If a company brings a product to market and tries to sell it based on mendacious data, that company will most likely get into trouble. Consider the case of Elizabeth Holmes, the founder of health tech company Theranos, who was charged by the SEC with fraud after making false claims about the utility of her company’s blood test devices. But what would have happened to Holmes had she included similar fanciful uses in a patent application for her company’s device? Would she have been fined or sent to jail? Hardly. In fact, the fictitious claims would have probably made her application stronger. While government law enforcement agencies and regulatory bodies don’t take kindly to false or misleading information, the U.S. Patent and Trademark Office allows pie-in-the-sky claims when it comes to the potential uses of a particular patent. Professor Janet Freilich studies these and other important curiosities of patent law. A magna cum laude graduate of Harvard Law School, Freilich served as a member of the Harvard Journal of Law and Technology and was a student fellow with the John M. Olin Center for Law and Economics. For her writing on patent law, she has won the Samsung-Stanford Patent Prize, in 2010, and two years later, the Irving Oberman Memorial Award in Intellectual Property. Prior to entering academia, she practiced at Covington & Burling LLP as a patent litigator and prosecutor. In this excerpt from “Prophetic Patents,” Freilich examines the fictional uses that pop up in patent applications and the effect these have on the protection of inventors and the promotion of innovation.
In May of 2005, a team of scientists made headlines after the prestigious journal Science published a report that they had cloned human embryos. Only a few months later, the team was making headlines for a different reason: the data in the paper had been faked; Science retracted the paper, and the team’s leader, Dr. Hwang Woo-Suk, was fired and spent two years in prison for violating bioethics rules. Almost ten years after the retraction, Dr. Hwang received a U.S. patent on his discredited technique. Other scientists were “shocked” by the news that Dr. Hwang obtained a patent for falsified data. The New York Times quoted Dr. Jeanne Loring, a stem cell scientist at Scripps Research Institute, saying that her first reaction was “You can’t patent something that doesn’t exist.”
Dr. Loring’s reaction is common, sensible, and intuitive—but wrong. The Patent and Trademark Office (PTO) and the courts explicitly permit made-up experiments and fictional data in patents. Far from fraudulent, fictional data is instead treated as equivalent to factual data. To illustrate, the fictional experiment below was published in a recently granted patent:
A 67-year-old male has pancreatic cancer. … He is provided with A. paucinervis pomel extract [the patented invention]for three years. The patient is examined later and … [h]is tumor is reduced in mass …
The supposed ability of the patented compound to cure cancer borders on miraculous—yet it is also highly improbable, as real experiments have found the compound to be extremely toxic.
There is little scholarship on these fictional experiments—commonly called “prophetic examples.” The articles that have mentioned prophetic examples do so only in passing, with no more than a few sentences dedicated to the issue. These articles are almost uniformly critical of prophetic examples—hinting at potential problems surrounding the practice. Despite the lack of scholarly attention, prophetic examples are common. It is possible that the PTO has been granting hundreds of thousands of patents based on fake, implausible, and unreplicable experiments—and we know nothing about it.
In this article, I set out to understand the history, prevalence, and impact of prophetic examples. I collected a unique data set consisting of all prophetic and non-prophetic examples from U.S. patents and applications published between 1976 and 2017. To identify prophetic examples, I exploited a PTO rule that requires prophetic examples to be written in the present or future tense, while non-prophetic examples are written in the past tense. I focused on chemistry and biology patents, as those are the only industries that commonly include experimental data (real or otherwise) in patents. I analyzed 2,214,551 patents and applications in those industries, a population that contains 12,300,156 examples.
I confirm that prophetic examples are indeed common; in chemistry and biology patents issued between 1976 and 2017, at least 17% of examples are prophetic, and, of patents with examples, at least 24% contain some prophetic experiments. This means that prophetic examples are widespread enough to seriously impact patent law—and that we need to know more about them.
At first glance, the practice of allowing prophetic examples in patents seems baffling—why would the PTO allow fictional data? The PTO has never explicitly stated its reasons, but it is possible to construct a strong theoretical case for prophetic examples and then test it empirically, which I do here.
