WIPIP Concurrent Session #3: Copyright Doctrine

Christopher Buccafusco (& Rebecca Tushnet), Base Rate Neglect in Copying-in-Fact

Comes out of an excellent Buccafusco paper about the failures of copying in fact, which led me to think about base rate neglect in cases where plaintiff’s expert claims that it’s not possible that these similarities arose in the absence of copying. Mostly music infringement.

To prove copying in fact, you show access and similarities probative of copying. Much of the litigation thus turns on what it means for a similarity to be probative of copying.

One must wonder how the plaintiff composed the song if it wouldn’t be possible to create that sequence without copying, but the real question is what we’d need to know to say that it’s unlikely that these similarities would arise without copying.

Given degree of similarity X, how often will a song have degree of similarity X to other songs in the corpus? How often will that similarity come because of copying, rather than independent creation or joint copying from a common source? At best, these experts glance at question 2 and not at question 1.

This is a problem of base rate neglect. Imagine a warehouse w/barrels of flour that may break loose and injure people. If negligent, barrels break loose 90% of time; if properly handled, 1% of time. If they properly handle barrels 99.9% of the time, then, given a loose barrel, what is the chance that it was because of negligence? Humans, including federal judges, get this wrong all the time, but the chance that a loose barrel is due to negligence is about 8.3%.

What if they handle barrels properly 99% of the time? Small change in underlying probability makes it over 47% likely that it is because of negligence.

So what we need to know but don’t: how often will a song have X similarity with other songs? And how often is that b/c of copying?

What should we do? For now, nothing: until experts can estimate either of these values, courts shouldn’t accept their testimony. Or Ds could introduce their own experts.

Is there anything we could start to do to get closer? Calculate total number of six note sequences in music? Some will occur more often than others due to how music works? We could then try to estimate how common it is for a 6-note sequence to show up in multiple songs. We might not be far from being able to do that.

But the best we’ve done is estimated the first question. We still need to know how likely it is that a 6 note sequence will show up in 2 songs due to copying v. not copying/independent creation. Questions: what are the other options?

Zahr Said: thinking about use of barrel, you’ve incorporated the normative “proper handling” conclusion—it produces a conclusion b/c you can rule out things like other causes. It allows us to eliminate plaintiff’s contribution to the harm. But it’s only applicable if we think it’s under the exclusive control of the D or can attribute to D.

A: OK, so do “HIV test everyone in the country.”

Said: In cases of copying, you might not find something like an HIV test—those other steps are present in other copying cases—it’s hard to rule out other inferences.

Wu: Expert says I’ve never seen this pattern before—doesn’t that say something about the base rate?

A: This is base rate misestimation, then. They’re not doing step 2 in any event. [Step 2 requires generalization—it’s not that I’ve never seen this person test positive for HIV, it’s about what the positive test/similarity means, and so the entire problem of base rase neglect

Vishnubhakat: is a jury any better at Bayes’ theorem than a judge?

A: no, but an expert would have to explain what you’d have to know to answer the question. Not worse to draw attention to the problem.

Neel Sukhatme: It’s not even a fixed target—the rate of copying is dependent on what the law is [e.g., who gets interpolation credits]—will change similarity going forward. And it will vary across fields. A bunch of people will announce their appropriation art; others won’t. But look how quickly the answer changes w/ very small changes in underlying math.

Rosenblatt: options: a) Daubert, b) counterexperts, c) impossible math. But what about presumptions and burdens of proof? We can’t presume from a combination of access and similarity that copying in fact took place. One possibility: Therefore, in the absence of identical copying or admission of copying, there should be a presumption against it.

Gena Feist: is infringement/copying binary in the same way? [I think this is in the P’s hands—if the expert says this is the copying, then that’s what we should be testing.]

Zvi S. Rosen, Fog and Fiction and Copyright Term Extension aka Who Framed Mickey Mouse?

In 1963, Disney expressed skepticism about monopoly aspects of extended term and “expressed doubt that Congress would approve a longer ownership period.” Authors’ groups were proponents. Big proponent: Gershwin family, unabashed about lobbying, saying it was less about money than about control, though of course it was also about money.

1993 hearing, transcript not public before. MPAA is there; “Coalition of Creators and Copyright Owners,” represented by songwriter (ASCAP, music, dramatists, BMI, SESAC, VLA, Writers Guild); National Music Publishers. Largely royalty-dependent groups.

In subsequent hearings, Quincy Jones, Henry Mancini’s widow, Bob Dylan, Don Henley, Alan Mencken, Stephen Sondheim—individual creators, especially composers. Disney didn’t send anyone individually, though MPAA does show up. Individual composers were the ones losing European revenues because of the rule of the shorter term.

But there was some attention to Disney’s interests as well in 1998—Disney did endorse term extension. Wondering if DMCA was also on their minds. It wasn’t that they opposed extension, but Bob Dylan was important too, and the DMCA was also tied to term extension.

