How did the first machines record sound, long before anyone could play it back?

The first machines to capture sound could not play it back at all. In 1857 a French editor and typographer named Édouard-Léon Scott de Martinville patented the phonautograph, the earliest known device for recording sound. Sound funnelled down a trumpet onto a thin membrane that stood in for an eardrum, and a stiff boar's bristle at its centre scratched a wavy line into a glass plate coated with soot. Scott expected people to learn to read those lines like a kind of natural stenography. Playing them back was not part of the plan.

Then came Edison. In 1877 he was working on ways to play back recorded telegraph messages when he conceived of recording and reproducing sound itself. His phonograph used a diaphragm and a stylus that dented a sheet of tinfoil wrapped around a cylinder; to play it back, a playback stylus followed the dents and was shaken by them, faintly reproducing the sound. Edison announced it in November 1877. At the offices of Scientific American, a visitor turned the crank and, to the astonishment of everyone present, the machine said: "Good morning. How do you do? How do you like the phonograph?"

So recording sound took two ideas: first write the vibration down as a line, then find a way to follow the line back into sound. The French poet and inventor Charles Cros is the first person known to have made that second leap. And the story has a twist. In 2008 scientists scanned some of Scott's soot traces made before 1861, a computer processed the scans into audio, and some of the oldest known recordings were heard at last.