Measuring Secrecy

33 JLEGST 59

(Cite as: 33 J. Legal Stud. 59)
Journal of Legal Studies
January, 2004
 
*59 MEASURING SECRECY: A COST OF THE PATENT SYSTEM REVEALED
Jeremy M. Grushcow [FNa1]
 
Copyright © 2004 by University of Chicago; Jeremy M. Grushcow
 
ABSTRACT
 
Patent laws worldwide require inventors to keep their inventions secret prior to patent filing so that proprietary claims do not issue on material that inventors appear to have placed in the public domain. The resulting secrecy is widely acknowledged to be one of the main costs of the patent system. This paper takes a novel approach to measuring scientists’ secrecy and shows that scientists seeking to patent their work do in fact withhold presentation of their data at scientific meetings. The data reveal a further effect—a more widespread change in the behavior of university scientists, who exhibit increased secrecy even when not seeking patents. The increased secrecy by those not seeking patents is a challenge to existing patent law, because it occurs without the compensatory benefits of patent-driven commercialization or patent-motivated invention. To minimize this unexpected cost, this paper suggests a combination of changes to the law, technology, and norms relating to academic scientists.
 
1.         INTRODUCTION
 
A patent will not issue in the United States if the invention claimed was “patented or described in a printed publication” or “in public use or on sale” more than 1 year before the inventor applied for the patent (35 U.S.C. sec. 102[b]). Similar rules apply in every industrialized nation, *60 the idea in each case being that once placed in the public domain, however inadvertently, inventions should not be privately appropriable. [FN1] Because of these laws, an inventor must be extremely secretive about an invention before a patent application is prepared and filed. An incautious inventor risks losing patent rights by describing or using the invention publicly. There is therefore a fundamental trade-off in the patent system between the benefits of patenting and the costs of the increased secrecy required to secure a patent in the first place. Although widely acknowledged to be one of the main costs of the patent system (see, for example, Eisenberg 1987, p. 216), secrecy is hard to measure.
 
This paper empirically measures scientists’ secrecy by measuring the delay between a scientist’s presentation of data at a scientific meeting and the formal publication of that work in a peer-reviewed journal. A short publication gap indicates that the scientist maintained data in secret—withholding the meeting presentation until the work was substantially complete. Conversely, a long publication gap indicates that the scientist readily shared data that were far from complete—presenting data at the meeting that were years from being ready for publication. I examined presentations from the 1980 and 1990 meetings of the American Association for Cancer Research (AACR), which because of its clinical relevance should be ripe for patent-seeking among participants. The AACR meeting allows for comparative statistics because it draws attendees from universities, government labs, and industry and because it covers a time period in which these different populations of scientists were differently affected by changes in the law regarding patentability. The data show that academic scientists seeking to patent their work did in fact respond to the economic incentives of patent law by withholding presentation of their data until they were substantially complete. However, the data also reveal a further effect—a more widespread change in norms among university scientists, which resulted in increased secrecy even among those scientists not seeking patents. The increased secrecy by those not seeking patents is a challenge to existing patent law because it occurs without the compensatory benefits of patent-driven commercialization or patent-motivated invention.
 
Section 2 describes the legal changes that allow a comparison of scientists’ behavior, specifically explaining how the Bayh-Dole Act (*61Public Law No. 96-517 [1980]), the Stevenson-Wydler Act (Public Law No. 96-480 [1980]), and the Supreme Court’s decision in Diamond v. Chakrabarty (447 U.S. 303 [1980]) changed incentives for federally funded scientists. It shows that affected scientists took advantage of the new opportunities created by these changes by patenting an increasing number of inventions, by entering into collaborations with industry, and by focusing their efforts on newly patentable subject matter. Section 3 measures the effect of Bayh-Dole, Stevenson-Wydler, and Chakrabarty on information sharing by scientists in three populations differentially affected by these legal changes: university scientists, scientists at the National Institutes of Health, and scientists working in private industry. Section 4 suggests legal, technological, and norms-based changes to combat at least that secrecy unassociated with patent seeking.
 
2.         REWARDS AND RISKS OF LEGAL CHANGES PROMOTING PATENTING BY SCIENTISTS
 
The incentives for scientists and institutions performing federally funded biology research were radically altered in 1980. Prior to 1980, policy regarding the patenting and licensing of government-funded inventions was determined at the agency level, and decisions were made on a case-by-case basis. Without the secure promise of patent rights or exclusive licenses, it was difficult to secure the investment needed for commercialization. As a result, most government-funded inventions either languished while subject to government-owned patents or were released to the public domain through publication (on the relationship between exclusive rights and commercial investment, see Kitch [1977]).
 
In 1980, the federal government passed two bills to promote the commercialization of government-funded research through a harmonized technology transfer policy. The first, the Bayh-Dole Act, allowed universities and small businesses to retain title to federally funded inventions and to retain the revenue generated by licensing such inventions to private corporations (Public Law No. 96-517). The second, the Stevenson-Wydler Act, made transfer of publicly funded inventions to the private sector a part of the mission of government labs and explicitly mandated the establishment of technology transfer offices in government labs to accomplish this goal (Public Law No. 96-480). These two bills significantly altered the focus of federally funded scientists—shifting their attention to the patentable and commercial aspects of their work. *62 They also ensured that institutional frameworks developed to support commercially oriented activities.
 
The year 1980 was a pivotal year for patents in another way as well: the Supreme Court’s decision to allow an inventor to patent a genetically engineered bacterium in Diamond v. Chakrabarty emphasized that “anything under the sun that is made by man” is patentable (447 U.S. 309). Chakrabarty was a pro-patent decision encouraging to all inventors, but particularly to those in the nascent biotechnology industry. [FN2] The biotechnology industry grew rapidly, in part fueled by the commercialization of inventions made in federally funded labs. [FN3]
 
These contemporaneous events—the Bayh-Dole Act, the Stevenson-Wydler Act, and Chakrabarty—opened a world to academic biological scientists that promised financial reward as a supplement to the more ephemeral rewards of intellectual accomplishment and peer approval. As I will show below, scientists took advantage of these opportunities and generated more patents and commercialized more inventions after 1980. These changes provide the backdrop for my empirical study. By examining the behavior of federally funded scientists before and after these changes, I can observe the effects of increased commercial activity on scientists’ secrecy.
 
