For more than a century, since he captured the spoken words “Mary had a little lamb” on a sheet of tinfoil, Thomas Edison has been considered the father of recorded sound. But researchers say they have unearthed a recording of the human voice, made by a little-known Frenchman, that predates Edison’s invention of the phonograph by nearly two decades.
Isabelle Trocheris
The audio historian David Giovannoni with a recently discovered phonautogram that is among the earliest sound recordings.
Audio: 1860 recording:
The Phonautograph Recording from 1860 of ‘Au Clair de la Lune‘
The 19th-century phonautograph, which captured sounds visually but did not play them back, has yielded a discovery with help from modern technology.
The 10-second recording of a singer crooning the folk song “Au Clair de la Lune” was discovered earlier this month in an archive in Paris by a group of American audio historians. It was made, the researchers say, on April 9, 1860, on a phonautograph, a machine designed to record sounds visually, not to play them back. But the phonautograph recording, or phonautogram, was made playable — converted from squiggles on paper to sound — by scientists at the Lawrence Berkeley National Laboratory in Berkeley, Calif.
“This is a historic find, the earliest known recording of sound,” said Samuel Brylawski, the former head of the recorded-sound division of the Library of Congress, who is not affiliated with the research group but who was familiar with its findings. The audio excavation could give a new primacy to the phonautograph, once considered a curio, and its inventor, Édouard-Léon Scott de Martinville, a Parisian typesetter and tinkerer who went to his grave convinced that credit for his breakthroughs had been improperly bestowed on Edison.
Scott’s device had a barrel-shaped horn attached to a stylus, which etched sound waves onto sheets of paper blackened by smoke from an oil lamp. The recordings were not intended for listening; the idea of audio playback had not been conceived. Rather, Scott sought to create a paper record of human speech that could later be deciphered.
But the Lawrence Berkeley scientists used optical imaging and a “virtual stylus” on high-resolution scans of the phonautogram, deploying modern technology to extract sound from patterns inscribed on the soot-blackened paper almost a century and a half ago. The scientists belong to an informal collaborative called First Sounds that also includes audio historians and sound engineers.
David Giovannoni, an American audio historian who led the research effort, will present the findings and play the recording in public on Friday at the annual conference of the Association for Recorded Sound Collections at Stanford University in Palo Alto, Calif.
Scott’s 1860 phonautogram was made 17 years before Edison received a patent for the phonograph and 28 years before an Edison associate captured a snippet of a Handel oratorio on a wax cylinder, a recording that until now was widely regarded by experts as the oldest that could be played back.
On a digital copy of the recording provided to The New York Times, the anonymous vocalist, probably female, can be heard against a hissing, crackling background din. The voice, muffled but audible, sings, “Au clair de la lune, Pierrot répondit” in a lilting 11-note melody — a ghostly tune, drifting out of the sonic murk.
The hunt for this audio holy grail was begun in the fall by Mr. Giovannoni and three associates: Patrick Feaster, an expert in the history of the phonograph who teaches at Indiana University, and Richard Martin and Meagan Hennessey, owners of Archeophone Records, a label specializing in early sound recordings. They had collaborated on the Archeophone album “Actionable Offenses,” a collection of obscene 19th-century records that received two Grammy nominations. When Mr. Giovannoni raised the possibility of compiling an anthology of the world’s oldest recorded sounds, Mr. Feaster suggested they go digging for Scott’s phonautograms.
Historians have long been aware of Scott’s work. But the American researchers believe they are the first to make a concerted search for Scott’s phonautograms or attempt to play them back.
In December Mr. Giovannoni and a research assistant traveled to a patent office in Paris, the Institut National de la Propriété Industrielle. There he found recordings from 1857 and 1859 that were included by Scott in his phonautograph patent application. Mr. Giovannoni said that he worked with the archive staff there to make high-resolution, preservation-grade digital scans of these recordings.
Listen: Au Clair de la Lune
Scott never dreamed of playing back his recordings. But this morning, the dream Scott never had will come true.
A cadre of audio historians, recording engineers, and scientists working in conjunction with the First Sounds initiative has transformed Scott’s smoked-paper tracings into sound. They will premiere Au Clair de la Lune at the annual conference of the Association for Recorded Sound Collections at Stanford University this morning. Examples of sounds evoked from French and American phon- autograph recordings made between 1857 and 1878 will also be played publicly for the first time.
Contrary to popular belief, Thomas Edison did not make the first recording in history.