Does it matter whether this was the Mickey Mouse Term Extension act? How do we tell stories about the law? Laws can be more complicated than popular narrative; legal narrative v. legal history.

Q: why not assume that Disney, like the music publishers, would send individual creators to front for them?

Bruce Boyden: can you interview people on the Hill at the time to see what they were talking about.

Hughes: it was the Fairness in Music Licensing Act, not the DMCA, which was intertwined. That was a great trade for restaurants.

A: Bono’s widow said he would have hated the final bill because of FMLA.

RT: (1) Important to have film in play b/c in Eldred incentivizing preservation rationale was all about film. (2) A name can be given you by your enemies: rhetorically important to opposition: “death tax”/this was a Frank Luntz move, which isn’t surprising. (3) “So what?” Opening up possibilities for change: new scholarship on FTC and kidvid serves to restore confidence in FTC’s ability to act. May suggest possibilities now.

Xiying Tang: European context—music ends up w/ a very different term than European term.

Lemley: opposition at time viewed it as Disney-driven, even if we weren’t right. Disney’s lobbying expenses for that year are relatively low and don’t spike up in that year. They definitely donated to many committee members but not unusually so.

Tyler Ochoa: 1993 is before European extension—does this happen w/o European extension?

A: hard to figure that out, including hard to figure out who pushed it in Europe.

Zahr Said, Shotgun Damages in Copyright

Multipliers used to ratchet up damages. Before 2016, appeared to be that these multipliers were impermissible punitive damages. Since 2016, shift either in practice or characterizations—quite a few different kinds of multipliers for exclusivity/scarcity/quality being applied to actual damages before final judgment separate from statutory damages. But their increased uptake hasn’t been noted in treatises/scholarship. Part of larger pattern of “shotgun damages,” working definition: damages characterized by inflationary demands for multiplied awards; coercion; scattershot approach. Why are they being awarded?

Older cases reject expert guesses about multipliers that weren’t willing buyer/willing seller rates. E.g., these images were good, so you should apply a quality multiplier; scarce, so you should apply a scarcity multiplier to the standard quoted rate; also P didn’t want to license them so should get an exclusivity multiplier. In 2016, Leonard v. Stemtech, 3d Cir, affirms that award! Cited over 130 times since then—cottage industry using scarcity, exclusivity, and quality multipliers; tons of default judgments. But the exclusivity multiplier is based on P’s own quoted rates; P may testify that it would have asked for more because it didn’t like cut of licensee’s jib.

Problematic that so many of these are done on default; actual damages that get ratcheted up then get funneled into statutory damages; a low showing of willfulness can yield a very high award.

Lemley: Patent law uses the actual license amount; there’s no reason to use a multiplier when we actually know. Policy question: do you want to make it punitive to punish not transacting? But that’s supposedly not the law we have, and also brings in punitive damages constitutional jurisprudence.

Jessica Silbey, An Empirical Study of the Copyright De Minimus Defense

How has it evolved? What works are most affected? Was Ringgold a turning point? Is de minimus about taking a little bit or using a whole work in an insubstantial way? Ringgold=aesthetic analysis of function of use in background; that’s interesting.

Only fragments and parts pre-1978; after, whole works sometimes; only direct competitors pre-1978, but more potential licensees after. Standard evluation confirms Oren Bracha’s critique: not de minimis = subtantially similar and vice versa, with caveats. More qualitative assessments than anticipated, especially since 1998. Works almost like a fair use defense. 30% in dataset prevail on independent de minimis defense, which is not just substantial similarity and not just fair use, contrary to Nimmer and previous study.

Hived off of fair use and became its own defense in early 20^th century, then became swallowed by substantial similarity, which is a huge problem b/c it was about market substitution and it means the derivative work rights ballooned.

Where is de minimis used elsewhere? Not in constitutional violations, crime, or trespass.

Brauneis: maybe changes with rise of concept of fragmented literal similarity, invented by Nimmer in 60s: maybe shows up with “literal” takings more?

A: yes, essentially all except for a couple of takings.

Q: some cases are now recorded that wouldn’t have been before WL. Hard to deal with that but must take that into account.

Guy Rub, Revisiting the Copyright-Contract Conflict

Most courts said that contracts were never equivalent to © rights. Second Circuit changed course in ways that weren’t clear. Universal, 2019: Software company A sells to B with a contract that maybe prohibits modification.  Genius, 2022: website sues Google for breaching TOU prohibiting commercial scraping (copying)—preempted.

The main issues: under this analysis, more restrictions on use are less likely to be preempted.

Conflict premption! 301(a) asks the wrong question. In re Jackson (2d Cir 2020): what values does state law, as applied to the case, protect? What and how meaningful is the harm to federal copyright policy? Could attend to contracts of adhesion in that way. Genius briefing doesn’t address the core question: are they undermining (c) policy and the scheme that Congress is trying to promote? Reverse engineering cases would come out differently if you used this framework as well.