2.1.      The 1980 Changes and the Increase in Patenting and Commercialization
 
It is difficult to separate the influence of the 1980 legal changes from the general boom in biotechnology facilitated by contemporaneous technological advances. The strongest evidence for a causal relationship between the legal changes and the increased commercial activity is the correspondence between the date of the legal changes and the date of the behavioral change by the affected population. For university scientists, *63 the establishment of technology transfer offices and increased patenting immediately followed the Bayh-Dole Act in 1980 (see Section 2.1.1). However, the increased commercial behavior of the National Institutes of Health (NIH) scientists does not emerge until subsequent changes to Stevenson-Wydler in 1986 added financial incentives for individual intramural scientists (see Section 2.1.2). The difference in timing between increased patenting by university scientists and by NIH scientists cannot be explained by technological changes, because these changes would have affected university and NIH scientists equally. [FN4]
 
2.1.1.   University Scientists—The Bayh-Dole Act. The increasingly obvious commercial potential of biotechnology inventions exacerbated universities’ frustration with the federal government’s unstable and restrictive license policies. In response to this frustration (Eisenberg 1996, p. 1691), Senators Birch Bayh and Robert Dole introduced a bill to grant nationwide permission for universities and small businesses to file for patents on their federally funded inventions and to grant exclusive licenses on those patents to U.S. firms (Public Law No. 96-517). [FN5] The Bayh-Dole Act states that “[e]ach nonprofit organization or small business firm may, within a reasonable time after disclosure [to the funding federal agency], elect to retain title to any subject invention” (35 U.S.C. sec. 202[a]). Although the act provides that the government agency funding the research be granted a “nonexclusive ... paid-up license” (35 U.S.C. sec. 202[c][4]), the funding agency may not require inventors to agree to grant licenses to third parties (sec. 202[f]). Finally, and most important for individual scientists, the act requires nonprofit organizations, including universities, to “share royalties with the inventor” (sec. 202[c][7][B]).
 
The effects of Bayh-Dole were felt almost immediately. First, universities moved at an unprecedented rate to establish technology transfer offices. Of the institutions responding to a survey by the Association of University Technology Managers (AUTM), only 28 had established technology transfer offices prior to 1980 (AUTM 1999, p. 20). In the 5 years following the passage of Bayh-Dole, an additional 44 respondent universities*64 established such offices. Second, between 1980 and 1985, the percentage of university research funded by industry increased more than in any other 5-year period in the 15 years before or after the passage of the act (Figure 1). Third, the number of academic patents filed in the 1980s was more than double the number filed in the 1970s, and the difference was even more pronounced in biology (National Science Board 2000, pp. A-479-A-480). In contrast, overall patenting rates increased by only 2 percent from the 1970s to the 1980s. [FN6] These data show a number of important changes occurring concomitantly with the passage of Bayh-Dole: an increase in technology transfer efforts, an increase in commercial investment in university research, and a clear shift in inventive activity toward universities generally and toward university biologists specifically.
 
2.1.2.   National Institutes of Health Scientists—The Stevenson-Wydler Act and the Federal Technology Transfer Act. The Stevenson-Wydler Act and the Federal Technology Transfer Act (FTTA) accomplished for researchers in government labs what the Bayh-Dole Act accomplished for university *65 researchers (Public Law No. 96-480). The Stevenson-Wydler Act, passed in 1980, was mainly hortatory, creating offices to encourage and monitor technology transfer. The 1986 FTTA made more substantive changes. It provided financial incentives for commercialization to scientists by mandating royalty and income sharing in addition to a “cash awards program” for inventions (Public Law No. 99-502, secs. 6, 14 [1986]). [FN7] The FTTA also allowed government labs and private firms to enter into Cooperative Research and Development Agreements (CRADAs) and allowed government labs to license or assign inventions developed under CRADAs to industry partners (sec. 2). [FN8]

 
The FTTA’s incentives for inventors were very effective. Beginning in 1986, there was a sharp increase in the overall patenting effort of NIH inventors. For example, the number of successful patents filed that named the Department of Health and Human Services (HHS) [FN9] (or its predecessor, the Department of Health, Education and Welfare) as the assignee [FN10] increased every year from 1986 to 1992 (Figure 2). [FN11] The number of patents with HHS as an assignee climbed from 190 between 1980 and the end of 1985 to 375 between 1986 and 1991.
 
The NIH also quickly took advantage of the FTTA’s permission to enter CRADAs: there were 22 active cooperative projects at the HHS in 1987, and nearly seven times that many (146) 5 years later (National Science Board 2000, p. A-108). The output of these CRADAs can be measured by counting the number of patents assigned jointly to HHS and an outside partner. [FN12] During the 6 years from 1980 to 1986, none *66 of the patents assigned to HHS were jointly assigned to a partner in industry. [FN13] In the following 6 years, 2.4 percent of the patents assigned to HHS were also assigned to an outside partner (Fisher’s exact p = .03). [FN14]
 
The fact that university scientists’ behavior changed in 1980, yet NIH scientists’ behavior did not change until 1986, is compelling because aside from the FTTA, NIH scientists functioned in largely the same technical and legal environment as their peers in universities. Therefore, *67 the coincident timing of the behavioral changes suggests (but, of course, does not prove) a causal role for the financial incentives.
 
2.1.3.   Biotechnology in General—Chakrabarty. In 1972, Ananda Chakrabarty filed a patent application on a genetically altered bacterium that was designed to help clean up oil spills by breaking down petrochemicals. The United States Patent and Trademark Office (USPTO) rejected his claims on the bacteria themselves, arguing that “as living things they are not patentable” subject matter (Chakrabarty, 477 U.S. 306). The Court of Customs and Patent Appeals upheld the USPTO’s determination (596 F.2d 952 [1979]). The Supreme Court reversed in a 5-4 decision, holding that because Chakrabarty’s bacterium was “not nature’s handiwork, but his own” it was patentable (477 U.S. 309).
 
The impact of Chakrabarty on inventors’ incentives to patent is particularly difficult to separate from the effects of other 1980 changes. However, one approach is to examine the number of patents in utility class 435, the utility class for Chakrabarty’s bacterium, [FN15] as a percentage of all patents in the years before and after the Supreme Court’s decision. These data serve as an indication of the effect of Chakrabarty because other contemporaneous influences affected all utility classes. From 1977 to 1980, patent applications in utility class 435 constituted 7.6 percent of all patents with academic institutions as assignees. This percentage increased to 12.2 percent for the period between 1981 and 1984. [FN16] This increase suggests that Chakrabarty caused at least a shift in universities’ inventive activity toward such newly patentable subject matter. However, Chakrabarty was in no way limited to federally funded inventors and would be expected to affect university, NIH, and industry scientists equally.
 
