myfl.uk · first

The first recorded voice, on a record that could not be played

On 9 April 1860 a machine in Paris scratched the opening lines of a French folk song onto a sheet of paper blackened with soot. The machine could write sound but not play it back, and the song stayed on the paper for 148 years. This page is the record’s release sheet, filed as if the record could be shelved and never heard.

Release sheet

ArtistÉdouard-Léon Scott de Martinville (1817–1879), editor and typographer of manuscripts at a scientific publishing house in Paris.
Title“Au Clair de la Lune” (By the Light of the Moon), opening lines.
Recorded9 April 1860, Paris.
FormatPhonautogram: a line scratched in lampblack on paper, wrapped round a hand-cranked cylinder.
PlaybackNone. The medium was a written document, not a record.
FiledWith the Académie des Sciences, 1861.
First heard2008, after optical scanning at the Lawrence Berkeley National Laboratory.
Words settledMay 2010, by the First Sounds project.

The page’s own accession number is the experiment ledger’s: 031. The sheet it describes is filed with the Académie des Sciences.

The machine

Scott built the phonautograph as a laboratory instrument for studying sound, not for listening to it later. He wanted speech to write itself, “le problème de la parole s’écrivant elle-même”. The idea arrived while he was editing a treatise on physiology and came upon an engraving of the anatomy of the human ear; he described it, in his own account, as “the imprudent idea of photographing the word”.

Sound entered an acoustic trumpet, moved a thin membrane, and a rigid boar’s bristle about a centimetre long scratched the vibrations into soot as the sheet slid past at one metre per second. Scott believed the traces would one day be read like writing, a kind of natural stenography, and he called them phonautograms. He patented the machine on 25 March 1857, patent 17,897/31,470. From 1860 he drew a 250 Hz tuning-fork trace alongside the sound, to correct for the hand-cranked cylinder’s uneven speed.

It did not occur to anyone before the 1870s that the tracings held enough information to be turned back into sound, and direct physical playback was impossible in any case: the recording was a two-dimensional line.

The song

The phonautogram holds the opening lines of “Au Clair de la Lune”. When the tracings were first scanned, the words were thought to be “Au clair de la lune, Pierrot répondit”. The playback released in May 2010 settled them: “Au clair de la lune, mon ami Pierrot, prête moi—”. The recording stops in the middle of a line.

The cylinder slowed, possibly to a complete stop, between “Pierrot” and “prête”; Scott appears to have paused to check how much unrecorded sheet remained. At the correct speed the voice is a man’s, thought to be Scott’s own, singing very slowly. He recorded the song at least three times.

Playback

The player below has one control and no output. Pressing it will not produce sound. The record was not made to be played; the machine that wrote it had no means of playing anything back. What it could do was write the sound down, and the player below does that instead.

No sound. The phonautograph could only write. Scott never built a means of playback, and the first playback did not come until 2008, when the tracings were scanned and converted to audio. The recording has been public since then, held by the First Sounds project.

A stand-in tracing, not a reproduction of the 1860 sheet. It is drawn the way the machine drew: left to right, by a line. The original slid under the bristle at one metre per second.

The rest of the catalogue

The 1861 deposit was not Scott’s only recording. The First Sounds catalogue lists the other phonautograms of 1860, among them:

What came after

The phonautograph also fed the gramophone: its inventor, Emile Berliner, worked with one while developing his own device.

Sources

Dates, figures and quotations are taken from those two sources and cross-checked between them. The machine cannot be consulted; it was a machine for writing, not for answering.