Unveiling the Secrets Behind the Lion’s Roar: Insights from AI

We recognize his cry among all the others. The lion's roar impresses, fascinates, and seems to speak aloud the strength of the king of animals. But this sound, which we thought we knew well, actually hides a greater complexity. With the help of artificial intelligence, researchers are now discovering that it says much more than we thought.

Artificial intelligence listening to big cats

For a long time, scientists considered the lion's roar as a simple signal of intimidation or contact between members of the same group. But recently, thanks to recordings made in African reserves, researchers from the University of Exeter used AI for the first time to automatically analyze these sounds, as reported by Eurekalert. Based on the study of more than a thousand roars, the algorithm identified nuances imperceptible to the human ear and, above all, revealed the existence of a second type of roar.

This vocalization, never previously identified, appears between the first low calls and the emblematic full-throated roar. Artificial intelligence was able to differentiate it with an accuracy of 95.4 percent, while reducing human errors. For specialists, this discovery marks a new stage in the use of bioacoustics, a discipline which is becoming central to observing the behavior of large predators.

What the complexity of the lion's roar reveals

Contrary to popular belief, the lion's roar is not only a manifestation of power, but a structured language whose codes are beginning to be revealed. According to a study published in Ecology and Evolution, there are actually two distinct types of roars. On the one hand, the full-throated roar, powerful and prolonged, serves to mark territory or to signal the presence of a group. On the other hand, the intermediate roar, shorter and less serious, seems to be emitted after the first calls, without having the same intensity.

By analyzing the frequency, duration and modulations of these vocalizations, researchers have shown that lions have unique sound signatures, comparable to voice prints. This individual dimension opens up new perspectives for recognizing a lion without seeing it. Scientists suggest that some variations could even reflect differences linked to the geographical origin, age or physical condition of the animal.

Towards more reliable acoustic monitoring of populations

With the discovery of two types of roars, associated with vocal signatures specific to each individual, eavesdropping becomes a reliable tool for monitoring wild lions. Researchers are now using autonomous audio sensors installed in African parks. They collect roars continuously, which machine learning models then analyze to identify lions and track their movements.

Using this method, researchers can estimate populations without disturbing the animals, which contrasts with camera traps or track surveys, which are often expensive and intrusive. Furthermore, it also makes it possible to identify isolated individuals, a crucial aspect to ensure their protection. Currently, the International Union for Conservation of Nature estimates the wild population at between twenty and twenty-five thousand lions. In response to this situation, researchers are opening a new way to protect this great symbol of African wildlife by letting lions signal their presence themselves.