Thursday, August 09, 2012

IPSC part 4

Copyright and Music
Robert Brauneis, Copyright and the Dilemma of Defining Musical Works in the Era of Fixed Sound

Initially in copyright: written score was the way music was disseminated and a human performer was necessary to transmit it to the audience—largely amateur performances.  Two-stage music.  Composers wanted to distinguish their activities from performers.  Thus: composition, a deliberative activity allowing rethinking and editing; performance: realtime, low-deliberation, no-editing rendering of score.  Composition: mostly discrete allographic notation; performance: continuous analog sound frequencies.  Stable visually perceptible representation in scores v. purely aural, evanescent phenomenon.

Change: the early sound recording.  Sound recording enables repeatable musical experiences in which notation plays no part: Robert Johnson composed but didn’t notate.  Early recordings include manipulation, even if relatively crude.
Federal copyright protection for scores; inconsistent common-law protection for sound recordings.

Then: changes in technology, production, consumption of music. Multitracking, splicing, reverb and echo effects, electronic sound synthesis, sequencing: fine-grained production of sound at every level.  Folks who do this often no longer use musical notation in any step of production.  Atomic Dog case: in a recording studio without a written score, composed spontaneously.  Burn (Usher song)—co-wrote and recorded over several days. 

Consumption: listeners usually exposed to new song as single integrated musical audio work, rather than live or recorded performances by different performers.  (Me: Not true for a subset: YouTube fans.  Pretty immediate versioning there! Maybe there was an RCA parenthesis, just as some argue for a Gutenberg parenthesis?  He says: but do we hear a single musical work out of that?  Or do we just get a cluster that coheres around a sort of common core?  Each person will choose slightly different things.  Not clear there’s one set that everybody imitates as a union in the center that counts as a musical work.)
1976: no longer need to fix musical work in a score; two works—musical work and sound recording.  Is there an identifiable subset of components of a musical audio work that can be identified as a score? Compositional elements versus performance elements?  Melisma: multiple notes sung on one syllable.  One case filters this out as to the musical work.  Yet, particular melisma in recorded music can be the subject of deliberation, editing, trial and error and even notation.  Even if a particular melisma originated in improvised performance, what makes that different from an improvisation that is then written down?  What about a transcription: what level of detail, and what method of notation?  Example of Born this Way as performed and as written—lots of detail missing from official sheet music, produced within/for a particular market.  What about the basic stuff everybody remembers: empirical question, but people tend to remember somewhat different basic stuff. Maybe there are a cluster of substantially similar derivative works with family resemblances rather than a single musical work core.

Q: another reason for persistence: embodied in int’l IP norms/instruments in copyright rights/neighboring rights. 

Chris Newman: why is music exceptional in this regard?  AV film also has a fixation of a literary work, maybe a choreographic work.  (Doctrine tends to treat the screenplay as subsumed, not a separate work with separate rights.)

A: Do think it’s important that we think of musical works as separate from sound recordings.  Literature is a separate field that doesn’t depend as much on sound.  Music is independent of its written expression as literature as not.  If an actor improvises lines in a film, would we say he has a separate copyright in a literary work?  Glynn Lunney points out that the joint authorship cases tend to deprive actors of any rights, but the theory is that there could be joint authorship at a certain point (though it would be joint authorship of the film then, not the literary work). (Plays: claims to stage directions.)

Jerry Liu, Copyright Complements

Downloading had a huge impact on the music industry: sales have decreased a huge amount.  Has drawn attention to alternative revenue streams.  Sell services and relationships, ancillary markets.  Complementary goods: typically involve two products consumers purchase and use together. Decreasing price of one increases demand for another.  Video games and consoles; printers and ink cartridges; left and right shoes.
First degree complements with music: Products that consumers have to buy and use together. iPod and iTunes.  iPod is priced very high.  Videogame manufacturers tend to follow the razor/blade model, selling consoles below cost.  Advantages: allows price discrimination; risk sharing.  But the iPod reverses the traditional model: sells iTunes music at cost but has a very high profit margin for the iPod.  Underpriced music to compete with piracy; overpriced hardware to recoup investment.  Increasing deadweight loss.  Pricing songs over marginal cost causes inefficient consumption.  But if you price iPod above marginal cost, leads to inefficient consumption of the hardware.  Deadweight loss can become dominant if the hardware market involves monopoly (does it) and the music market involves oligopoly.

