It’s frequently the first question asked. And a forensic musicologist’s unsatisfying short answer is, “substantially so.”
There’s no percentage. No threshold. No rule that says 30% similar means infringement or 10% means you’re safe. There is no completely reliable rule as to how many notes in a row, or such. If there were, all those silly copyright infringement checkers would be worthwhile, but they’re not. The question courts ask is whether it was copied, whether the similarity involves expression that’s actually protectable under copyright law and to a degree that it shouldn’t be allowed.
The Question Isn’t Similarity — It’s What’s Similar
Copyright protects original expression. It may or may not protect chord progressions, rhythmic patterns, tempos, grooves, genres, or general stylistic approaches. I intentionally listed things often said to be unoriginal and unprotectable. Those elements are the shared vocabulary of music — no one owns them, and any song can use them freely, unless they can’t. And there are always degrees to which they can’t; there have to be; it’s physics. So, we have to live with the uncertainty, for the purposes of this page at least.
What copyright protects is specific original expression: like a melody with enough distinctiveness to be protectable, or an involved harmonic sequence that goes beyond common practice, lyrics beyond banalities. And even within those categories, the protection only extends to what’s actually original. If a melody resembles equally something that came before both songs, the available analysis is that neither song owns it.
So the real question isn’t merely how similar the songs are, such that we can just count things, although we do that. It’s also whether the similarity is located in something that’s (a) protectable and (b) not already in the prior art. You can see how immediately circular and interdependent the law chooses to be. It wants structure. Courts are trying to arrive at both truth (did it happen) and wisdom (is this what copyright is for)? Two songs can be substantially similar in overall feel and have no viable claim between them. Two songs can have a relatively small area of overlap that still constitutes infringement if that overlap is in a distinctive, protectable element.
What Courts Actually Look At
The legal test for infringement, with some subtleties from geography to geography, has two elements: copying and unlawful appropriation.
Copying means the later work derived from the earlier one, either through direct access we can establish or through striking similarity so obvious (to a forensic musicologist) it implies copying and therefore access (here again, the circularities). Independent creation of an identical melody is a defense, though an increasingly difficult one to establish. If it was independent, it wasn’t copying. It’s a coincidence.
Unlawful appropriation means the copying involved protectable expression and to a certain degree, as opposed to a trifling amount too small for copyright law to be bothered. This is where most infringement claims should succeed or fail because it’s in my opinion where the wisdom the courts seek is most vested. A would-be defendant who can show that the similarities in the songs, said to have been copied, are actually things like common chord progression, standard rhythmic patterns, material that exists in earlier songs, or other.
Forensic musicologists’ analysis addresses this: not whether similarity exists but whether it is in protectable material, whether that material was original to the plaintiff’s work or drawn from a commonly owned pool, whether it was copied, and ultimately whether society should care. Copyright has a noble intent. Expert analysis is just a piece of the process. The courts have a job to do that serves the greater good. People want art. Art is enjoyable, right?
Percentage Scores From Tools Miss the Point, badly, embarrassingly, and annoyingly.
Automated similarity tools produce numbers. The numbers measure overlap. Maybe they count shared intervals, rhythmic pattern events, or other structural similarities. They don’t measure whether any of that overlap is in protectable expression. They can’t do the “significance” or “substantiality” work.
A tool might flag 40% melodic similarity between two songs that share a common phrase drawn from a folk melody that predates both. Neither song infringes the other; both drew, in a sense, from common source material. The 40% number is accurate as a measurement and irrelevant as a musicological or hopefully legal determination. It would be tragic to find one of these in litigation.
The reverse is also true. A song might show relatively low overall similarity to another work but share the specific four-bar hook that makes the earlier song distinctive and identifiable. Small quantitative overlap, serious legal exposure.
The percentage doesn’t answer the question, not remotely.
Here to help.
If you’re worried about similarity, being too close, however you might think of it; say, your song and another song share something that you can hear as the same thing, the question is worth getting answered properly rather than guessing.
Musicologize offers the best and most helpful originality clearance analysis, taking the similarities that are actually there and evaluating them against what copyright law treats as protectable in terms of substantial similarity. The analysis will say if you’re too close or not. And you’ll have a written opinion that documents the assessment and demonstrates your care and concern.
To discuss a matter or ask a question — call, email, or schedule a preliminary call.
✆︎ Call: (212) 217-9512 ✉ Email: brianmcbrearty@gmail.com 📅 Schedule: Book a free preliminary call