UC Berkeley and Project CETI find sperm whale calls mirror human vowels, hinting at a structured “language”
A UC Berkeley and Project CETI study finds sperm whale calls contain sustained acoustic shapes resembling human vowels, suggesting continuous, modulated signals rather than simple click codas—pointing to structured, dialectal communication and new paths for interspecies study.
- Continuous, vowel-like acoustics: Researchers detected elongated, resonant patterns in sperm whale calls that resemble human vowel shapes, challenging the view of calls as only discrete click codas.
- AI-driven analysis at scale: Project CETI reports click extraction at ~99.5% accuracy, 23 click-pattern classification at ~97.5% accuracy, and dialect identification near 95.3%.
- Context and dialects: Call patterns vary by social context and group, indicating dialectal structure and possible encoding of identity or intent.
- Open science and conservation aims: Teams plan to publish data and tools to enable verification, collaboration, and conservation-informed policy.
Study details and how it changes thinking about whale sounds
For decades, sperm whale vocalizations were framed primarily as codas—short trains of clicks treated as discrete signals. That view led many to liken whale communication to a simple “Morse code.” The UC Berkeley and Project CETI study reframes that understanding by examining the fine spectral shape and temporal structure of calls rather than counting clicks alone.
Linguist Gašper Beguš of UC Berkeley co-led analyses that treat clicks as components of continuous acoustic events. The team reports many whale vocal elements show elongated resonant patterns that can be modulated over time—acoustic behavior analogous to vowel sounds in human speech. The authors emphasize this is not evidence whales “speak” like humans, but that their signals are richer and potentially organized in ways that support complex meanings and social functions.
Project CETI AI research: scaling up analysis with machine learning
Project CETI (Cetacean Translation Initiative) combines marine biology, linguistics, robotics, and deep learning to study whale communication at scale. Their AI systems extract clicks from noisy ocean recordings, classify call types, and detect large-scale patterns across thousands of hours of sound.
Project CETI’s internal benchmarks report ~99.5% accuracy extracting clicks, ~97.5% accuracy classifying 23 click patterns, and ~95.3% accuracy identifying dialectal differences—showing machine learning can reveal recurring acoustic structures missed at smaller scales.
Field teams use non-invasive tools—drones, hydrophones, and logging devices—to build paired datasets linking sound with behavior and context. This scaling allows researchers to search for consistent associations between vocal patterns and social activities across individuals and groups.
Fieldwork and data collection off Dominica
The project concentrates fieldwork on sperm whale populations around Dominica. Teams deploy aerial drones and underwater microphones to observe social behavior while recording acoustics, enabling linkage between vocal events and actions such as socializing, foraging, or mother-calf interactions.
Non-invasive monitoring and long-term tracking let researchers identify individual repertoires, compare groups, and test whether particular acoustic features recur in comparable contexts across time and populations.
Dialectal variation and social meaning
Beyond vowel-like features, the team documents consistent variation across groups: codas and continuous call features vary between social groups, suggesting regional or social “dialects.” AI models flagged patterns tied to contexts—some call contours recur during socializing, others during foraging or mother-calf interactions.
Researchers caution that identifying structure and dialects is not the same as producing a semantic translation. Instead, they describe this work as a blueprint—a systematized map of sounds, contexts, and patterns that could support future translation efforts.
Why vowel-like acoustics matter for language comparisons
Human language uses vowels and consonants as building blocks; vowels often serve as carrier sounds that change meaning when combined or modulated. Finding vowel-like continuous features in whale calls suggests sperm whales may use analogous building blocks—continuous signals that can be varied and combined, which is a hallmark of complex communicative systems.
If modulated, continuous signals function like linguistic units, researchers can move beyond cataloguing call types to modeling compositional structure, testing whether specific combinations reliably correlate with behaviors or events.
Open data, tools, and long-term aims
UC Berkeley and Project CETI intend to make recordings, annotated datasets, and AI tools public. Open release aims to accelerate independent verification, invite fresh methods, and encourage ethical, conservation-focused applications.
Near-term goals include reliably linking vocal patterns to social behaviors, building shared vocabularies for simple concepts, and using translation tools to inform conservation decisions. Project leaders also highlight the potential for meaningful two-way interaction or sustained communication in the longer term, while stressing careful, ethical development.
Implications for the world
Policy and conservation: Demonstrated complexity and culturally transmitted dialects could strengthen calls for reduced ocean noise, protected marine areas, and regulation of shipping lanes, with economic implications for fishing, shipping, and offshore development.
Science and technology: Project CETI’s AI work exemplifies how machine learning can accelerate biological discovery. Shared datasets and methods could foster global collaboration to study other species and regions.
Ethics and culture: Evidence of structured communication in whales raises ethical questions about human use of marine environments, captivity, and whaling, potentially shifting public opinion and policy in many regions.
Practical tools: Reliable acoustic indicators could enable monitoring systems that detect stress, prey scarcity, or human disturbance in real time—informing rapid conservation responses.
Sources and further reading
- UC Berkeley news release: “UC Berkeley and Project CETI study shows sperm whales communicate in ways similar to humans”
- Project CETI official site
- Project CETI profile on The Audacious Project (grantee overview and technical claims)
This report draws on the UC Berkeley release and materials published by Project CETI and its funders describing methods, fieldwork, and AI benchmarks. The research proposes testable approaches to sperm whale–human language comparisons and advances an open-science effort to decode another intelligent species.