Second degree complements: might not consume together: live performance, merchandising. Pre digital era: tickets often underpriced. Scalper prices were much higher than face value.  Because they wanted hard-core fans to come to the concert. Digital era: 82% increase in ticket prices, 5x faster than inflation.  $1.3 billion to $2.1 billion 1996-2004.  Bigger share of income.  Again, not surprising given piracy.  Today people put out albums to promote their tours.  David Bowie: copyright won’t exist in 10 years (said in 2002)—touring will be all that’s left.  New kind of label: Live Nation, a concert promoter established in 2005. 360 deals with U2, Madonna, etc.

Can this make up for recorded music losses?  No.  2007: live performance $2.6 billion, but recorded $6.1 billion even after a big drop.  Top 25 tours are .76% of the number of tours but account for more than half of total revenue. Relying on performance means scaling down the industry. Advantage to established artists, legacy acts with large repertoire and stable fan base. Disadvantages new artists—limited large venues, extensive investment, long-term recoupment: higher risk.

Third degree: not provided by the same firm, but still more valuable for others: ad-supported music and sponsorship.  China: a popular tactic.  “Break out of the Window,” written for a software commercial; “I’m Loving It,” for food; “Sweet and Sour, Just Like Me,” for yogurt.
In many areas, people believe greed outsmarts economists, but people also believe that greed causes stupidity by music companies.  When it comes to music, sometimes we should resist the temptation to overreact.  Little things: preserving traditional contours of the music market/facilitating voluntary transactions/allowing market players to decide what revenue streams would be best.  (Note: these do not appear to be little things to me.)
Sprigman: Joel Waldfogel, And the Bands Played On: a bunch of data, concluding that quality of music, amount and diversity has gone up since disintermediation; theme contrasts sharply.   Apple has market power, for reasons that are probably independent of the fact of piracy.  Piracy increases the value of the iPod, but is that deadweight loss?  People exposed to more music may value music more.  Preferences shift = makes iPod more valuable. One problem with live performance is that DOJ stood by while Live Nation constructed an effective monopoly.  Now, because of piracy, people are exposed to recorded music, and that shifts preferences for going to concerts, not a deadweight loss but a shift in demand.
A: agree, there’s no contradiction.

Lunney: paradox may be explained by declining cost structure: we’d expect that when costs radically fall, for example in manufacturing and shipping to thousands of stores across the country.  If you put up slides of revenue, have to do it in constant dollars.

Q: recorded v. live: in most cases, recorded and live are two different companies (exc. Live Nation) so different companies get the revenue streams; the artist is unlikely to see much revenue from recorded music and might not notice the difference though the label would.  The other big piece: sync licensing.  A much larger portion of the pie for indie music.
A: but those opportunities mostly go to established artists.
Q: seem to have a dim view of third party patronage. What about first party patronage/Kickstarter?

A: like venture capital—wonderful new business model.
Jamie Lund, Getting it Wrong: Juror Assessments of Similarity in a Music Copyright Experiment

178 participants: how much can we sway jurors with superficial performance similarities?  Composition: rhythm, melody, harmony, lyrics.  Perfomance: tempo, key, genre, orchestration/style: computer program generated the sound samples.  Used Carey v. Swirsky music.

Answer: yes, people can be influenced strongly by performance elements.  When similarly performed, 86% said they were similar, 85% said they were different when played differently.

What about the songs caused this?  Answers from listeners: tempo was primary, then instruments, then melody (this and following were below 40%), then beat/feel/rhythm.  Performance elements win out: melody was third, then rhythm/beat were sixth, just over 20%; elements/song structure/harmony were not common.  Predominantly focused on performance, 71% at least when asked.

Other issues: deputizing “experts” with relatively greater apparent musical expertise.  They would say things like “I thought this but then I was persuaded by another guy who used musical terms/sounded like he knew what he was talking about.”  Familiarity of instrumentation: I would have been able to understand better if it wasn’t this crappy electronica.  People familiar with genres are able to make better distinctions.  A lot of people complained about the absence of lyrics.  Some said they only listened to lyrics. Context: People sometimes don’t have the vocabulary to describe what they think was similar.  Some said ‘sounds like a lullaby/sounds like it would be a movie/sounds like I’d listen to it on a beach’: put it into a context.  Some wrote of how the work made them feel.  People may not even hear distinctions.  Some people really can’t pick out a melody or harmony (that would be me).  “[T]he first song sounded like puppies and kittens and what society tells me to be and the second one—I thought if hell exists this would be what was played.”  One response was “what about the untrained ear?”

Next project: can you actually help jurors?

Current instructions aren’t helpful. Asked them: what was original expression and quantitative/qualitative similarity.  In terms of how well they understood: they said original expression was much more difficult for them to understand.  E.g., “it means having not related to the music they are trying to produce, hence creating a differen type of genre.”  People who wrote about feelings might be trying to get at what they thought “expression” meant.  They were better at figuring out what quantitative/qualitative meant.
How do jurors consider evidence of copying versus evidence of substantial similarity?  Do we think that they’re so similar one must have copied?  72% said there was copying and 86% said there was substantial similarity—is that illogical?

