A new study led by the University of California, Berkeley’s Linguistics Department and Project CETI (Cetacean Translation Initiative) has found that sperm whales may communicate using sounds similar to human vowels. This research challenges previous assumptions about whale communication and suggests a complexity in their vocalizations that could be closer to human language.
“In the past, researchers thought of whale communication as a kind of morse code,” said Berkeley Linguistics Professor Gašper Beguš, who is the linguistics lead at Project CETI. “However, this paper shows that their calls are more like very, very slow vowels. This suggests a complexity that approaches human language.”
The study, titled “Vowel- and Diphthong-like Spectral Patterns in Sperm Whale Codas,” identified two main patterns—an ɑ-vowel and an i-vowel—as well as several diphthong-like combinations in sperm whale sounds. Beguš explained that these patterns appear to be exchanged between whales in what resembles dialogue.
“The whales’ production of the ɑ-vowel, i-vowel and diphthongs is likely controlled,” Beguš said. “This is true across almost all whales. We don’t understand the meaning yet, but we know that whales produce these sounds intentionally and we know that they differentiate between them.”
Project CETI brings together experts from fields such as artificial intelligence, marine biology, cryptography, robotics and underwater acoustics to study how sperm whales communicate. The research team observed similarities between the acoustic properties of sperm whale calls and those found in human speech. In both cases, vowel characteristics such as length, timing, frequency and trajectory can carry meaning.
“The spectral properties we discovered are very similar to human vowels. They correspond so closely that we can use human letters to describe them,” Beguš said. “Even the production of those sounds, which mirrors human vocal tract pulses, is similar to humans.”
Beguš noted that these findings not only advance efforts to decode sperm whale communication but also raise broader questions about animal cognition and rights.
“We’re thinking deeply about what finding these human-like structures means for the legal rights of animals,” said Beguš. “This paper prompts questions like, for example, what is language? Is there anything uniquely human about language, or is it just a continuum? What does that mean for the law?”
Project CETI’s broader goal is to translate sperm whale communication by observing their behaviors with technology such as tags, buoys and drones over five years of research efforts.
To analyze whale calls’ acoustic features more effectively than before—when most studies focused on clicks—Beguš used generative adversarial networks (GANs), a type of machine learning model capable of identifying patterns within large datasets by mimicking learning processes similar to those seen in children acquiring language.
“GANs can discover words and meaningful structure. When designing the model, we asked whether they could do that in whales as well,” Beguš said. “We still need human researchers to analyze the details, but they help us look in a specific direction.”
This approach enabled researchers to identify vowel-like elements and diphthongs among sperm whale vocalizations—a significant shift from earlier work focusing mainly on high-frequency clicks or their timing sequences.
“Before, researchers focused primarily on whale clicks and inter-click timing,” said Beguš. “Analyzing vowels adds a completely new dimension that brings much more complexity.”
Beguš emphasized how this interdisciplinary work could reshape both linguistics and our understanding of animal intelligence: “This work is so important because it helps you relativize your own position as a human,” he said. “We exchange inner worlds through speech, through vowels and consonants. This is a small step towards understanding the inner worlds of animals, their cultures and their intelligences.”
