The standard length of a popular song — somewhere between two and a half and four minutes — feels like a natural unit. Songwriters write to it. Radio is built around it. Streaming rewards it. It is also, very specifically, the run time of one side of a ten-inch shellac disc spinning at around 78 revolutions per minute. (Shellac is a natural resin; the 78 rpm speed itself was only standardized in the late 1920s, and earlier discs ran at a range of speeds.) That format dominated commercial music through the first half of the twentieth century, into the early 1950s.

This is the central, easily-missed point of recorded-music history: every major change in playback technology has quietly rewritten the rules of what gets made. The 78 forced songwriters to fit a complete musical statement into about three minutes. The LP, introduced by Columbia in 1948, opened a twenty-two-minute side, and within a decade jazz musicians were stretching solos to lengths that would have been unthinkable on a 78. Each new way to play music back rewrites the unspoken rules of the music made for it — and this article follows that chain forward, through the cassette, to the CD's 74 minutes, and on to streaming's thirty seconds.

None of these decisions started with composers. They started with engineers, executives, and patent attorneys. The music caught up.

The recordings of Enrico Caruso, the great Italian tenor, all sound as if every note is belted at full volume. That is not only because Caruso was an operatic singer. Before 1925, no other vocal style could be recorded at all — that ringing sound was itself an artifact of the machine.

Until 1925, recording used no electricity at all; it was a purely physical process. Performers crowded around a large horn, the loudest instruments closest, and the sound vibrated a diaphragm that cut a groove directly into a wax master. The system favored brass and percussion. It punished string sections: a violin was too quiet to register, so players switched to the Stroh violin, a special instrument fitted with a metal horn to amplify the sound. And it punished singers, who had to push their voices to fill the horn. It was inside this mechanical logic that Caruso's discs made him — by the most commonly cited account — the first recording artist to sell a million copies.

In 1925 the Western Electric system married the recording lathe — the machine that cuts the sound groove into the disc — to a condenser microphone and an electronic amplifier. The change is sometimes filed under "better fidelity," but the real shift was psychological. A singer could now stand back from the mic, hold a long phrase at conversational volume, lean into intimacy. Within five years a young Bing Crosby was building a career on a vocal style that would have been physically impossible to record in 1924 — a relaxed, microphone-aware baritone known as crooning. By the mid-1930s Crosby was among the most successful and highest-earning entertainers in the United States, Frank Sinatra was studying his records, and the whole of American popular song had quietly reorganized itself around the assumption that singing was something you did close to a microphone.

Classical recording followed the same path, but the change came more slowly. Conductors discovered that with electrical recording they could capture an orchestra in a real hall rather than a cramped, makeshift studio. The 78 still imposed its three-minute limit, so the symphonic repertoire was sliced across multiple sides — six discs for a single Mahler symphony, packaged in what was literally called an "album": a bound book of pasteboard sleeves. The word stuck — which is why we still call a collection of songs an album, long before any single disc could hold a whole work.

Columbia's 12-inch microgroove LP, introduced in 1948, did two things at once. It extended a side to about twenty-two minutes, and it let a jazz musician build a long-form statement without splicing. Kind of Blue (1959) gets to be a single forty-five-minute Miles Davis statement only because Columbia had been selling LPs for a decade by then. A 78 would have demanded that "So What" be a 3-minute single, with the modal vamp truncated to a hook.

That twenty-two-minute side did more than lengthen solos; it let an album be designed as a single long experience. Brian Eno's "Music for Airports" (1978), composed to hang in the air of an airport lounge, is the extreme case — music that works whether you focus on it or let it wash past, which only the long uninterrupted side of an LP made possible.

The Philips compact cassette, launched in 1963 as a dictation format, accidentally became the dominant music carrier of the 1970s and 1980s. Its decisive feature was not fidelity but writability. Listeners could record what radio was playing, exchange mixtapes with friends, and circulate music outside the control of record companies. The early hip-hop economy in 1970s New York ran on cassettes of park-jam recordings; the 1980s underground metal scene that birthed extreme genres ran on cassette trading by mail. A whole generation's listening habits were shaped by a format originally designed for office memos.

Sony's Walkman in 1979 added the final piece: privacy. Headphone listening on the street, on the subway, in bed, alone. The producer's mix could now assume an audience of one, hearing the record at conversational level rather than across a room. Pop production from the early 1980s onward — the close-mic'd vocal, the precisely panned reverbs, the way the bass and kick drum sit forward — is recognizably designed for headphone listening. It is no coincidence that the first hit records to fully exploit this aesthetic, like Michael Jackson's "Thriller" (1982), arrived three years into the Walkman era.

The compact disc, jointly developed by Sony and Philips and launched commercially in 1982, looked at first like nothing more than cleaner sound. The deeper change was that music was now a stream of numbers. Any device that could read the numbers could play the record. Any device that could copy the numbers could clone the record. The CD made digital audio the default; everything since has been a consequence.

First came three devices that changed how music traveled. The MP3 (standardized within MPEG-1, ISO/IEC 11172-3, and published in 1993) made an audio file small enough to send over a slow 1990s connection; Napster (1999) made sending one between strangers trivial; the iPod made it cheap to carry thousands of them in a pocket. Music could now be sent, copied, and carried by the thousand. But the two that changed how songs were actually made came next. The iTunes Store, in 2003, sold individual tracks for ninety-nine cents and quietly ended the album-as-commercial-unit. Spotify, founded in Sweden in 2006 and launched in 2008, replaced ownership with rental and replaced the ninety-nine-cent sale with a fraction of a cent per stream, counted only after thirty seconds of listening — though that thirty-second rule is not Spotify's own. It is an industry-wide counting standard shared by Apple Music and the chart bodies, and the payout is a pooled, pro-rata share of total streams rather than a fixed per-stream price.

That thirty-second threshold is the contemporary three minutes. A stream that drops out at twenty-nine seconds pays next to nothing, so a 2025 pop song typically front-loads its hook into the first eight bars — the first fifteen seconds or so — eliminates the long instrumental intro, and puts its biggest chorus up front, before listeners can drop off. By various analyses, between 2000 and 2020 the average pop song lost somewhere around twenty to thirty seconds or more, and intros shrank sharply — the dramatic drop from roughly fifteen or twenty seconds down to a few is most strongly documented across a longer window, from the 1980s onward.

The pattern across a century is consistent enough to read as a law. Whatever shape the playback device takes — wax cylinder, shellac disc, vinyl LP, magnetic tape, optical disc, network packet — the music being made at the time will eventually conform to its constraints. Composers and listeners experience these constraints as taste. The engineers who designed the formats experienced them as tolerances. They are, in the end, the same thing.