Draft 9. September 23, 2004.
Introduction. This policy implements the two most important steps that foundations can take to promote open access. First, it puts a condition on research grants, so that all publications resulting from the funded research will be made available in some open-access form. The policy offers three forms of open-access dissemination, among which the author may choose the most convenient. Second, if the author's choice is an open-access journal that charges an upfront processing fee, then the foundation will pay the fee. The policy treats the cost of open-access dissemination as part of the cost of research.
Everything else in the policy is detail for handling different situations, e.g. when a grant is only one of many supporting a particular project, when the grantee is not the author or copyright holder of a work based on the funded research, and when the policy is violated.
The main reason for writing the draft is to have worked-out language ready for any foundation that might be considering an open-access policy. The detail shows some of the complexities that could arise and one way to handle them. It should save time for any foundation that adopts an open-access policy, even if its terms differ widely from those in this draft.I welcome comments and suggestions on this draft.
- By accepting this grant, the applicant agrees to provide open access to any report, article, thesis, dissertation, or publication (hereafter "work") directly based in whole or part on the research funded by this grant. (Call this the "open-access condition" or the "open-access requirement".)
- This condition does not apply if the applicant produces no such work at all.
- This condition does not apply to patentable discoveries, books, software, or other creative products for which creators traditionally receive royalties or revenue. Hence nothing in this policy should deprive grantees of revenue based on the research funded by this grant.
- This condition only applies to works directly based on the funded research, not to works based on published accounts of the funded research. Hence nothing in this policy should constrain authors from writing about any published research or results.
Annotation. The condition might apply only to works "substantially" based on funded research. But that would require line-drawing decisions (while shirking the responsibility of making them) and invite quarreling and litigation. The only clear and clean rule is a "whole or part" rule.
To apply this condition to software would require that all software based on the funded research be open source. While that would be consistent with the principles underlying this policy, I omitted it from the policy in order to focus on scholarly research publications, the traditional focus of open-access initiatives.
- By "open access" to the work we mean its availability (2.1) on the public internet, (2.2) free of charge for everyone with an internet connection, (2.3) in an archive, repository, database, or other collection compliant with the Open Archives Initiative (OAI) metadata harvesting protocol, and (2.4) with some label indicating that the copyright holder has consented in advance to unrestricted reading, downloading, copying, sharing, storing, printing, searching, linking, and crawling.
- Posting the work to a personal web site or journal site will not normally suffice, because it will not normally meet Term 2.3.
- We would be glad to provide you with help in finding archives and journals that meet these conditions. Request our Pamphlet X or see Web Page Y.
Annotation. More on Term 2.3. Compliance with the standards of the Open Archives Initiative makes distributed archives interoperable, and makes their contents visible to users who don't know which collections exist, where they are located, or what they contain. It also means that third-party developers may develop tools (search engines, indexers, classifiers, summarizers, etc.) that apply to all compliant archives without more ado. Many open-access archives are not OAI-compliant; we hope this condition will nudge them toward compliance.
This requirement is almost costless. First, many OAI-compliant institutional and disciplinary archives already exist and new ones are launched almost every week. Second, there are three open-source programs to build and maintain OAI-compliant archives: Eprints from Southampton, DSpace from MIT, and CDSWare from CERN. A funding agency could even launch its own OAI-compliant archive to distribute the works based on the research it has funded.
More on Term 2.4. Foundations may want to specify the language that should appear on a work to indicate this consent, or they may want to specify one of the lawyer-written and machine-readable licenses provided for this purpose by the Creative Commons.
- There are three ways to provide such open access. Grantees may pick the one most convenient for them.
3.1. The author may deposit the work in an open-access archive or repository.
- If there is a difference between the preprint and the postprint, then the applicant must deposit the postprint (more in Term 4).
- For the purposes of this policy, depositing a preprint is neither necessary nor sufficient, although it is very welcome as a step along the way toward postprint archiving.
3.2. The author may publish the work in an open-access journal.
- We will pay any reasonable fee that an open-access journal may require to evaluate submissions or publish accepted articles. We will want evidence that a submission fee will not distort the peer-review process or that a dissemination fee will only be charged for articles already approved by the peer-review process.
Annotation. Some journals will meet most but not all our definition of open access in Term 2. For example, not all will make their open-access articles available on an OAI-compliant form (Term 2.3). If not, then authors will have to deposit the work in an open-access archive (Term 3.1) in order to satisfy this policy.
Instead of promising to pay any reasonable fee, foundations might put a dollar cap on the fee they will pay. For reference, BioMed Central (BMC) asks for $500 per accepted paper. The forthcoming Public Library of Science (PLoS) journals will initially ask $1500. A recent review of the literature puts the average cost of peer review at about $400 per article. In judging whether the BMC figure makes a good benchmark for this fee, consider that, on the one hand, BMC is a for-profit publisher, and that its processing fee is presumably a bit higher than a break-even fee. On the other hand, BMC is still building market share, will not rely on processing fees alone to cover its costs, and is not yet (apparently) making a profit. Mike Eisen defended the much higher PLoS fee in a January 2003 posting to a discussion forum. He also makes clear that the fee may go down as journal management software improves, and may approach zero for journals not using professional editors.
