Kindle and the Copyright Conundrum

Ars Technica reports:

The Authors Guild has come in for a fair amount of ridicule since their executive director, Paul Aiken, claimed that the speech-to-text feature of Amazon's new Kindle 2 violated copyright law, telling the Wall Street Journal: "They don't have the right to read a book out loud." On Wednesday, Guild president Roy Blount Jr. took to the pages of The New York Times to defend his group's much-mocked position, arguing that the device would "swindle" penurious writers out of precious audiobook revenue. Whether or not you think that's likely, however, probing the Guild's objections reveals one more way that advancing technology may blur traditional categories in copyright law.

The short version of Blount's argument is that the right to create an audiobook version of a text is generally more valuable than the right to distribute an e-book, and that the Kindle effectively gives Amazon a free twofer: having bought the cheaper e-book right, it gets to offer customers a multimedia package that includes an audiobook rendered on the fly.

While proponents of this argument have had some trouble explaining why it's OK for a human to privately read a bedtime story out loud, but not for a machine to do it, there is at least this potential distinction: private human reading is not a for-profit product that presents a realistic alternative to publisher-authorized audiobooks. The Kindle may not produce the kind of rich dramatic reading a human actor might record, but advances in text-to-speech technology make it a far more passable alternative than the sort of stilted Cylon centurion delivery that represented the state of the art just a few years ago.

For the moment, the Authors Guild seems to be hoping to rely on contract law rather than copyright litigation, urging authors and publishers to condition their licensing of e-book rights on Amazon's blocking the read-out-loud feature unless a separate performance license is purchased. It's not hard to see why: under the standard established in the seminal Betamax case, the Kindle's out-loud feature would almost certainly be deemed to have "substantial noninfringing uses," leaving the company in the clear, and publishers with the unrealistic recourse of trying to sue individual owners who "perform" their books without permission.

How to create a derivative work

But there is an implicit copyright claim lurking in the Guild's objections: US law assigns creators and publishers the exclusive right to prepare (or authorize) "derivative works" based on their creations—such as audiobooks, translations, and dramatic adaptations. But that raises a surprisingly tangled question: What does it mean to create a derivative work? As Sherwin Siy of Public Knowledge points out, the Guild has somewhat misleadingly spoken of "audio rights"—but there's no such thing as an "audio right" per se, only the right to prepare a "work," such as a particular (independently copyrightable) audio recording. But does the Kindle create a "work"?

Michael Kwun of the Electronic Frontier Foundation argues that it doesn't, for two reasons. First, he says, a "derivative work" must be a work of creative authorship: He cites the copyright statute's definition of "derivative work" as "a work based upon one or more preexisting works . . . which, as a whole, represent[s] an original work of authorship."

But there's ambiguity hiding in that ellipsis. The full text is:

A "derivative work" is a work based upon one or more preexisting works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted. A work consisting of editorial revisions, annotations, elaborations, or other modifications which, as a whole, represent an original work of authorship, is a "derivative work".

The second sentence imposes an originality requirement as the criterion for determining when revisions and annotations amount to the creation of a derivative work, not as a general requirement of any derivative work. Kwun invokes the hoary maxim "No creativity, no copyrightability," but this conflates the standard for independent copyrightability of a work with the criteria for infringement of the derivative work right. As legal scholar Paul Goldstein notes in his definitive essay on derivative works, "the [Copyright] Act does not require that the derivative work be protectable for its preparation to infringe." In other words, a work insufficiently original to be eligible for an independent copyright may still be an infringing derivative.

Consider, for instance, the Ninth Circuit Court of Appeals' decision in the 1984 case Lone Ranger Television v. Program Radio. Program Radio edited for broadcast old recordings of Lone Ranger teleplays that had fallen into the public domain. Thanks to a quirk of the law, however, the original scripts on which the programs were based remained under copyright. Though the edited versions clearly fell far short of the originality that would be required for an independent copyright, the Court held that Program Radio had violated the authors' exclusive "derivative works" rights. While the record is mixed, subsequent cases have at least been open to the category of "unoriginal" derivative works.

And Kwun concedes that his argument has some counterintuitive results: he told Ars that, on this theory, a translation of a text into another language—a paradigmatic example of derivative work—would not count as such if the translation were produced by an algorithmic software program like Google Translate. While no author has yet raised this claim, as far as I know, it might be argued that Google's on-the-fly translations of Web pages constitute large-scale production and dissemination of "derivative works." One interesting, and equally counterintuitive, wrinkle is that it might actually make an enormous legal difference whether the translation is performed on Google's servers and transmitted to the user, or by plug-in software on the end-user's home computer.

That's OK, Kwun has a backup argument with more solid support in case law. In the 1992 case Lewis Galoob Toys v. Nintendo of America, the Ninth Circuit found that Galoob's Game Genie—a device interposed between a game cartridge and the Nintendo console to allow users to tweak gameplay—did not create infringing "derivative works" of Nintendo's games because it did not "incorporate a portion of a copyrighted work in some concrete or permanent form." In other words, Game Genie altered the gaming experience, but did not actually create a new modified copy of the game. The Court codified that principle in Micro Star v. FormGen Inc., making the existence of some kind of nontransitory copy of a work a prerequisite for the creation of a "derivative."

