A Song Is Two Copyrights

If you learn one thing about music law, learn this. A recorded song is not one piece of property. It is two, and they have different owners.

The musical work is the composition and the lyrics: the song as written. It exists whether or not anybody has recorded it, and it belongs to the writers, usually administered by a publisher who takes a share for collecting the money.

The sound recording is one fixed performance, captured on one occasion. It belongs to whoever paid for the session, which historically means the record label. Record the same song again next week and you have created a second, entirely separate property.

Two properties, two owners, two negotiations.

The practical consequences are everywhere once you see the split.

A cover version copies the composition and not the recording, so it needs the publisher and not the label. And in American law it does not even need the publisher's agreement: a compulsory mechanical licence lets you record your own version of any published song at a statutory rate, and the writer cannot refuse. That provision dates from 1909 and it is the reason the cover version exists as a normal commercial activity rather than as a favour.

A sample copies the recording, so it needs the label. It almost always copies something of the composition too, so it needs the publisher as well. Two clearances, either of which can be refused for any reason or no reason, which is most of why sample clearance is slow, expensive and frequently abandoned.

A performance in a venue licenses the composition. A film using the record licenses both. A karaoke track licenses the composition and makes its own recording. Every one of these is legible from the split.

Now follow the streaming money, because the structure explains an arithmetic that confuses almost everybody.

A subscription service pools its revenue, takes its share, and divides what remains among rights holders pro rata by total plays across the whole service. Note what that means: your payment depends on everybody else's listening, not on yours. If total plays across the platform rise and your plays stay flat, your payment falls, because you are being paid a share of a pool rather than a price per play. There is no per-stream rate, only a quotient, which is why published figures for one vary so much and why they are all averages of something that is not a rate.

Then, within the money that reaches a track, the recording side and the publishing side are paid separately, through different pipes, at very different rates, and typically the recording side receives several times what the publishing side does.

The alternative that gets proposed is user-centric: split each subscriber's fee among the artists that subscriber actually played. It is more intuitive and it is a redistribution rather than an increase. Artists with devoted, low-volume listeners gain; artists who do well in passive, high-volume playlists lose. It changes who gets the money and settles nothing about how much money there is, and the arguments for it are usually made as though it would do both.

Which is the general shape of this whole area. Almost every dispute is about the division of a pool or about which of two properties was taken, and almost none of them is about the thing the headline says it is about.