Lunney: a lot of music sounds similar without copying. 

Lund: found these results heartening: they should be separate and independent inquiries. The fact that they could say that there was substantial similarity without copying, maybe by chance or from third party source, is great.  To what extent does the jury perceive these as independent, and how does presentation to the jury affect this?  When there’s strong evidence of copying, maybe jury will think substantial similarity is more plausible, but these should be independent.

RT: One possibility: these really were parts of the musical work, so people aren’t being tricked?

A: if you agree with Brauneis that these are parts of the musical work, then yes.  Justin Bieber’s covers sound different than popularizers’ versions.  Musicologists don’t agree on how to define melody etc.  Maybe harmony is just background.  Could have expert testimony, for example on how all cover versions include this particular element—a pause or a melisma.)
RT: this is beginning to sound like a TM problem.  With trade dress, we’ve long known that there’s a problem with figuring out what the plaintiff claims to own, because the contours of what P owns is vital to figuring out what the D can do/what the P can stop D from doing.  Registration is helpful there (though you can proceed without registering if you can sufficiently define your trade dress)—are there solutions in copyright like examination (or some other kind of definition)?
A: ex ante and ex post preferences of breadth are different.
Pam Samuelson: would a more precise jury instruction about protectable elements of a musical work help?  Are your juries getting a broader context about what infringement looks like?
A: maybe if we can’t define the song we can at least tell the jury to ignore certain things, like tempo or key.  (Of course if Brauneis is right, maybe they shouldn’t ignore those things!)  (I also have the reaction: if you want to claim it as yours, you should probably be able to identify it, so even if it’s hard the burden to identify protectable elements can be fairly put on the plaintiff.)  Maybe context is prejudicial instead of helpful.
Q: probative similarity to prove copying v. substantial similarity for infringement terminology might help.
David Morrison, Substantial Similarity and Psychological Similarity: Perfect Strangers

Other work in similarity in other disciplines.  Different ideas of how to model similarity.  Subset of research: Tversky’s feature matching approach and contrast model. Contrast model challenged prevailing geometric models of similarity.  Mathematics of set theory: similarity between two objects is a function of three distinct sets: similarity increases as shared feature set increases and decreases as non-shared features increase. Salience matters: intensity, frequency, familiarity, goodness of form, informational content. 

Case law: tends to describe individual features: duration of a sample, pitch, tone, other aspects of the work that can be counted as features.  Thus, this model offers a relatively compelling account of perceived similarity in copyright.  There’s a distinction between perceptual and conceptual features of an object.  Features can either be attributes or relations of an object.  Attributes are features that can be described as predicates that take a single argument, where relations can only be ascribed by reference to other attributes or relations (predicates take two or more arguments). 

Beyond the commonality and difference distinction in sets in the contrast model, there’s an important distinction between differences related to a common structure (alignable) and differences unrelated to a common structure (nonalignable).  Relational similarity is important: structure mapping asks what counts as a commonality and a difference.  Showed two images: Subjects asked different questions would map the first woman (getting food) to the squirrel in the second picture or the woman to the second woman (giving food).  The woman is a perceptual match but is out of place in the relational structure.

How does this fit into what we know about copyright?  Similarity according to types of predicates shared: mere appearance similarity versus comprehensive nonliteral similarity (analogy). 

For music, the literature suggests that determining not merely whether copying took place but whether the resulting similarity is substantial would ask about role similarity.  A brief sample of an extended improvisation, looped and used to create an accompaniment in another song would be cross-mapped into the new work, suggesting lower judgments of similarity than if the sample were placed in a context that was relationally similar.

RT: use this as a guide for the plaintiff’s lawyer and the defendant’s lawyer: what should they be pointing us toward?  Framing in this way could help translate the literature.

A: yes, this is in large part a descriptive project.

Q: does Content ID play into this?

A: hasn’t noticed any direct relation, though might be.  (I think there are claims now to find cover versions of music.  This story claims YouTube can ID a melody.)  A lot of overlap between AI and cognitive science.

Frischmann: do the studies cover music or other genres?

A: tends to be of relatively uncomplex objects.

Buccafusco: what happens if you have retinal trackers on people looking at pictures?  Ways in which questions are framed will affect how they look and length of time they look at particular features. 

A: there is research, including from Tversky: how does asking a question frame answers?  Asking people to judge similarity v. asking them to judge dissimilarity affects the overall perceptions of similarity. 

No comments: