] compared it to people. "It's sort of like what humans do in conversation, where you almost start to adopt the speaking patterns or accents of whoever you're interacting with," she said [6]. She also said convergence may help elephant families build a plan of action [8]. Joyce Poole of Elephant Voices [14] said the animals "appear able to negotiate, agree and disagree" [9].

I think the within-chorus matching is the firmer finding. The idea that choruses carry plans rests more on interpretation. Keen said the choruses vary in a way that suggests they encode what a family is doing together and what happens next [7]. The thing this doesn't tell you is whether the elephants use that information. A model can find the pattern because the acoustics line up with situations that people defined [3]. Whether a listening elephant gets the same message from a chorus is a separate experiment.

Rumbles, often too low for human ears, can travel up to 10 kilometres [10]. Andrew McLean of the IUCN's Asian Elephant Specialist Group called the work "a fascinating finding" and added that "it's not really surprising, given how intensely social elephants are" [12].

What to watch

  • Publication of per-situation chorus counts, effect sizes and the baseline the convergence was tested against.
  • Playback experiments testing whether elephants respond differently to choruses whose rumbles have converged.
  • Whether the separation method works on recordings from other elephant populations or from Asian elephants.

Clarity's read

What the record supports and how the coverage leans. The claims behind it follow.

Reality

Evidence62
Adoption
Insufficient
Hype gap+25
Incentives30
Confidence66
Why these scores

Claim ledger

Ranked by verification strength, evidence, and original report placement.

  1. [1]

    A machine learning study tracking rumbles of individual elephants in the same chorus found that in some cases the calls became more acoustically alike as the chorus progressed.

    ReportedSupportedSource: New Scientist report on the study by Keen, Poole and colleagues2 sources— create a free account to open themView cited source
  2. [2]

    The researchers used 902 recordings of rumbles collected across four decades from wild female African savannah elephants (Loxodonta africana) from dozens of family groups in Kenya's Amboseli National Park.

  3. [3]

    The calls were grouped into six social situations: staying in contact while separated; preparing to move; coordinating what to do next; comforting or reassuring calves; greeting close family members; and participating in large family reunions.

Sources

2 independent publishers whose own reporting we read for this story.

  1. newscientist.com

    1 article · October 1, 2026

    Elephants imitate each other when they produce deep rumbling sounds
  2. phys.org

    1 article · October 1, 2026

    Learning to speak elephant: Choruses of rumbles help groups coordinate

Share your take

Let Clarity write the post for you.

Signed-in readers get a short post drafted on this story in the register they choose — narrative, analytical, or a direct position — editable to the last word before it goes anywhere. The share buttons at the top of this story work without an account.

Topics and entities

Follow any of these and your For You feed starts watching them — no settings page required.

Entities

Loading related stories