It is hard to think of a patent doctrine—or indeed any doctrine in IP law as a whole—that has received more critical attention over the past decade than patentable subject matter. In a series of four cases from 2010 to 2014, the Supreme Court sparked an ongoing controversy by imposing sharp limits on patenting inventions such as medical diagnostics, human genes, software, and business methods. In the wake of these decisions, an invention may be groundbreaking and important, but nonetheless unpatentable if it falls into the Court’s implicit exceptions to patent eligibility for “laws of nature,” “natural phenomena,” and “abstract ideas.” Under the “Alice/Mayo” test, if a patent claim is “directed to” one of these ineligible concepts, it is patentable only if it also contains an “inventive concept” such that the claim “amounts to significantly more” than a patent on the ineligible concept.
These subject matter limits led to a barrage of criticism, much of which focused on the test’s unpredictability. The criticisms have come from judges, academics, practitioners, the DOJ, and even memes. Bipartisan legislation has been repeatedly introduced to address the “confused, constricted, and unclear” law. Senators have also tasked the USPTO with collecting public comments on the issue.
Yet, according to a new draft article by Nikola Datzov and Jason Rantanen, the narrative of unpredictability is on “shaky empirical ground.” Surprisingly, they argue that over the past decade, disputes about patentable subject matter have been more predictable than disputes about other patent doctrines. Datzov and Rantanen analyzed all patent eligibility decisions by the Federal Circuit between 2012 through 2022. This court hears appeals both from all district court patent litigation and from decisions by examiners at the USPTO. They found that the Federal Circuit not only affirmed lower tribunals almost 90% of the time, but also almost always affirmed on the same basis. These cases also involved fewer dissents than on other patent law issues. The authors are not arguing that the Alice/Mayo test is easy to apply, or that it is necessarily good policy. But their results should shift the focus of the ongoing patentable subject matter debates.
Datzov and Rantanen are not the first scholars to argue that handwringing about patent eligibility’s unpredictability is unwarranted. For example, Jason Reinecke surveyed over 200 patent attorneys in 2017 about actual litigated claims and concluded that the Alice/Mayo test is “not as amorphous as many commentators have suggested.” Ben Dugan was able to train an AI classifier in 2018 based on USPTO eligibility decisions. Mark Lemley and Samantha Zyontz’s empirical study of judicial decisions on patent eligibility through 2019 also found a high rate of affirmance at the Federal Circuit.
Nevertheless, the conventional wisdom remains that the doctrine is a quagmire. Datzov and Rantanen tackle this narrative with a thoughtful analysis of why evidence cited in favor of the unpredictability thesis does not indicate that the sky is falling. They also offer a more detailed and updated empirical study of their preferred predictability indicators that includes both methodological innovation and an admirable commitment to transparency and data availability.
Datzov and Rantanen used the Federal Circuit Dataset Project, an open-source dataset of all Federal Circuit decisions. This dataset includes a large number of summary affirmances under Rule 36, which the Federal Circuit frequently and asymmetrically used in patent-eligibility cases. They found 368 eligibility decisions between 2012 and 2022. About one-quarter of these appeals came from USPTO and three-quarters from district courts.
The Federal Circuit affirmed the USPTO’s decisions over 95% of the time. This is perhaps unsurprising given that only ineligibility decisions are appealed from the USPTO. Moreover, the agency’s guidance to examiners is more eligibility-friendly than the Federal Circuit’s caselaw. But the Federal Circuit affirmed district court decisions 85% of the time. In almost all cases, it did so in procedural postures in which the district court receives no deference. This is higher than its affirmance rates for claim construction and obviousness. Despite some high-profile fractured decisions, there are nonetheless fewer dissents in patent eligibility decisions than in decisions on all other patent law issues, whether one focuses on all appeals (6.5% vs. 8.1%) or just appeals from district courts (8.6% vs. 10.7%).
Reversal rates and dissent rates are established metrics for evaluating the predictability and uniformity of a particular doctrine. Datzov’s and Rantanen’s data suggest that patentable subject matter isn’t out of step with other patent doctrines. They also look to a third metric that is an important innovation for empirical work: whether the Federal Circuit identified a legal error in the lower tribunal’s analysis even if it affirmed the result. Error rates provide an additional perspective on whether the law can be predictably applied because they can show whether judges get the right results for the wrong reasons. Three coders reviewed each of the 153 affirming opinions (excluding Rule 36 affirmances). They found that the Federal Circuit identified errors in only seven opinions—less than 5% of the cases.
Datzov and Rantanen argue that the very high rate of error-free decisions “significantly undermines the assertion that the law cannot be predictably applied by judges.” Because of the novelty of this empirical analysis of error rates, this result cannot be compared with other patent doctrines, but it should serve as a benchmark for future studies.
What do these results mean for patent-eligibility reform efforts? Datzov and Rantanen note at the outset that their goal is not to defend the Alice/Mayo test. They explicitly cabin normative issues such as whether limiting patents in certain areas has increased or decreased innovation. This is a question that existing evidence cannot answer. They also acknowledge that the doctrinal test could be clarified further. Some of their results may stem from patent eligibility being used largely as a shortcut to quickly dispose of clearly invalid patents early in litigation.
If judges do not feel confident resolving eligibility motions at the pleadings stage, patent challengers may shift their efforts to doctrines like obviousness that are easier to explain to a jury. These considerations do not undermine Datzov and Rantanen’s conclusions. Rather, this would suggest that in practice, eligibility is being used in the easier cases where the results are more predictable, perhaps reducing litigation inefficiencies.
Of course, the surprising predictability of patent eligibility may offer little solace to law students who struggle with applying the Alice/Mayo test to borderline claims in casebook practice problems. But every patent doctrine has thorny borderline cases. Just because law professors love to live in the grey areas doesn’t mean we shouldn’t remind students that in practice, the answer isn’t always “maybe.”