2.2.      The Risk of Rent Dissipation due to Increased Secrecy
 
2.2.1.   The Risk of Early Data Sharing. It is undisputed that the 1980 changes resulted in a tremendous increase in private-sector investment in universities and government labs and in a corresponding dividend of technology transfer and product development (see, for example, Cockburn and Henderson 2001). However, scientists seeking to patent their *68 inventions face a conflict between the incentives created by patent law to keep their data secret until they are substantially finished and the norms of the scientific community, which dictate that data should be shared at an early stage (on the nature and scope of this conflict, see Eisenberg [1987]). Under the Patent Act, disclosure of an invention starts the running of a statutory clock that bars U.S. patentability 1 year later (35 U.S.C. sec. 102[b] [West 2000]). [FN17] The law is presently unclear as to whether conference presentations constitute “printed publication[s]” such as would bar patentability under the Patent Act. In MIT v. Fortia, for example, the federal circuit held that oral presentation of a paper to 50-500 people at a conference along with distribution of the full paper to at least six individuals does constitute prior publication (Massachusetts Institute of Technology v. AB Fortia, 774 F.2d 1104 [Fed. Cir. 1985]). Since biologists generally distribute abstracts (but not full printed copies) of their papers at conferences, it is unclear whether the holding of MIT v. Fortia extends to their practice. Therefore, biologists planning to patent should (Palladino 1999) (and do—see Section 3) withhold early data from conference presentations until the patent application is filed or until the work is substantially complete and the patent application is ready to be filed.
 
2.2.2.   The Benefits of Early Data Sharing. Conference presentations of early data are not only consistent with the communalist norms of scientists, they also serve an important economic function by helping scientists avoid wasteful effort both by sending signals to other scientists and by receiving information from other scientists. First, early data sharing may avoid the expenditure of duplicative effort by allowing scientists working on similar projects to identify each other. They then have the opportunity to either diversify their efforts, collaborate on the remainder of the project, or abandon one of the projects. Second, early data sharing can reduce wasted effort by allowing other scientists to shed light on a project that is not duplicative but merely misguided. This information may be available because another scientist has already attempted the experiment and discovered a reason it is bound to fail or because a combination of other scientists’ data shows that a project will not be *69 fruitful or is moot. Therefore, the increased financial reward for inventive activity increases the risk of rent dissipation in the form of wasted expenditures both from duplicative races to invent and from misguided research that continues unchecked. [FN18]
 
3.         THE EFFECT OF THE 1980 CHANGES ON INFORMATION SHARING AT CONFERENCES
 
Because the Bayh-Dole and the Stevenson-Wydler Acts encourage publicly funded scientists to seek patents on their work, they increase the risk that the prefiling secrecy mandated by the Patent Act will prevent valuable early data sharing at conferences. To determine whether this occurred, I examined the data presented in abstracts at the meetings of the AACR in 1980 and 1990. [FN19] The AACR organizes a large meeting with domestic and international participants from industry, universities, and the NIH. [FN20] I selected this meeting not only for its size but also because cancer research is a field that has obvious potential for commercialization and is therefore ripe for patent-seeking behavior.
 
3.1.      Assessing Early Data Sharing
 
3.1.1.   The Publication Gap as a Metric for Data Sharing. As a proxy for the stage at which data were presented by scientists at the AACR meetings, I determined the time that elapsed between presentation of the abstract and formal publication in a peer-reviewed journal. The time lapse was determined by searching the MEDLINE database (http://www.ncbi.nlm.gov) for a formal publication that matches the meeting abstract and that is authored by the principal investigator (last author) of the meeting abstract. For example, if the MEDLINE search revealed a publication in 1981 that matches an abstract from the 1980 meeting, it was scored as a “+1”—a 1-year lag between meeting abstract and formal *70 publication. This process cannot be automated because the abstracts do not exist in electronic form and because it requires expertise in the subject matter in order to match meeting abstract to publication when there has been some change in the title or in the text of the abstract. Ideally, and in most cases, the title of the abstract was identical or very similar to the title of the formal publication and the match was unambiguous. However, in any case of multiple possible matches between an abstract a subsequent formal publication, I selected the earliest possible matching publication.
 
A long gap between the meeting abstract and the formal publication indicates early data sharing. A short gap, or a formal publication of the data before the AACR meeting, indicates that the data presented in the AACR abstract were substantially complete and ready for patenting. Finally, I searched the USPTO database (http://www.uspto.gov) for the 5 years preceding and the 10 years following the AACR meeting to determine whether the data in the abstract were the subject of a patent grant. [FN21]
 
3.1.2.   Alternative Explanations for the Publication Gap. It is formally possible that a shorter gap does not represent a delay of data sharing, although the data do not support any specific alternative explanations. One possibility is that a short time from abstract to publication could represent faster formal publication or faster research over the entire life of the project. This alternative explanation might create biases in my data if labs that were seeking patents were also doing more “important,” higher priority work, in which case a shorter gap among patent seekers could arise because their research moved faster, not because they were more secretive. To examine this possibility, I scored each formal publication by the Institute for Scientific Information (ISI) impact factor of the journal in which it was published. [FN22] This standard metric is a good *71 proxy for the “importance” of the work. [FN23] Table 1 shows that for academic scientists, publications associated with patents are, in fact, higher impact than those not associated with patents (p = .03). [FN24] However, the correlation between the “importance” of the work and a short publication gap does not hold universally true. Most important, the short publication gap among nonpatenting university scientists in 1990 occurs despite the relatively low mean impact of their formal publications (Tables 1 and 2). Furthermore, the shorter time to formal publication observed among non-patent-seeking university scientists in 1990 was not observed for their peers at the NIH (Table 3) despite the significantly greater “importance” of the NIH scientists’ work (p = .05, Table 2). Therefore, the shorter gap between conference presentation and formal publication is not correlated with the importance of the work and is not likely an artifact of faster publication caused by better funding or higher productivity. [FN25]
 
Table 1. Impact of Formal Publication by Affiliation and Patent Behavior
-
No Patent Patent p-Value
(t-Test)
 University 2.44 +- .23 4.63 +- 1.00 .04
 National Institutes 3.39 +- .21 3.57 +- .56 .38
of Health
 Subtotal (academic) 2.94 +- .16 3.94 +- .50 .03
 Industry 1.70 +- .35 2.46 +- .58 .13
 Total 2.63 +- .15 3.39 +- .40 .04
 
Note. Values presented are mean Institute for Scientific Information impact
factor +- standard error of the mean.
 
