Transforming Fair Use

Glynn S. Lunney, Jr. *

Download a PDF version of this article here.

In the United States, the fate of fair use, and by extension, copyright itself, hang in the balance. In recent years, courts have disagreed over fair use’s proper scope. On the one side are appellate decisions that interpret fair use broadly. While these decisions do not go so far as to suggest that every reuse is fair, they typically use the talismanic phrase “transformative use” to give fair use a generous and flexible interpretation. On the other side are appellate decisions that interpret fair use restrictively. While again they do not go so far as to suggest no reuse is fair, they typically either reject the transformative use rubric outright or give it a miserly construction.

In Andy Warhol Foundation v. Goldsmith, the Court largely sided with the courts that favored a narrow and restrictive view of transformative use specifically and fair use more generally. In reaching its conclusion, the Court used a variety of interpretative approaches: realism, textualism, and purposivism. This article critically re-examines each of these interpretative approaches and demonstrates that none supports the Court’s reasoning and outcome.

While courts are bound by the Court’s Goldsmith decision, it is a judicial decision, not legislation. Courts should treat it as such. Each of the Court’s statements on various issues should not be treated as independently binding, but as a reflection of a unified whole, tied to the specific facts of the Goldsmith case. In particular, courts should confine the decision’s application to other instances where a commercial use that was previously licensed is now claimed as a fair use. Wendy Gordon has previously suggested that where licensing is very likely to fail, fair use should be more readily found. Goldsmith represents the opposite side of the coin: Where licensing of a commercial use is very likely to succeed, and indeed, where the copyright owner has previously licensed the precise use at issue, a court should tend to find infringement to reinforce the licensing market in place.

Introduction

Thirty years ago, in Campbell v. Acuff-Rose Music, the Court introduced transformative use into copyright’s fair use analysis.¹ When an author copies from a copyrighted work to create a new work, the Campbell Court held, whether the copying constitutes fair use depends, inter alia, on “whether and to what extent the new work is ‘transformative.’”² A new work is transformative if, rather than “merely ‘supersed[ing] the objects’ of the original creation,”³ the new work “instead adds something new, with a further purpose or different character, altering the first with new expression, meaning, or message.”⁴ If present, such a transformative use weighs, perhaps decisively, in favor of finding the use fair and hence non-infringing under the first statutory fair use factor—“the purpose and character of the use.”⁵

In the decades since, appellate courts have split over the role that transformative use should play in the fair use analysis. On one side, a series of appellate decisions embraced transformative use and relied on it to expand the scope of fair use.⁶ On the other, a series of appellate decisions expressed skepticism regarding the analytical utility of transformative use, rejected the rubric’s broad or general application, and ruled in ways that narrowed the fair use doctrine.⁷

In 2021, the Court had its first opportunity to resolve the split over the proper role for transformative use, specifically, and the direction of fair use, generally, in Google v. Oracle.⁸ In its decision, the Court held that Google’s verbatim copying of 11,500 lines of code from the copyrighted computer program, Java, to create a new operating system for the mobile environment, Android, was fair use as a matter of law.⁹ In its reasoning, it identified Google’s use as transformative, specifically,¹⁰ and approved a broader and more flexible approach to fair use, generally.¹¹ However, the Court cautioned that its decision might apply narrowly—only within the computer programming context and perhaps only to the specific facts of the Google v. Oracle dispute itself.¹²

Just two years later, in Andy Warhol Foundation v. Goldsmith, the Court had a second opportunity to resolve the split over the proper role for transformative use and to set the direction of fair use going forward.¹³ Rather than build on its decision in Google v. Oracle, however, the Court reversed course. At issue were a series of sixteen silkscreens and drawings that artist Andy Warhol had created based on professional photographer Lynn Goldsmith’s black-and-white photograph of the musical artist Prince.¹⁴ After the Second Circuit held that the sixteen Warhol works were not transformative as a matter of law under the first fair use factor,¹⁵ the Andy Warhol Foundation (“the Foundation”) petitioned for certiorari and asked the Court to determine whether “a work of art is ‘transformative’ when it conveys a different meaning or message.”¹⁶ After granting the petition for certiorari, the Court re-framed the question presented. Instead of asking whether a work of art is transformative generally, the Court asked whether the specific use at issue—the Foundation’s licensing of one of the Warhol Prince prints (known as “Orange Prince”) for use on a magazine cover—was transformative.¹⁷ As is often the case, the framing of the question dictated the Court’s answer.¹⁸ As the Court noted, Goldsmith had also licensed the use of her black-and-white photo for a magazine cover.¹⁹ So, in a sense, Warhol’s artwork and Goldsmith’s photo were both vying for the same money. Both artists wanted their works on magazine covers, and both wanted money for such use. As a result, the Court held that the Foundation’s licensing of Warhol’s art for a magazine cover was not transformative under the first fair use factor.²⁰ Because the Foundation agreed that the other three fair use factors weighed in Goldsmith’s favor, Warhol’s use was unfair.²¹

It is a startling decision. Startling not as to outcome, on which reasonable minds can disagree, but on the route by which the Court held Warhol’s use unfair. The Court could readily have said that to whatever extent Warhol had transformed the meaning, message, or aesthetic of Goldsmith’s original black-and-white photograph under the first statutory fair use factor, that transformative character was outweighed: (i) by the potential lost licensing revenue to Goldsmith under the fourth factor; or (ii) by some combination of the other three fair use factors.²² Indeed, given the Court’s repeated insistence in its opinion that the issue required balance and was a question of degree, such an approach to resolving the case would have been more consistent internally.²³ Yet, that is not how the Court resolved the fair use issue. Instead, the Court resolved the case by concluding that Warhol’s use was not transformative at all under the first factor.²⁴ No balance. No question of degree. Despite the artistic choices Warhol made in creating his silkscreens and drawings, in the Court’s view, Warhol’s Orange Prince, at least when used for a magazine cover, was no more transformative than an exact, mechanically or digitally re-created copy of Goldsmith’s black-and-white photograph. That conclusion is startling. By reaching it, the Court effectively equated Warhol, probably the most celebrated and influential American artist of his generation, with a copy-shop employee.²⁵

It was probably this implicit characterization of Warhol that provoked Justice Kagan to write her blistering dissent.²⁶ The sharp disagreement between Justice Sotomayor, writing for the Court, and Justice Kagan, in dissent, hearkens back to the equally sharp disagreement between Justices Stevens and Blackmun in Sony Corp. v. Universal City Studios.²⁷ As in Sony, the viciousness with which the majority and the dissent assailed each other’s position in Warhol likely reflects a combination of factors. Among them, two in particular stand out for me. First is Sayre’s Law.²⁸ Like academic politics, copyright politics are vicious because the stakes are so small. Whether we agree with Kagan or Sotomayor, people will still go to bed every night hungry, thirsty, and at risk of death from war, poverty, and disease. The climate crisis will continue towards its seemingly inevitable conclusion. And the end of real economic growth in the United States will still loom.²⁹ Against these truly important stakes, whether copyright is slightly broader or slightly narrower so that one wealthy copyright owner, such as the Foundation, will or will not have to pay a token licensing fee to some other wealthy copyright owner, such as Goldsmith, is hard to get worked up about. Second, we will never know for certain who was right. The counterfactual world in which Justice Kagan’s position prevailed does not exist, or at least, is unavailable for our inspection. We cannot therefore readily compare outcomes in the world where the majority prevailed to outcomes in a world where the dissent prevailed to know who was right and who was mistaken.

As a result, whether the Court’s decision will lead to more creative works or fewer, whether those works will be higher or lower quality in terms of their emotional and aesthetic impact, and whether they will be distributed more or less widely, is difficult to know. The only thing we can say for sure is that the Court’s decision will make fair use litigation more expensive. By requiring fair use to be decided for new works on a use-by-use basis, courts going forward will have to decide questions of fair use not once for each new work, but repeatedly, for each use of each new work. The need for repeated rounds of fair use litigation threatens to increase the cost of fair use litigation exponentially. That might not have been so bad in the analog world in which copyright was born. Because of the high costs entailed in the distribution of creative work through the printing press, broadcast radio and television, theater chains, and the other distribution mechanisms available in the analog world, only the privileged few could share their creative efforts with the many in any event. Whether copyright was long, broad, and complicated, or short, narrow, and simplistic made little difference. In the analog world, having a high-cost copyright system was not a barrier to entry in itself. It merely reinforced the preexisting high-cost structure of the available analog distribution technologies. Today, however, we live in a digital world. In today’s digital world, social media and other avenues of digital distribution have made it trivially easy for anyone to share even their most idle creative impulse with the many. In today’s digital world, the high cost of copyright itself has become the defining barrier to authoring and distributing creative work. Precisely at a time when digital technologies have democratized authorship and made the fair use privilege available to the many, the Court’s Goldsmith decision threatens to restrict the availability of the fair use privilege to those few that can afford the expense of repeated litigation.

As justification for gating fair use behind repetitive litigation, the Court offered a variety of reasons reflecting a mixture of interpretative approaches. From a realist perspective, the Court wrote that substitution is copyright’s “bête noire,”³⁰ and at least when used for a commemorative magazine cover, Warhol’s Orange Prince is not transformative because it is a substitute, even if “not a perfect substitute,” for Goldsmith’s photograph.³¹ From a textualist perspective, the word “transform” cannot simultaneously: (i) define activity as infringing, as part of the definition of a derivative work, and (ii) define activity as non-infringing, as part of the definition of fair use. From a purposivist perspective, finding Warhol’s use to be fair would require the Court to find many films based upon novels to be fair as well and thus frustrate Congress’s intent to grant the copyright owners of novels the exclusive right to control the making of their novels into films.³²

Insisting that its prior decision in Campbell v. Acuff-Rose tied its hands, the Goldsmith Court reduced the first fair use factor to a three-part, mechanical checklist.³³ First, does the defendant’s specific use at issue serve the same general purpose as the plaintiff’s work seeks to serve? Second, is the defendant’s specific use at issue commercial? Third, does the defendant’s work criticize or otherwise comment on the plaintiff’s work?

Unfortunately, the Court’s analysis and resulting checklist are fatally flawed. The Court’s realism is unrealistic. Its textualism ignores the statutory text. And its purposivism frustrates Congress’s purpose in elevating fair use from common law exception to statutory privilege. As for its reading of Campbell, the Goldsmith Court mistakes dicta for holding. The resulting mechanical checklist is contrary to the statutory text, is contrary to the Court’s own precedent, and frustrates Congress’s purpose in enacting copyright.

Under the Constitution, Congress has the authority to enact copyright legislation for a singular purpose: “the Progress of Science.”³⁴ To achieve that purpose, Congress did not define copyright as all rights with no limits. To the contrary, to balance the competing interests at stake, between earlier authors and later authors, between authors and consumers, between what may be freely copied and what may not be freely copied, Congress defined copyright to include both rights and limits, as the Goldsmith majority recognized.³⁵ Yet, Justice Gorsuch, in his concurrence, insisted that interpreting the fair use doctrine to help strike that balance is not the Court’s proper role. As he chided Justice Kagan:

Worry not. This case does not call on us to strike a balance between rewarding creators and enabling others to build on their work. That is Congress’s job.³⁶

While Justice Gorsuch is correct that it is Congress’s job to balance appropriately the various interests at stake in defining the optimal scope of copyright, Congress may do that job in, at least, two ways. First, Congress can do so directly through statutory language. Congress could, for example, set a shorter or longer copyright term or define the copyright owner’s exclusive rights more broadly or more narrowly—not that courts, including the Goldsmith Court itself, have paid much attention to the statutory language Congress used to define the exclusive rights.³⁷ Second, Congress can do so indirectly, by delegating the task of striking the appropriate balance to others, such as the federal courts. When it came to fair use, by expressly leaving the analysis open-ended, Congress delegated the task of defining fair use in a way that optimally balances the competing interests at stake to the courts.³⁸ Pretending otherwise merely guarantees that the Court will perform its congressionally delegated task poorly.

Consider, for example, the language that Congress used to define the fourth fair use factor. That language requires a court to consider “the effect of the use upon the market for or potential value of the copyrighted work.”³⁹ While the statute identifies this factor and requires a court to consider it, the statute does not tell a court how to evaluate this factor or how to balance it against the other three.⁴⁰ For example, the statute does not say, at one extreme, that any effect on the market will make a use unfair. Nor does it say, at the other extreme, that only a complete destruction of value will make a use unfair.⁴¹ Instead, the statute leaves to the courts the question how much or what sort of loss in potential value will weigh in favor of or against fair use.

The same is true for the other three fair use factors. The statute identifies them. The statute requires a court to consider them.⁴² But the statute does not say how to weigh each factor on its own or how to balance them against the others. The statute does not even provide a comprehensive definition of each.⁴³ Moreover, the statute directly states that the four statutory factors are not the only factors a court may consider in evaluating fair use. The statutory language setting forth the four fair use factors is open-ended–inclusive, rather than exclusive. It expressly leaves courts the leeway to develop and consider factors other than the four factors the statute expressly sets forth.⁴⁴

In short, when it came to fair use, Congress did not strike the appropriate balance between the competing policy concerns—whether “between rewarding creators and enabling others to build on their work,”⁴⁵ or otherwise—that copyright and fair use can implicate. Instead, Congress attempted to codify a judicially created and judicially developed doctrine. But any such attempt is necessarily fraught: fraught with the risk that reducing a common law doctrine to a simplistic set of factors may omit or misstate something important and fraught with the risk that codifying the doctrine at a specific point in time may risk stifling further judicial development and common law evolution. Congress codified fair use despite these risks. Not to bar further judicial development of fair use or to reduce the complexities of balancing the interests at stake to a mechanical checklist,⁴⁶ but to give fair use express statutory recognition. Where before the Copyright Act of 1976, fair use was a judicially created exception to the legislatively drafted and executively approved statutory exclusive rights, after the Act, fair use became statutory too. Codifying fair use ensured that fair use carried a dignity and stature equal to the copyright owner’s exclusive rights.

