From struck sticks to generated sound.

The oldest surviving instruments are bone flutes from caves in southern Germany, around 40,000 years old. But simpler instruments — struck sticks, rattles, stones — are almost certainly older, and archaeologically invisible: wood rots, and a percussion stone looks like a stone.

So the honest opening is that music's first chapter is mostly missing. What survives is the durable end of a much longer story: the moment someone realised that a hollow bone, blown just so, could sing.

In 1914, Erich Moritz von Hornbostel and Curt Sachs published the classification system musicologists and museums still use, building on Victor-Charles Mahillon's instrument catalog for the Brussels Conservatory. It sorts every instrument by how it makes sound — not by how it looks or what it's made of. Those categories are the sticks-to-strings arc, laid out in order:

• Idiophones

  The instrument's own body vibrates — struck, shaken or scraped. A clave, a rattle, a xylophone bar, a bell. The oldest and simplest class: a stick struck against a stick is already an idiophone.

• Membranophones

  A stretched skin vibrates. Every drum, from a frame drum to a kettledrum. The membrane is the sound source; the shell only shapes and projects it.

• Chordophones

  A string under tension vibrates — the harp, the lute, the piano, the violin. The word shares its root, chord, with this network's name: a sounding string is where our whole family of sites starts.

• Aerophones

  A column of air vibrates — the flute, the reed, the trumpet, the pipe organ. The bone flutes in the museum's opening are aerophones; so is the human voice, by most readings.

• Electrophones

  Sound generated or amplified by electrical means — added to the scheme later, commonly credited to Sachs in 1940. The theremin and the synthesizer belong here: sound with no vibrating body at all.

A fifth class, electrophones, was added to the scheme later — commonly credited to Sachs in 1940 — for instruments that make sound electrically. Everything after this point in the museum is the story of that class being born.

Edison's phonograph scratched sound into a groove — the first time a sound could be kept and replayed. Before it, every sound was gone the instant it was made; after it, a performance could outlive the room it happened in.

Every audio format since is a descendant of that one idea: the wave, written down. The groove became magnetic tape, became a pit on a disc, became a number in a file — but the move Edison made, capturing a wave in a physical trace, is the move underneath all of them.

Leon Theremin's instrument (1920) made electricity itself sing — played without touch, the pitch shaped by the space around two antennas. Robert Moog's modular synthesizer (1964) made voltage-controlled sound practical for musicians, turning electrical signals into an instrument you could actually compose with.

This is the electrophone class arriving in earnest: sound no longer had to start as vibration in a physical thing. It could start as a voltage.

The Nyquist–Shannon sampling theorem says a wave can be perfectly captured by sampling it at twice its highest frequency. Human hearing tops out near 20 kHz — which is why digital audio's standard rates sit just above 40,000 samples per second. The compact disc standardised 44,100 Hz in 1982, and it has been the quiet baseline of recorded sound ever since.

This is where the museum meets the rest of the site. Every frequency page here describes a wave — and these samples are how that wave gets carried, exactly, into a file and back out to a speaker. The number on a note page is the thing the sampling theorem promises to preserve.

Modern machine-learning systems generate audio directly — synthesizing voices, instruments and whole recordings from learned patterns, rather than from captured vibration. It is the newest exhibit, and it is worth stating plainly rather than as a verdict:

The phonograph captured sound. The synthesizer shaped it. These systems produce it.

Where that leads is not the museum's to say. What the arc shows is only this: for forty thousand years, making a sound meant making something vibrate. That is no longer the only way.