Table 2. Impact of Formal Publication by Affiliation and Year
University National Institutes of p-Value (t-Test)
Health
 1980 2.48 +- .27 3.59 +- .32 .005
 1990 2.38 +- .38 3.19 +- .33 .05
 p-Value .58 .19
 
Note. Values presented are mean Institute for Scientific Information impact
factor +- standard error of the mean.
 
Two further arguments suggest that the shorter publication gap represents increased secrecy rather than faster formal publication. First, the *72 degree to which the text and data change between a meeting abstract and formal publication is a qualitative proxy for the stage at which data are presented at a meeting. For example, if the abstract of a formal publication is worded identically to the abstract presented at the conference, it would imply that the conference abstract represented work that was at a late stage—virtually or entirely ready for publication. Conversely, if there were major changes in the text of an abstract or if substantial amounts of new data were added before the formal publication, it would indicate that early-stage data were presented at the conference. This measure is inherently subjective, but the degree to which the abstracts changed from conference to formal publication does seem to correlate with the time-to-publication metric adopted by this paper. Second, the timing of a formal publication is much less flexible than the timing of a conference presentation. A scientist must produce enough data to survive peer review, so moving formal publication earlier is difficult. Delaying formal publication is feasible, but there is considerable academic pressure to publish as soon as possible. [FN26] In contrast, the timing of a conference presentation can be changed at will because there is little to no external review of conference submissions and little academic pressure to rush conference presentations. Since there is more flexibility in the conference submission than in the formal publication, a short publication gap more likely represents delayed sharing at conferences than faster formal publication.
 
3.2.            Patenting and Industry Collaboration at American Association for Cancer Research Meetings Are Representative
 
 
The general trends toward increased patenting activity and increased industry collaboration observed in Section 2 are reflected in the behavior *73 of scientists at the AACR meetings. In 1980, 4.5 percent (7/157) of meeting abstracts examined were associated with a patent, whereas by 1990, 19.2 percent (44/229) of abstracts examined were associated with a patent (p = . 0001, x2 test). The number of joint projects between industry and publicly funded researchers rose from 1.6 percent (20/1,269) in 1980 to 2.8 percent (54/1,960) in 1990 (p = .03, x2 test). [FN27]
 
Table 3. Mean Time to Publication in 1980 and 1990
University National Institutes of Health Industry
 1980 1.37 +- .23 1.31 +- .18 1.53 +- .65
 1990 .83 +- .19 1.28 +- .23 1.60 +- .40
 p-Value .04 .47 .46
 
Note. Values presented are mean +- standard error of the mean.
 
3.3.      Conference Presentations Are Delayed by Patent Seekers
 
The Patent Act’s 1-year clock provides an incentive to withhold publication until work is substantially complete and ready for patent filing (or until the patent application has already been filed). Consistent with this incentive, 37 out of the 51 patents (73 percent) associated with abstracts had a filing priority prior to the conference presentation. Interestingly, of the 14 whose filing priority was later than the conference presentation, only five had filing dates that predated formal publication, which indicates that of the 42 scientists withholding data for patent filing, 88 percent refrained even from presenting a conference abstract. Two more scientists had filing dates within 1 year of publication, but these probably represent strategic timing of patent filing rather than strategic restraint of publication. [FN28] The subsequent analysis treats all 51 patent-associated abstracts as part of the patent-seeking group, since all *74 51 examples still represent patent-seeking behavior on the part of the scientists who published the conference abstracts. [FN29]
 
Consistent with the incentive to withhold data when seeking patents, the lag between abstract presentation at the meeting and the formal publication in a peer-reviewed journal was shorter for university and NIH scientists who sought patents than for their peers who did not seek patents (Table 4). University scientists who sought patents presented meeting abstracts only on work that was complete, on average publishing formally in the same year as the meeting abstract, whereas university scientists who were not seeking patents published on average 1.21 years after their data were presented as a meeting abstract (p = .00003, Table 4). The NIH scientists who were motivated by patenting likewise presented abstracts on work that was substantially complete, with formal publication coming on average .75 years after the meeting, whereas those not seeking patents presented abstracts that were 1.36 years from formal publication (p = .036, Table 4). This shortened lag time between meeting presentation and formal publication demonstrates that the sharing behavior of federally funded scientists is significantly different when patents are involved.
 
Table 4. Effect of Patent Seeking on Early Data Sharing: Average Years to Publication after Meeting Presentation
 
 University National Institutes Industry Scientists of Health Scientists
 Scientists Not seeking  1.21 +- .16 1.36 +- .16 1.21 +- .35
patent
 Seeking patent .0 +- .19 .75 +- .28 2.89 +- .95
 p-Value .00003 .036 .066
 
 
 Note. Values presented are mean +- standard error of the mean.
 
Interestingly, industry scientists seeking patents formally published their work substantially later than those who did not seek patents--2.89 years after the AACR meeting compared with 1.21 years (p = .066, Table 4). This behavior may reflect the fact that once a patent application has been filed, formal publication is no longer necessary to protect the inventor’s economic interest. Presumably, having satisfied their primary goal of patent filing, formal publication becomes very low priority for *75 industry scientists. [FN30] Fifty percent of industry scientists’ AACR meeting abstracts were not formally published, and in one striking example, an industry scientist’s abstract that did result in a formal publication did so only after a gap of 9 years. In contrast, academic scientists followed up on over 80 percent of AACR meeting abstracts with a formal publication. Even among patent-seeking academic scientists, 90 percent of conference abstracts were followed by a formal publication. This indicates that for academic scientists, the kudos market retains its importance. Historically, publication has been the primary metric of academic success. [FN31] Delaying publication at best means a corresponding delay in kudos and at worst means that a rival publishes first (Eisenberg 1987, pp. 197-98). Therefore, even when academics withhold early data, they present the completed work at meetings and formally publish it as soon as possible.
 
3.4.      Impact of the 1980 Changes on Scientific Norms
 
The impact of the 1980 changes went beyond a direct response to the changed economic incentives. Even university scientists who did not seek patents refrained from early data sharing in 1990. No such change was observed among NIH or industry scientists.
 