The Court once proclaimed: “[I]t should not be forgotten that the Framers intended copyright itself to be the engine of free expression.”⁴⁷ Yet, to achieve that lofty goal, copyright cannot be all rights and no limits. Balance is required. Some of the limitations necessary to ensure balance Congress dictated directly in the statute. Even though technology may chance over time and circumstances may vary from case-to-case, some limits—limits such as the term of copyright⁴⁸—can be defined adequately in advance and applied uniformly, even mechanically, across cases. But some limits necessary to achieve the optimal balance cannot be defined in advance nor can they be applied uniformly or mechanically. Changing technology and changing circumstances from case to case may require a more flexible set of standards intelligently applied. In codifying fair use, Congress expressly recognized fair use as one of the limits necessary to achieving the appropriate balance. But Congress did not reduce fair use to a mechanically applied checklist. Nor did Congress balance the competing interests at stake itself. Instead, it provided some general guides and left to the courts the task of applying those guides, or developing new ones, to ensure that the two sides of copyright, rights and limits, optimally balance the competing interests at stake.⁴⁹

In the remainder of this article, I will explore and develop each of these issues. I begin in Section I with a brief history of transformative use in fair use doctrine. Section II then explores the Court’s Warhol decision in more detail. As part of this exploration, Section II critically reexamines the Court’s decision on its own terms and finds the decision fundamentally lacking. The Court’s realism is unrealistic. The Court’s textualism ignores the statutory text. The Court’s purposivism defeats the purpose for which Congress enacted fair use.

To chart a path forward for fair use, Section III returns to the statutory text and the Court’s own precedents. In both, we find the need for balance. Through threats of litigation and unduly burdensome licensing requirements, too much copyright will stifle the very creativity and widespread dissemination of original works that copyright seeks to encourage. Too little, and there is a risk that the public may receive too few original works of authorship. Congress gave fair use express statutory recognition in the Copyright Act of 1976 to recognize its essential role in striking that balance. With fair use, however, rather than strike that balance itself, Congress left the balancing to the courts. By appropriately drawing the line between fair and unfair use on a case-by-case basis, courts can ensure that copyright continues to “promote the Progress of Science” as the technology, markets, and economics associated with authorship change. The final Section concludes.

To begin, we turn to the origins of transformative use.

I. A Brief History of Transformative Use

Transformative use was introduced into fair use innocently enough. The Court set the stage in its first two decisions applying fair use after Congress formally codified the doctrine in the Copyright Act of 1976. In the first, Sony Corp. v. Universal City Studios, the Court addressed whether a consumer’s use of a Betamax to copy broadcast television programs for the purpose of time shifting was a fair use.⁵⁰ The Ninth Circuit and Justice Blackmun in dissent argued that it could not be a fair use because it was not “productive.”⁵¹ A Betamax consumer merely copied the television program for its intrinsic purpose, not to facilitate the creation of some new work of authorship, they argued. Such a consumptive or ordinary use could not be fair, they insisted.⁵² The Sony majority disagreed.⁵³ And although the Sony majority acknowledged the continued relevance of the productive nature of the use in appropriate fair use contexts,⁵⁴ its rejection of the Ninth Circuit’s and the dissent’s “productive use” arguments had the effect of eliminating or sharply limiting the relevance of “productive use” to the fair use analysis. As part of its analysis, the Sony majority also seemed to suggest, at least in dicta, a presumption that every commercial use of a copyrighted work was unfair.⁵⁵

In the second, Harper & Row v. Nation Enterprises, the Court considered whether copying and publishing a 2,200-word essay—that quoted a total of 300-400 words directly from President Gerald Ford’s at the time forthcoming biography—was fair use.⁵⁶ In concluding that it was not, the Court emphasized that the biography from which the defendant had copied was, at the time of the defendant’s use, unpublished.⁵⁷

With only these two Supreme Court decisions to go by, appellate courts in the late 1980s took the Court’s reasoning much further than the facts and holdings of each case would require. By 1990, a trial judge, Judge Pierre N. Leval, had grown frustrated with what he saw as the Second Circuit’s overly narrow approach to fair use, particularly with respect to reuse of unpublished materials for biographical research.⁵⁸ For such research in particular, the Court’s decisions in Sony and Harper & Row, taken together, would seem to leave very little fair use room. Sony both: (i) made it hard to rely on the productive nature of the use; and (ii) suggested that such use was presumptively unfair as long as the researcher intended to sell the resulting biography.⁵⁹ Harper & Row reinforced the conclusion that the use was unfair because the scholarly researcher was copying from unpublished materials.⁶⁰ But Judge Leval felt that, at least some, such uses were fair. The researchers were combing through obscure historical archives and bringing new information to light. In Judge Leval’s view, copyright should reward, not punish, these researchers’ creation and dissemination of new knowledge. Yet, despite his careful and thorough analysis explaining why the uses at issue in the cases before him were fair, the Second Circuit would reverse.⁶¹ No one likes to be reversed. So Judge Leval wrote a law review commentary explaining why he was mostly right and why the Second Circuit was mostly wrong on the fair use issue.⁶² As part of that explanation, Judge Leval suggested that fair use in copyright law “turns primarily on whether, and to what extent, the challenged use is transformative.”⁶³

Four years later in Campbell v. Acuff-Rose Music, the Court embraced Judge Leval’s suggestion and incorporated transformative use into the fair use analysis.⁶⁴ In Campbell, the Court faced the question whether 2 Live Crew’s rap song Pretty Woman, was a fair use of Roy Orbison’s song, Oh, Pretty Woman. 2 Live Crew had copied the opening phrase and a distinctive bass riff from Orbison’s song. After that, however, the meaning and message of 2 Live Crew’s song diverged quite dramatically from Orbison’s.⁶⁵ Although the Sixth Circuit accepted the district court’s finding that 2 Live Crew’s song was “a criticism in the nature of a parody in the popular sense,” the Sixth Circuit nevertheless reversed the district court’s conclusion of fair use.⁶⁶ Citing language from the Court’s opinions in Sony and Harper & Row, the Sixth Circuit held “that the admittedly commercial nature of [2 Live Crew’s] derivative work … requires the conclusion that the first factor weighs against a finding of fair use.”⁶⁷

After granting certiorari, the Supreme Court reversed.⁶⁸ With respect to the first fair use factor, the purpose and character of the defendant’s use,⁶⁹ the Campbell Court rejected the rule, that the Sixth Circuit had, “ostensibly culled from Sony, that ‘every commercial use of copyrighted material is presumptively … unfair.’”⁷⁰ As the Court noted, any such “presumption would swallow nearly all of the illustrative uses in the preamble paragraph of § 107.”⁷¹ Instead of a presumption, the Campbell Court held that commercial use was merely “a separate factor that tends to weigh against a finding of fair use.”⁷² As for how heavily commercial use will weigh against fair use in any given case, “even the force of that tendency will vary with context.”⁷³

Rather than rely on the commercial nature of a use as the central touchstone under the first factor, the Campbell Court directed courts to consider the transformative purpose or character of the use. Specifically, the Campbell Court wrote that the analysis of the first factor must evaluate:

whether the new work merely “supersede[s] the objects” of the original creation, or instead adds something new, with a further purpose or different character, altering the first with new expression, meaning, or message; it asks, in other words, whether and to what extent the new work is “transformative.”⁷⁴

If the defendant copies to create such a transformative work, that weighs strongly in favor of fair use under the first fair use factor. Leaving room for such transformative works promotes one of copyright’s legitimate ends—the creation of new works of authorship—directly.⁷⁵

Moreover, once the Court concluded that 2 Live Crew’s use was transformative, that shaped, if not dictated, the remainder of the Court’s analysis.⁷⁶ In analyzing another consideration under the first fair use factor, the Court held that, although 2 Live Crew’s use was commercial, the commercial nature of 2 Live Crew’s use did not weigh against fair use because 2 Live Crew’s use was transformative.⁷⁷ Under the second fair use factor,⁷⁸ although 2 Live Crew copied from an entertaining, rather than useful, work, the Court held that did not weigh against fair use because 2 Live Crew’s use was transformative.⁷⁹ Under the third fair use factor,⁸⁰ although 2 Live Crew copied the “heart of the work,” at least according to the Sixth Circuit,⁸¹ the Court held that did not weigh against fair use because 2 Live Crew’s use was transformative.⁸²

As for the fourth factor, “the effect of the use upon the potential market for or value of the copyrighted work,”⁸³ the transformative nature of 2 Live Crew’s use ensured that 2 Live Crew’s Pretty Woman would not serve as a market substitute for Roy Orbison’s Oh, Pretty Woman.⁸⁴ No ordinarily prudent consumer looking for the Roy Orbison version would buy the 2 Live Crew version instead. They were quite different songs, each with its own message and aesthetic appeal. Moreover, because the 2 Live Crew song could be perceived as a parody,⁸⁵ and hence a criticism of the Roy Orbison song, the Court reasoned that the copyright owner would not ordinarily license the use.⁸⁶ Thus, the transformative nature of 2 Live Crew’s use also weighed against an effect upon “the potential market for or value of” Orbison’s song under the fourth fair use factor.⁸⁷ Nevertheless, the Court held that the transformative nature of 2 Live Crew’s use did not fully resolve the fourth factor. The Court left opened the possibility that 2 Live Crew’s transformative use might somehow interfere with the licensing market for a less critical, more sympathetic rap version of Oh, Pretty Woman.⁸⁸ Seizing on a concession that 2 Live Crew’s attorney made in oral argument—that fair use is an affirmative defense—the Campbell Court placed the burden on the defendants to introduce evidence on that issue and remanded for further development of the record.⁸⁹

In this initial proposal and adoption, there were few hints of what transformative use would become. In Judge Leval’s analysis and in Campbell itself, transformative use seemed merely a new label for an old fair use doctrine, “productive use,” – a doctrine that had fallen out of favor after Sony. Whether under the new “transformative use” label or the old “productive use” label, the question, as traditionally understood, was whether, and to what extent, the alleged infringer used what was copied to create a new or different work. If so, the productive or transformative nature of the use weighed in favor of fair use. If not, then not.

That changed in 2003. On July 7th of that year, the Ninth Circuit responded to a petition for rehearing by issuing its second opinion in Kelly v. Arriba Soft.⁹⁰ Judge Thomas G. Nelson wrote the opinion and held that, as a matter of law, a search engine’s copying and display of thumbnail-size versions of full-size copyrighted images found elsewhere on the Internet was a fair use.⁹¹ There was nothing transformative about the thumbnail versions of the full-sized images, at least not in the traditional sense of the word. No new work was created. No new aesthetic meaning explored. In the words of the copyright owner Kelly: “because Arriba reproduced his exact images and added nothing to them, Arriba’s use cannot be transformative.”⁹² The Ninth Circuit rejected Kelly’s interpretation of the word “transformative” as overly restrictive.⁹³ The thumbnail may have merely reduced the size and resolution of the original image, yet by doing so, the resulting thumbnail served a different purpose from Kelly’s original. Rather than the aesthetic enjoyment the originals provided, the thumbnails were intended to “improv[e] access to information on the internet.”⁹⁴ The Ninth Circuit held that because the later use served a new purpose it was transformative. Because it was transformative, the use was fair.⁹⁵

While both Kelly and Campbell used the same word “transformative,” they otherwise had very little in common. The question in Campbell was one that copyright has struggled to answer from its inception: how much of an earlier work can a later author reuse and for what purposes, given that all authorship inevitably entails some borrowing.⁹⁶ It was not a new question nor did it arise due to new authorship or distribution technology. In contrast, the question at the heart of Kelly was entirely new and entirely the result of a new distribution technology, the Internet. The closest parallel to the Arriba Soft copyright issue in the analog world would be whether the description of a book in a card catalog in a library infringed the copyright in the book.

While some of the language in both Judge Leval’s commentary and the Court’s ruling in Campbell could be read to encompass a different purpose as a transformative use,⁹⁷ it is not clear that either intended or foresaw the extension of the transformative use rubric to encompass creating thumbnail images for an Internet search engine. Nor is it clear that either considered “purpose and character” to be separate and distinct categories—“purpose or character.”⁹⁸ The trial judge in Kelly, Judge Taylor of the Central District of California, and on appeal, Judge Nelson of the Ninth Circuit, deserve the credit (or blame) for being the first to apply Campbell’s language outside of its original context.⁹⁹ They also deserve credit (or blame) for effectively dividing transformative use into two distinct categories.¹⁰⁰ First, for cases like Campbell where an author copied from a copyrighted work to create a new work, the question is whether the new work has a transformative character. Second, for cases like Kelly that involve a new use of an existing work, the question is whether the new use has a transformative purpose.

The two types of transformative use are not only different in terms of the types of follow-on uses to which they apply but also because the two types of uses serve fundamentally different ends.¹⁰¹ Courts applied the transformative character rubric when a defendant used what was copied to create a new work, one “with a different new expression, meaning, or message” compared to the earlier work.¹⁰² This approach seeks to advance the first legitimate end for which Congress may enact copyright: the creation of new works of authorship.¹⁰³ In contrast, courts applied the transformative purpose rubric when a defendant used what was copied to expand access to existing works. Uses such as an image search engine or the Google Books project do not create new works, but they give us new forms of, and easier, access to existing works. Finding fair use in transformative purpose cases thus seeks to advance copyright’s second legitimate end: ensuring widespread dissemination of, and access to, existing works of authorship.¹⁰⁴

Yet despite their differences, both the original “transformative character” aspect and the more recent “transformative purpose” aspect of the transformative use rubric followed similar trajectories. For both, beginning in the early 2000s, courts relied on the transformative use label to justify increasingly expansive fair use outcomes. These decisions reached their apogee in Cariou v. Prince in the traditional transformative character context¹⁰⁵ and in Authors Guild v. Google in the transformative purpose context.¹⁰⁶ Some thought these decisions went too far, provoking scholarly and judicial responses that questioned and criticized their approach to fair use.¹⁰⁷ This led some courts to embrace a narrower view of fair use. In decisions such as Andy Warhol Foundation v. Goldsmith, a pair of decisions involving Dr. Seuss Enterprises, and Fox News v. TVEyes,¹⁰⁸ judicial panels rejected claims of fair use on facts arguably similar to those on which judicial panels had found fair use just a few years earlier.

