Arnstein v. Porter: The Two-Step Framework Behind Every Substantial Similarity Test

Few copyright decisions have aged as strangely, or as durably, as Arnstein v. Porter, 154 F.2d 464 (2d Cir. 1946). The plaintiff was Ira B. Arnstein, a composer and serial litigant who had already lost five similar suits against other defendants. The defendant was Cole Porter, then at the height of his fame. Arnstein claimed that Porter hits including “Begin the Beguine,” “Night and Day,” and “My Heart Belongs to Daddy” had been plagiarized from his own compositions, among them “The Lord Is My Shepherd” and “A Mother’s Prayer,” and he supported the claim with allegations that Porter had employed “stooges” to shadow him and ransack his apartment for manuscripts.

It is the sort of complaint a district court dismisses without much thought, and the district court did exactly that, granting summary judgment for Porter. Yet when Judge Jerome Frank wrote for the Second Circuit reversing that judgment, he produced the analytical framework that still organizes copyright infringement litigation in the United States: the separation of infringement into two distinct questions, copying and improper appropriation, each with its own evidence and its own decision-maker. Every modern substantial similarity test, from the Second Circuit’s own refinements to the Ninth Circuit’s extrinsic and intrinsic analysis, descends from this case.

At a glance

• Case: Arnstein v. Porter, 154 F.2d 464 (2d Cir. 1946)
• Decided: February 11, 1946, opinion by Judge Jerome Frank, Judge Charles Clark dissenting; summary judgment for Porter reversed and remanded for trial
• Holding: Infringement requires proof of copying (which may rest on access plus similarity, aided by expert dissection) and separately proof of improper appropriation, judged by the response of ordinary lay listeners; disputed credibility on those issues cannot be resolved on summary judgment.
• Significance: The foundational two-step infringement framework from which all modern substantial similarity tests descend; its summary judgment holding, by contrast, has been repudiated.

The problem the court was solving

Under the Copyright Act of 1909, as under 17 U.S.C. § 106 today, the copyright owner held the exclusive right to copy the protected work. But “copying” conceals two very different inquiries. A defendant may have taken something from the plaintiff as a factual matter and still not be liable, because what was taken was too little, or too unoriginal, to matter. Conversely, two works may resemble each other closely without any copying at all, because both drew on common sources or because composers working in the same idiom converge on similar phrases.

Before Arnstein, courts tended to run these questions together into a single impression of similarity. Judge Frank pulled them apart. As he framed it, the plaintiff must prove two separate elements: “(a) that defendant copied from plaintiff’s copyrighted work and (b) that the copying (assuming it to be proved) went so far as to constitute improper appropriation.”

On the first element, copying as a factual matter, the plaintiff may offer direct evidence such as an admission, but almost never has it. Instead, copying is usually inferred from circumstantial evidence: proof that the defendant had access to the plaintiff’s work, combined with similarities between the works that make independent creation improbable. The two inputs work together. Where there is evidence of access and similarities exist, the trier of fact decides whether those similarities are sufficient to prove copying. Where evidence of access is absent, Frank wrote, the similarities must be “so striking as to preclude the possibility that plaintiff and defendant independently arrived at the same result.” Similarities are indispensable either way: if there are none, no amount of evidence of access will suffice.

On the second element, improper appropriation, the question changes character entirely. Copying alone is not infringement; the defendant must have taken material protected by the copyright, and enough of it to matter to the audience the work was written for.

Experts for one question, lay ears for the other

The most distinctive feature of Arnstein is its allocation of evidence between the two steps. On the copying question, the court welcomed technical analysis: “analysis (‘dissection’) is relevant, and the testimony of experts may be received to aid the trier of the facts.” Musicologists may line up the scores, count the matching intervals, and testify about the probability of coincidence, because the factual question of copying is exactly the kind of question expert comparison can illuminate.

