The Science

What do we know about dolphin communication?

Forty years of field research. Decades of cognitive experiments. And now — AI tools powerful enough to find patterns in vocalisation data that no human ear could detect. Here is what the science actually says.

01 — Identity

Signature Whistles — Dolphins Have Names

Every bottlenose dolphin develops a unique signature whistle within the first few months of life. This whistle functions as an individual identifier — effectively a name. Dolphins use their own signature to announce themselves, and they respond specifically when they hear their own whistle played back.

More remarkably, dolphins copy each other's signature whistles when separated, apparently as a way of calling to a specific individual — not just vocalising into the void. This is referential communication: a sound that stands for a specific entity.

In 2013, Jason Bruck at the University of Chicago demonstrated that bottlenose dolphins recognise the signature whistles of former companions after more than 20 years of separation — the longest social memory ever demonstrated in a non-human animal.

These are not simple learned responses. The specificity, retention, and cross-individual copying of signature whistles suggests a level of social cognition that rivals great apes.

Key finding

"Dolphins not only have names — they use each other's names. That is a very short list of animals."

— Based on work by Stephanie King & Vincent Janik, University of St Andrews (2013)

02 — Cognition

The Cognitive Evidence

Mirror Self-Recognition

2001 · Reiss & Marino

Atlantic bottlenose dolphins pass the mirror test — recognising their own reflection as themselves rather than another animal. Only humans, great apes, elephants, and magpies have reliably demonstrated this.

Syntax Comprehension

1984 · Herman, Richards & Wolz

Dolphins Akeakamai and Phoenix learned an artificial gestural language with defined grammar rules. They correctly interpreted novel sentences — demonstrating sensitivity to word order and syntactic structure.

Social Memory

2013 · Jason Bruck

20+ years of companion recognition via signature whistles — the longest social memory ever demonstrated in a non-human animal, surpassing chimpanzees.

Tool Use & Culture

1997–present · Mann, Krützen et al.

Bottlenose dolphins in Shark Bay use marine sponges as foraging tools. The behaviour is culturally transmitted through maternal lines — learned, not instinctual.

Referential Gestures

2011 · Xitco, Gory & Kuczaj

Dolphins spontaneously direct human attention toward objects using gaze and body orientation — triadic attention-sharing analogous to pointing, a key marker of theory of mind.

Vocal Learning

Ongoing · Multiple labs

Dolphins are one of a small number of mammals capable of vocal learning — producing novel sounds through imitation. This places them in an exclusive group alongside humans, elephants, and some birds.

03 — Research

The Projects Pushing the Frontier

Wild Dolphin Project

Dr. Denise Herzing · Since 1985

WDP

Little Bahama Bank, Atlantic Ocean

The longest-running study of free-ranging dolphins in the world. Dr. Herzing has studied the same community of Atlantic spotted dolphins for over 40 years — following individuals across their entire lifetimes. The WDP dataset is one of the most valuable acoustic archives in cetacean science. The project also developed CHAT (Cetacean Hearing and Telemetry) — a wearable underwater computer designed for real-time two-way communication with wild dolphins.

Species studied: Atlantic spotted dolphin (Stenella frontalis)

Project CETI

David Gruber, Pratyusha Sharma, Tom Mustill et al. · Since 2020

CETI

Dominica, Eastern Caribbean

The Cetacean Translation Initiative applies transformer-based machine learning to sperm whale click communications called codas. CETI has deployed underwater microphone arrays off Dominica to record thousands of hours of sperm whale social interactions. Their 2024 paper in Nature Communications identified previously unknown structural complexity in coda sequences. The project is explicit: they have not "translated" whale communication — they are mapping its structure.

Species studied: Sperm whale (Physeter macrocephalus)

Herman Laboratory

Dr. Louis Herman (1930–2023) · Since 1970s

KBMML

Kewalo Basin Marine Mammal Laboratory, Hawaii

The foundational experimental work on dolphin cognition. Herman's laboratory produced the definitive evidence that dolphins comprehend artificial language with grammatical structure. His subjects Akeakamai and Phoenix demonstrated sensitivity to word order, semantic content, and novel sentence construction. This work established that dolphin intelligence is not just social and emotional — it includes abstract representational capacity.

Species studied: Bottlenose dolphin (Tursiops truncatus)

04 — AI

What AI Is Changing

For decades the bottleneck was not recordings — researchers have thousands of hours of them. The bottleneck was analysis. Human experts could listen, categorise, and annotate only so fast. That is now changing.

Large-Scale Acoustic Analysis

ML models can process years of acoustic recordings in hours — clustering vocalisations, detecting rare call types, and surfacing structural patterns that manual analysis would never find.

Sequence Modelling

Transformer architectures trained on vocalisation sequences can predict what comes next, detect anomalies, and identify higher-order structure — the same approach that powers large language models.

Real-Time Interaction

Faster inference means underwater devices can now classify incoming vocalisations in near real-time and generate appropriate playback responses — moving from passive recording toward active exchange.

DolphinGemma

Google DeepMind and Georgia Tech trained a Gemma-based model on 40 years of Wild Dolphin Project data. The first LLM built specifically for cetacean communication research — and the upstream AI for the next generation of CHAT devices.

What this is not

None of this amounts to translation. We are mapping structure, not meaning. The honest position is that we do not know what dolphins are saying to each other — and we may never know in any human sense of "knowing." What we can say is that their communication is structured, learned, referential, and socially complex. That is already extraordinary. The science does not need embellishment.