Although the reasons for the rejection varies somewhat from decision to decision, these restrictive fair use decisions tend to focus on the two substantive concerns that the Second Circuit panel identified in Andy Warhol Foundation: (1) a textual concern that overly broad fair use threatens to eviscerate the copyright owner’s exclusive rights generally and the derivative work right specifically; and (2) an instrumental concern that overly broad fair use might reduce the incentives for original authorship.¹⁰⁹ The first concern leads panels to narrow subtly or to reject outright the application of the transformative use rubric outside the specific fact pattern the Court addressed in Campbell. The second concern leads panels to weigh the fourth factor heavily against fair use if there was any chance that the use, if found infringing, could become a possible source of licensing revenue. They would weigh such a possibility heavily against fair use almost without regard to how unexpected or remote that possibility was.

In the next section, we turn to the Court’s decision in Andy Warhol Foundation v. Goldsmith as an illustration of this restrictive fair use trend and critically re-examine the proffered textual and instrumental justifications for narrowing fair use.

II. A Critical Re-examination of Narrowing Fair Use

Although the dispute just came before the Court last year, the foundations for the Andy Warhol Foundation v. Goldsmith litigation were laid forty years ago, back in 1981. At that time, Prince Rogers Nelson—or as he became commonly known, Prince—was just starting his musical career, and Newsweek hired Lynn Goldsmith to take photographs of him.¹¹⁰ Among other photographs, Goldsmith took a black-and-white frontal portrait of Prince.¹¹¹ Three years later, in 1984, Prince had released his album Purple Rain and had become a pop culture icon.¹¹² Condé Nast wanted to run a story on Prince in Vanity Fair and paid Goldsmith a $400 license fee to use the black-and-white photograph as an artist’s reference.¹¹³ Unbeknownst to Goldsmith, Condé Nast hired Andy Warhol as the artist who would use Goldsmith’s photograph to create the artwork to accompany the Prince story, and Warhol did so.¹¹⁴ The article ran in Vanity Fair and Warhol’s Purple Prince accompanied it.¹¹⁵ For thirty-two years, nothing further happened. But then, in 2016, Prince died of an accidental fentanyl overdose.¹¹⁶ Condé Nast decided to do an issue commemorating Prince’s life and so approached the Andy Warhol Foundation to use Purple Prince once again.¹¹⁷ However, Warhol had actually created sixteen different Prince works based upon Goldsmith’s photograph (known collectively as “the Prince series”).¹¹⁸ After learning that the Andy Warhol Foundation had additional images from the Prince series, Condé Nast obtained a license for a different Prince series image, Orange Prince, and ultimately used the image for the commemorative issue’s cover.¹¹⁹ When Goldsmith learned of the use, copyright litigation ensued.¹²⁰

On cross motions for summary judgment, the district court held that each of the sixteen art works in Warhol’s Prince series was transformative.¹²¹ Applying the standard the Court had established in Campbell, the district court asked not whether the work was solely or primarily transformative, but whether a transformative character “may reasonably be perceived.”¹²² The district court also applied the standard for transformative use that the Campbell Court had established: “The central purpose of this investigation is to determine ‘whether the new work merely supersede[s] the objects of the original creation or instead adds something new, with a further purpose or different character, altering the first with new expression, meaning, or message.’”¹²³ The district court noted the specific alterations Warhol made to the earlier photograph, and based upon those alterations, concluded:

These alterations result in an aesthetic and character different from the original. The Prince Series works can reasonably be perceived to have transformed Prince from a vulnerable, uncomfortable person to an iconic, larger-than-life figure.¹²⁴

After ruling that Warhol’s Prince series was transformative, the court concluded that Warhol’s series was a fair, and hence non-infringing, use.¹²⁵

On appeal, the Second Circuit reversed.¹²⁶ Although the appellate panel acknowledged that the Warhol Prince series embodied a different aesthetic from Goldsmith’s photograph, the panel held that Warhol’s use was not transformative.¹²⁷ Following the lead of earlier decisions questioning the transformative use rubric, the Warhol panel, first, worried that an overbroad fair use doctrine would undermine the derivative work right generally¹²⁸ and the right to control the making of a film from a novel specifically.¹²⁹ Second, the panel also held that Warhol’s use, despite the admittedly different aesthetics, served the same purpose, at least, at a sufficiently high level of generality.¹³⁰ Both Goldsmith’s photograph and Warhol’s silkscreens “are portraits of the same person.”¹³¹ Third, the panel emphasized that Warhol’s use was commercial.¹³² Although the panel acknowledged that the Court itself in Campbell had rejected the presumption that the Sony decision had seemingly endorsed—that a commercial use is presumptively unfair¹³³—the panel nevertheless seemed to embrace such a presumption.¹³⁴ For these three reasons, the panel held that the first statutory fair use factor weighed against Warhol’s use being fair. Based upon that conclusion and its review of the other fair statutory use factors,¹³⁵ the panel held that Warhol’s use was unfair as a matter of law.¹³⁶

The Foundation petitioned for certiorari and asked the Court to decide a single question:

Whether a work of art is “transformative” when it conveys a different meaning or message from its source material (as this Court, the Ninth Circuit, and other courts of appeals have held), or whether a court is forbidden from considering the meaning of the accused work where it “recognizably deriv[es] from” its source material (as the Second Circuit has held).¹³⁷

After granting certiorari, the Court refused to answer the question presented–whether Orange Prince was transformative as a new work of art. Instead, the Court answered a different one: Whether the licensing of Orange Prince for a magazine cover of a commemorative issue about Prince was a transformative use.¹³⁸

As it often does, the question’s framing dictated the Court’s answer. The Court adopted a mechanical three-question checklist to determine that the specific use of Orange Prince at issue was unfair under the first fair use factor. The first question was whether Orange Prince, when used as a magazine cover for a story about Prince, served the same purpose, broadly defined, as Goldsmith’s photograph. The second was whether the use at issue was commercial. The third was whether Warhol had a particularly compelling justification for his copying in that he was criticizing, parodying, or commenting on the Goldsmith photograph.¹³⁹ Working through this checklist, the Court held that Orange Prince served the same purpose, broadly defined, as Goldsmith’s photograph.¹⁴⁰ Both works depict Prince. The Court also held that the Foundation’s use was commercial.¹⁴¹ Condé Nast paid the Foundation a licensing fee to use Orange Prince as the cover of the commemorative magazine issue. Finally, the Court held that Warhol did not have a compelling justification for the copying.¹⁴² Orange Prince did not parody, comment on, or criticize Goldsmith’s photograph. Having mechanically worked through its checklist, the Court concluded that each item weighed against fair use and held accordingly.¹⁴³

In adopting a checklist approach, the Court relied on an ad hoc mixture of interpretative approaches. First, from a realist perspective, it insisted that substitution is copyright’s “bête noire.”¹⁴⁴ Reframing the analysis from Warhol’s works themselves, as set forth in the original question presented, to the licensing of Orange Prince for use on the cover of a magazine commemorating Prince’s life enabled the Court to offer, at least, the pretense that for that use specifically, Warhol’s Orange Prince was a substitute for Goldsmith’s photograph.¹⁴⁵ Second, from a textualist perspective, the Court undermined the transformative use rubric generally by noting that the word “transform” is not included in the statutory text of fair use, but is included in the statutory text defining a derivative work.¹⁴⁶ From a textual perspective, relying on the same word “transform” to identify a use as non-infringing, as part of the definition of fair use, fatally conflicts, or so the Court seemed to suggest, with Congress’s use of the same word to identify a use as an infringing derivative work. Third, from a purposivist perspective, finding Warhol’s use to be fair would require the Court to find many films based upon novels to be fair as well. It might also allow thousands of follow-on creative works to flourish.¹⁴⁷ This apparently is a bad thing because it would frustrate Congress’s intent to grant the copyright owners of novels the exclusive right to control the making of their novels into films specifically, and the intent to grant copyright owners a broad derivative work right generally.¹⁴⁸ Fourth, the Goldsmith Court insisted that its own precedent mandated the Court’s outcome.

As stated in the introduction, however, the Court’s reasoning on each of these approaches is flawed. The following sections discuss the flaws in the Court’s reasoning for each approach.

A. Realism without Reality

The Court’s realism is unrealistic. From a realist’s perspective, the Court’s analysis gets off to a great start. The Court correctly identified substitution as copyright’s “bête noire.”¹⁴⁹ From the outset, copyright’s purpose has been to address the risk of market failure that arises from the sale of unauthorized, and hence, lower priced copies of a book. As the Stationers Company long ago articulated, without copyright, a second printer will offer unauthorized copies of the same book for less. Consumers will buy the lower priced unauthorized copies instead of the higher priced authorized copies.¹⁵⁰ As a result, the first printer will not recoup its investment. Foreseeing that result, the first printer will not publish the authorized copies in the first place. Unless copyright intervenes, these copying competitors will cause the book market to fail. Without copyright, many books that society values will go unauthored and unpublished.

For that risk of market failure to be realized, however, it is not enough that two works generally serve the same purpose. By that definition, 2 Live Crew’s version of Pretty Woman would not be transformative because it serves the same general purpose—listening pleasure—as the Roy Orbison original, Oh, Pretty Woman. For competitive substitution to occur and the risk of market failure to be realized, a second work must be so similar to the first that, given a small, but non-transitory price increase on the first, a consumer will actually switch to the second.¹⁵¹ By that standard, 2 Live Crew’s Pretty Woman and Roy Orbsion’s Oh, Pretty Woman are not substitutes.¹⁵² No ordinarily prudent consumer, looking for the Roy Orbison version, would buy the 2 Live Crew instead. But by that standard, Orange Prince and Goldsmith’s black-and-white photograph are not substitutes either. Indeed, by that standard, Orange Prince and Goldsmith’s photo are not substitutes even if we focus on the specific licensing of the Warhol work for Condé Nast’s commemorative magazine cover, as the Court insisted we must.¹⁵³

Even the Goldsmith Court is careful never to state outright that Orange Prince and Goldsmith’s black-and-white photo are substitutes. Instead, the Court approaches the issue in a backhanded way and merely asserts that the two works “shared the [same] objectives, even if the two are not perfect substitutes.”¹⁵⁴ But to say that two products are not perfect substitutes is to say nothing at all, at least outside the realm of economic theory. And the Goldsmith Court’s approach to defining this issue—do the two works serve the same purpose broadly defined—is not likely to differentiate competitive substitutes that may pose a risk of market failure, from everything else. For example, using the Court’s approach, one might conclude that salt supersedes the demand and hence serves as a substitute for pepper, at least when both are used to season food. Just like Warhol’s print and Goldsmith’s photo, while salt and pepper are not perfect substitutes, both serve the same purpose for that specific use—to season food. Just like Warhol’s print and Goldsmith’s photo, both salt and pepper are sold commercially and thus vie for the same consumer dollars. Therefore, salt supersedes the purpose of pepper, and thus, salt is a substitute for pepper—or so the Goldsmith Court’s reasoning would lead one to conclude.

Yet, that conclusion is nonsense, and the reasoning by which the Court reached it misses the point entirely. Even though salt and pepper are used for the same purpose, they have different flavor and texture profiles and are therefore not substitutes.¹⁵⁵ More likely, even when both are used to season food, they are unrelated goods.¹⁵⁶ As a result, the introduction of pepper into a marketplace, where formerly only salt was available, does not lead to market failure. To the contrary, enabling consumers to choose, against the backdrop of a budget constraint, whether to buy salt or pepper, and if so, how much of each, is precisely how markets ensure a Pareto optimal allocation of resources. It allows each consumer to maximize their utility by making purchases that reflect their preferences subject to their budget constraint. That is not market failure but market success. The substitution that has long been the bête noire of copyright is the competitive substitution of copies that both: (i) duplicate the appeal of the original so closely that consumers will buy them instead of the original; and (ii) do so at a lower price because they were copied. Only when both conditions are satisfied does a risk of market failure arise that would justify government intervention, such as copyright. Even if “the two are not perfect substitutes,” neither salt and pepper, nor Warhol’s print and Goldsmith’s photo, pose such a risk of market failure. Orange Prince neither duplicates the appeal of Goldsmith’s preexisting photo so closely that consumers will substitute one for the other nor was Orange Prince available at a lower licensing fee because of copying.¹⁵⁷

Moreover, as the Court expressly recognized in Google and Campbell, whether two products are substitutes is ultimately a factual issue. In both Google and Campbell, the Court addressed this issue under the fourth fair use factor, the effect of the use “upon the market for or potential value of the copyrighted work,” rather than under the first factor, “the purpose and character of the use.”¹⁵⁸ Yet, whether addressed under the first fair use factor or the fourth, the question of whether if one of the Warhol works had not been available or if the licensing fee for using Orange Prince had been just a little higher, Condé Nast would have used the Goldsmith photograph instead, remains factual, not legal.¹⁵⁹ It cannot ordinarily be resolved on summary judgment.¹⁶⁰

More importantly, however, the emphasis on substitution feels, in part, like an anachronistic throwback, and, in part, like an intentional misdirect. As historical anachronism, the Court’s test under the first fair use factor, asking whether the defendant’s use “‘supersede[s] the objects’ of the original creation” or “‘supplant[s]’ the original,” traces back to Justice Story’s opinion in Folsom v. Marsh in 1841.¹⁶¹ However, in 1841, the copyright act then in force provided no derivative work right at all.¹⁶² The 1831 Copyright Act gave the copyright owner “the sole right and liberty of printing, reprinting, publishing and vending such map, chart, book or books.”¹⁶³ No dramatization right. No translation right.¹⁶⁴ No abridgement right. At that time, competitive substitution was not merely something copyright particularly disliked—the bête noire of copyright, as it were—but was copyright’s sole focus and defined the full extent of the copyright owner’s exclusive rights.