3.4.1.   University Scientists. University scientists as a group shared less early data in 1990 than in 1980. When Bayh-Dole was implemented in 1980, university scientists presented data that were on average 1.37 years from publication. By 1990, they presented significantly fewer early data, with abstracts being on average only .83 year from publication (p = *76 .04). This decrease in early data sharing by university scientists extended beyond those seeking patents and therefore cannot be explained as a direct response to Bayh-Dole’s economic incentives. Even excluding those abstracts associated with patents, early data sharing by university scientists decreased between 1980 and 1990, with mean time to formal publication declining from 1.43 to .97 year (p = .07, Figure 3).
 
3.4.2.   National Institutes of Health and Industry Scientists. In contrast to the increased secrecy among university scientists, there was no significant change in the behavior of NIH or industry scientists between 1980 and 1990. When abstracts associated with patents are excluded, NIH scientists presented data that were on average 1.31 years from publication in 1980 and 1.43 years from publication in 1990 (p = .36, Figure 4). Similarly, there was no significant change in the behavior of industry scientists. Their abstracts were on average 1.1 years from publication in 1980 and 1.3 years from publication in 1990 (p = .37, Figure 5).
 
3.4.3.   Analysis. The absence of a change in the behavior of industry scientists is expected, since incentives related to patenting did not change for industry scientists between 1980 and 1990. Likewise, the absence of *77 a change in the behavior of non-patent-seeking NIH scientists may simply reflect the fact that the legal changes affected only patent seekers. However, secrecy did increase even among those university scientists not seeking patents, despite the fact that no legal change affected them directly either. Furthermore, since the FTTA made the same changes for NIH scientists in 1986 that the Bayh-Dole Act made for university scientists in 1980, any possible explanation for the short publication gap among non-patent-seeking university scientists would have to explain why increased secrecy is observed only in the university community and not at the NIH. [FN32]
 
One possibility is that tenure incentives changed in a way that encouraged secrecy—since NIH investigators are hired with tenure, but university scientists are not, such changed incentives would differentially affect the two groups. However, this explanation seems unlikely because *78 there is no increase in the impact factor of university professors’ publications between 1980 and 1990 (Table 1), so if they had adopted increased secrecy as a strategic response to changed tenure incentives, the response was unsuccessful in improving the quality of their work. Furthermore, secrecy would be a bad strategy for tenure seekers because tenure decisions are made in part on the basis of the recommendations of outside faculty, who would likely be sensitive to (and resentful of) their colleagues’ increased secrecy.
 
Another possible explanation for the increased secrecy by university professors not seeking patents is a shift in norms. While direct benefits should still accrue to those who do choose to share early data, the norm of data sharing could erode as the sense of reciprocity and collegiality is lost. Consistent with this theory, a recent survey reports decreased sharing of materials among academic scientists and supports the hypothesis that a broad shift in norms has occurred in the university community. [FN33] Notably, there are two mechanisms by which changing norms *79 could account for the difference between scientists not seeking patents at universities and at the NIH. [FN34] First, because the increased commercial activity at universities started in 1980 (see Section 2.1.1) but did not start at the NIH until 1986 (see Section 2.1.2), the difference between university and NIH researchers could be a result of the time required for a more secretive norm to propagate to those scientists who are not seeking patents. In a related manner, it is possible that the persistent sharing of early data by NIH scientists reflects a more durable communalist norm, which would not be surprising considering that unlike their counterparts in universities, the intramural scientists at NIH form a large, undispersed community that may be more able to enforce its norms.
 
4.         COUNTERACTING THE SHIFT TOWARD SECRECY
 
The 1980 changes to incentives for university and NIH scientists were implemented with the goal of increasing patenting and commercialization of publicly funded inventions, an effort that has met with tremendous success (see Section 2). [FN35] However, this paper suggests that increased secrecy is associated with patent seeking [FN36] and that increased patenting in a scientific community can cause the entire community to become more secretive (see Section 3 and Campbell et al. [2002, pp. 476-78]). Increased secrecy by those seeking patents has been assumed, and is at least implicitly accepted, to be a cost with corresponding benefits. In contrast, the unwillingness of those who are not seeking patents to share early data causes rent dissipation that is not offset by the benefits *80 of patenting and commercialization. It is therefore important to consider whether there are ways to regenerate the pre-1980 scientific norm that encouraged early data sharing. [FN37] A less secretive norm could be propagated by simultaneously reducing the risk of early data sharing by implementing an “experimental disclosure” exception in patent law and increasing the rewards of sharing to the disclosing and to the receiving parties.
 
One way to reduce the risk of sharing would be to amend the Patent Act to explicitly allow early data sharing at conferences without jeopardizing patentability—an “experimental disclosure” exception to section 102(b). A similar exception to section 102 already excuses some public uses and sales. [FN38] For example, in the paradigmatic case of City of Elizabeth v. American Nicholson Pavement (97 U.S. 126 [1877]), an inventor installed a new type of pavement on a public street, where it remained for 6 years before he applied for a patent. Since the inventor had not profited from the installation, and since he monitored its performance throughout the testing period, the Court allowed a patent to issue. [FN39] The rationale in that case was that the only way to test the pavement’s durability was to subject it to public use (97 U.S. 135). Similarly, providing a safe harbor for the publication of conference abstracts would allow scientists to validate their inventions, and a provision could include a similar restriction forbidding interim profit. Although the data would be out of the inventors’ control, peer review of some *81 sort is the only way to test the experiments that led to the invention. [FN40] By recognizing a section 102(b) exception for experimental disclosures at meetings, Congress or the courts could encourage renewed sharing by federally funded scientists.
 
