Everything the Dolphin Point laboratory claimed about Peter rests on quarter-inch tape. The tapes were the instrument; the weekly reports, the transcripts, the photographs and the book were readings taken from them or beside them. This piece is about the instrument: what it recorded, how its output was turned into marks on paper, what the one measure that could be checked by a stranger measured, and what became of the reels.

What was recorded

The recording studio was a corner of the flooded inside room, twenty feet by fifteen, hung with carpets on the walls and ceiling to kill the echo of a concrete box full of water. The microphones hung from the ceiling in plastic bags against the splash. The tape machines were in the dry room next door, with a glass window between, and when there was a second person at the laboratory he or she watched through the glass and set the levels. Mostly there was not a second person. Margaret Howe set up the apparatus herself, having learned to in October 1964 when she was first left in charge, and ran the lessons alone.

The lessons were scheduled three times a day, at 8:00, noon and 3:00, each half an hour, each with five pounds of butterfish. Sessions on the balcony could not be recorded at all. Underwater sounds were picked up separately by hydrophones and were kept out of the transcripts; the whole published analysis is of what was said in air. In the earlier mimicry work at the Institute, from 1957 to 1961, the dolphin’s microphone was placed over a transparent sidearm of its tank where its blowhole came above the surface, and the human operator wore a lavalier microphone in front of his mouth, protected from sea water, and the two were recorded on separate channels of a two-track machine so that they could be told apart. The flooded room was less tidy. Margaret’s voice and Peter’s blowhole were in the same air.

The schedule said that Saturday was for tapes: “organizing tapes of the previous week and setting up tapes for the following week.” In the first week she found she was “much more encouraged” listening back than she had been in the lesson. By the third week she wrote that she had not begun re-recording the good parts because “I am so lousy at the mechanics involved that it will take forever,” and because she did not yet have enough good material, which, when Lilly asked, meant good pronunciation. By the fourth she had resolved to keep making the tapes and go over them as best she could, and “at a later date I will have to take several days and perhaps get someone in who can properly re-record.”

How it was written down

Lilly’s chapter of tape analysis in The Mind of the Dolphin opens with the admission that print cannot carry sound: “To date there is no way that we know of to print something so that you can hear it.” A footnote offers the next best thing: “Arrangements can be made to obtain a copy of a tape of samples of sessions showing Margaret and Peter’s vocal progress in a year’s work.” The segments chosen for transcription were to be copied and put in calendar order as part of the record, and some samples were to go to the sound spectrograph, so that sonograms of Peter and Margaret “pronouncing the same word again and again” could be compared. No sonogram is printed.

The transcription itself was Howe’s method. Two lines are read together, like a staff: MH and PD. The letter c is any click, whistle or other delphinese, and a row of them is not a count; “cccccc does not mean six clicks.” The letter x is a humanoid sound “not a clear enough humanoid to accurately describe phonetically.” A humanoid clear enough to write is spelled as best it can be, “oie” for an attempt at “boy.” Margaret’s instructions are in lower case and the words to be copied in capitals. Each sample ends with a few numbered notes.

Two things follow from the key. The first is that the transcripts can show the proportion of x to c, and do: it rises from November 1964 to August 1965 until there are no c left. The second is that they cannot show what any x was. The symbol was defined as the sound that could not be written down. The report’s claims about pronunciation, inflection and the beginnings of “ball,” “hello” and “Margaret” therefore rest on the ear of the transcriber, and the transcriber was the teacher. Lilly’s own standard, stated earlier in the book, was that “independent judges not working directly with the dolphin make their judgment on the recorded material on tapes.” The book does not say that any did.

The checkable measure

Lilly had met this problem before, and his answer to it is the best science in the story. Since 1957 he had played tapes of dolphins making sounds that he heard as copies of words, and found that his colleagues did not hear them so. He wrote that most scientists were not trained to pick a foreign accent out of a badly distorted transmission, so that “my demonstrations of direct tape recordings of the phenomenon have been unconvincing to many types of persons and scientists.” So he devised “demonstrations of mimicry which did not involve hearing and psychophysical judgments.”

The method was this. Lists of nonsense syllables were drawn up by shuffling cards, in groups of one to ten syllables, so that neither the operator nor the dolphin could predict a group’s length. A dolphin already trained to vocalise in air was given a tank with a transparent sidearm holding twelve to eighteen inches of water, which it could enter or leave at will; when it entered, its blowhole was above the surface. The operator stood beside it with the list in one hand and a fish in the other, said “Hello,” waited for a humanoid reply, read a group, held out the fish, and gave it when the dolphin answered in air. Within minutes the dolphin’s answers began to contain the same number of bursts as the group it had just heard. Five for five, four for four, three for three. The syllables were read at about one every seven tenths of a second; each burst lasted about four tenths of a second with four tenths of silence between; and the dolphin’s reply began about four tenths of a second after the human stopped, which is the same as the human’s own gap, so that it could not have been waiting out a fixed silence to know the group had ended. One thousand bursts were analysed from inkwriter records made directly from the tape. In 92 per cent of the exchanges the dolphin’s count equalled the human’s, give or take one. One source of error was the human: a syllable like “it” is physically two bursts, “i” and then “t,” and the dolphin answered the physics rather than the spelling, so some of its “errors” were the operator’s miscount and had to be corrected. Lilly drew the moral himself: choose “physically specifiable variables so that one is not confused by syllables or words which have ‘human meaning.’”