But that is no longer true. Beginning in 1870 with congressional recognition of the translation right,¹⁶⁵ copyright has expanded its focus beyond competitive substitution to encompass the opportunity to license an original work for uses that are not competitive substitutes. A film based upon a novel, for example, increases on average sales of the underlying novel, and is thus a complement, not a substitute. Yet, as the Goldsmith Court itself recognized, the Copyright Act today grants the copyright owner in a novel the exclusive right to make such a film.¹⁶⁶ While competitive substitution remains a central concern of copyright, it does not define its limits, as it did in 1841. So, the question becomes, why the pretense? Given the express statutory recognition of the derivative work right, why bother pretending that Orange Prince is a competitive substitute for Goldsmith’s black-and-white photo? There was no evidence in the record that the use of Orange Prince for the magazine cover cost Goldsmith a sale, in the sense that but-for Orange Prince, Goldsmith’s black-and-white photo would have been used instead. But it may have cost her a “lost opportunity to license,” as I have used that phrase previously.¹⁶⁷ Today, that should be enough, and yet, it is not the Court’s focus.

For me, this is where the feeling that the Court’s emphasis on substitution is an intentional misdirect arises. Focusing on substitution enables the Court to distinguish its fair use holding in Google v. Oracle.¹⁶⁸ In Google, the later work, Android, served as an operating system for smart phones.¹⁶⁹ The earlier work, Java, in contrast, operated in the laptop and desktop markets.¹⁷⁰ As a result, the two served “distinct and different” “environments.”¹⁷¹ Adopting a narrow focus on substitution allowed the Goldsmith Court to assert, accurately, that Android did not “supersede the objects” of the original work, Java, on which it was based.¹⁷² But that begs the question whether Android superseded the objects of a licensed Java derivative designed for the smartphone market. The Google Court addressed that issue directly, ruling under the fourth fair use factor that a jury could permissibly find that “Android did not harm the actual or potential markets for Java SE.”¹⁷³

Thus, the Goldsmith Court’s insistence that substitution is the focus of the inquiry enabled the Goldsmith Court to dodge two potential impediments to its desired result. First, it enabled the Goldsmith Court to decide the fair use issue as a matter of law. Where the Court in both Campbell and Google required factual evidence on whether the later use affected “the potential market for or value” of the earlier work, the Goldsmith Court simply offered its own opinion on the issue. No pesky evidence. No expensive trial. Simple judicial fiat. Second, the focus on substitution enabled the Court to avoid the need to balance what is likely to be gained and what is likely to be lost from finding or rejecting the fair use privilege on the facts presented.

Such balancing is the quintessential heart of any realist approach to fair use. In Google, the Court addressed this balancing directly. The Google Court began by acknowledging that:

Google copied portions of the Sun Java API precisely, and it did so in part for the same reason that Sun created those portions, namely, to enable programmers to call up implementing programs that would accomplish particular tasks.¹⁷⁴

For the Goldsmith Court, this statement would mark the end of the analysis into whether the defendant’s use was transformative. But for the Google Court, it marks the beginning. From that starting point, the Google Court goes on: Android is a new program; it expands the utility of Android-based smartphones; it offers a “highly creative and innovative tool” for smartphones.¹⁷⁵ In short, Google used what it copied “to create a new platform that could be readily used by programmers.”¹⁷⁶ For the Google Court, that made Google’s verbatim copying of 11,500 lines of Java code to create Android transformative.¹⁷⁷ The copying gave us a new program and thus, “was consistent with that creative ‘progress’ that is the basic constitutional objective of copyright itself.”¹⁷⁸ As for the other side of the balance—the risk that a finding of fair use will reduce creative output by undermining incentives—the Court relied on the jury’s finding of no harm to “the potential market for or value of” Java.¹⁷⁹

In Warhol, no such balancing is evident.¹⁸⁰ Despite the self-evident differences between Orange Prince and the black-and-white photograph, the Goldsmith Court rejects the argument that these differences establish a transformative purpose or character:

Although the purpose could be more specifically described as illustrating a magazine about Prince with a portrait of Prince, one that portrays Prince somewhat differently from Goldsmith’s photograph (yet has no critical bearing on her photograph), that degree of difference is not enough for the first factor to favor AWF, given the specific context and commercial nature of the use.¹⁸¹

As for why that degree of difference is not enough, the Court explains that if such a difference in degree were enough, that would allow a thousand creative flowers to bloom:

To hold otherwise would potentially authorize a range of commercial copying of photographs, to be used for purposes that are substantially the same as those of the originals. As long as the user somehow portrays the subject of the photograph differently, he could make modest alterations to the original, sell it to an outlet to accompany a story about the subject, and claim transformative use.¹⁸²

The Court does not explain exactly why allowing a thousand creative flowers, each with modest alterations, to bloom is undesirable.¹⁸³ Just as leaving room for the substantially similar but different Android operating system did, leaving room for additional substantially similar but different photographs would seem “consistent with that creative progress that is the basic constitutional objective of copyright itself.”¹⁸⁴ A thousand creative flowers would seem to allow everyone to find the exact version that best satisfies their aesthetic and other preferences. Allowing each consumer to purchase the version that best satisfies their preferences, subject to an exogenous budget constraint, is the very definition of market success, not market failure. Instead of explaining why that is not true, the Court merely repeats itself:

Copying might have been helpful to convey a new meaning or message. It often is. But that does not suffice under the first factor.¹⁸⁵

Again, the Court offers no explanation for why copying to convey a new meaning or message does not suffice. Perhaps, the Court believes the reason is self-evident. Perhaps, the Court believes that repeating the statement will make it true.

Most likely, however, these assertions reflect a fundamental misunderstanding as to how markets, particularly those for copyright-protected goods, work. For example, the Court might be implicitly assuming that licensing for magazine covers about Prince (the “market”) is a zero-sum game and that each additional “substantially similar” creative flower reduces the revenue for existing market participants proportionally. Under those assumptions, Goldsmith’s photograph might stand to lose directly as copyright allows more and more creative flowers to bloom. Indeed, under those assumptions, as each additional “substantially similar” flower enters the market, Goldsmith loses proportionally. At some point, an overly generous fair use doctrine might allow enough creative flowers to enter the market that Goldsmith no longer earns enough to cover her persuasion costs and the market, as a result, fails.

Among other difficulties, this implicit model of the market for copyrighted works misunderstands how markets operate. If “substantially similar” meant that two works: (i) were exact copies of each other; (ii) were as indistinguishable as two grains of rice; and (iii) were otherwise competitive substitutes; the additional creative flowers may add no value beyond that created by Goldsmith’s photo. But there is a wide gap between substantially similar and indistinguishable. In that wide gap, there is plenty of room for differences in aesthetic, meaning, or message between an earlier and a later work that matter to consumers.¹⁸⁶ In the Goldsmith case, Orange Prince may be substantially similar to Goldsmith’s photo, for purposes of copyright law, but the two works are not the same. The differences between them matter in two ways. First, some consumers may prefer Goldsmith’s photo, while others prefer Orange Prince. As a result, Orange Prince adds value. Orange Prince may add value by increasing the licensing fees available within the existing market. For example, the available license fees may increase in that market from $1,000 to $2,000. Or, more likely, Orange Prince may add value by creating a new market for licensing a Warhol original depicting Prince. Second, the differences between the Goldsmith photo and Orange Prince matter because the differences mean that Orange Prince is not a competitive substitute. The availability of Orange Prince without a copyright license would not lead to market failure, but to market success, as consumers choose which work better satisfied their preferences.

For a realist, there are, in the end, three possibilities that might justify the Court’s holding that Warhol’s use is insufficiently transformative. First, the Goldsmith Court could be saying, as the above discussion suggests, that Orange Prince conveys no new meaning, message, or aesthetic to anyone beyond Goldsmith’s black-and-white photo. In other words, we, as a society, are no better off: (i) with Orange Prince and the photo than we would be (ii) with the photo alone. Second, the Goldsmith Court could be admitting that Orange Prince adds something new, and thus society is better off with both works than society would be with the photo alone. Nevertheless, a license is both practicable and readily available. Society, therefore, has nothing to lose by requiring the Foundation to pay Goldsmith her pound of flesh for the use of Orange Prince on a magazine cover. Or, third, the Goldsmith Court could be saying that, on balance, shifting rents from Warhol to Goldsmith will lead to more and better original works going forward.

From a realist perspective, however, each of these possibilities has fundamental flaws. The first possibility—that Warhol’s artwork adds no value to society—is untrue. The Court acknowledges that we, as a society, are better off with Warhol’s Prince series than we would be without them.¹⁸⁷ And as the Court long ago recognized, art remains art whether hung on the walls of a museum, or used commercially as an advertisement for a circus or on the cover of a magazine.¹⁸⁸ So the particular use made should not matter. Moreover, as the Goldsmith Court itself acknowledges, courts should not be making these sorts of artistic value judgments “outside of the narrowest and most obvious limits” in any event.¹⁸⁹ If Orange Prince were a mere Xerox copy of Goldsmith’s black-and-white photo, then a court might be justified in holding, as a matter of law, that Warhol’s copy did not add value to society through the creation of a new work with a new aesthetic, a new message, or a new meaning.¹⁹⁰ But whatever one thinks about the merit or value of Warhol’s art, I think that we can all agree that Orange Prince is not a mere Xerox copy of Goldsmith’s black-and-white photo.

The second possibility—that society has nothing to lose by requiring a license—is also untrue. Society always has something to lose by requiring a license. Here, a realist must distinguish, as the Court did not, between circumstances where requiring a license will increase the incentives for authorship and circumstances where requiring a license will not. The paradigmatic example of the first scenario is copyright’s prohibition on the unauthorized reproduction of a copyrighted work. By requiring a license to print copies of a novel, for example, copyright can ensure a monopoly in the market for those copies.¹⁹¹ Only the authorized copies will be available. That monopoly will increase the producer surplus associated with those copies for both the publisher and the author. Copyright thus creates a larger pie for the publisher and author to share.

Nevertheless, although prohibiting competitive substitution can increase the incentives for authorship, we should not misunderstand what competitive substitution of the sort that may lead to market failure requires. Competitive substitution does not occur merely because a thousand “substantially similar” flowers bloom, even if each has only “modest alterations.” In recent work, I have shown that a hardcopy, a paperback copy, and an e-book version of the same novel are not competitive substitutes for each other.¹⁹² While how close two versions of the same work have to be for competitive substitution is ultimately an empirical question, it is an empirical question that research has largely answered. For direct competitive substitution to occur, something like unauthorized file sharing of exact copies of the same work is necessary. Even that sort of file sharing is not a perfect substitute¹⁹³—a reminder of how useless the Court’s use of that characterization was. But file sharing can and does reduce the sales of authorized copies of, for example, a song. The file sharing of a digital copy of a song is, thus, a competitive substitute for the purchase of an authorized copy of that same song. But Orange Prince is not. In the record before the Court, there was no evidence that but-for the availability of Orange Prince at a lower price, Condé Nast would have used Goldsmith’s black-and-white photo instead.¹⁹⁴

But the Court might say Goldsmith could derive, and did derive, revenue by licensing Warhol’s use. That’s true, but again, in balancing what society has to gain and what it has to lose in terms of promoting “the Progress of Science” from requiring such a license, a realist would again recognize that requiring a license for such a derivative use will sometimes increase and sometimes decrease the incentives for authorship. Requiring a license to make a film from a copyrighted novel, for example, likely increases the incentives for both the novel’s author and the film’s producer. By requiring a license, copyright ensures that instead of twelve “motion picture versions” of Harry Potter and the Sorcerer Stone, there will be only one—at least at any given time. By doing so, copyright ensures a monopoly for the authorized film. That monopoly will enable the one licensed film producer to capture more producer surplus than the twelve film producers would have captured, combined, in a competitive market. More producer surplus means a larger pie, and hence more revenue, for both the novel’s author and the film producer to share. Thus, requiring a license for a motion picture version of a novel means more incentives for both.

However, that is not true in the Goldsmith case. In the Goldsmith case, requiring a license case will not increase the incentives for authorship. Unlike the authorized publisher or the authorized motion picture version, the Foundation will not earn more from Warhol’s artwork by obtaining a license from Goldsmith. The price of a Warhol original in the market derives not from a license from Goldsmith, but from Warhol’s own market power and his fame as an artist. Even if a thousand other artists used the Goldsmith black-and-white photo to create their own works of art, the value of the Warhol originals would remain unchanged.¹⁹⁵ Requiring a license does not therefore increase the size of the pie associated with the Warhol Prince series. It simply requires Warhol to share his pie with Goldsmith.

