A Primer On Measuring Music Copyright Infringement

Guillaume Laroche is an LLM candidate at Osgoode Hall Law School and a Graduate Student Member of IP Osgoode. Prior to entering law, he completed studies in music performance, composition and theory at the University of Alberta, the Schulich School of Music of McGill University, and in England.

At the Society for Music Theory’s 2011 Annual General Meeting, held October 27-30, 2011, in Minneapolis, I gave a poster presentation, “Striking Similarities: Quantifying Music Copyright Infringement”.  At the heart of my presentation was the problem that courts have of articulating a consistent standard of what exactly constitutes music copyright infringements, whether using the test of “substantial similarity” or not.

Music copyright infringement actions are said to be among the most unpredictable cases for lawyers. Little wonder, too; except in cases of glaringly obvious copying, the line between using a common musical device and borrowing a substantial portion of a prior work is anything but clear. Is a tune that sounds a little bit like another one copying and thus infringing copyright, or do both tunes just make use of elements common to the style? Such distinctions are not always immediately clear. As an additional barrier, our language, and thus the courts’, tends to be imprecise in terms of outlining different degrees of similarity on a spectrum. We can describe two things (musical or otherwise) as being similar, comparable, homogeneous, indistinguishable or somewhat alike, or we can talk about how they are dissimilar, rather distinct, or unalike; but there isn’t much middle ground available in terms of describing degrees of similarity or dissimilarity without using additional (and usually only vaguely helpful) qualifiers.

My project attempts to tackle this the above thorny legal problem by using various music analytical techniques and relating them to the application of copyright infringement. My approach starts with borrowing some techniques for measuring musical similarity from the field of music informatics—i.e., that discipline best known for bringing you Shazam, Pandora, Grooveshark, etc.—and seeing how some of its metrics for doing so could be useful in assessing music copyright infringement actions. Of course, there are some technicalities to work out; the above services tend to compare sound signals (i.e., recordings) rather than musical features in and of themselves, which stay constant across different recordings/performances of a work. With a little technical know-how, though, it’s possible to adapt the ideas of informatics to this new context. So, in sum, instead of trying to qualitatively describe (dis)similarity between two musical works, I propose we measure it.

I test my method by using a wide variety of prior caselaw on music plagiarism decided between 1910 and 1990 in the US. Each case supplies a pair of allegedly similar works, and a verdict by the court on whether the similarities of one work are substantial enough so as to infringe the copyright of the other work. Using these data, I apply my model for measuring similarity to the works under scrutiny, and correlate the results to the probability of a finding of infringement. My initial results show that while there is no ‘bright line’ standard of infringement, there is nonetheless a good degree of consistency in the caselaw as to what degree of similarity between two musical works constitutes infringement. With this degree of similarity numerically codified, it could potentially be much easier to make future judgments about the infringing nature of pairs of other vaguely similar musical works, getting some idea beforehand of if the similarity shown is nowhere near the infringement line, flirting with it, or clearly far beyond it.

That being said, this approach, if indeed it is applied to new music copyright infringement actions, should serve mainly to supply courts with additional information, as opposed to decide the case in and of itself. While language is functionally incapable of precisely distinguishing between two pairs of works that share 60% of their features in one case and 70% in the other, measurements succinctly offer this kind of information in an easy-to-understand format. By no means should this information entirely replace a circumspect examination of the plaintiff’s and defendant’s cases, a critical evaluation of the evidence, and expert testimony offered by qualified music scholars which is specific to the style of the works under consideration; rather, it should supplement it. Said otherwise, similarity metrics should exist as tools to aid the court in its work, not as arbiters of infringement in and of themselves. But if using them helps the court make more informed decisions, then perhaps we ought consider what their proper role in a court of law ought be.

The chance to engage in discussions with scholars who study musical similarity and relatedness for purely musical reasons as part of the Society for Music Theory’s conference was a welcome opportunity to reflect on further improvements to my model in a community well-versed in the problems of musical similarity. Although I am temporarily setting this project aside because my new research on music and copyright at Osgoode Hall Law School takes on more of a qualitative aspect than a quantitative one, I hope to publish a detailed version of my initial findings on this matter soon.