On the improper appropriation question, the court excluded the same evidence. The test, Judge Frank wrote, is “the response of the ordinary lay hearer,” and on that issue “dissection” and expert testimony are “irrelevant.” The reason is economic rather than aesthetic. The value of a popular song lies in its appeal to the public that buys it, so the wrong occurs only if the defendant took “so much of what is pleasing to the ears of lay listeners, who comprise the audience for whom such popular music is composed, that defendant wrongfully appropriated something which belongs to the plaintiff.” The plaintiff’s protected interest is the lay audience’s approbation, and only that audience’s reaction can measure whether it was taken.

This division has proved remarkably persistent. The Ninth Circuit’s extrinsic test (objective, dissection and expert testimony permitted) and intrinsic test (subjective, the ordinary observer’s impression) announced in Sid & Marty Krofft Television Productions, Inc. v. McDonald’s Corp., 562 F.2d 1157 (9th Cir. 1977), are a direct translation of Arnstein’s two steps. The Second Circuit itself later sharpened the vocabulary, using “probative similarity” for similarities that help prove factual copying and reserving “substantial similarity” for the appropriation question, but the architecture is unchanged.

The summary judgment holding, and its fall

What made Arnstein notorious in its own time was not the doctrinal framework but the disposition. Arnstein’s access story was, as the majority conceded, “fantastic”: spies, shadowing, a ransacked apartment. Porter denied ever hearing the compositions or knowing Arnstein existed. The district court found the whole thing incredible and granted summary judgment.

Judge Frank reversed, reasoning that Porter’s denials put his credibility in issue, and credibility is for the jury. A jury observing Porter’s demeanor on cross-examination might disbelieve his denial of access, and the similarities between the works, after the court itself listened to the recordings, were not “so trifling” that a trial judge could properly direct a verdict on improper appropriation. In the opinion’s most quoted procedural passage, the court indicated that summary judgment should be denied where there is “the slightest doubt as to the facts.”

Judge Charles Clark, a principal draftsman of the Federal Rules of Civil Procedure, dissented sharply, warning that the majority’s approach would gut summary judgment and force defendants to trial on claims no reasonable factfinder could accept. History has sided with Clark on this point. The “slightest doubt” standard was steadily eroded and is now flatly inconsistent with the Supreme Court’s summary judgment trilogy, including Anderson v. Liberty Lobby, Inc., 477 U.S. 242 (1986), which requires a genuine dispute of material fact that a reasonable jury could resolve for the nonmovant. Courts today routinely grant summary judgment in copyright cases, on both copying and substantial similarity, where no reasonable juror could find for the plaintiff.

The trial Arnstein won on appeal ended the way the district court predicted: on remand, the jury found for Porter, and the Second Circuit affirmed that judgment later the same year. Arnstein v. Porter, 158 F.2d 795 (2d Cir. 1946) (per curiam).

Why the framework endures

The durability of Arnstein owes something to the problem it isolates. Infringement doctrine has to police two different errors at once. It must not let defendants escape by pointing to everything they did not take, and it must not let plaintiffs monopolize ideas, styles, and commonplace building blocks by dressing an ordinary resemblance up as theft. Splitting the inquiry lets courts use rigorous, technical evidence where rigor helps (did copying happen?) and audience reaction where audience reaction is the very thing the statute protects (was the taking wrongful?).

The framework also has well-known stress points. Scholars, notably in Mark Lemley’s critique of what he calls our bizarre system for proving infringement, have argued that Arnstein got the allocation backwards: juries hear expert dissection on the technical copying question, where they may be least equipped to weigh it, and are then forbidden analytic help on the appropriation question, where filtering out unprotected elements is most needed. The Ninth Circuit’s modern music cases, which permit expert testimony to identify objective similarities in protectable expression under the extrinsic test, are in part a response to that criticism. But no court has abandoned the basic two-step structure.