If forcing that wealth transfer were costless, there would be no associated efficiency loss. But like most forced wealth transfers, this one is not costless. To force the wealth transfer, the Court interprets copyright law to require the Foundation to obtain a license. Negotiating a license imposes transaction costs. Those transaction costs come out of the producer surplus available. As a result, forcing Warhol to share reduces, rather than increases, the incentives for authorship. Some of the incentives that would otherwise go to an artist will now go to lawyers. Because it reduces the incentives for authorship, requiring a license in the Goldsmith case cannot address the risk of market failure long thought to justify copyright. Redistributing wealth from the Foundation to Goldsmith may feel emotionally satisfying to some, but it does not and cannot promote “the Progress of Science.”¹⁹⁶

In addition, requiring a license also opens the door to a failed licensing negotiation. Even where: (i) a license is generally practicable because there are only two parties to the negotiation and the gains from trade far exceed the transaction costs of negotiating the license; and (ii) the copyright owner is generally willing to license the use at issue, a license negotiation may still fail. Oracle and Google, for example, tried to negotiate a license for the creation of Android from Java, but were unable to agree to terms.¹⁹⁷

Thus, under the second possibility, from a realist perspective, concluding that Warhol was an infringer reduces the incentives for authorship. It also creates a risk of licensing failure. In the Goldsmith case, society does have something to lose by requiring a license.

The third possibility—that shifting rents from Warhol to Goldsmith would somehow increase creative output—would justify the Court’s ruling if true. Unfortunately, the Court offers no reason to believe it is true. The Court does note that “licenses, for photographs or derivatives of them, are how photographers like Goldsmith make a living. They provide an economic incentive to create original works, which is the goal of copyright.”¹⁹⁸ However, as the Court acknowledges, “[t]his copyright case involves not one, but two artists.”¹⁹⁹ Even if one believes, despite the empirical evidence to the contrary,²⁰⁰ that additional revenue for authors leads to increased creative output, the Court offers no reason to believe that merely shifting revenue from Warhol to Goldsmith, or from artists like Warhol to artists like Goldsmith, will increase, on balance, creative output, either in this case or going forward. Even if the “more-revenue-equals-more-creative-output” assumption holds, requiring a license in this case would increase Goldsmith’s creative output at the expense of decreasing Warhol’s, or more generally, increase the creative output from artists like Goldsmith at the expense of decreasing the creative output of artists like Warhol going forward. The Court offers no reason to believe that society would be better off, on balance, from that trade-off. Indeed, as previously noted, the Court expressly disclaims its authority to make any such value judgment.²⁰¹

In short, the Court’s realism is unrealistic. In the Warhol case, more incentives for Goldsmith necessarily mean less for Warhol. From a realist perspective, there is no reason to believe that such rent redistribution will increase authorial output, on balance, or otherwise “promote the Progress of Science.” Realism is not, however, the sole, or perhaps even primary, interpretative approach the Goldsmith Court relied on to justify its outcome. The Court also seems to suggest that the statutory text requires it. To the textual justifications for the Court’s outcome, we now turn.

B. Textualism without Text

The Court’s textual analysis ignores the statutory text. The Court’s principal textual argument follows the lead of Judge Easterbrook, who is not exactly well-known for his textualist approach to statutory interpretation.²⁰² In Kienitz v. Sconnie Nation, Judge Easterbrook, writing for a panel of the Seventh Circuit in 2014, expressed skepticism about transformative use.²⁰³ He cautioned that relying on transformative use to find a use fair seemed to conflict with the statutory definition of a derivative work, writing:

asking exclusively whether something is “transformative” not only replaces the list in § 107 but also could override 17 U.S.C. § 106(2), which protects derivative works. To say that a new use transforms the work is precisely to say that it is derivative and thus, one might suppose, protected under § 106(2).²⁰⁴

In Warhol, the Court embraced the same argument:

But Campbell cannot be read to mean that § 107(1) weighs in favor of any use that adds some new expression, meaning, or message. Otherwise, “transformative use” would swallow the copyright owner’s exclusive right to prepare derivative works.²⁰⁵

In particular, both courts point to the word “transformed” in the definition of a derivative work to undermine transformative use in the fair use analysis and to confine its potential scope.²⁰⁶ Of course, both courts are right that the word “transformed” does appear in the statutory definition of a derivative work.²⁰⁷ From a textualist perspective, there would seem, then, to be a conflict in using the same word to define a use both as infringing, as a derivative work, and as non-infringing, under fair use.²⁰⁸

However, this supposed conflict is more apparent than real. Properly interpreted, the word “transform,” when used in the fair use context, carries a different meaning from the same word when used in the derivative work context. Indeed, the word “transform” carries precisely the opposite meaning in the two contexts. For fair use, the Campbell Court defined “transformative” as copying to create a work with “new expression, meaning, or message.”²⁰⁹ In contrast, for a derivative work, “transformed” means copying an existing work into a new language or artistic medium, while retaining the same expression, the same meaning, and the same message.

Both the Seventh Circuit in Keinitz and the Court in Warhol mistake the meaning of the root word “transform” in the derivative work context because they pluck the word “transformed” out of the definition of a derivative work and attempt to define it in isolation. For a textualist, such an approach is improper. After all, “transformed” is not the only word Congress used in defining a derivative work. We should not ignore the rest.²¹⁰ If we look at the statutory text in its entirety, Congress defined a derivative work, in relevant part, as:

… 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.²¹¹

In this statutory definition, the word “transformed” appears as part of a catch-all phrase at the end of a list of specific examples. As a result, for a textualist, the principle that the specific defines the general—or in Latin, ejusdem generis—governs.²¹² In other words, to define the catch-all phrase “any other form in which a work may be … transformed,” we must look to the specific examples listed.

If we look to the specific examples the definition of a derivative work includes, Orange Prince is not a derivative work. Sure, it might be thought to be “based upon” the Goldsmith black-and-white photo in some general sense. It might also be thought to be “any other form in which [the Goldsmith photo] may be … transformed,” again in some general sense. And if “based upon” and “transformed” were the only words found in the statutory definition, such a broad and general interpretation of those words might be appropriate. However, they are not the only words Congress used to define a derivative work. Congress included not just these general phrases but specific examples to illustrate what the general phrases mean. For art such as Warhol’s, Congress used the phrase “art reproduction” to identify the specific type of later works based upon an earlier work that would be considered derivative works.²¹³ “Art reproduction” is a term of art and has a very specific meaning. It does not merely refer to a later work that is based upon, or transforms, an earlier work in some broad or general sense; Rather, “art reproduction” describes a later work that attempts to duplicate the earlier work as closely as possible.²¹⁴ Under that literal meaning, Orange Prince is not an art reproduction.

Nor is Orange Prince any of the other specific examples in the statutory text. It is not a translation or a musical arrangement or a dramatization or a fictionalization or a motion picture version or a sound recording or an abridgment or a condensation. And because it is none of the specific examples listed, Orange Prince also does not fall within the general language, whether “based upon” at the start of the list of examples, or “any other form in which a work may be recast, transformed, or adapted” at the end of the list. For a textualist, the specific defines the general.

For a textualist, both the general phrase “based upon” at the start of the list and the catch-all phrase “[otherwise] transformed” at the end of the list should be interpreted in the light of the specific examples set forth in the statute. Interpreted in the light of the specific examples set forth in the statute, both “based upon” and “any other form in which a work may be … transformed” mean based upon or transformed in the same way as the listed examples are based upon or transform an earlier work.

When we look at the specific examples, it stands out that they all refer to later works that copy from an earlier work in order to tell the same story, share the same aesthetic, convey the same message, albeit in a new language or artistic medium, as the earlier work. If I take the characters from Romeo and Juliet and put them in a movie that tells a different story, that is not a “motion picture version” of Romeo and Juliet. A motion picture version of the play Romeo and Juliet tells the same story, shares the same aesthetic, and conveys the same message as the play. Similarly, if I take the characters from Romeo and Juliet and tell a different story based upon those characters in German, that is not a “translation” of Romeo and Juliet. A translation of Romeo and Juliet from English to German tells the same story, shares the same aesthetic, and conveys the same message as the English original. While some changes are inevitable when a novel is brought to life on the screen or when an English-language novel is translated, we should not mistake those incidental changes for the intentional changes we see in Orange Prince. Nor should we overlook the fact that every one of the specific examples Congress listed in the statutory definition of a derivative work involves a later work that tells the same story, shares the same aesthetic, or conveys the same message, albeit in a different language or artistic medium, as the work they are based upon. Because that characteristic is common to all the specific examples Congress listed, the phrases “based upon” and “transformed” should be interpreted to cover additional uses only if they do likewise. The specific defines the general.²¹⁵

Thus, put in context and interpreted properly, the word “transformed” in the derivative work right carries exactly the opposite meaning from the one the Campbell Court gave “transformative” in the fair use context. Where “transformative” for purposes of fair use means copying to create a new work with a new meaning, a new message, or a new aesthetic, “transformed” in the derivative work right means copying to create the same work with the same meaning, the same message, and the same aesthetic, albeit in a new language or artistic medium.

Rather than interpret the statutory text as written, the Goldsmith Court both ignores the inconvenient parts of the statutory text and rewrites the statutory definition. Thus, when the statutory definition of a derivative work does not include the words “sequel” and “spinoff,” the Court rewrites the text so it does.²¹⁶ This is not textualism.

From a textualist perspective, the Goldsmith Court makes other mistakes as well. For example, the Court acknowledges that the statutory text expressly makes the derivative work right (and all the other exclusive rights as well) “subject to” fair use.²¹⁷ On its face, the phrase “subject to” means that some uses that would otherwise qualify as infringing derivative works are nonetheless non-infringing because they are a fair use. The phrase “subject to” also expressly subordinates the derivative work right to fair use. If the textual phase “subject to” is not plain enough, the text of the statutory fair use section begins: “Notwithstanding the provisions of section 106 and 106A, the fair use of a copyrighted work … is not an infringement of copyright.”²¹⁸ Like the phrase “subject to,” the “notwithstanding” text expressly defines fair use as non-infringing activity even when it would otherwise fall within the scope of one of the copyright owner’s exclusive rights— stating directly that when fair use applies, it trumps the derivative work right. Nevertheless, the Court rejects the plain meaning of the statutory text and reverses the hierarchy the statutory text establishes. Instead of following the statutory language, the Court writes:

To preserve [the derivative work] right, the degree of transformation required to make “transformative” use of an original must go beyond that required to qualify as a derivative.²¹⁹

In this sentence, the Court both reverses the hierarchy the statutory text imposes and renders fair use impotent and irrelevant. Because the statutory text makes the derivative work right “subject to” fair use, the Court’s role is not to preserve the derivative work right from fair use. Rather, it is to identify those uses that fall within the scope of the derivative work right that are nonetheless non-infringing because they are fair. More problematically, if fair use applies only to a use that “go[es] beyond” the derivative work right, as the Goldsmith Court suggests,²²⁰ then it will never apply at all. If the use at issue “go[es] beyond” the derivative work right, then the later work is so far different from the earlier work as to be non-infringing in any event. At that point, fair use becomes unnecessary.

From a textualist perspective, we see the same “pro-exclusive rights, anti-fair use” bias in the Court’s willingness, on the one side, but refusal, on the other, to add to the open-ended language in the text of the Copyright Act. For example, both the definition of the derivative work and the fair use privilege include open-ended statutory language.²²¹ In the definition of a derivative work, the statute uses the open-ended phrase “such as” before the list of examples.²²² Similarly, in the preamble, section 107 offers a list of uses that might qualify as fair use and precedes the examples with the open-ended words “including” and “such as.”²²³ In setting forth the four fair use factors, the statutory text states that an analysis of fair use “shall include” the four factors.²²⁴ Again, the word “include” is open-ended and expressly leaves room for courts to develop additional factors. In defining “the purpose and character of the use” under the first fair use factor, the statutory text is again open-ended: “including whether such use is of a commercial nature.”²²⁵

Such open-ended statutory text gives the Court not complete freedom but at least some leeway to add to the statutory text. The Court probably went beyond the permissible bounds of that leeway in its zeal to add “sequels” and “spinoffs” to the definition of a derivative work.²²⁶ But regarding whether the Court went too far in expanding the derivative work right or not, the Court should presumably have the same leeway to add to the examples of fair use, to develop additional fair use factors, and to define what constitutes a fair purpose or character under the first factor of the fair use analysis. Yet, the Court eschewed any such authority. For example, rather than balance what the public has to gain and what the public has to lose, generally, from a defendant’s use, as the Court did in Google v. Oracle, the Goldsmith Court pretended that the statutory text required a different balance under the first fair use factor. Because the first fair use factor expressly mentions “commercial nature,”²²⁷ the Goldsmith Court insisted that any transformative character or purpose must be balanced solely against the work’s commercial nature.²²⁸ Despite the open-ended nature of the statutory text defining the first fair use factor, nothing else may be considered, at least according to the Goldsmith majority. In effect, the same open-ended statutory text must be interpreted to expand the derivative work right but to narrow the fair use privilege, according to the Court. Nothing in the statutory text supports that approach.

Just as realism cannot justify the Court’s decision, neither can textualism. The statutory text of the Copyright Act does not support the Court’s outcome, its reasoning, or its approach. That leaves purposivism as the last possible interpretative approach that might justify the Court’s decision, and to that interpretative approach we now turn.

C. Purposivism without Purpose

The Court’s purposivism defeats the purpose for which Congress enacted copyright. Like its principal argument from a textualist perspective, the Court’s principal purposivist argument also focuses on the risk that an overbroad interpretation of transformative use in the fair use context will gut the derivative work right generally and the exclusive right of the copyright owner of a novel to prepare a film based upon the novel specifically. Here, the argument is not a realist’s argument that society will be better off, and the Progress of Science better promoted, if the Court holds Warhol’s use unfair. Nor is it a textualist’s argument that the text of the Copyright Act requires the Court to hold that Warhol’s use is unfair because the derivative work right trumps the fair use privilege. As discussed, the text of the Copyright Act expressly states the opposite: fair use, if present, trumps the derivative work right. Instead, the argument seems to be that Congress intended the owner of a copyright in a novel to have the exclusive right to transform that novel into a motion picture and ruling in favor of Warhol would frustrate that intent.

