Researchers at Tel Aviv University, led by Professor Yosi Yovel, have used artificial intelligence to analyze the vocalizations of Egyptian fruit bats and identify patterns in their calls. The study, conducted over 75 days in a cave in Israel housing more than 22,000 bats, recorded over 15,000 distinct vocalizations alongside video of the bats’ behavior.

The research team adapted a voice recognition algorithm, the same type of technology used in speech recognition software, and trained it on the recorded bat calls. By linking each sound to a specific observed behavior from the video footage, the AI learned to identify the context of the calls. It correctly identified the subject of a dispute, such as a disagreement over food or a sleeping position, with 71 percent accuracy.

The algorithm also demonstrated the ability to distinguish individual bats by their vocal signatures, identifying the specific bat making a call with over 60 percent accuracy. In many cases, the AI could also identify the bat to which the call was directed. This suggests the sounds were not random noise but targeted communications between specific individuals.

The findings indicate that bat vocalizations carry information about the sender, the receiver, and the reason for the interaction. The study provided evidence that bats have individual vocal identities and a communication system capable of managing complex social relationships.

Bat communication is part of a broader sensory toolkit. Bats use echolocation, emitting ultrasonic clicks and interpreting returning echoes to build a detailed acoustic map of their surroundings. This process allows them to navigate and hunt in complete darkness. The auditory cortex in a bat’s brain is substantially larger and more complex than in most other mammals, reflecting the importance of sound processing in their lives.

Young bats also learn to vocalize in a manner similar to human infants, babbling and imitating adult calls over several months. This vocal learning is a rare trait in the animal kingdom, found in only a small number of species including songbirds, whales, and humans. It suggests bat communication is at least partly a learned skill rather than purely instinctual.

Bats also exhibit complex social behaviors that require sophisticated communication. Vampire bats, for example, share food with roost mates that have previously shared with them, a system of reciprocity that requires long-term memory and recognition of individual bats within a large group.

The bat study serves as a proof of concept for a broader field of bioacoustics, in which AI is being applied to analyze animal communication across species. One notable project is the Cetacean Translation Initiative, which focuses on the vocalizations of sperm whales. Sperm whales communicate using rhythmic click patterns called codas, which can travel long distances through the ocean. Different whale clans have their own dialects, suggesting these are learned cultural traditions. Researchers are building a database of whale clicks with the goal of understanding the content of these communications.

Other research projects are applying similar techniques to different animals. Studies of prairie dogs have found that their alarm calls can include descriptive details, such as the color of a person’s clothing. AI has also been used to analyze bird songs and to determine the emotional state of pigs based on their vocalizations.

The bat study provides evidence that animal communication is more complex than previously assumed, suggesting the line between human and animal communication is less distinct than once believed. The findings do not, however, indicate that researchers can hold conversations with bats or fully understand their experience of the world. The AI has identified patterns and correlations between sounds and behaviors, but it has not revealed the internal experience behind those sounds.

The research represents a change in how scientists approach animal intelligence. It suggests that the sounds animals make may contain more information than previously recognized, and that the tools used to analyze human language can be applied to other species. The broader implication is that researchers are only beginning to understand the complexity of communication occurring across the animal kingdom.