This page is a ledger, not an argument. It lists what was achieved, and how each effort ended. You will reach your own conclusion before the bottom of the page.
In December 1967, an acoustic engineer named Dwight Wayne Batteau delivered a final report to the U.S. Naval Ordnance Test Station. In it, two bottlenose dolphins named Maui and Puka responded correctly to thirty-five spoken commands, translated in real time into whistles. One reproduced four trained whistle contours following a repeat cue.
The last paragraph describes what the team intended to do next: extend the vocal channel and make the human a responding participant. We have not located a public record showing that the full controlled experiment was completed.
What follows is every major attempt since, what each one demonstrated, and — the column that matters — how each one ended.
Contract N00123-67-C-1103. Eighty-eight pages. Read in full.
The work ran at a lagoon on Coconut Island in Kāneʻohe Bay, Oahu. A houseboat held the equipment. A ring of floats defined the working area. Two dolphins, a male and a female, worked with trainers who spoke English into a microphone; a device called the Man-to-Dolphin Translator Mark 5 converted the vowel sequences into frequency-modulated whistles and put them into the water through a hydrophone.
Thirty-five command strings were in use by 1 November 1967. They were not isolated cues. Every command took the same four-part frame — NAME · IMUA · WORD · OK — an addressee, an imperative marker, the content word, and a terminator. The animals honoured the addressee slot: a command aimed at Maui did not move Puka.
| Command | Response |
|---|---|
| BIP | Hit ball with pectoral |
| BAIEP | Swim through hoop |
| UWEIAP | Roll over |
| BEIAP | Raise flukes |
| BAEP | Make sonar sound — a vocal response |
| RETRIEVE | Retrieve bottle from anywhere on the float circle, return it to station |
| BURRAP | Swim farther seaward than present location |
| MAUKA | Come to instation |
| MAKAI | Go seaward, outside the float circle |
| PORT | Swim to his left |
| STARBOARD | Swim to his right |
| REPEAT | Vocalizes the word — the dolphin says it back |
Puka’s set also included WAVE (wave pectoral), PLOP (slap flukes) and YUMP (jump out of water).
The team did not simply report that the animals obeyed. They ran controls, and several of them are the controls a linguist would ask for.
This report is not a promotional document. It fences its own results harder than most modern press releases do.
Command sequencing failed: given two commands in a row, “the second behavior associated word released the first behavior.” When the hoop and ball were swapped, performance inverted — the team concluded the animals had been conditioned “to associate a command word with a location rather than an object.” And on the vocal channel, the sentence that governs everything:
“Attention was paid specifically to gross body motions since it was felt that the vocalizations were not, at that time, under adequate control.”
When Puka correctly performed two behaviours she had never been taught, moments after watching Maui perform them, the authors declared their own most exciting result inconclusive.
In the last quarter of 1967, the team inverted the loop. The dolphins were allowed to start or extend a session themselves by offering a behaviour the trainer had not asked for. The report describes this as establishing “a command-response point in a set assigned to the trainer” — and says plainly what it meant:
“This rule permitted the dolphin to test the responding mechanism, the trainer.” Man/Dolphin Communication, final report, page 83
A dolphin-initiated channel was running in a Hawaiian lagoon in 1967. It was restricted to body movement rather than sound — for the stated reason that the vocal channel was not yet reliable enough to use.
Read the second half of that sentence again. It is not a proposal to teach a dolphin more words. It is a proposal to build a system in which the human answers — where the animal is not a subject responding to commands but a participant in an exchange.
That is the proposed next experiment. The report shows that equipment and trained animals existed. We have not yet established whether the same funding continued or whether a later public or unpublished project completed the proposed work.
Dwight Wayne Batteau drowned on 26 October 1967, during his regular morning swim. Reported death date — sourcing pending, see the note at the foot of this page
It was the day the two dolphins were to be moved from Coconut Island to the Oceanic Institute. The report confirms the move: “On 1 November 1967, both animals were placed in a newly constructed training tank at Oceanic Institute.” Six days later.
The report was carried to completion by Peter Markey and delivered that December. The sentence describing the next experiment was published in a document finalised after its lead author was dead.
It would be easy, and wrong, to write that the Navy shut this down. The Navy funded it — across at least four separate contracts, including a parallel contract that paid for the facility, the animal transport, the veterinary support and the food fish. A newly constructed tank was waiting for the animals.
The documents located so far show plans for continued work and the animals moving to a new tank. They do not yet establish what happened to the research branch after the final report. Batteau’s death is a documented lead in that history, not a verified sole cause.
The left column is what was achieved. The right column is why it stopped. Read the right column.
The book that created the field and the field’s reputational problem in a single volume. Lilly predicted interspecies conversation within a decade; his later work discredited much of what surrounded it.