Yet, the Court is mistaking a question for an answer. Let us assume that Congress intended in 1976 to recognize that making a film from a copyrighted novel was an infringing rather than fair use. The question is whether Congress intended that placement of the line between infringing and fair use to remain forever in place no matter how the technology, markets, and economics of creating and distributing original works of authorship changed. If Congress did so intend, then the film-from-a-novel example may suggest an answer to the proper placement of the line between infringing and fair uses, at least in those cases that are similar. But if Congress did not so intend, then the film-from-a-novel example is not an answer, but a question to be answered: Have the technology, markets, and economics associated with such films changed sufficiently such that what was once infringing should now be fair?

Congress, after all, made the derivative work right, along with every other right, “subject to” fair use.²²⁹ Thus, even a use that would seem to fall squarely within the scope of one of the exclusive rights—indeed, even a use that was once the paradigmatic case of infringement—may become fair as the technology, markets, and economics of creating and distributing original works change. Consider, for example, the mechanical duplication of an entire copyrighted work. Such a use falls squarely within the reproduction right.²³⁰ Moreover, at one time, mechanically duplicating a copyrighted work and offering the resulting unauthorized copy as a competitive substitute for the original at a lower price represented the paradigmatic case of unfair, infringing use. While such mechanical copying increased dissemination of an existing work, the use was thought to be on balance unfair because allowing lower priced substitute copies was thought to pose an existential threat to authorship. As the Stationers Guild argued in a 1586 petition to the Star Chamber: If a publisher does not have the right to stop such unauthorized mechanical duplication, “no books at all should be [printed].”²³¹ This sort of mechanically perfect copy is precisely the use that would seem most likely to “prejudice the sale, or diminish the profits, or supersede the objects, of the original work,” as Justice Story wrote in creating and explaining the first fair use factor in Folsom v. Marsh.²³²

Yet, even if true, that conclusion is not a truth for all time. It is a function of the technology for authoring and distributing original works and the associated economics and markets of a given time. It may be during the analog era and the days of the printing press that exact duplication of an entire work created a material risk of market failure. However, exact duplication was not a threat to authorship in the pre-printing press era, when books were copied by hand. And exact duplication is not inevitably a threat to authorship today.²³³ Just as the introduction of the printing press gave rise to the need for copyright,²³⁴ so too, today, new and ongoing changes in the technology and economics of authorship have shifted and will continue to shift where the line between fair and unfair uses should fall to optimally balance copyright’s constitutional purpose. As a result, uses that were once unfair can become fair as the technology and economics of authorship change. Thus, in Sony Corp. v. Universal Pictures, the Court held that the mechanical duplication of an entire work for a consumer’s later viewing—a use that, in Justice Story’s day, represented the paradigmatic example of an unfair use—had, as the technology and the associated markets changed, become fair.²³⁵ For Justice Story, the relevant reproduction and distribution technology was the printing press. At that time, only a competing publisher could make such a use. As a result of the technology available at that time, allowing such a use raised the specter of lower priced substitute copies that would deprive the original author and publisher of any return on their investment. But by the time of Sony, technology had changed. With the Betamax, an ordinary consumer could make a copy. Together with the change in the associated markets from a direct charge for access, that is, $9.95 for a copy of the book, to an indirect charge, that is, watch embedded advertisements, the balance between: (i) the threat to future creative output and (ii) the potential expansion in access to existing creative works had shifted in favor of allowing such copying, at least under some of the circumstances the facts in Sony presented²³⁶

Fair use thus allows a court to incorporate flexibility into the Copyright Act. As the technologies, markets, and economics associated with authorship change, courts can adjust the boundaries copyright sets between fair and unfair uses accordingly. Even a use that at one time represented the paradigmatic case of infringement may become fair as technology changes and the associated economics of, and markets for, authorship change, as the Sony case itself well illustrates.²³⁷ We should therefore be cautious before enshrining a historical relic, such as the rule requiring a license before making a movie from a book, as the touchstone for drawing the line between fair and unfair use for all time.

When we turn to the film-from-a-novel context specifically, the technology, markets, and economics of such uses have all changed dramatically since 1976. In particular, the shift from the analog creation and distribution technologies of the 1970s to digital has radically changed the economics of, and associated markets for, audio-visual works based upon novels. In the analog era, making a film or audio-visual work for public performance was expensive. Distribution was limited. As a result, only the privileged few could share their creativity with the public. In the digital era, the costs of creation have fallen dramatically, and the avenues for distribution have increased exponentially. Through social media platforms, such as YouTube, Instagram, Twitch, and TikTok, and with an iPhone in their hand, anyone can make and share their audiovisual creativity with the world. As of 2022, on YouTube alone, more than 500 hours of new audiovisual content was added every minute.²³⁸ That is 30,000 hours of new audiovisual content every hour. At first glance, the risk of market failure in the digital era is not that we may have too little original content but that we will have too much.

In terms of the balance copyright strives to achieve, digital technologies have shifted that balance sharply in favor of fair use for two reasons. First, the rise of digital technologies has sharply reduced the cost of producing and distributing original works of authorship. The need for copyright to provide an incentive for such authorship, and as a result, the benefits from copyright’s exclusive rights, have fallen correspondingly. Second, the rise of digital technologies has also radically expanded the avenues for building on earlier works. The costs of copyright’s exclusive rights have thus risen accordingly.

If we focus on filmmaking specifically, during the analog era, filmmaking exhibited economic characteristics, including high fixed and low marginal costs, that may have led to natural monopoly even in the absence of copyright. In a natural monopoly setting, whether copyright protects broadly against unauthorized film versions of a novel or not, the market itself would only support one such work. In the digital era, that is no longer true. Consider the slew of unauthorized but professionally produced Sherlock Holmes audio-visual works. After the copyright on the Sherlock Holmes stories finally expired,²³⁹ the public received at least three English-language, professionally produced film or television works based upon the novel more or less simultaneously. Robert Downey Jr. starred as the titular character in a pair of movies, Sherlock Holmes and Sherlock Holmes: A Game of Shadows, released in 2009 and 2011, respectively.²⁴⁰ Benedict Cumberbatch starred as the titular character in four seasons of the BBC series, Sherlock, with the first season released in 2010.²⁴¹ Jonny Lee Miller starred as the character in the CBS television series, Elementary, that aired for seven seasons, beginning in 2012.²⁴²

It seems self-evident that multiple Sherlock Holmes works better promote “the Progress of Science” than does just one.²⁴³ With several to choose from, fans can find the version that matches their preferences more perfectly. True fans can enjoy all three. Copyright should not, after all, seek to achieve creative monopoly—a single creative work controlled by a single creative voice. After all, individual tastes and preferences may vary. So your idea of the perfect song and mine may be quite different. Instead, copyright should expressly encourage creative competition—different authors presenting their own variations on any given theme. If courts continue to place the line between fair and infringing uses in the digital age as they placed it in the analog age, we will have to wait for copyright to expire to enjoy the full benefits of the creative competition that digital technologies enable. Recognizing that the introduction of digital technology has changed the economics of authorship—and embracing a more vibrant fair use doctrine and adopting a correspondingly narrower scope to the reproduction, derivative work and public performance rights—would allow for more creative competition during a work’s copyright term.

Even if derivative uses were not inevitably natural monopolies in the analog era, derivative uses during the analog era were both fewer and generally more predictable. For example, the predictability that a best-selling novel would be made into a film allowed licensing revenue from that use to become part of the ex ante incentives for authors to write best-selling novels. Moreover, in the analog era, each derivative work had to earn sufficient revenue to cover the already high costs of analog distribution. As a result, the transaction costs for licensing derivative use were trivial, relative to the other costs of authoring and distributing a derivative work, and licensing failure were correspondingly unlikely.

But the rise of digital creation and distribution has changed this, too. In the digital era, reuses have become both many and unpredictable. Rather than become an expected part of the ex ante incentives, requiring licenses for follow-on uses will more likely create unpredictable and unexpected windfalls. Precisely to the extent that such windfalls are unpredictable and unexpected, they will not serve to encourage authorship of the earlier work. Moreover, in the digital era, a derivative work need not earn much, if any, revenue to make its creation and distribution economically rational. Thus, the risk that the transaction costs of negotiating a license will exceed the gains in trade, and hence cause licensing failure, rise accordingly. Failing to adjust the line between fair and unfair uses accordingly to reflect these differences between the digital and analog eras, makes it less likely that copyright will strike an appropriate balance between earlier and later authors.

Rather than rely on the novel-to-movie copyright rule as a touchstone to expand copyright for other types of works, courts should be re-visiting whether, if tested against a fair balancing of its costs and benefits today, the rule should continue to exist in our new digital world at all. At the very least, it suggests that applying an evidentiary-based, case-by-case balancing of whether a finding of infringement or a finding of fair use would better promote “the Progress of Science,” courts should not use the novel-to-film rule as a model for expanding similar protections elsewhere in copyright.

That is the approach the Court adopted in Google v. Oracle, where the Court recognized the need to avoid expanding the film-from-a-novel rule into a general rule for other types of copyrighted works.²⁴⁴ Although the Google Court acknowledged the broad derivative work right in the novel-to-film context, it refused to rely on that rule as a touchstone for resolving whether Google’s copying was fair or unfair.²⁴⁵ In rejecting Oracle’s argument that Google’s use had harmed the “potential market for or value of” Java, the Court began its analysis with three facts it believed a jury could find. First, a jury could find that the copyright owner of Java “was poorly positioned to succeed in the mobile phone market.”²⁴⁶ Second, a jury could find that the markets in which Java and Android were marketed—desktops and laptops for Java, smart phones for Android—were separate and distinct.²⁴⁷ Third, the jury could find that Google’s use of the lines of code from Java to create Android would benefit Java’s copyright owner “as it would further expand the network of Java-trained programmers.”²⁴⁸

Yet, all these facts are just as true in the novel-to-film context. First, popular authors, such as J.K. Rowling, are poorly positioned, and lack the technical skills and resources, to make their own films based upon their novels. Second, movies and books are separate markets, just as laptops and mobile smartphones are. And third, the release of a film based upon a popular novel increases demand for the novel, by increasing the network of those interested in it and adding to the novel’s fanbase.²⁴⁹

Despite these similarities, the Google Court refused to extend the broad derivative work right from the novel-to-film context to computer programs. Of course, the Court did not address the real differences between licensing Android and licensing a film. The first difference is that the practice of licensing a film version of a novel has been so long established and is so predictable that the expected licensing fees have become a significant and systematic part of the ex ante incentives for authors to write popular novels. In contrast, in Google v. Oracle, no such practice was, as yet, established. At the time the original copyright owner, Sun, wrote Java, Sun had no reason to expect that Google would copy 11,500 lines from Java to build Android. Sun had no reason, therefore, to incorporate the expectation of any licensing fees from such a use into its cost-benefit analysis of whether to author and publish the Java computer program at the time that decision was made. That is true as far as it goes. But the Court’s ruling not only resolved the dispute between Google and Oracle, it also established a legal rule going forward. In evaluating that legal rule, the more difficult questions to answer are: first, whether, had the Court found infringement in Google v. Oracle, such licensing would become the norm, just as licensing novels to be made into films is the norm today; and second, if licensing became the norm for computer programs, would that outcome better promote “the Progress of Science” than a finding of fair use.

Fortunately, the Google Court did not need to answer those more difficult questions to determine whether requiring a license or finding fair use was more likely to promote “the Progress of Science” because of the second difference between licensing Android and licensing a film. As previously discussed, requiring a license to prepare a motion picture version of a novel tends to ensure exclusivity for the film in the marketplace. Requiring a license to make such a film thus generates market power. That increases the producer surplus associated with the film. Thus, in the film context, requiring a license rather than finding fair use increases the incentives for both film and underlying novel. That is not true for Android. The revenue for and surplus associated with Android are not the result of a license from Oracle. They are the result of Google’s market power and associated network effects. As in the Goldsmith case, requiring a license would simply require Google to share some of its surplus with Oracle. Requiring a license would thus convert some of the surplus associated with Android into transaction costs. Requiring a license would thus reduce the incentives for authorship accordingly.

In sum, the Goldsmith Court’s argument that holding Warhol’s use to be transformative would frustrate Congress’s intent to require a license to make a film from a novel is unpersuasive for two reasons. First, Congress intentionally made the exclusive right to make a film based upon a novel as subject to fair use as it made every other exclusive right. As a result, in an appropriate case, courts will need to decide whether the technology, markets, and economics of film production and distribution have changed sufficiently that fair use should cut back on the derivative work right accordingly. The film-from-a-novel example is not an answer to where the line between infringing and fair uses lies. It is a question waiting for a suitable case for courts to answer. Second, there is no need to revisit the scope of the film-from-a-novel rule in the Goldsmith case. The associated markets for the later use are simply different. For a film based upon a copyrighted novel, requiring a license guarantees exclusivity, increases the producer surplus associated with the film, and thus increases incentives for both film and novel. That is not true for Orange Prince, and it is not true for Android. Ignoring such material differences only ensures that the Court will misbalance the interests of earlier and later authors and misplace the line between fair and unfair uses.

Just as realism and textualism cannot support the Goldsmith Court’s reasoning or outcome, neither does purposivism. Congress intended the derivative work right to be “subject to” fair use, just like every other exclusive right.²⁵⁰ Whether making a film from a copyrighted novel is an infringing use and requires a license is not the answer to the proper placement of the line between fair and infringing uses. Even if, at one time for a given technology, such a use represented the paradigmatic case of copyright infringement, time and technologies change. The question is whether those changes suggest that a different placement of the fair use line would better promote “the Progress of Science” today. That leaves the Goldsmith Court’s final justification: Campbell made us do it.²⁵¹ To that justification, we now turn.