The theoretical case for prophetic examples rests on benefit to patentees. The Patent Act requires inventors to describe how to make and use their invention. Inventors often do this by writing experimental protocols and results in the patent. For example, a patent on a diabetes medication might include an experiment showing how to synthesize the molecule and another showing that the molecule can be given to humans to reduce the need for insulin injections. The broader the patent, the more experiments are required. A patent covering one molecule might only need to include one synthesis protocol, whereas a patent covering a family of one hundred molecules might need to include many more experiments. It is always faster and cheaper to make up data than to conduct real experiments, so if the experiments disclosed in the patent can be fictional, inventors will be able to file broader patents more easily. This should be particularly useful for small companies, which have small budgets and cannot afford extensive real experimentation. For companies of all sizes, broader patents provide a greater reward to the inventor, which might incentivize more innovation.
I test this hypothesis empirically and find there is surprisingly little evidence that prophetic examples actually help patentees. Patents with more prophetic examples are narrower than patents with fewer prophetic examples—despite the prediction that prophetic examples help patentees get broader patents. Second, there is no evidence that patents with more prophetic examples are filed earlier than those with fewer prophetic examples—again, contrary to prediction. Finally, although small companies should benefit disproportionately from the ability to use prophetic examples, they do not. I find that small companies are significantly less likely to use prophetic examples as compared to their larger counterparts. In sum, the case for prophetic examples does not fit with the empirical evidence.
Evidence for the benefits of prophetic examples is weak, but evidence for their harms is much stronger. Patents with prophetic examples are frequently abandoned, which suggests that the inventor is not commercializing their invention. The problem is that because of the patent, neither is anybody else. While in force, the patent prevents others from working in that area. Even after the patent has been abandoned and no longer has legal force, a chilling effect remains. Because patents are granted only if an invention has not been previously disclosed, it is difficult for any subsequent inventor to get a patent in an area previously described by a prophetic example. This is true even if the prophetic example is incorrect and the subsequent inventor was the first to actually make a functioning prototype. Essentially, instead of incentivizing innovation, prophetic examples may create an innovation dead zone.
Prophetic examples also lead to a second type of harm: they mislead scientists. In their patent, inventors must disclose a detailed description of their invention that can be used by other scientists to build further upon the technology. This disclosure function of patents has long been recognized as a crucial element of innovation—allowing downstream innovators to see further by metaphorically standing on the shoulders of giants. However, the disclosure function breaks down if scientists are misled by the disclosed information.
I analyzed how prophetic examples were cited in scientific publications and found that 99 percent of citations to prophetic examples incorrectly cited the example as if it represented work that had actually been done. This would not necessarily be a problem if the prophetic examples were good predictions, but, as I demonstrate in this article, many prophetic examples are probably wrong. False information is infiltrating the scientific community by way of prophetic examples.
My empirical findings have implications for several core debates in patent theory, including the disclosure function of patents, theories about constructive reduction to practice, and the optimal timing of patent filing. For disclosure, the misinformation spread by prophetic examples adds strength to widespread accusations that disclosure functions poorly and furthers a line of the literature emphasizing the gap between scientific writing and “patentese.” For constructive reduction to practice—a doctrine that allows inventors to obtain a patent without having physically created the invention—scholars argue that it disincentivizes physical reduction to practice. However, my evidence suggests that, surprisingly, there may be more advantages to physical invention than previously realized. For the optimal timing of patent filing, I show that while proponents of early filing might be expected to favor mechanisms that contribute to earlier filing, some such mechanisms—such as prophetic examples—do not fit with the traditional justifications for early filing.
All of this strongly argues for reform. Banning fictional experiments in patents is an attractive solution, given the findings herein, but likely too drastic an institutional change (for now). Instead, I argue that we should not think about fictional experiments as prophecies—a name that carries a ring of accuracy and infallibility—but rather as hypotheses—testable predictions that may or may not turn out to be correct. The shift from prophecies to hypotheses has several practical implications. First, it would reverse the current legal presumption that prophetic examples are enabled (i.e., that they work as written), since, as I show empirically, there is simply no reason to assume accuracy. Second, we should give inventors a grace period after filing to test their hypotheses and update prophetic examples with real results. Finally, we should require prophetic examples to be clearly labeled and to include some explanation about why the inventor expects the experiment to work. These changes all reflect that the predicted results are possibilities, not inevitabilities, and the shift can preserve what value such speculation has, while mitigating the clear harms that now prevail.
