The throaty roar of an African lion has long inspired wonder or struck fear, even for those who have never seen a wild lion. In the 1980s, MGM Studios finally trademarked the majestic lion’s roar that preceded its movies since the company’s founding in 1924. But the lion’s roar we’re all familiar with is only one of its call types, according to a study published today in Ecology and Evolution. Using machine learning models, researchers have classified three other calls, including a newly named “intermediary” roar.
In 2023, researchers deployed recording devices in Nyerere Park, Tanzania, to capture video and audio data from a resident population of African lions (Panthera leo) there. Now, using the 2023 recordings, 1,216 male lion vocalizations were classified into categories with the help of a machine learning model. In addition to the “moans” and “grunts” typical of the big cat species, the scientists learned that a roar previously thought to be a single type actually consists of two types. The new intermediary roar is lower in max frequency and shorter-lasting than the iconic full-throated roar.
When the researchers applied the AI model trained on the Nyerere Park lion sounds to 1,733 male vocalizations from another population of lions in the Bubye Valley Conservancy, Zimbabwe, it classified lion call types at an accuracy of 95.4 percent. The model was also better at identifying individual lions from their sounds than human observers.
Still, researchers had to lend the AI a hand. They report that they needed to give the machine learning model assistance for it to achieve high classification accuracy. This help included determining start-stop points of roars and eliminating the moans that preceded roaring bouts.
It’s particularly important to recognize individual lion roars, since African lions are ranked “Vulnerable” on the International Union for the Conservation of Nature’s Red List. Their roars can serve as signatures that allow researchers to track individual lions and estimate population sizes. “Until now, identifying these roars relied heavily on expert judgment, introducing potential human bias,” explained lead author Jonathan Growcott, a doctoral student at the University of Exeter, in a statement.
Read more: “How to Tell a Leopard From Its Roar”
The bioacoustics techniques pioneered in this study set the stage for better accuracy in monitoring not just lion populations, but other African wildlife of concern for conservation, such as leopards and hyenas, which also produce loud, identifiable vocalizations.
In their paper, the study authors suggest that their methods are easy to understand and implement. Added Growcott, “We believe there needs to be a paradigm shift in wildlife monitoring and a large-scale change to using passive acoustic techniques,” added Growcott.
Lions may be joined by other species caught making previously unappreciated vocalizations if researchers can deploy such high-tech eavesdropping in more ecosystems. 
Enjoying Nautilus? Subscribe to our free newsletter.
Lead image: Courtesy of Growcott, J., et al. Ecology and Evolution (2025).
