Strict adherence to Arnstein is inefficient and potentially corrupting.
Arnstein v. Porter (That’s Cole Porter) put in place the two-pronged test for infringement. The idea is that we first ask if someone copied a work from a pre-existing work, and then, secondly and separately, whether that copying constitutes improper appropriation, which is comprised of substantial similarity and originality.
But forensically, these are not independent or sequential questions, or at least it’s worth considering whether they ought to be. The former question, that of copying, is about access and similarity. The latter is similarity and originality. One analysis informs the other.
To illustrate, let’s say a forensic musicologist might transcribe two works and hold the velum up to the light to see how alike two melodies are. They might count up the notes in common, divide them by notes in total, and thus substantiate a client-serving percentage of identical notes, and make a declaration of how this “could never have occurred except as a result of copying.” And that’s when Arnstein is supposed to continue on to consider the question of originality. This is, at best, inefficient. It is the originality of the notes, phrases, harmonies, etc, that qualify or weaken the case for copying in the first place. If, for example, we find eight notes in a row in common and they both happen to be Sol Mi Do Sol Mi Do Sol Mi Do, instead of thinking, “wow, that’s 9 identical notes in a row!!” A thoughtful musicologist thinks, “that’s an arpeggio; the same one, three times.” That’s not so suspicious. If it’s a string of neighboring notes, such as Do Re Mi Re Do Ti Do Ti La Sol, I’m not only going to think, “that’s stepwise motion along a scale (all the notes are adjacent to one another in a major scale), and therefore not so suspicious, but I’m also going to think, “that’s how ‘Penny Lane’ begins,” and so, not original.
“Nine identical notes in a row” sounds compelling to a layperson, and maybe even more so to a statistician! If your thinking is at all anchored to the question of “what would simple probability expect?” then nine notes in a row is very unlikely, but if you’re a musicologist, it’s a nothing-burger for a variety of reasons. Did I need to get to the end of the process, consider all the similarities, and then filter for arpeggios and scales, and conduct a prior art search to uncover Beatles tunes? Of course not, such similarities were effectively grayed out in my mind from the start, and since access generally needs to be inferred partly via similarity, such lack of originality feeds back and negates that inference. So originality isn’t at all the second or final thing you need to assess. As a forensic musicologist, after identifying the elements most at issue, originality is integral to similarity analysis from the start.
Let’s look at it another way.
Defendants, of course, rarely admit either access or copying gladly. Denial is expected, met with scrutiny, and the courts are left to infer from the forensic evidence whether access and copying happened. Forensic musicologists look at the evidence and try to help the courts substantiate that inference. Given Arnstein, we might consider the matter thusly:
Hypothetical question: Did Composer B have access to Song A?
Hypothetical train of thought: Song A was recorded in Composer A’s basement, never released, never posted to YouTube, never performed outside of Saugerties, NY, childhood home to Jimmy Fallon. There’s seemingly no access, so we don’t move on to the supposedly subsequent question of similarity? That makes no sense. We must look at the notes and consider whether the songs are so similar that IT’S THE ONLY POSSIBLE WAY these two songs could share such astonishing similarity. And indeed, this is how access is established quite often — a forensic musicologist concludes that the high degree of similarity demands a presumption of access. It turns out Jimmy Fallon’s cook has a side hustle as a producer and he knows Justin Timberlake’s cook, also a producer. There was access after all. Access in the age of YouTube and Soundcloud is impossible to completely dismiss. You can’t entirely prove a negative, and often the evidence is just enticing enough that the case proceeds.
Related to that, there is, or was, the “Inverse Ratio Rule” which recommended that triers of fact in infringement cases should be accepting of less compelling evidence of similarity where there’s very strong evidence for access, and vice versa. In other words, a preponderance of one helps to compensate for a paucity of the other. It’s intellectually reasonable, but it perhaps asks too much reasonableness of the system. I’d have liked it better were it renamed the “Inverse Ratio Consideration,” because as a so-called “rule” it was always bound to be poorly applied. Suppose the author of “Mary Had A Little Lamb” wanted to sue Run D.M.C. because they think “Mary Mary (Why Ya Buggin?)” sounds too much like their nursery rhyme song. They can pretty easily establish one hundred percent absolute comprehensive familiarity with their song. But this would have absolutely no bearing on whether the songs were at all the same. No amount of access makes dissimilar music more similar. On the other hand…
There’s a more favorable light and it’s interesting that the rule was famously thrown out by the Ninth Circuit following a case where it might well have been applied thoughtfully and meaningfully. In the “Stairway To Heaven” case, Jimmy Page was found to have a copy of the plaintiff’s record in his personal collection, and the song had been performed by the warm-up band Spirit while Zeppelin was touring. That’s considerable access, but perhaps not enough. Page had a huge record collection and denied ever having heard “Taurus.” Could we listen to Stairway and infer that he’d probably heard it despite his denial? That’s the forensic task after all. Perhaps if we applied the inverse ratio rule we’d say, “pretty good similarity, enough to beef up the access probability satisfactorily.” Similarity analysis gets over the access hump. Musicologize, it so happens, doesn’t agree that Stairway is similar such that copying can be inferred, even if we assumed or believed in the access. But while Stairway doesn’t get any more similar if Page listened to it every day with his breakfast, does some access with some similarity increase the likelihood of “copying?” It has to.
On the other hand, in my example, the songs were so much the same that we inferred the far-fetched Timberlake chef story really did happened. There the inverse ratio rule made perfect sense, logically at least. The question then becomes, “is the forensic musicolgy correct about the probative value of the similarity?”
“Often wrong but never in doubt,” I often review musicologist reports that express no doubt where considerable doubt was readily available. But that’s beside the point. We endeavor to ask and argue the right questions and accept that the adversarial nature of things is partly unavoidable, and needs only to be reasoned through.