The sound files of Édouard-Léon Scott de Martinville’s phonautograms released in 2008 by the First Sounds collaborative were created using Lawrence Berkeley National Laboratory’s virtual stylus technology. Unfortunately, as these phonautograms were not made to be played back, they do not adhere to the most fundamental technical requirements of sound recording; their tracings are “malformed.” Because modern audio processing software cannot handle such malformations, these precious phonautograms have remained mute. In the quest to better understand the work of pioneer phonautogram makers, Dr. Patrick Feaster of Indiana University, Bloomington, has devised an alternate approach to playback. Although it must necessarily ignore or misinterpret information contained in malformations, this approach is sufficiently robust to let us hear something from recordings that are otherwise too compromised to process. The following two recordings were played by Dr. Feaster using this approach.
Chi crederia che sotto forme umane e sotto queste pastorali spoglie fosse nascosto un Dio? Non mica un–["Who would believe that under human form and under this pastoral garb there would be found a God? Not only a...."]. As of mid-May 2009, this phonautogram of the opening lines of Torquato Tasso’s pastoral drama Aminta is the earliest audible record of recognizable human speech–at least, recognizable enough to follow if you already know the words. (The April 9, 1860 recording of Au Clair de la Lune appears to be earlier, but it is sung, not spoken.) Édouard-Léon Scott de Martinville recorded it for the physicist Henri Victor Regnault, probably in April or May 1860, as a “study of the tonic accent,” so he was more interested in capturing the intonation than the words anyway. But there’s a mistake in the recorded recitation. “I was wrong,” Scott wrote at the bottom: “it should be umane forme.” By apologizing for reversing the word order, Scott indirectly identifies himself as the speaker.
Listen: Tasso’s Aminta
Vole, Petite Abeille – Fly, Little Bee (undated, probably September 1860)
This is the only phonautogram Édouard-Léon Scott de Martinville identified as made with an “amplifying lever,” his last known phonautographic design change. It is therefore presumably also his last known phonautogram, dating from late September 1860 or maybe even later. Equally noteworthy is the content: a lively rendition of a song that is as much performance as experiment. When First Sounds first published this recording in May 2009, the selection was unidentified, except that the final syllables seemed to be the same as the inscribed title–”vole, petite abeille,” or “fly, little bee.” It turns out to be “La Chanson de l’Abeille” from the comic opera La Reine Topaze by Victor Massé, first performed in 1856. Thanks to Peter L. Goodman for his help in the search, as well as to David Lasocki of the William & Gayle Cook Music Library.
Listen: Vole, Petite Abeille
Vole, Petite Abeille – Fly, Little Bee (September 15, 1860)
The first of Scott’s alternate takes we’ve heard, this phonautogram again features “La Chanson de l’Abeille” from Massé’s La Reine Topaze but was recorded without the amplifying lever. The sound quality is markedly different, even though both phonautograms were played back using identical methods. On the other hand, the adaptation of the song itself is nearly identical in both cases. Years later, performers would famously have to learn to cut pieces of music to fit the duration of a phonograph cylinder. By comparing Scott’s two phonautograms of “La Chanson de l’Abeille,” however, we learn that a consistent abridgement of one song had already been worked out for recording purposes back in 1860.
Listen: Vole, Petite Abeille (alternate take)
REINTERPRETED SOUNDS – RELEASED MAY, 2009
When we released “Au Clair de la Lune” and “Gamme de la Voix” in 2008, we chose a playback speed that presumed the singer was a woman or an adolescent. Discrepancies in Scott’s notes suggested several possible playback speeds (each consistent with one of his inconsistencies) and the singing sounded “right” to us at this speed. But early in 2009, upon hearing “Aminta” in which Scott indirectly identifies himself as the speaker, Dr. Feaster realized that this speed was too fast. When “Au Clair” and “Gamme” are slowed, their singer sounds very much like the speaker in “Aminta.” We have adjusted their playback speed in accordance with this reinterpretation. Great performances they’re not. But Scott made his recordings to be seen, not heard. He sang purposely into his instrument to reveal the shape of his sounds and the frequency of his notes. In listening to these phonautograms we eavesdrop on scientific experiments wafting imperfectly through briefly-opened windows in time.
Au Clair de la Lune – By the Light of the Moon (April 9, 1860)
Scott recorded the French folksong “Au Clair de la Lune” on April 9, 1860, and deposited the results with the Académie des Sciences de l’Institut de France in 1861. As with all four 1860 phonautograms on this page, the existence of a tuning-fork calibration trace allows us to compensate for the irregular recording speed of the hand-cranked cylinder. The sheet contains the beginning line of the second verse–”Au clair de la lune, Pierrot répondit”–and is the earliest audibly recognizable record of the human voice yet recovered.
The romantic image of a woman singing to us through the veiled curtain of time was at the heart of Scott’s allure as we introduced his work in 2008. As we are loathe to let her go, we maintain her two sound files below: the first version as released in March, and a restoration from September that applied more advanced technologies.