A new study finds that lions from different African regions roar in distinct “dialects,” with machine learning revealing clear geographic differences—including one Botswanan lion whose unusual roar patterns confused AI classifiers—raising important implications for conservation monitoring.
A new study published in Ecology and Evolution shows that African lions from Tanzania and Zimbabwe produce roars with noticeably different frequencies and durations, suggesting they develop regional “accents” shaped by environment, social learning, and possibly genetics. Researchers identified two types of lion roars—full-throated roars, which are “individually unique,” and newly named intermediary roars—using machine learning to classify calls with up to 87% accuracy.
One Botswanan male’s roars were so atypical that algorithms misclassified 83% of them, echoing the study’s description of his vocalizations as akin to “a thick New York accent in Boston.” While the findings show promise for using acoustics to identify individual lions for conservation without cameras or capture, they also reveal challenges, as regional variation may cause automated systems to overlook nomadic lions with unfamiliar roar characteristics.