It's worth noting, however, that this is in some tension with the actual language of the statute. As the Court noted in Game Genie, the Copyright Act uses "fixation" in some concrete form to establish the point at which a new work is first "created," but the definition of "derivative work" omits any reference to fixation. Indeed, the House Report on the bill explicitly asserts a legislative intention to cover "derivative works" for which no fixed copy is produced:

The exclusive right to prepare derivative works, specified separately in clause (2) of section 106, overlaps the exclusive right of reproduction to some extent. It is broader than that right, however, in the sense that reproduction requires fixation in copies or phonorecords, whereas the preparation of a derivative work, such as a ballet, pantomime, or improvised performance, may be an infringement even though nothing is ever fixed in tangible form.

The Ninth Circuit established its own "concrete form" requirement—and if there's a distinction between "embodiment in concrete form" and "fixation," it borders on the metaphysical—for largely practical reasons. After all, without some such bright line, it would suddenly become rather thorny to determine the point at which a derivative had been "prepared." If I begin thinking or talking about writing a Star Wars novel before I've secured permission from George Lucas, have I already "prepared" a derivative?

But it's also worth noting the context of the reasoning behind that rule in the Game Genie case: The Court's fear was that too broad a construction of "derivative work" would chill innovations that enhanced or improved existing works, such as software add-ons. The ephemeral nature of the modified games produced by the Game Genie was significant in part because the device did not "supplant demand for Nintendo game cartridges": you still had to buy Nintendo's original game to plug it into Galoob's device.

The Authors Guild, by contrast, is claiming that the Kindle's read-aloud feature, though requiring the purchase on an e-book, would displace sales in the distinct market for audiobooks. And one of the criteria for a "derivative work" is whether it entails a contribution to a distinct market. In the Micro Star case alluded to above, a company that sold user-generated add-on maps for the game Duke Nukem 3D ("embodied in concrete form" on a CD) infringed FormiGen's right to profit from expansions of the Duke Nukem story, even though the add-on maps did not directly displace the original game, which users still had to buy in order to make use of the maps.

The significance of the "concrete form" rule would likely have been tested in 2005 thanks to a company called ClearPlay, which outraged Hollywood by offering a service that would produce family-friendly versions of popular films on the fly using a special player that would mute or fast forward through the naughty bits. Congress, however, preempted a test case by passing the Family Movie Act to explicitly immunize such services against copyright liability.

In the Kindle case, it's not clear how much any of this matters. As noted above, Amazon is probably protected by the Betamax rule, and as the recording industry has discovered at some length, trying to sue your customers for infringement is a losing proposition—not to mention logistically infeasible in this case. Even if we allow the ephemeral computer reading of an e-book to count as a "derivative work," private use of text-to-speech software likely counts as a "fair use"—although strictly speaking, the copyright statute creates a very narrow safe harbor immunizing such technology only when used "exclusively" for the benefit of the sight-impaired.

Bigger questions loom

There's a larger philosophical question here, however, highlighted by the flexibility of digital technology: what, exactly, is the "work" that is "embodied" in a computer file? (I considered a related question in the context of conceptual art back in November.) Consider, for instance, that normally a musical composition and a particular recording of that composition are regarded as distinct "works" for copyright purpose.

But the rendering of a "work" from its stored form is a function of both the information in the copy and the player: an MP3 isn't a song without software that interprets it. I can take a file representing a musical composition, open it with one program, and see it represented visually as sheet music—notes and rests on a staff. I can feed the very same string of bits into another program and produce an actual "performance" of the piece—or rather, an indefinite number of performances, depending on my settings. So is that string of bits "really" a "concrete form" of the composition itself, or does it also encompass some or all of the particular sonic manifestations of that composition? Recall, after all, that a particular interpretation of a piece—Glenn Gould's performance of Bach's "Goldberg Variations," say—can be under copyright long after the composition itself has fallen into the public domain. And with the advent of electronic music, a "performer" may ultimately be executing a piece by feeding a file into a piece of software.

If all this sounds a little academic, consider some more specific ways computing power diminishes the relevance of the "concrete form" standard. When an artist like Girl Talk takes two or ten recordings and creatively recombines them into a new piece of music, then whether or not we think that constitutes a transformative "fair use" of the original pieces, it is pretty unambiguously a "derivative work." Now consider an alternative method of producing those mash-ups. Instead of sending you an MP3 of the final product, Girl Talk sends you a list of the 20 tracks that form the raw material of his song "Once Again." Then he sends you an algorithm specifying how precise chunks of each track must be spliced, looped, altered, and remixed to generate an on-the-fly reproduction of Once Again. To be sure, the RIAA might not quibble with this model, as it would surely spur demand for the original recordings. But as a strictly legal matter, it seems bizarre that whether "Once Again" is a "derivative work" should turn on the specifics of how the song is encoded.

Apply the same principle, mutatis mutandis, for on-the-fly translations of texts or recombination of videos. Language software, of course, is still far too clunky to serve as a substitute for proper human translation. But it should also be possible in principle to create text-specific "tweak" files that modify the operation of a generic translation ending, so that (for instance), when the Spanish text of Love in the Time of Cholera is fed into the translation software, the program has a better idea of how to parse it to create a fluid, readable English text. The "tweak" file itself would not actually need to "embody" any of the copyrightable text of the original; it might be sufficient to tell the engine to err on the side of alliteration, or to prefer one English rendering of a particular Spanish word in its dictionary when in close proximity to certain other words. Following the Game Genie model, the software could be designed to pipe the translation to a screen without creating a "fixed" file.

These examples may sound fanciful, but as increases in computing power render it ever easier to transform and recombine digitally stored works in real-time, courts are likely to find themselves confronted with a growing body of real-world cases posing tricky questions about where a "work" resides and when and how a "derivative" is created—questions the old rules and boundaries of copyright may be ill-equipped to answer.