D. Campbell: Mistaken and Misapplied

In insisting that Campbell required its outcome, the Goldsmith Court made a mistake common to the writings of first-year law students. It mistook dicta for holding. In particular, the Goldsmith Court relied on Campbell’s distinction between parody and satire to insist that a new meaning, message, or aesthetic alone is insufficient to make a work transformative for purposes of the first fair use factor.²⁵² But satire was not before the Court in Campbell, only parody. As a result, that distinction is unnecessary to the Campbell Court’s decision. The distinction is dicta.

Because satire was not before the Court, we should not expect the Court to have balanced the equities satire implicates as carefully as it balanced the equities for parody.²⁵³ That is one reason to distrust dicta. There are others.²⁵⁴ But the satire-parody distinction is not just dicta; it is “backhanded” dicta. Whatever the reasons to distrust dicta generally, backhanded dictum is even less trustworthy. Backhanded dictum effectively says, “Sorry, Acuff-Rose you lose here. But if only the use at issue had been satire, you would have won.” Such a statement offers the Court a chance: (i) to make its analysis seem more reasonable; and (ii) offer the losing side something. Moreover, it does so at no cost to the Court—it is in not binding. As a result, through such backhanded dicta, the Court can appear reasonable and offer each side something without any need to find infringement based upon the distinction the Court articulates.

Generally, one of the reasons we distrust dicta is that the adversary system will often be less effective at bringing the critical trade-offs to the Court’s attention. With backhanded dicta, the adversary system fails entirely. The attorney for 2 Live Crew, Bruce Rogow, was there to defend the interests of his clients, not satirists. If it helped his clients win, he was not only perfectly willing to throw satirists and every other non-parody use under the bus, ethically he was required to do so. And he did.²⁵⁵

In his brief to the Court and at the outset of his oral argument, Rogow adopted a broad view of parody.²⁵⁶ Parody, he argued initially:

can poke fun at the original, or it can poke fun at something else using the original work. There are two aspects of the criticism. One would be criticism of the original work, the other would be criticism of society using the original work as a means of conveying that criticism.²⁵⁷

Yet, as soon as Justice O’Connor challenged his initial position as overbroad and unnecessary to the resolution of the case, Rogow abandoned it:

Justice O’Connor, for this case it is true that the parody in this case only poked fun at the original. And one could limit this case to just those facts and that would be quite fine.²⁵⁸

If testing the pros and cons of a given rule in the fires of the adversary crucible is an essential part of the judicial process, no such testing occurred in Campbell with respect to whether parody and satire should receive differing treatment under fair use. As soon as he was challenged on the issue, Rogow made no attempt to explain why satire should also qualify as a transformative use. As he was ethically required to do, he abandoned any attempt to defend satire to increase the likelihood of a fair use holding for his clients.

Because 2 Live Crew’s use presented the question, the Campbell Court could and did decide that a later work that copies from an earlier work in a way that “can reasonably be perceived” as parody is transformative.²⁵⁹ But 2 Live Crew’s use was not satire.²⁶⁰ As their attorney quickly conceded at oral argument, 2 Live Crew’s song did not use what was copied to comment on or criticize some aspect of society more generally. As a result, the Campbell Court had no power under Article III to resolve the question of whether a satiric use was transformative and hence, should weigh in favor of or against fair use under the first fair use factor.²⁶¹

The Goldsmith Court could properly rely on Campbell for its holding that parody, because it comments on the earlier work from which it borrows, is transformative. The Goldsmith Court could also treat Campbell’s dicta regarding satire not being transformative as persuasive and follow it. But to treat that distinction as binding was and is a mistake.²⁶² As I have explained elsewhere, courts are structurally ill-suited to law-making through the common law process.²⁶³ Adversary testing of the precise question at issue is one of the judiciary’s very few mechanisms that helps ensure that the common law process works at all.²⁶⁴ The Goldsmith Court should not have been so quick to abandon that mechanism by pretending that the Campbell Court had resolved an issue not before it.

Yet, even if we overlook the holding-dicta distinction, the Goldsmith Court failed to follow the very rule from Campbell it purports to retain: in evaluating the purpose of the later work, the test is whether criticism or commentary “can reasonably be perceived.”²⁶⁵ As the Goldsmith Court expressly held, Warhol’s subjective intent does not determine whether Orange Prince had a critical purpose.²⁶⁶ Instead, whether a critical purpose is present must be determined by an objective examination of the two works.²⁶⁷

Applying that test to Orange Prince, its criticism of and commentary on Goldsmith’s photo is immediately apparent. Orange Prince re-frames and re-colors Goldsmith’s photo.²⁶⁸ In doing so, Orange Prince proclaims loudly and in unmistakable terms that Goldsmith’s photo is “Boring!”²⁶⁹ By re-framing and re-coloring the Goldsmith photo,²⁷⁰ Orange Prince exclaims: “A black-and-white photo exhibiting the garden variety creativity of a high school yearbook photo is no way to portray a pop icon. Here’s how you portray a pop icon.” The criticism of and commentary on the Goldsmith photo can readily be perceived by the differing creative choices Warhol made, just as my students can readily perceive the criticism and commentary I am making when I provide a model answer. All you have to do is look at what is different to see what you did wrong. Of course, both Orange Prince and Goldsmith’s photos still depict Prince. But Goldsmith has no copyright over Prince himself. Prince’s features are his own. They are not original to Goldsmith. By making different creative choices on the issues each of them respectively controls, while still portraying Prince, Orange Prince can reasonably be perceived to criticize, even ridicule, Goldsmith’s photo.

If the Goldsmith Court could not see the changes Warhol made to Goldsmith’s creative choices as both critical of and commenting on Goldsmith’s “white-bread original,”²⁷¹ it was not really looking.

No more than realism, textualism, or purposivism could, deference to its own precedent cannot support the Goldsmith Court’s reasoning or outcome.

III. Fair Use: The Path Forward

That leaves us with a question, however: Where does fair use go from here? Whether the Goldsmith decision is right or wrong in its analysis—indeed, even if the Court just made up its approach—the Goldsmith decision remains the law. But it is the law only in a particular sense. Judicial power is always limited. Sometimes, courts acknowledge the limited nature of their power. For example, in Google v. Oracle, Justice Breyer expressly cautioned that the Court was deciding only the case before it.²⁷² But such an express statement is unnecessary. That limitation is inherent in judicial power. Courts always resolve only the parties’ dispute before them. Nothing more. Courts have neither the competence nor the authority to act as a legislature.²⁷³ Thus, not everything the Goldsmith Court said is binding law on other parties or in other cases going forward.

Moreover, even I, as troubled as I am by much of what the Goldsmith Court did, have to concede that the Goldsmith majority was unmistakably right on one issue: The path forward lies in the statutory language and the Court’s fair use decisions. Both statutory language and the Court’s previous fair use cases provide a single guide for fair use going forward: balance. Not the stilted, constrained balance suggested by the Goldsmith Court—that is, balancing any transformative purpose or character exclusively against commerciality under the first factor. Instead, a broad balancing of the competing interests at stake in the placement of the infringement-fair use line, including balancing the interests of earlier authors against later authors, of authors against distributors, of authors and distributors against consumers. In short, the statutory language, the legislative history, and the Court’s own precedent all require a broad inquiry into whether finding fair use or infringement in a given case will better promote the “Progress of Science” in the circumstances presented under all the statutory factors and potentially additional considerations as well.

The statutory language requires courts to balance what the public has to gain and what it has to lose in terms of “the Progress of Science” from a court’s placement of the line between fair and infringing uses. The statutory language does not define an exclusive set of factors to be mechanically understood and applied. As previously discussed, the statutory language defining fair use is repeatedly open-ended. First, in the preamble, it sets forth the types of uses and different purposes that might qualify as fair use.²⁷⁴ In setting forth these lists, Congress used the word “including” and the phrase “such as” to expressly make both inclusive and open-ended, allowing for courts to add to each list when they see fit.²⁷⁵ Second, Congress stated that a fair use analysis “shall consider” the four fair use factors listed.²⁷⁶ While the word “shall” requires a court to consider the four statutory factors, the phrase “shall consider” allows the court to consider additional factors in its fair use analysis as well. Third, and finally, in defining the first fair use factor, the statutory language provides two examples of the sort of purpose and character that might weigh for or against fair use. But again, Congress preceded the two examples with open-ended language—“including.”²⁷⁷ That open-ended language again leaves the Court free to add to the statutory examples even within a single fair use factor.

Similarly, and as previously discussed, while the statute states four of the factors a court “shall consider,” the statute does not define those factors or explain how they are to be weighed against one another. For example, the first factor directs a court to consider the “purpose and character of the use.” Yet, the statute does not provide a comprehensive definition of either purpose or character.²⁷⁸ In setting forth the first factor, the statute does offer two exemplars of a use’s “purpose and character”: “whether such use is of commercial nature or is for nonprofit educational purposes.”²⁷⁹ The lack of a definition of these words, along with the open-ended nature of the examples given, leaves courts considerable leeway to define “purpose and character” as appropriate to balance what the public has to gain and what it has to lose under the specific circumstances presented. Moreover, the statute provides no guidance on how heavily any given “purpose and character” shall weigh in the overall analysis.

The statute defines the second factor as “the nature of the copyrighted work,” and the third as “the amount and substantiality of the portion used in relation to the copyrighted work as a whole.”²⁸⁰ But again, the statute gives no indication of how heavily any given nature or amount shall weigh for or against fair use. The statute expressly leaves those issues for a court to resolve on a case-by-case basis under the circumstances presented.

The only portions of the statute that are clear and unambiguous with respect to fair use are those portions defining the relationship between fair use and the exclusive rights. As discussed, in section 106, the statute expressly states that the exclusive rights are “subject to” the fair use doctrine.²⁸¹ In section 107, the statute expressly states that “the fair use of a copyrighted work … is not an infringement of copyright,” the exclusive rights set forth in sections 106 and 106A “notwithstanding.”²⁸² This language expressly subordinates the exclusive rights to the fair use privilege. The language also provides that fair use expressly carves out uses that would otherwise fall within the literal scope of the exclusive rights and defines them as non-infringing. But as to whether fair use should be the exception or the rule, or something in between, the statute does not expressly say. Precisely how to balance the exclusive rights and fair use so as to appropriately balance the competing interests at stake, the statute leaves to the courts.

If we look to the legislative history accompanying the enactment of the Copyright Act of 1976 and the statutory codification of fair use, we find the same story. Congress did not intend for the statutory codification to “freeze” the common law development of fair use.²⁸³ Nor did it intend to reduce the complexities of balancing the interests at stake to a mechanical checklist.²⁸⁴ Instead, Congress provided “a very broad statutory explanation of what fair use is and some of the criteria applicable to it.”²⁸⁵ Otherwise, Congress intended to leave “courts … free to adapt the doctrine to particular situations on a case-by-case basis.”²⁸⁶ As for whether fair use should be the rule or the exception, or something in between, the legislative history suggests something in between. The House Report accompanying the codification of fair use describes it as “an equitable rule of reason.”²⁸⁷ Describing fair use as a “rule of reason” suggests courts need to balance whether a finding of fair use or infringement for the use at issue will better promote “the Progress of Science,” just as they balance what the public has to gain and what the public has to lose in terms of competition in antitrust cases where the “rule of reason” phrase originated.²⁸⁸

When we look to the Court’s previous decisions on fair use, balance is precisely what the Court has done. It has treated the statutory codification not as freezing the doctrine in place, nor as setting forth a mechanically-applied checklist, but as a living common law doctrine by which the Court may balance whether, in terms of more and better original authorship and its wider dissemination, the public has more to gain by allowing or prohibiting the use at issue. In its first case applying the newly codified fair use privilege, the Court in Sony found that time-shifting constituted fair use.²⁸⁹ Time-shifting is not listed in the preamble of section 107, and indeed, is quite different from the examples the preamble lists.²⁹⁰ Nevertheless, the Court found the use fair.²⁹¹ On balance, the public gained more from the increased access to existing original works that the Betamax and time-shifting enabled, than it might lose from the merely speculative losses in the “potential market for or value of” the copyrighted works at issue.²⁹² In the second case, Harper & Row Publishers v. Nation Enterprises, the Court found the Nation’s scooping of excerpts from Gerald Ford’s then unpublished biography to be an unfair use.²⁹³ The defendant’s use at issue was one of the examples listed in the preamble of section 107—news reporting.²⁹⁴ But the Court, on balance, found that despite the news-reporting purpose, the public had more to lose from the harm to the actual market for the biography than it had to gain from the Nation’s use.²⁹⁵ In striking that balance, the Court was both realistic and flexible. In terms of realism, the infringement finding in the case did not deprive the public of the Nation’s thoughts on the Nixon pardon. It simply required the Nation to wait a few more weeks for the biography and the officially licensed excerpt in Time magazine to be published before it published its own essay.²⁹⁶ In terms of flexibility, the Court added a consideration to the second fair use factor, whether the plaintiff’s work was unpublished or published,²⁹⁷ and rewrote the third fair use factor to focus on the “amount and substantiality of the portion used” in relation to the defendant’s work,²⁹⁸ rather than in relation to the plaintiff’s work, as the statute provides.²⁹⁹

Then, in Campbell, as previously discussed, the Court added transformative use to the first fair use factor.³⁰⁰ Holding 2 Live Crew’s song to be transformative meant the public had something to gain, that is, a new song, from allowing the use. But that was not the end of the issue. The Campbell Court did not hold that the use was fair because it was transformative. Instead, it remanded to allow the defendants to present evidence on whether 2 Live Crew’s version interfered with the potential market for a licensed derivative.³⁰¹ If it did, presumably the trial court would need to balance the transformative character of the song against that harm to the potential market for Orbison’s original.