Open questions

Arnstein left, and its successors still debate, several unresolved problems. First, who is the “lay listener” when the audience for a work is itself specialized? Some circuits have adjusted the ordinary observer standard to an “intended audience” standard for works aimed at experts, such as software or choral arrangements. Second, how should the appropriation step filter unprotected material? If lay listeners react to unprotectable elements (a common chord progression, a shared public domain source), an unfiltered audience test risks protecting what copyright does not cover, which is why the Second Circuit developed the “more discerning observer” variant for works that mix protected and unprotected elements. Third, the relationship between access and similarity remains contested. The Ninth Circuit abrogated its “inverse ratio” rule, under which a strong showing of access lowered the standard of proof of substantial similarity, in Skidmore v. Led Zeppelin, 952 F.3d 1051 (9th Cir. 2020) (en banc), on the ground that the rule “is not part of the copyright statute, defies logic, and creates uncertainty for the courts and the parties.” That rule was a Ninth Circuit construct first articulated in 1977 rather than an Arnstein holding, and the Second Circuit had itself rejected it in Arc Music Corp. v. Lee, 296 F.2d 186 (2d Cir. 1961).

Implications for creators and businesses

• Access evidence is discoverable and decisive. Because copying is usually proven circumstantially, records showing who received a demo, a manuscript, or a pitch (and who did not) often determine whether an infringement claim survives. Submission logs and clean-room procedures exist because of this framework.
• Expert reports win the first step, not the second. A musicologist can establish or destroy the inference of copying, but the ultimate appropriation question belongs to ordinary ears. Litigants should build both cases: technical dissection for copying, and a presentation that lets the factfinder simply listen.
• Independent creation is a complete defense. Arnstein’s structure means a defendant who never copied cannot infringe no matter how similar the works are. Documenting the creative process contemporaneously is the cheapest infringement insurance available.
• Do not rely on the “slightest doubt” language. The two-step test survives; the near-absolute right to a trial does not. Weak claims are now regularly resolved on summary judgment under modern Rule 56 standards.

Frequently asked questions

What are the two elements of copyright infringement under Arnstein v. Porter? First, the plaintiff must prove the defendant actually copied from the protected work, which can rest on direct evidence, on access plus similarities, or on similarities so striking that they rule out independent creation. Second, the plaintiff must prove the copying went far enough to be an improper appropriation, judged by the reaction of ordinary lay listeners rather than experts.

Can expert testimony be used to prove copyright infringement in music cases? Under Arnstein, experts may dissect and compare the works on the first question, whether copying occurred at all. On the second question, improper appropriation, the Second Circuit held that expert dissection is irrelevant because the test is the spontaneous response of the lay audience for whom the music was written.

Is Arnstein v. Porter still good law? Its two-step structure remains the foundation of infringement analysis in every circuit, including the Ninth Circuit’s extrinsic and intrinsic tests. Its separate suggestion that summary judgment should be denied whenever there is the slightest doubt about the facts has been rejected under the modern summary judgment standards of the Federal Rules.

Authorities and sources

• Arnstein v. Porter, 154 F.2d 464 (2d Cir. 1946): full opinion (PDF); case background and works at issue (GW Law, Music Copyright Infringement Resource). Subsequent history: Arnstein v. Porter, 158 F.2d 795 (2d Cir. 1946) (per curiam) (affirming judgment for Porter after the jury trial on remand).
• 17 U.S.C. § 106 (exclusive rights): Cornell LII.
• Sid & Marty Krofft Television Productions, Inc. v. McDonald’s Corp., 562 F.2d 1157 (9th Cir. 1977); Skidmore v. Led Zeppelin, 952 F.3d 1051 (9th Cir. 2020) (en banc): Skidmore opinion (Ninth Circuit).
• Mark A. Lemley, Our Bizarre System for Proving Copyright Infringement: Stanford Law School (SSRN PDF).
• Case analyses: Casebriefs; Studicata case brief.