In a breakthrough for wildlife bioacoustics, researchers from the University of Exeter have utilized artificial intelligence to uncover a hidden vocalization in African lions. The study, published in Ecology and Evolution in late 2025, reveals that lions possess a second, “intermediary” roar that had previously gone unnoticed by human experts.
The Discovery: Decoding the “Intermediary Roar”
For decades, biologists treated the middle of a lion’s roaring bout—which begins with moans and ends with grunts—as a single, undifferentiated sound. The Exeter team’s AI model proved otherwise.
Full-Throated Roar: The iconic, loudest part of the bout. It follows a clear acoustic arc, rising in pitch before trailing off.
Intermediary Roar: The newly identified sound is shorter, lower-pitched, and has a much flatter pitch profile than the full-throated version.
The “Fingerprint” Effect: While both roars are distinct, the AI found that the “full-throated” roar contains the most unique vocal signatures, acting much like a human fingerprint to distinguish one lion from another.
AI vs. Expert Intuition
The research involved analyzing over 3,000 vocalizations captured via remote recorders in Tanzania’s Nyerere National Park and acoustic collars in Zimbabwe.
| Method | Identification Accuracy | Key Limitation |
| Expert Human Judgment | High, but subjective | Prone to bias and fatigue; difficult to scale. |
| Exeter AI Model | 95.4% | Requires high-quality initial recordings. |
By automating the identification process, the AI removes human subjectivity and allows for Passive Acoustic Monitoring (PAM). This means conservationists can count individuals in a landscape simply by listening, rather than relying on more intrusive or difficult methods like camera traps or tracking “spoor” (footprints).
Conservation Impact: A Race Against Time
The timing of this discovery is critical. African lion populations are estimated to have dropped from roughly 45,000 in the early 2000s to fewer than 23,000 today.
Habitat Loss & Conflict: Lions have lost 94% of their historic range, and human-wildlife conflict remains a leading cause of death.
Precision Tracking: Identifying individuals (e.g., “Jeronimo” vs. “Sandra”) allows researchers to derive exact population estimates and identify “conservation hotspots” that require urgent on-the-ground protection.
Future Outlook: Lead researcher Jonathan Growcott suggests this marks a “paradigm shift” in monitoring, moving toward a future where bioacoustics can protect various threatened species across the continent.
“Lion roars are unique signatures. Until now, we relied on expert judgment, but AI promises a more accurate, less subjective way to monitor these dwindling populations.” — Jonathan Growcott, University of Exeter