1 Gina Kolata, Koreans Report Ease in Cloning for Stem Cells, N.Y. TIMES (May 20, 2005).
2 Andrew Pollack, Disgraced Scientist Granted U.S. Patent for Work Found to Be Fraudulent, N.Y. TIMES (Feb. 14, 2014).
3 Id. The patent in question is U.S. Patent No. 8,647,872 (issued Feb. 11, 2014).
4 Pollack, supra note 2 (“‘Shocked, that’s all I can say,’ said Shoukhrat Mitalipov, a professor at Oregon Health and Science University who appears to have actually accomplished what Dr. Hwang claims to have done. ‘I thought somebody was kidding…’”).
6 U.S. PATENT & TRADEMARK OFFICE, MANUAL OF PATENT EXAMINING PROCEDURE [hereinafter “MPEP”]§ 608 (9th ed. 2015); Atlas Powder Co. v. EI Du Pont De Nemours, 750 F.2d 1569, 1577 (Fed. Cir. 1984).
7 DONALD CHISUM, CHISUM ON PATENTS § 2212 (2015).
8 U.S. Patent No. 8,003,137, Example 15 (issued Aug. 23, 2011).
9 Frederic D. Debelle, Jean-Louis Vanherweghem & Joelle L. Nortier, Aristolochic Acid Nephropathy: A Worldwide Problem, 74 KIDNEY INT’L 158, 158 (2008). These experiments were conducted before the patent was issued, so they did not infringe on the patent.
11 Smith Kline & French Laboratories v. Teva Pharmaceuticals USA, Inc., 2006 WL 6331923 (D.Del. 2012), Expert Report of Egon E. Berg (Sept. 22, 2006). (“Based on my experience as a patent attorney and patent examiner … prophetic examples are also common in patents.”).
12 Part III.A, infra.
13 MPEP § 608.
14 Part III.B, infra. Note that the problem of fictional experiments is certainly present in other industries. See HAROLD FULLMER, PATENT PROSECUTION 277 (2017). Further, the theory and policy discussed herein also apply across industries. The empirical study focuses on chemistry and biology because the methodology is best suited to those industries. However, the implications of this article are not so limited.
15 Table 1 and accompanying text, infra.
16 35 U.S.C. § 112.
17 In re Wands, 858 F.2d 731, 737 (Fed. Cir. 1988).
18 U.S. Patent No. 6,916,848 col. 13, ll. 23-45, col. 67, ll. 30-67 (issued July 12, 2005).
19 See, e.g., ALZA Corp. v. Andrx Pharmaceuticals, LLC, 603 F.3d 935, 942 (Fed. Cir. 2010).
20 See, HAROLD C. WEGNER, FIRST TO FILE PATENT DRAFTING § 8:5 (2016).
21 Part II.A.1, infra.
22 Irwin Aisenberg, The Patent and Present of Working Examples, 23 IDEA 25, 30 (1982).
23 E.g., Bonito Boats, Inc. v. Thunder Craft Boats, Inc., 489 US 141, 146 (1989).
24 Part III.D.1.b, infra.
25 Part III.D.1.c, infra.
26 Part III.D.1.d, infra.
27 Part III.D.1.a, infra.
28 35 U.S.C. § 271.
29 35 U.S.C. § 102-103.
30 To anticipate a subsequent patent, the prior prophetic example must be enabled. MPEP § 2121.01. However, this is not a requirement for obviousness. Id. Further, prophetic examples in granted patents are presumed to be enabled, so proving otherwise involves a legal battle and is sufficient to dissuade others from working in an area. Id.
31 35 U.S.C. § 112.
32 E.g., Graham v. John Deere, 383 US 1, 6 (1966).
33 Part III.D.2.b, infra.
34 Part IV.B.1, infra.
35 Part III.D.2.b, infra.
36 Part IV.B.2, infra.
37 Part IV.A.2.c, infra.
38 Part IV.A.2.b, infra.
39 Part IV.C, infra.