Sperm whale calls show vowel-like structure, UC Berkeley and Project CETI find — a step toward decoding a deep-sea “language”
A UC Berkeley and Project CETI study finds sperm whale vocalizations contain steady, vowel-like acoustic shapes, suggesting more complex, structured communication than simple clicks — a major step toward decoding deep-sea ‘language’ using AI and linguistic analysis.
Key takeaways
- UC Berkeley study and Project CETI found whale clicks have acoustic features resembling human vowel sounds.
- The team used unsupervised translation techniques and machine learning to analyze clicks without a direct reference point.
- Project CETI reports AI accuracy up to 97.5% for click types and 95.3% for dialects, enabling detailed structural analysis.
- Researchers used non-invasive sound and behavior data from whales near Dominica to preserve social and environmental context.
Main content
What the study found and how it was done
Researchers long believed sperm whale vocal patterns were mostly simple pulses. The new analysis — led by UC Berkeley linguistics professor Gašper Beguš in collaboration with Project CETI — reveals that many whale clicks show steady acoustic shapes that closely match how human vowels are formed.
“very, very slow vowels,”
— Gašper Beguš, UC Berkeley
Human vowels are characterized by sustained changes in sound energy across frequencies. The study identifies long, slow click envelopes that act like vowel analogues — not identical to human speech but structurally similar. To uncover these patterns, the team applied deep machine learning pipelines and large annotated datasets collected with non-invasive equipment around Dominica.
A crucial methodological advance was using unsupervised translation approaches that compare unknown communication systems without needing a Rosetta Stone. This allows detection of structural parallels between whale calls and human speech even when meanings remain unknown.
Why the vowel-like finding matters
Vowels act as core building blocks in many human languages. Finding **vowel-like acoustic properties** in sperm whale calls suggests their vocal system could support combinatorial signaling, meaning discrete sound units might be assembled to convey varied information.
If whales have phoneme-like or vowel-like units, their communication could encode richer information about behavior, family relationships, and environmental context than previously assumed.
Project CETI’s broader mission
Project CETI aims to map and decode the deep structure of sperm whale communication by combining long-term behavioral observation with robotics, acoustic engineering, and AI. The group focuses on family-oriented populations near Dominica that show clear dialect patterns — an ideal setting for studying social communication.
Project CETI’s AI systems classify click types and dialects with very high reported accuracy, giving researchers confidence the units they measure are meaningful rather than random noise. The project emphasizes non-invasive data collection to preserve natural behavior while gathering synchronized audio, video, and context.
What researchers are not yet saying
The study does not claim sperm whales use words like humans or that we can yet converse with them. The findings reveal structural similarities in sound patterns — an important step toward translation — but not proof of human-like language or understood meanings.
Implications for United States
Economic and industry effects
- Fisheries and coastal economies: Regulators could adopt new noise-management measures or seasonal restrictions to protect key whale behaviors, affecting fishing schedules and coastal business operations.
- Offshore energy and development: Developers may face targeted mitigation if whale communication is shown to be vulnerable at specific times and locations.
- Technology and jobs: AI, robotics, and acoustic engineering needs could drive research grants and contracts to U.S. tech firms, universities, and contractors supporting low-impact monitoring.
National security and navigation
- The U.S. Navy and other maritime users may refine sonar exercises and timing to reduce interference with whale social signals.
- Improved acoustic mapping can aid shipping and search-and-rescue by helping detect and avoid whale groups and reduce ship strikes.
Conservation and legal actions
- Stronger evidence of complex communication could support calls for expanded protections, buffer zones, and stricter permitting for noisy marine projects.
- Community-based monitoring and citizen science could be integrated into non-invasive data collection programs for coastal stakeholders.
Education and public interest
- Findings provide material for schools, museums, and outreach linking AI and marine science to workforce training and STEM programs.
- Whale-watching and coastal tourism may see increased interest or new operational guidelines to avoid disrupting whale communication.
Practical steps and next actions
- Support non-invasive monitoring with robotics and passive acoustic systems to minimize disturbance while maximizing data quality.
- Hold local stakeholder meetings with coastal counties, fishing groups, and researchers to craft pragmatic noise-management plans.
- Invest in education and workforce programs for acoustic monitoring, data analysis, and eco-friendly maritime operations.