There are some problems with an experimental disclosure exception. First, it does nothing to protect scientists from the laws of most foreign countries, which bar patentability upon publication, including publication at conferences. For those scientists motivated by foreign patent rights, changes in U.S. law would not be sufficient. [FN41] Second, merely lifting the disincentive may not be sufficient to restore early-data-sharing norms once they have already been abandoned. Third, and most importantly, it conflicts with the policy choices behind the statutory bar of section 102 to begin with—the reluctance to allow inventions seemingly within the public domain to be reclaimed by the inventor, particularly in cases where others have begun to use the invention without notice of any proprietary rights. [FN42]
 
Changes in patent law can reduce the risk of early sharing, but technological and norms-based changes could increase the rewards of sharing even under current law. One approach would be to encourage the use of online fora for sharing early data. If such fora were widely used and cited, early data sharing online would provide significant opportunities for collaboration and criticism. Note that even without legal protection for early online publication, online fora might still allay the harms of increased secrecy. For example, the ready availability and searchability of online materials could result in better dispersion and utilization of the posted data to avoid rent dissipation. Some online dissemination of results already occurs. One notable effort is the Los Alamos e-print archive arXiv (http://www.arXiv.org), now at Cornell, which has become “a major forum for dissemination of results in physics and mathematics” (Ginsparg 1999). Early efforts at online data sharing in biology include *82 the NIH’s PubMed Central (http:// www.pubmedcentral.nih.gov), the American Association for the Advancement of Science’s Signal Transduction Knowledge Environment (STKE; http:// stke.sciencemag.org), and the prepublication of papers in press by the Journal of Biological Chemistry (http://www.jbc.org/pips/pips.0.shtml). However, these sites are structured much more like traditional journals than like arXiv’s early-data-sharing community. Online fora that present particular results rather than entire publications have a further advantage because they could allow some early data sharing even by patent-seeking scientists. Disclosures have to render the subsequent invention obvious to activate the statutory bar to patentability. In many cases isolated data could be published that would ameliorate some of the harms of secrecy but that would not be sufficient to trigger the statutory bar (for a discussion of which online publications should trigger the sec. 102[b] bar, see Pierotti [2002]).
 
Norms-based changes could also increase the reward of early data sharing in the kudos market. Awards could be given for novel work presented at meetings, meeting abstracts could be cited in future publications (which they currently are not), or academic institutions could weight meeting presentations more heavily in merit decisions affecting scientists’ careers.
 
Any approach that increases the significance of informal early data sharing also increases the likelihood that, under current patent law, such sharing will constitute publication and hinder the acquisition of patent rights. Therefore, the optimal solution would combine rewards for early data sharing with risk-reducing changes in patent law that provide safe harbors for early data sharing.
 
5.         CONCLUSION
 
This paper confirms that granting patent rights to scientists and their institutions succeeds in generating increased industry participation in publicly funded research and increased patenting and commercialization activity. However, the paper also shows that scientists who seek patents are more secretive, withholding publication or presentation of their data so as not to jeopardize patentability. Notably, this paper also observes that even those university scientists not seeking patents became more secretive in response to the 1980 changes. This unanticipated increase in secrecy increases the risk of wasteful duplication caused by the patent *83 system. The norm of data sharing may best be restored by increasing the rewards for early data sharing while providing an experimental disclosure exception to reduce the risk that early data sharing will jeopardize patentability.
 
REFERENCES
Association of University Technology Managers, Inc. (AUTM). 1999. AUTM Licensing Survey: FY 1999. Survey Summary. http:// www.autm.net/surveys/99/survey99A.pdf (visited January 20, 2004).
Blumenthal, D., E. G. Campbell, M. S. Anderson, N. Causino, and K. S. Louis. 1997. Withholding Research Results in Academic Life Science: Evidence from a National Survey of Faculty. Journal of the American Medical Association 277:1224-28.
Campbell, Eric G., Brian R. Clarridge, Manjunsha Gokhale, Lauren Birenbaum, Stephen Hilgartner, Neil A. Holtzman, and David Blumenthal. 2002. Data Withholding in Academic Genetics: Evidence from a National Survey.
Journal of the American Medical Association 287:473-80.
Cockburn, Iain M., and Rebecca M. Henderson. 2001. Publicly Funded Science and the Productivity of the Pharmaceutical Industry. NBER Innovation Policy and the Economy 1:1-34.
Eisenberg, Rebecca S. 1987. Proprietary Rights and the Norms of Science in Biotechnology Research. Yale Law Journal 97:177-231.
-----. 1989. Patents and the Progress of Science: Exclusive Rights and Experimental Use. University of Chicago Law Review 56:1017-86.
-----. 1996. Public Research and Private Development: Patents and Technology Transfer in Government-Sponsored Research. Virginia Law Review 82:
1663-1727.
Ginsparg, Paul. 1999. Some Comments on e-biosci. HMS Beagle: The Bio-MedNet Magazine 61:5. http://news.bmn.com/hmsbeagle/61/viewpts/page5 (visited July 25, 2001).
Heller, Michael A., and Rebecca S. Eisenberg. 1998. Can Patents Deter Innovation? The Anticommons in Biomedical Research. Science 280:698-701.
Kitch, Edmund W. 1977. The Nature and Function of the Patent System.
Journal of Law and Economics 20:265-90.
Murphy, Sean D. 2001. Biotechnology and International Law. Harvard International Law Journal 42:47-139.
National Science Board. 2000. Science and Engineering Indicators--2000.
Arlington, Va.: National Science Foundation.
Palladino, Donald R. 1999. Comment: The Publication Bar: How Disclosing an Invention to Others Can Jeopardize Potential Patent Rights. Duquesne University Law Review 37:353-70.
*84 Patent Office (UK). 2002. UK Consultation on Patents Grace Periods. http://www.patent.gov.uk/about/consultations/grace/index.htm (visited October 25, 2003).
Pierotti, Neil P. 2002. Does Internet Information Count as a Printed Publication? IDEA 42:249-78.
Rai, Arti Kaur. 1999. Regulating Scientific Research: Intellectual Property Rights and the Norms of Science. Northwestern University Law Review 94:77- 152.
Straheilevitz, Lior Jacob. 2003. Charismatic Code, Social Norms, and the Emergence of Cooperation on the File-Swapping Networks. John M. Olin Law and Economics Working Paper No. 162, 2d series. University of Chicago Law School. http://www.law.uchicago.edu/Lawecon/index.html.
United States Congress. Office of Technology Assessment. 1995. Federal Technology Transfer and the Human Genome Project. Washington, D.C.: U.S. Government Printing Office.
United States Patent and Trademark Office (USPTO). 2002. Table of Issue Years and Patent Numbers, for Patents Issued since 1836. http:// www.uspto.gov/web/offices/ac/ido/oeip/taf/issuyear.htm (visited November 25, 2002).
 
