How to reconstruct a split sheet for a song from five years ago
20 · · 12 min read · Español
You cannot recreate a document nobody signed. What you can do is agree now, write down what you agree, and record inside the document when the agreement was originally made — which is a different job from the one you missed, and it is the one that works. The song’s ownership was never waiting on the paperwork. Only everybody’s memory of the room was.
- A split sheet records an allocation the law already made. Reconstructing one means agreeing now on what that allocation was, not creating it five years late.
- Read what is already on file before writing anything down. A registration filed at release already asserts an answer about who wrote the song.
- Timing changes what a copyright registration is worth: a certificate of a registration made before or within five years after first publication is prima facie evidence of the facts stated in it.
- Sign and date the sheet on the day it is signed. The date the writers originally agreed belongs inside the document as a stated fact, never on the signature line.
- A civil action under the US Copyright Act must be commenced within three years after the claim accrued, and between co-authors that period begins at a plain and express repudiation.
What are you actually reconstructing?
You are reconstructing the agreement, not the ownership.
The song acquired an ownership position the moment it existed, because the law splits a song whether anyone writes it down or not. Five years of silence did not move that position an inch. It only removed the cheapest moment anyone was ever going to have to write it down, which was the afternoon it happened.
So the document is not a container the rights have been sitting outside of. It is the writers' own record of what they agree the split is, and there is no rule that says it has to be made on the day. Late is a worse document. Late is not a void one.
One thing to settle before you start assigning numbers: whether a given contributor is a writer at all. That is a test with its own machinery and its own two halves, and it is the question most reconstructions actually founder on. Nobody argues about 40 versus 35. They argue about whether the guy who suggested the key change is on the sheet.
What still counts as evidence five years later?
Anything dated and made at the time. Start there, before anybody’s memory gets a vote.
Session files and their timestamps. The DAW project, its bounced versions, and the folder they sit in. Voice memos on somebody’s phone. Texts and email from that week. The studio’s booking record, if there was a studio. Invoices, e-transfers, and whatever you paid the engineer.
Then the release itself, which is the piece most people forget they own. The credits your distributor delivered are a statement somebody made about this song and dated. So is the ISRC on the recording, and so is every writer name that travelled with it.
None of that proves who wrote the bridge. What it does is turn a conversation about what people remember into a conversation about what can be checked — and the second one ends, which is the whole reason to have it.
What do the registries already say about your song?
Whatever the person who registered it told them, which may be nothing and may be wrong.
Two places are worth opening before you write a single number. The MLC’s Public Work Search is searched by work title and shows a registered work’s details, including publisher information. The US Copyright Office’s public records cover registrations from 1898 to 1945 and from 1978 to the present, so a song from five years ago is in there if anybody filed one.
Do this first, because a registration filed at release is already a public answer to the question you are about to answer privately. If the two disagree you now have two answers in the world rather than one, and the second one does not quietly replace the first.
Where each register sits and what it actually pays is its own subject. Changing numbers that are already registered is a procedure with an order to it, and filing a competing claim over a disagreement that is still live is how the money stops moving. Read first. File second.
Does a copyright registration settle who wrote it?
No, but an early one is considerably harder to argue with than a late one.
This is where the five years in the title stops being a figure of speech, because section 410(c) draws a line at exactly that mark:
In any judicial proceedings the certificate of a registration made before or within five years after first publication of the work shall constitute prima facie evidence of the validity of the copyright and of the facts stated in the certificate. The evidentiary weight to be accorded the certificate of a registration made thereafter shall be within the discretion of the court.
Read the dates in that against the dates on your song. A registration filed when the record came out sits inside the window, and the facts stated on it — including the writers it names — arrive already carrying weight. One filed today, for a song published five years ago, sits outside it. Still admissible. Worth whatever a judge decides it is worth.
Prima facie evidence is not the last word. It is the evidence the other side now has to answer, which is a different position to be standing in than the one where you have to prove your own case from scratch. That is the practical difference between arriving with a document and arriving with a story.
How do you ask four people to remember one session?
Separately, in writing, and only after you have written down and dated your own version first.
Write yours before you send anything. Not because your memory is better than theirs — because the number you send first is the number you get back, and then you own one account with four signatures on it rather than four accounts you can lay side by side.
Ask what each person did. What they wrote, what they remember writing, who else was in the room. Do not open with 40/30/30 and ask whether everyone is comfortable, because everyone is always comfortable right up until the sync request arrives.
Then take the four answers, find what all of them agree on, and put the disagreements in a list. That list is the actual work. Everything else is transcription.
Do you date it five years ago, or today?
Today. Sign it on the day you sign it.
The date the writers originally agreed goes inside the document as a stated fact — a line saying that these writers agreed these shares in March of that year and are recording that agreement now. That sentence does real work, and it does it without anybody pretending about anything.
Do not backdate it. Not out of scrupulousness and not out of caution, but because a document carrying a false date is a document you now have to explain, and the entire fucking point of writing something down was to stop explaining it.
Whether a later writing can carry an earlier transfer is a real question and the courts do not all answer it the same way; the piece on what makes a split sheet enforceable has the disagreement and its limit. What matters here is the plain half of it: a later writing that confirms an agreement which genuinely happened is an ordinary thing, and a later writing pretending to have been made earlier is not.