That paper, “Vocal Mimicry in Tursiops: Ability to Match Numbers and Durations of Human Vocal Bursts,” appeared in Science on January 15, 1965. The fuller version, with Alice Miller and Henry Truby, was published in the Journal of the Acoustical Society of America in June 1968 as “Reprogramming of the Sonic Output of the Dolphin: Sonic Burst Count Matching,” the title now saying exactly what had been shown. It is a real finding. A bottlenose dolphin can be trained to make loud sounds in air and to match the number and roughly the duration of a sequence of sounds it has just heard, up to ten. It is also, as the title of 1968 concedes, a finding about counting, and the flooded room was built to go beyond it to meaning. The one measure that needed no ear was the one that could not carry the claim, and the claim was carried instead by the measure that needed an ear.

Where the reels are

Lilly’s papers were bought by Stanford University Libraries in five accessions between 1996 and 2015: 242 linear feet, 154 boxes in the processed part alone. One series of the finding aid is headed “Audiotape reels (dolphin research).” It says that “Cartons 67-115 contain audiotape reels with recordings of John C. Lilly’s dolphin experiments conducted between 1961-1968,” that the box lists give the names of the dolphins and rough dates, and that the reel cases usually hold Lilly’s handwritten notes naming the activity or experiment, among them “vocal output; language; visual acoustics of sound production; repeated sounds; voice spectrum transfer; LSD; Ultrasonic analysis; telephone system.” Most reels are seven or twelve inches; some are three or five.

The St Thomas animals can be followed through the box list. Box 77: Pam, Peter, Sissy, thirty-eight seven-inch reels, May and June 1963. Box 79: Pam, Peter, Sissy, twenty-one reels, April 1963. Box 81: Baby, Cheechee, Pam, Peter, Sissy, 1962 to 1965. Box 83: Pam, Peter, Sissy, 1962 to 1964. Box 106: Peter and Sissy, forty seven-inch and seven five-inch reels, 1965 to 1966. Box 107: Pam, Peter, Sissy, forty-six reels, 1965 to 1966. Box 108: Pam, Peter, Sissy, 1964 to 1966. Box 105: Jobe, Lady, Notch, Pequot, Peter, Sissy, forty-six reels, 1965 to 1966, the one carton in which Peter’s name sits beside dolphins that were never on St Thomas. The library notes that audio “has been digitally captured from reels in Box 79, 82, 99, 105, 106, 108 and 120,” and that rights in this series of recordings belong to Stanford. Elsewhere in the collection are the folders “Report on Work at St. Thomas July 1963–September 1964,” “Report on work at St. Thomas and Sea Life Park, Oahu, 1963-1964,” the Institute’s audit reports of December 1966, and “papers to be completed for publication, 1968.” The “Progress Report, St. Thomas” by Lilly and Howe, dated 1965, which Lilly’s book lists in its bibliography, does not appear by that name in the processed listing.

What an independent analysis would need is now in one place: the lessons in calendar order, Howe’s own transcripts to compare them with, and listeners who did not give the fish. The 2014 documentary used some of the digitised reels for their sound. Nobody has yet published the count.

Sources

• John C. Lilly, The Mind of the Dolphin: A Nonhuman Intelligence, Doubleday, 1967, chapters 9, 10, 12 and 15: the recording studio and microphones, Howe’s schedule and her remarks on listening back, the earlier two-channel method and the sidearm tank, the nonsense-syllable experiments and their timings, the transcription key, and the footnote offering a sample tape. In copyright; quoted in short phrases. archive.org
• John C. Lilly, “Vocal Mimicry in Tursiops: Ability to Match Numbers and Durations of Human Vocal Bursts,” Science 147 (3655), 300–301, January 15, 1965, from the published summary. pubmed.ncbi.nlm.nih.gov
• John C. Lilly, Alice M. Miller and Henry M. Truby, “Reprogramming of the Sonic Output of the Dolphin: Sonic Burst Count Matching,” Journal of the Acoustical Society of America 43 (6), 1412–1424, June 1968. pubmed.ncbi.nlm.nih.gov
• John C. Lilly Papers, M0786, Stanford University Libraries, finding aid: the extent and acquisition, Series 12.1 with its scope note and box list, the preservation note on the digitised reels, and the folders on the St Thomas reports and audits. oac.cdlib.org, archives.stanford.edu
• Christopher Riley, “The dolphin who loved me,” The Observer, June 8, 2014, for the use of the archived quarter-inch tapes in the documentary. theguardian.com