Lilly’s later work became controversial. The effect on subsequent funding and the end point of this research branch have not been established in this audit.
Branch continuation unansweredThree consecutive U.S. Navy contracts. Thirty-five commands on a four-part syntactic frame. Speaker independence, place independence at 75 yards, minimal-pair discrimination, elicited reproduction of four trained whistle contours, and a dolphin-initiated body-movement channel.
The final report named continued work, and the animals moved to a new tank. Whether the branch continued, and why it ultimately stopped, remain unresolved.
UnresolvedThe most rigorous artificial-language work ever done with dolphins. Akeakamai and Phoenix demonstrated comprehension of word order, novel sentences, and reference to absent objects — receptive language, tested to a standard the field still cites.
Akeakamai and Hiapo died in 2003 and 2004, and the laboratory closed in 2006. These events do not establish why the artificial-language branch was not continued elsewhere.
Branch continuation unansweredDiana Reiss’s keyboard programme: dolphins operated a submerged interface that produced computer-generated whistles paired with objects and activities.
The Redwood Shores facility closed and the park moved to Vallejo. Whether the keyboard programme continued after the move, and why this branch stopped publishing, remain unresolved.
Branch continuation unansweredAn underwater keyboard at a Disney park. Dolphins learned not only to respond to human key activation but to activate keys themselves — sometimes moving toward the associated object or location ahead of the human. Published 2001 and 2004.
No termination result has been located. Later related publications and the animals’ continued presence do not establish why the keyboard branch stopped publishing.
UnexplainedDolphins spontaneously imitated computer-generated whistles and increasingly produced artificial whistle facsimiles of their own accord — production without a command prompting it.
Reiss later worked at WCS Osborn Laboratories, the National Aquarium, and Hunter College. This career history does not establish why the keyboard branch ceased or whether it continued elsewhere.
Branch continuation unansweredA wearable underwater computer, worked with wild Atlantic spotted dolphins. Imitation of computer-generated signals was observed, including delayed attempts. Functional understanding of the artificial object labels was not demonstrated.
Primary records verify continuation through the summer 2025 field season; 2026 activity has not been independently confirmed. Functional understanding of the artificial labels has not been demonstrated; subject engagement and interface performance are both live explanations.
OngoingAfter thirty-two years on site, the animals leave the park.
The 2024 animal relocation occurred roughly twenty years after the last keyboard publication identified in this review. It does not establish when or why the research branch ended.
Branch continuation unansweredThe located record contains promising results, serious experimental limitations, and several incomplete termination histories. It does not supply a final verdict on a shared acoustic interface.
The next historical question is therefore narrower:
Which branches continued, which stopped, and what evidence explains the change?
These branches left useful methods and published results, but they did not all answer the same question or reach the same evidentiary standard. The open task is to learn whether their strongest components can be combined under modern controls.
The literature reports each component below with different subjects, methods, and levels of support. We have not located one study that combines them under a single controlled protocol.
That is the whole argument, and it does not require anyone to have done anything wrong. The pieces appeared across different programmes. The 1967 report proposed extending the vocal channel while making the human a responding element; no located public record shows a controlled completion of that combined design.
Twenty years ago, every component in that grid except the last one already existed. The last one now exists too.
It is worth saying clearly what we do not believe, because a page like this invites the assumption.
We looked for evidence of suppression and did not find it. The working census names more than four hundred institutions, while the population estimate of roughly three thousand dolphins remains advocacy-sourced.Figures pending verification Our bounded search identified one current two-way collaboration; it did not establish a complete global count.
You do not suppress three hundred and forty-five facilities in fifty-six countries.
Access to dolphins is only one requirement. A useful interface programme also needs animal welfare oversight, longitudinal participation, communication researchers, engineers, funding, and a protocol that separates learning from cueing. Our census suggests those elements are rarely coordinated; it does not yet establish why.
Worse, because nobody is accountable for neglect. There is no decision to overturn, no memo to find, no one who can be asked to reverse themselves. A thing that never became anyone’s job simply doesn’t get done, year after year, and no one is at fault.
Better, because suppression would have to be defeated. Neglect only has to be answered. One institution deciding this is worth a year of someone’s time would close a gap that has stood open since 1967.
The reviewed literature reports relevant components, while a controlled combination has not been located. Missing public evidence leaves several possibilities open:
The design may have been attempted without a discoverable publication; it may have been limited by scientific, welfare, access, funding, or technical constraints; or it may not have been attempted.
Resolving that history and specifying a controlled modern protocol are separate tasks. Better equipment reduces some engineering constraints but does not settle the scientific question.
We are not trying to prove that dolphins have language. We are trying to establish, with enough historical resolution to design a defensible test, exactly how far this got and exactly where it stopped. If you know of work that closes any branch on this page — especially work that contradicts it — that is the most useful thing you could send us.