Although some may want to confine its relevance to computer programs, as if Congress drafted a different derivative work right or fair use privilege for computer programs specifically, the Court’s decision in Google v. Oracle reflects the same balancing. Google’s copying gave the public a new, original work of authorship: Android.³⁰² Finding that copying to be fair use, rather than an infringing derivative work, ensured the public broad and ready access to Android. And it did so without undue harm to “the potential market for or value of” Java.³⁰³

The statutory text, the legislative history, and the Court’s own decisions all point to a single unifying guide to resolving fair use cases: balance. All three suggest the need for fair use to balance, flexibly and realistically, what the public has to gain and what it has to lose in terms of the “Progress of Science” from an infringement or fair use holding in the circumstances presented. Only through such a balance can a court determine whether a fair use or infringement outcome will better promote “the Progress of Science” going forward.

Yet, if balance is the key, that still leaves the question of how to calibrate that balance. Congress has the authority to enact copyright for a single purpose: “to promote the Progress of Science.” The Court has defined that purpose to include two legitimate ends: (i) increasing creative output; and (ii) increasing its dissemination.³⁰⁴ As a result, if a court could know, before it made its decision, whether fair use or infringement would lead, going forward, to: (i) more or better original works; and/or (ii) their broader dissemination, then striking the appropriate balance between fair use and infringement would simply require the court to rule accordingly. In such a case, that the use at issue fell squarely within one of the exclusive rights, as the copying in Sony did, is not a reason to reject fair use. It is the reason fair use applies.

Yet, in practice, the issue is not quite so simple. Courts, like the rest of us, lack perfect information. That may explain why Justice Story, rather than having set forth the relevant balance directly, instead set forth a number of proxies. In the absence of perfect information regarding whether fair use or infringement will better promote the Progress of Science going forward, looking to factors such as “the nature and objects of the selections made, the quantity and value of the materials used, and the degree in which the use may prejudice the sale, or diminish the profits, or supersede the objects, of the original work,”³⁰⁵ may serve as useful proxies that cast light on whether “the Progress of Science” is better served by allowing or prohibiting the use at issue.

For me, the Goldsmith Court is best read as embracing an additional proxy. As the Court noted, Condé Nast paid a licensing fee for the initial use.³⁰⁶ Later, in responding to the dissent’s concerns, the Court insisted that “[i]t will not impoverish our world to require [the Andy Warhol Foundation] to pay Goldsmith a fraction of the proceeds from its reuse of her copyrighted work.”³⁰⁷ Taken together, these statements suggest an inverse principle to Wendy Gordon’s market failure approach to fair use.³⁰⁸ Specifically, these statements suggest that, in the absence of more perfect information and at least for commercial uses, where a licensing market already exists and licenses are routinely obtained for the use at issue, a court should tend to find infringement, rather than fair use. Indeed, the facts in Goldsmith go beyond a market situation where licenses are routinely obtained. In Goldsmith, it is not just that such uses could be, and in theory would be, easily licensed. The previous use of the Warhol Prince series for an identical commercial use was licensed.³⁰⁹

In short, we should read the Court’s decision as a court decision. In recent decades, the Court has decided fewer and fewer cases.³¹⁰ In response, academics, lawyers, and judges have sought more meaning in the fewer decisions the Court has given us. Every line is carefully parsed. Every sentence becomes its own holding—true and valid not just on the particular facts before the Court, but generally for all facts and for all time. This is unfortunate. For reasons that I have explained elsewhere, courts work best when their dispute resolution function is primary and their lawmaking function secondary or incidental. The more closely we tie a court’s decisions to the precise facts before the court, the more the decision makes sense. The less closely, the less sense. The Court’s authority in Goldsmith is to resolve the dispute before the Court, not to make broad proclamations about the meaning of fair use going forward. When we read the Court’s definition of transformative character and see the balance the Court articulated, we must recognize that the Court’s analysis of those issues is inevitably colored by the facts before it. On those facts, the first and nearly identical commercial use of the Warhol’s Prince series was licensed; yet no attempt to license the second such use was made. This “failure to license a previously licensed use” matters—not just for the overall resolution of the case, but for understanding each and every aspect of the Court’s decision, including the Court’s interpretation of transformative character. The use of Orange Prince on a magazine cover commemorating the life of Prince was not transformative for the reasons the Court articulated in defining the nature of transformative character, but also because the identical use has previously been licensed and no attempt was made to license the second such use. If we try to rely on the Court’s language and reasoning outside of the specific factual context in which Goldsmith arose, we risk missing the point of the decision.

We see support for the use of this “failure to license a previously licensed use” principle as a fair use proxy in the Court’s resolution of the fair use claim in its 1990 Stewart v. Abend decision.³¹¹ In that case, Jimmy Stewart and Alfred Hitchcock had set up a film production company, Patron, Inc., and obtained a license to make a film based upon a copyrighted short story, “It Had to Be Murder.”³¹² Under the license, Patron made and publicly exhibited the film Rear Window.³¹³ Because of the special rules for vesting of the renewal term under the 1909 Copyright Act,³¹⁴ that license expired at the end of the primary copyright term for the short story, sometime in the late 1960s.³¹⁵ The film was nonetheless broadcast on the ABC television network in 1971.³¹⁶ In the ensuing lawsuit, the principal issue was whether the creator of a derivative work that was properly licensed at the time it was created, as with Rear Window, could continue exploiting the derivative work during the renewal term after the license had expired.³¹⁷ The Court held that they could not.³¹⁸ As an alternative, the defendants argued that they could broadcast their film under the fair use privilege.³¹⁹ In rejecting that argument, the Stewart Court relied primarily on Sony’s presumption that “every [unauthorized] commercial use of copyrighted material is presumptively an unfair exploitation of the monopoly privilege that belongs to the owner of the copyright.”³²⁰

But even without Sony’s presumption, requiring a license to make or publicly perform a motion picture version of a copyrighted short story makes more sense, at least in the absence of perfect information, than a fair use determination. The licensing of such uses is not merely possible; in theory, it already occurs routinely. Indeed, as in Goldsmith, such a use is not just easy to license in theory, the actual defendants in the case had licensed such use when they previously made it.³²¹ It may well be that with perfect information, a court could discern that requiring licenses in this situation does not, on balance, better promote “the Progress of Science” than does a fair use result. In the real world, however, courts do not have perfect information. With the limited information available as to whether such a license will or will not better promote “the Progress of Science,” leaving a working market in place is probably the better bet.

In the introduction, I suggested that reasonable minds can differ as to whether the Foundation’s use should be fair or infringing. For me, this is precisely the argument by which I can see holding Warhol’s use unfair. Condé Nast obtained a license for the first use of Warhol’s Purple Prince. There seemed to be a practical and working licensing market in place. It may be that with perfect information, a court could see that fair use would better promote “the Progress of Science” than the Court’s infringement outcome. But with the information available, it is plausibly better to reinforce the licensing market in place. Thus, going forward, I would read the Goldsmith decision as holding that where the use at issue is commercial and there is an established licensing market, and particularly where the defendant has licensed the identical use previously, then a court should tend to find infringement rather than fair use.

The only consideration that gives me pause in both Stewart and Goldsmith are the heightened risks of licensing failure when the parties move from an ex ante to an ex post licensing negotiation. In these markets, ex ante bargains, where neither side yet knows the value of the licensed work or use, are routinely stuck. Uncertainty over whether the short story will actually be made into a film, for example, or whether the film will be a hit or a dud, keep royalty and other demands reasonable. When the size of the licensed pie, and indeed, the question of whether there will even be any licensed pie, remain unknown, both sides have an incentive to cooperate to ensure that there is a licensed pie and to maximize its size. In addition, when ex ante, neither side is locked into bargaining with the other. If a given short story author makes royalty or other demands that are out of line with market norms, for example, the film production company can turn to other authors. The same is true on the other side. If the film production company offers an unreasonably low royalty or other terms inconsistent with market norms, the short story author can shop around.

When we move from ex ante to ex post licensing, however, the framework for negotiations changes. Now, each side is locked into negotiating with the other. The value of the licensed work has also been revealed. A game of chicken or bilateral monopoly hold-up may ensue as the would-be licensor and the would-be licensee each try to maximize their respective share of the now known and baked pie. The negotiations switch from cooperative, trying to maximize the size of an unknown and not-yet-baked pie, to antagonistic, trying to maximize my share of an existing pie. Even if licensing is routine in a market ex ante, courts should not too readily assume that licensing will be equally routine in the same market ex post. Moreover, when the two parties are forced to negotiate with each other in the ex post market, they do so in the shadow of the law–with the likely legal outcome if the parties were to litigate setting boundaries on the licensing negotiations. In the ex post context, where the alternative is litigation, each party will agree to a license only if they expect to do, at least, as well from the license as they expect they could do litigating. Unfortunately, the dual threats of statutory damages and injunctive relief may unfairly bias the ex post licensing negotiations in favor of the licensor. Holding the possibility of a fair use resolution over the parties can help re-balance the licensing negotiations in this context and encourage both sides to be reasonable.³²²

Reading Goldsmith as turning on the ease of licensing also helps us understand why the Goldsmith Court distinguished Campbell. In Campbell, 2 Live Crew had actively sought a license before the litigation, but the copyright owner refused. In refusing, the copyright owner specifically stated: “[W]e cannot permit the parody of ‘Oh, Pretty Woman.’”³²³ As the Campbell Court noted, “‘People ask … for criticism, but they only want praise.’”³²⁴

Campbell represents the other side of relying on licensing as a proxy for the balance fair use strives to establish—one that Wendy Gordon originally suggested: Where we expect the licensing market to fail, a court should tend to find fair use, rather than infringement.³²⁵ Licensing failure risks barring the defendant’s use directly. When licensing failure occurs, the public does not receive the benefit of the defendant’s use (because it does not occur). The defendant does not receive any revenue (because the defendant’s use does not occur). And the plaintiff does not receive any additional licensing revenue or attribution from the defendant’s use (again, because it does not occur). That’s a lose-lose-lose proposition. If there is good reason to believe that the licensing market will fail, allowing a use that directly promotes “the Progress of Science” – whether because the use at issue gives us a new original work or because the use at issue expands access to existing works–is the better choice.

Moreover, for some uses, requiring a license may distort rather than bar the use at issue. If the point of a digital database is to house a complete and comprehensive historical record of audio-visual works, requiring the permission of the copyright owner, as a panel of the Second Circuit did in TVEyes,³²⁶ holds an accurate historical record hostage to the whims of each copyright owner. A copyright owner may choose not to license those clips that portray them, with the benefit of hindsight, in an unflattering manner.³²⁷ Requiring a license may distort what is included in such a digital archive, just as requiring a license for a film or book review may distort what is supposed to be a neutral or objective review. In addition, the public may be better served by competing comprehensive archives, as we have had historically with libraries, rather than siloed archives, such as one for CNN, one for Fox, and one for MSNBC. Thus, even where licensing is practicable, courts should consider whether imposing a licensing requirement will distort what society receives, whether what society receives is an audiovisual archive, a movie or book review, or something else.

At the same time, courts must be cautious and not assume licensing failure is inevitable too readily. Even where the cost of individually licensing each use appears impractical, class action litigation and other collective resolution mechanisms may make seemingly impractical licensing practical. Given a choice between leaving money on the table and being creative, potential licensors and licensees in copyright markets can be creative. They can sometimes find ways to overcome even seemingly insurmountable difficulties to license uses.³²⁸

Nevertheless, at least as a first cut approximation, whether licensing a defendant’s use is likely to prove trivially easy or frustratingly hard can provide a helpful guide. In the absence of perfect information as to whether fair use or infringement will better promote “the Progress of Science,” the ease and certainty of obtaining a license can provide a useful proxy for courts trying to resolve a fair use case. In cases involving commercial use where a successful license negotiation is almost certain and will not distort what society receives, courts should tend to find infringement. This is particularly true where an identical previous use by the same party had been licensed. On the other hand, in cases involving commercial use, where a successful license negotiation is unlikely or will distort the use at issue, courts should tend to find fair use.

When a case falls somewhere in between these two ends of the licensing success-licensing failure spectrum, such as in Google v. Oracle, where both parties negotiated in good faith but were unable to agree to licensing terms,³²⁹ then a court should balance what the public has to gain and what it has to lose in terms of copyright’s constitutional objective from allowing or prohibiting the use at issue.

Conclusion

To achieve its constitutional purpose, copyright requires balance. Congress elevated fair use from common law exception to statutory privilege to help achieve that balance. In codifying fair use, Congress provided some general guidelines, but largely left it to courts to apply those guidelines and to develop additional guides, as necessary, to determine whether, in any given case, a finding of fair use or infringement would better advance copyright’s constitutional purpose.³³⁰ Until Goldsmith, the Court recognized that making this determination required not a mechanical checklist, but a realistic balancing of whether allowing or prohibiting the use at issue is likely to better promote “the Progress of Science” going forward.

To give fair use the role Congress intended it to play in promoting that purpose, courts should narrowly read Goldsmith to hold that where the use at issue is commercial and there is an established licensing market for the use at issue, a court should tend to find infringement rather than fair use. In the absence of perfect information as to whether a finding of fair use or infringement will better promote “the Progress of Science” going forward, reinforcing an existing and well-functioning licensing market is the better bet. Particularly where an identical commercial use had previously been licensed, a certainty of licensing success in this setting provides a workable but inverse proxy to Gordon’s market failure approach for resolving fair use cases.

Courts should be careful not to read more into the Goldsmith decision than Article III authorized the Court to decide. Even where the Court’s language and reasoning seems to apply, the further divorced a district or circuit court’s application of Goldsmith is from the specific facts of the Goldsmith case itself, the less reliable the Court’s analysis becomes.