Whales May Have Language — What Happens If We Prove It
AI is transforming how scientists decode sperm whale communication. If researchers crack it, the implications for law, conservation, and our understanding of intelligence run far deeper than linguistics.
The Sound of Birth
On July 8, 2023, off the coast of Dominica, a sperm whale entered labor inside a pod of eleven. Researchers listening through hydrophones heard something remarkable unfold.
As the mother pushed the calf toward the surface, the rhythm of her podmates’ clicks shifted. After the newborn breached, several females circled beneath it, supporting its body. The vocalizations continued changing at each key moment — delivery, first breath, the awkward early swims. David Gruber, founder of Project CETI, put it plainly: communication doesn’t happen in isolation. These animals adjust their speech around events humans can barely perceive without instruments.
What they were communicating remains unknown. But the question driving Project CETI — whether sperm whales possess something approaching language — sits at the center of one of the most heated debates in modern marine science. And it matters far more than linguistics alone.
Two Meds, Two Accents
Long before AI entered the picture, researchers had already mapped dialect boundaries. A study published in June in Proceedings of the Royal Society B analyzed over five thousand codas — short click sequences — recorded across the Mediterranean between 2003 and 2021.
Western populations near Spain’s Balearic Islands favored slow, deliberate patterns. Eastern groups along Greece’s Hellenic Trench clicked faster and employed a richer repertoire. The differences were audible to unaided ears, as lead author Taylor Hersh confirmed. Genetic evidence suggests both populations descended from whales that entered the Mediterranean roughly twenty thousand years ago through the Strait of Gibraltar, eventually splitting into largely isolated groups.
Critically, the eastern whales still occasionally used older western patterns — cultural memory preserved across millennia. The western whales never adopted the newer eastern styles. These are not mere sound variations. They are living traditions passed between generations, a form of culture measurable in frequencies and timing.
Different groups only socialize with whales speaking their dialect. This creates social boundaries as rigid as any human linguistic barrier.
Finding Vowels in Clicks
Project CETI, funded partly through a TED Audacious Prize, began systematically studying sperm whales off Dominica in 2020. Their approach combines hydrophone arrays, drone surveillance, and increasingly sophisticated computational models.
In 2024, the team analyzed 8,719 codas. Whales vary pause durations in ways the researchers call rubato after the musical term. Others append extra clicks, termed ornamentation. When accounting for tempo, rhythm and these micro-variations, the team identified at least 143 commonly used combinations — enough to constitute what they described as a phonetic alphabet.
Gašper Beguš, CETI’s linguistics lead at UC Berkeley, accelerated recordings until individual clicks became tractable. What emerged surprised even conservative analysts. Frequency bands in whale clicks map onto two broad categories resembling human vowel qualities — akin to “ah” and “ee.” Some transitions slide between them like diphthongs. A follow-up paper published in April showed these categories behave systematically: the “a”-like sounds run measurably longer than the “i”-like sounds, and quality bleeds between adjacent codas the way coarticulation works in human speech.
Begus compared the task to deciphering Latin or Mandarin — structural complexity first, meaning later. You learn the alphabet before you read the text.
The Backlash Is Real
Not everyone agrees the field is advancing.
Luke Rendell, who published the first study describing sperm whale vocal clans twenty-three years ago, calls CETI’s claims about alphabets, vowels and conversations “massively overhyped clickbait.” His objection cuts to the core definition of language. Symbols must stand for shared ideas. Rules must allow speakers to generate potentially unlimited novel messages from finite elements.
“We are very, very, very far from that,” Rendell said.
Laela Sayigh at Woods Hole Oceanographic Institution worries about the stakes beyond academic rigor. In the 1960s and 70s, neuroscientist John Lilly claimed dolphins possessed advanced language and could be taught to speak. When the evidence collapsed under scrutiny, decades of serious cetacean research lost funding and credibility. “So we really don’t want history to repeat itself here,” she said.
Gruber acknowledges the concern but pushes back against the framing. He says the word conversation describes what ongoing research aims to determine, not something already claimed proven.
Who Wins If Whales Speak
The scientific debate matters. But the practical consequences of resolving it matter more.
If sperm whales demonstrate systematic, rule-governed communication接近 language, existing legal frameworks fracture. Current international wildlife protection laws treat species as discrete units deserving protection from hunting, pollution and habitat destruction. None envision creatures whose inner lives include dialect, tradition, and contextual negotiation.
A legally recognized non-human language would trigger obligations no treaty currently anticipates. Shipping lanes crossing known whale territories could face litigation if vessels alter vocal behavior — which Roee Diamant’s team documented conclusively last July. Whales shorten codas and tighten click gaps when ships approach. The acoustic signature is consistent enough that researchers can detect vessel presence from vocalizations alone.
That finding stands independently of whether whales have words. It proves, however, that anthropogenic noise fundamentally disrupts their communication ecology. Courts evaluating endangered species protections would need to account for acoustic habitat — a concept virtually absent from current environmental law.
New biodiversity treaties could mandate silence zones during calving seasons, migration corridors defined by dialect boundaries rather than geographic convenience, and impact assessments measuring not just mortality but communicative disruption.
What Machines Teach Us About Ourselves
The AI angle deserves separate attention. Project CETI’s computational pipeline does not simply classify sounds. It detects subtleties — rubato shifts of milliseconds, ornamental additions invisible without temporal resolution measured in fractions of a second. Human listeners miss these patterns entirely.
This raises uncomfortable questions about anthropocentrism in AI ethics. Researchers training language models on human data assume syntax emerges from social necessity, not biological accident. Whale communication challenges that assumption directly. Social structure clearly exists among sperm whales — matriarchal pods, maternal kin networks, inter-pod alliances bounded by dialect. Yet the pressure to develop productive symbolic language appears absent or negligible.
Intelligence need not converge on human-like grammar. Machines trained to find it may simply be projecting familiar patterns onto alien data.
Or they may be discovering structure too elegant for human intuition to parse. The birth recording offers a clue. Every acoustic shift coincided with observable behavioral milestones, suggesting real-time information exchange rather than environmental reflex. Whatever the whales encode, the system handles it competently.
What Comes Next
CETI’s immediate goal is prediction: can a model identify whether a whale is about to dive or sleep solely from vocalizations? Success would demonstrate genuine syntactic or semantic knowledge, moving past pattern description toward comprehension.
“If CETI’s model consistently predicts what a whale is doing in situations it has never encountered, that’s a strong indication we’re uncovering a pattern,” Gruber said.
Even failure carries value. Diamant’s vessel-detection finding already provides actionable intelligence for maritime policy. Whales modify their speech around ships regardless of meaning. That modification is measurable, reportable and legally significant.
The Mediterranean dialect study similarly offers near-term utility. Dialect maps can inform marine protected area design, prioritizing corridors that connect culturally distinct populations rather than treating ocean space as uniformly distributed habitat.
The broader question — whether animals we share the sea with possess language — forces humanity to confront its own assumptions. If the answer is yes, the implications extend from courtroom hearings to classroom curricula to dinner-table conversations about what makes us uniquely human.
If the answer is no, the structural complexity documented so far still demands respect. Sperm whales maintain cultures across twenty thousand years. They teach their young accents their mothers learned from theirs. They adjust their voices around births, ships and possibly each other.
Understanding why requires technology we are only now developing. Respecting what we discover should require wisdom we may not yet possess.