Once the numbers are agreed and the date is straight, you are filling in an ordinary document. Ours is here, and there is nothing to sign up for.
Has a clock already run out?
Probably not, and what decides it is whether anybody ever plainly refused your claim.
The US limitation period is short and looks fatal at first glance. Section 507(b):
No civil action shall be maintained under the provisions of this title unless it is commenced within three years after the claim accrued.
Three years, against a song that is five. The reason that is usually survivable is the phrase at the end: after the claim accrued. In Garza v. Everly, the Sixth Circuit stated the rule for co-authors — such a claim “accrues only once,” at the time of repudiation, and the three-year period “begins to run between co-authors when ‘there is a “plain and express repudiation” … by one party as against the other.’”
Silence is not repudiation. Five years of nobody mentioning the song is five years of nothing starting. In that case what the court was looking at was a letter and a telephone call, later memorialised in a signed release — somebody being told, in terms, that they were not an author. Whether anything said to you five years ago amounts to that is fact-specific, and it is a question for a lawyer rather than for us.
Canada runs three years too, from a starting line it states outright. Section 43.1 of the Copyright Act gives a remedy only where proceedings are commenced:
within three years after it occurred, in the case where the plaintiff knew, or could reasonably have been expected to know, of the act or omission at the time it occurred; or … within three years after the time when the plaintiff first knew of it, or could reasonably have been expected to know of it
Two systems, two routes, and neither of them starts the clock at the session. Both start it at the moment you knew, or should have. Which is worth understanding before you send the email that tells four people something they did not know.
When does this stop being reconstruction?
When somebody disagrees rather than forgets.
Reconstruction assumes everyone wants the same document and nobody can remember the numbers. The moment one writer says no instead of I don’t recall, the job changes underneath you, and what you can do without their signature is a different piece with a different order of operations. If the argument is about whether somebody is a writer at all, that is the authorship question and it is worth settling on its own terms before anyone signs anything.
And if a co-writer has died or cannot be found, this is not the page for it. There is no clever workaround in here, because an estate is a real legal entity and a missing person is a real search, and both are somebody’s actual job.
Everything else is one afternoon with old files, three awkward phone calls, and a document nobody wanted to make. Make it anyway. The next person who asks about this song will be a publisher, a supervisor, or a buyer, and not one of them accepts we always meant to.
FAQ
Can you make a split sheet for a song that is already released?
Yes. Nothing about a release closes the door, and a sheet signed afterwards binds the people who sign it. Changing numbers that are already registered somewhere is a separate job with its own order of operations.
Should you backdate a split sheet to the date of the session?
No. Sign and date it on the day it is signed, and record the date the writers originally agreed as a fact stated inside the document. A document carrying a false date is a document you have to explain.
What if the other writers remember the session differently?
Write each version down separately before trying to reconcile them. A disagreement about whether someone is a writer at all is an authorship question, and a refusal to sign anything is a different problem again.
Does a split sheet signed now cover royalties already paid out?
Mostly not. A new agreement fixes the position going forward, and how far a change reaches backwards through a society’s accounting is largely unpublished by the societies themselves.
Sources
- US Copyright Act, 17 U.S.C. § 410(c) — that a certificate of a registration made before or within five years after first publication is prima facie evidence of the validity of the copyright and of the facts stated in the certificate, and that the weight of a later one is at the court’s discretion.
- US Copyright Act, 17 U.S.C. § 507(b) — the three-year limitation period for a civil action, running from the accrual of the claim.
- Garza v. Everly, No. 21-5530 (6th Cir., 10 February 2023) — that a co-authorship claim accrues only once, at the time of repudiation, quoting Everly v. Everly, 958 F.3d 442, 450 (6th Cir. 2020). The Sixth Circuit’s own PDF of the opinion could not be extracted on the machine this piece was written on; this is the full opinion text, read end to end.
- Copyright Act (Canada), section 43.1 — the three-year limitation or prescription period, and the knowledge test that decides when it starts. Consolidation current to 2026-08-06.
- The MLC, finding publisher information for a registered song — that the Public Work Search is searched by work title and shows a work’s details, including publisher information.
- US Copyright Office, public records — the coverage of the Copyright Public Records System, 1898–1945 and 1978 to the present.
This is not legal advice, and limitation periods in particular are unforgiving of general answers — they turn on specific facts, specific dates and the jurisdiction you are actually in. Every page cited here was read on the date at the top of this piece. If there is real money or a real disagreement in this, talk to a lawyer in the relevant territory before relying on any document.
Keeping the register
Half of this job is finding out what the outside world already says about a record you made five years ago, and that is the part CatalogTracker does by itself: it imports a discography from a name, an Apple Music URL or a streaming link, and pulls what iTunes, MusicBrainz and Credits.fm hold — identifiers, credits, songwriter data — so you are comparing your memory against a record rather than against nothing. It is slow on purpose, because MusicBrainz is throttled to one request a second, and when it cannot match something it says so instead of reporting a tidy success. Signed agreements can be uploaded against the tracks they cover, with a content hash and a warning when the splits have moved since the day it was signed. In development for iPhone.