[FNa1]. JEREMY M. GRUSHCOW is an associate with Kirkland & Ellis, LLP. He received his Ph.D. in 2000 from the University of Chicago and his J.D. in 2003 from the Law School at the University of Chicago. The author thanks Audrey Fried-Grushcow, William Landes, Saul Levmore, Douglas Lichtman, Eric Posner, Robert Sitkoff, and an anonymous reviewer for invaluable comments on earlier drafts. Thanks also to Mike Beverage at the American Association for Cancer Research and to David Stewart at the Cold Spring Harbor Laboratory for gracious assistance in obtaining meeting abstract publications. The views and opinions herein are the author’s own and are not necessarily the views or opinions of Kirkland & Ellis LLP or of any client.
 
[FN1]. Most foreign countries do not even allow a 1-year grace period, so disclosure before filing an application results in an immediate bar to patentability. See, for example, Patent Office (UK) (2002), which describes different approaches to grace periods around the world.
 
[FN2]. “At its most general level, biotechnology concerns techniques for using the properties of living things to make products or services” (Murphy 2001, p. 47). See Section 2.1.3.
 
[FN3]. In 1953, James Watson and Francis Crick first described the helical structure of DNA on the basis of data provided by Rosalind Franklin and Maurice Wilkins. The tools to translate and parse the genetic code were discovered in the 1960s and 1970s. In 1966, scientists cracked the genetic code and were able to decipher DNA for the first time. In 1962, Werner Arber discovered restriction enzymes that others later purified and used to cut DNA, and in 1967, Martin Gellert discovered ligase, an enzyme that pastes DNA back together. At Stanford and the University of California in the early 1970s, Stanley Cohen, Herbert Boyer, and Paul Berg used these enzymes to allow individual genes to be copied, moved, and studied at will. The universities realized over $200 million in income from licensing the Cohen-Boyer patents, and in 1976 Boyer cofounded Genentech, which reported sales of over $1 billion in 2000.
 
[FN4]. In fact, it is difficult to imagine nonlegal changes that would have affected these closely related communities differently. However, a more extensive discussion of alternate possibilities is presented in Section 3.1.
 
[FN5]. The Bayh-Dole Act left the relationship between federal funding agencies and large business contractors unchanged, but large businesses had generally been able to negotiate more favorable arrangements even under the earlier regime.
 
[FN6]. By patent number, 693,397 utility patents were issued in the 1970s compared with 709,468 in the 1980s (United States Patent and Trademark Office [USPTO] 2002).
 
[FN7]. When a research collaboration between NIH and a private company results in royalties, the inventor is eligible to receive 25 percent of the first $50,000 earned, 20 percent of the second $50,000 earned, and 15 percent of any amount in excess of $100,000 (United States Congress 1995, p. 15).
 
[FN8]. The National Competitiveness Technology Transfer Act of 1989 extended these provisions to government-owned, contractor-operated facilities (Public Law No. 101-189, secs. 3131-33 [1989]).
 
[FN9]. The Department of Health and Human Services is the federal agency responsible for NIH.
 
[FN10]. Almost all patented NIH inventions are assigned to HHS.
 
[FN11]. My source for these data is a USPTO patents database search, May 11, 2001. Because the pre-1986 patenting behavior of the HHS is fairly consistent, I believe it provides the appropriate control to assess post-1986 behavior. Of course there is a general increase in patenting activity throughout the time period in question, but choosing an appropriate control group outside the HHS is more problematic than using the pre-1986 data as a control.
 
[FN12]. Searches were performed on Westlaw’s US-PAT database using the string “apd(YYYY) & pas((health /2 human /1 services) h-h-s)”. Counting jointly assigned patents following the FTTA is probably an underrepresentation of the impact of CRADAs for three reasons. First, there is a time lag between the initiation of a joint research project and a resultant patent filing, so projects initiated in 1986 were unlikely to bear fruit until several years later. Second, as shown above, the overall number of HHS patents doubled in response to the FTTA’s incentives for individual inventors in federal labs, so the increase in joint assignments is diluted by the more general increase. Third, the FTTA allowed the outright assignment of CRADA-funded inventions to the outside partner, instead of requiring joint assignments, but these patents were not counted because they are not easily identifiable as originating at a federal lab.
 
[FN13]. One patent out of the 190 assigned to HHS from January 1980 to December 1985 was jointly assigned to the Institut Pasteur, the French research institute that codiscovered the HIV virus. Another Institut Pasteur coassigned patent was present in the 1986-91 sample. Both were discarded because they represent compromise in a political and scientific dispute and were not the result of joint research with industry partners.

 
[FN14]. Nine out of 375 HHS-assigned patents had a coassignor.
 
[FN15]. Chakrabarty’s patent number 4,259,444 issued March 31, 1981, in utility class 435 (Chemistry: Molecular Biology, Microbiology).
 
[FN16]. My USPTO search of May 11, 2001, showed that 42 out of 551 patents with application dates between 1977 and 1980 that listed “university,” “college,” or “technical institute” as assignee were in utility class 435. For patents filed between 1981 and 1984, the corresponding ratio was 139 out of 1,189.
 
[FN17]. The statutory clock can even be started by presentation at a meeting or conference under some circumstances (Palladino 1999, pp. 361-62). “Publication” for Patent Act purposes is distinct from scientists’ definition of publication. Scientists define publication as publication in a peer-reviewed scientific journal, which this paper refers to as “formal publication” to distinguish it from the Patent Act meaning.
 
[FN18]. Other accounts of costs from increased academic patenting have focused on the increased transaction costs associated with commercialization. These costs arise because of the potential need to license many component patented inventions to successfully commercialize a single new product. See Heller and Eisenberg (1998) and Rai (1999).
 
[FN19]. For each of the 1980 and 1990 meetings, I surveyed 50 abstracts presented by university scientists, 50 abstracts presented by NIH scientists, and 50 abstracts presented by industry scientists, except that only 20 abstracts were presented by industry scientists in 1980.
 
[FN20]. In 1980 there were 1,269 abstracts presented, and in 1990 there were 2,684 abstracts presented.
 
[FN21]. In attempting to match conference abstracts to issued patents, I searched for patents that named an author of the meeting abstract as an inventor, and then I focused primarily on the patent abstract and the description.
 
 
[FN22]. The journal impact factor is described as “a measure of the frequency with which the ‘average article’ in a journal has been cited in a particular year .... The impact factor is calculated by dividing the number of current citations to articles published in the two previous years by the total number of articles published in the two previous years” (http:// jcr8.isiknowledge.com/www/help/hjcrgls2.htm#Impact_Factor_def, visited Feb. 17, 2003). Abstracts that were never formally published were scored as having zero impact.
 
[FN23]. According to ISI, “The impact factor will help you evaluate a journal’s relative importance, especially when you compare it to others in the same field.” Because publication in a certain journal may be based partly on the reputation of the principal investigator, the impact factor of the formal publication’s journal will incorporate some general measures of lab quality and productivity external to the particular piece of work and may be a particularly apt proxy for “importance,” in the sense of available funding and other publication-speeding resources.
 
[FN24]. This might indicate that the higher priority given to scientifically interesting projects results in faster formal publication. Alternatively, or additionally, the shorter publication gap is caused by the secrecy mandated by the patent laws rather than by the faster formal publication that results from increased productivity in important labs. This hypothesis could be specifically tested by comparing the publication gap between patent-associated and non-patent-associated work within individual labs.
 
[FN25]. Also, because both pools cover a wide mix of journals, it seems unlikely that any tendency of a particular journal to publish quickly or slowly will affect the data.
 
[FN26]. While there are countervailing economic pressures to delay publication pending patent filing, these incentives do not apply to scientists who are not seeking patents.
 
[FN27]. This increase is not merely an artifact of an overall increase in industry participation, which rose in absolute terms but remained the same proportionately, accounting for 2.4 percent of abstracts in 1980 (30/1,269) and in 1990 (50/2,110).
 
[FN28]. There are two possibilities regarding the remaining seven patents: either they are genuinely invalid or they are not barred by the conference abstract or formal publications that preceded them, likely because while there was sufficient overlap between publication and patent for me to match the topics, the publications may not have disclosed or enabled precisely the invention claimed in the patent.
 
[FN29]. Furthermore, reanalysis of the data from the opposite perspective-- counting only the 37 abstracts preceded by patent filings as patent seeking—actually strengthens the following results, so the choice to include all 51 errs on the side of caution.
 
[FN30]. This raises the question of why industry scientists would bother to present abstracts at meetings once a patent application is filed. One possibility is that they seek the kind of valuable feedback on their own work that comes from interactions at meetings (see Section 2.2.2). Another possibility is that they use meeting presentations to promote their products or generate interest in their clinical trials. Whatever the motivation for the meeting presentations, the falloff in industry participation between meeting presentation and formal publication can probably be explained by the time and effort required to generate a formal publication, which is significantly greater than that required to generate a conference presentation.
 
[FN31]. Interestingly, publication and peer recognition may be increasingly important as a metric for private-sector scientists’ success. According to Cockburn and Henderson (2001, p. 7), “As the techniques of drug discovery evolved and it became increasingly important to be able to take advantage of the findings of public science, the most productive pharmaceutical firms began to reward their researchers on the basis of their standing in the eyes of their peers.” This may explain why industry scientists who produce publishable work that is not patented do publish that work in a time frame similar to that of their publicly funded peers.
 
[FN32]. Otherwise, for example, one might imagine that a short publication gap could be caused by decreased conference presentation slots compared with journal publication slots or by some change in the technical nature of experiments that yielded no results until they produced substantially complete results. These factors, however, would be expected to affect both university and NIH scientists equally and therefore cannot explain the different levels of secrecy between those two groups.
 
[FN33]. Campbell et al. (2002, pp. 476-78) report that surveyed faculty reported encountering refusals to share materials and that 35 percent of surveyed faculty felt that sharing had decreased in the previous decade.
 
[FN34]. Straheilevitz (2003, pp. 39-53) explains how Richard McAdams’s and Eric Posner’s theories address norm enforcement in close-knit groups and cites social science literature on a “norm of reciprocity” to develop a theory of the maintenance of norms in loose-knit groups.
 
[FN35]. According to Cockburn and Henderson (2001, p. 18), “The quantitative evidence suggests that the rate of return to public sector research as measured by its effect on the private sector, may be as high as 30 percent. There are a number of reasons for believing that this figure is in fact a quite conservative estimate of the overall social return to publicly funded research in this sector of the economy.”
 
[FN36]. Consistent with the findings in this paper, data from a 1997 survey suggest that scientists “among the most productive and entrepreneurial faculty” have sometimes withheld formal publication for strategic reasons, including to preserve patent filing (Blumenthal et al. 1997, p. 1224). But Campbell et al. (2002, p. 478) report that of those who had withheld research materials, 21 percent cited a “need to protect the commercial value of the results” as a motivating factor in their decision to withhold but also report no admitted correlation to patent seeking.
 
[FN37]. One possibility to consider is that scientists’ behavior will self-correct. If the cost of withholding early data is significant, then the norm should revert to its previous state when those costs become obvious. However, if the benefits of early data sharing are realized only when the entire community participates, then the collective action problem will not self-correct even if sharing would be advantageous. In that case, the early-data-sharing norm would not be restored without outside intervention.
 
[FN38]. For examples see Seal Flex, Inc. v. Athletic Track & Court Constr. (98 F.3d 1318 [Fed. Cir. 1996]), where experimental use of an outdoor running track was not considered a public use within sec. 102(b); Manville Sales Corp. v. Paramount Systems (917 F.2d 544 [Fed. Cir. 1990]), where the experimental sale of a lighting system was not “on sale” within sec. 102(b). An “experimental use” exception also exists as a defense to patent infringement, but despite significant academic support (see, for example, Eisenberg 1989), courts have steadily cut back on its scope, most recently in John M.J. Madey v. Duke University (307 F.3d 1351, 1362 [2002]), where the federal circuit reiterated that the experimental use defense is “very narrow and strictly limited.”
 
[FN39]. “The use of an invention by the inventor himself, or of any other person under his direction, by way of experiment, and in order to bring the invention to perfection, has never been regarded as such a use” (97 U.S. 134).
 
[FN40]. Obviously the analogy is imperfect. The intellectual testing of peer review is not precisely the same as the physical testing that characterizes cases under the experimental use exception.
 
[FN41]. This problem could be overcome if the United States was able to negotiate similar changes in foreign patent regimes.
 
[FN42]. Section 102(b) also serves to motivate inventors to file patents quickly. This goal is met in most jurisdictions by a first-to-file rule, whereby the first inventor to file a patent application gets priority. In the United States, however, patents are granted under a first-to-invent rule, under which a diligent early inventor who files second maintains priority over a later inventor. The statutory bar provides extra incentive to file quickly in a first-to-invent system.
END OF DOCUMENT