Also consider that the fee must compensate the journal for reviewing articles that it eventually rejects, not just for publishing accepted articles. Journals or disciplines with higher rejection rates will need higher fees. Further, consider that journals can split the fee, and charge a modest fee to evaluate submissions, and a more substantial fee to publish accepted articles. Finally, open-access journals will typically waive the fee entirely in cases of financial hardship. Both BMC and PLoS do so.
I would have suggested a cap in 3.2, but for the difficulty of specifying its amount. So I've retreated to the "reasonable fee" language and pointed to some of the issues in setting the best cap. One might well worry that staying with the "reasonable fee" language may encourage publishers to ask for large fees. Either the foundations will pay them or pay for staff time to listen to explanations of "reasonableness" and negotiate the fees downward. A cap would be more efficient if only the amount could be set in a way that was affordable to foundations and defensible to journals.
While a proper cap may depend on the journal (its rejection rate, its use of professional editors, etc.), it would be very difficult for foundations to publish different caps for different journals or different kinds of journals. Apart from triggering protests from journals, such a practice would require a large amount of empirical research into the business details of thousands of journals, seeking some data not normally made public, and repeating the process every year or so to keep the data current. An alternative is to live with the rough justice of an approximation. While it would be desirable to pay all of the processing fee that the author would have to pay, it would also be desirable to pay a significant portion of it. Some of the problems of setting the best cap disappear if we are willing to let the funded portion of the fee range from (say) 50% to 100%, rather than make it 100% for every journal. While this may reduce an author's ability to publish in the journal of his or her choice, it will also steer authors toward journals with lower processing fees and introduce a note of competition into a journal's fee-setting decision.
3.3. The author may publish the work as an open-access article in a conventional or non-open-access journal.
- Some non-open-access journals will provide open access to individual articles if the costs of evaluating, preparing, and disseminating them are paid in advance. We will pay any reasonable fee for this purpose. Again, we will want evidence that a submission fee will not distort the peer-review process or that a dissemination fee will only be charged for articles already approved by the peer-review process.
- Most conventional journals will not publicly offer to publish individual articles in open-access form, and some may never have considered doing so. We encourage authors to approach the journals of their choice and make this offer. If it would help in their negotiations with journals, they may cite our willingness to pay a reasonable fee to cover the journal's costs and our requirement that work based on funded research be made available in an open-access form. Authors should feel free to photocopy this portion of our grant agreement, or link to this section of our grant agreement, to use in their discussion with publishers.
Annotation. It's not likely that conventional journals that publish occasional articles in open-access form will meet every part of our definition of open access in Term 2. For example, they are not likely to make their open-access articles available on an OAI-compliant form (Term 2.3). If not, then even authors who successfully negotiate a "more open" edition of their work will have to deposit the work in an open-access archive (Term 3.1) in order to satisfy this policy.
Authors who approach journals to suggest the method sketched in Term 3.3 might be helped by knowing that this strategy has been proposed by David Prosser, building on ideas by Thomas Walker, and implemented in Florida Entomologist and the journals of by the Entomological Society of America. While the "3.3" method of experimenting with open access is not yet widespread, it is far from unprecedented.
Instead of promising to pay any reasonable fee, foundations might put a dollar cap on the fee they will pay. Here the considerations differ slightly from those that applied to Term 3.2. On the one hand, in the normal course of things the journal will conduct the peer review of the article, prepare the manuscript, and make a well-formatted electronic copy. In that sense, the fee need only cover the labor of putting a copy of the file in a different directory of the journal's web site and the present cost of the future revenue lost because this individual article is openly accessible. Both costs are trivial. On the other hand, this would leave the costs of peer review and manuscript preparation to be covered by subscription fees or, it would use toll-access fees to subsidize open access, which distorts the economics of both models. To make the article open-access from beginning to end, rather than subsidize only the last step of open-access, foundations should be willing to pay roughly the same fee for this form of open access that they are willing to pay under Term 3.2.
One purpose of having Term 3.3 in the policy, even though very few journals presently satisfy its conditions, is to allow funded authors to publish in any journal of their choice; to distribute the effort of negotiating open-access to all interested authors; and to get conventional journals to experiment with open-access in a piecemeal and entirely revocable way, and to invite this experimentation as a direct business proposition free from confrontational rhetoric. Journals curious about open access and willing to experiment needn't face the daunting prospect of converting their entire operation to open-access, and can count on receiving reasonable compensation for each individual open-access article. This converts the open-access option from an all-or-nothing proposition to a more-or-less proposition, and from a public battle at conferences to a rational calculation inside a journal's editorial office. If a journal publishes enough open-access articles to become familiar with the process, and the economics, then it can easily stop the practice whenever it wishes. Or it can just as easily make its collected open-access articles accessible in an OAI-compliant and form. Since this method of experimenting with open access is not widely known, this policy might be accompanied by a link to David Prosser's article (above) or another public statement of the benefits of this alternative. We would be happy to write that statement and put it online where foundations could link to it or borrow from its language in writing their own adaptations of it.
- The open-access version of the applicant's work must be the final edition of the full text.
- The open-access version must reflect all final revisions to the text and include all tables, charts, illustrations, appendices, and so on. If the text is a peer-reviewed journal article, then the open-access version must be the version approved by the journal editors for publication (the postprint).
- If possible, it should include the publisher's pagination, though perhaps after some delay for it to become available. It needn't include any of the publisher's formatting or "look and feel".
- If the publisher adds reference links to cited sources, the open-access version may but need not include those links.
- If an open-access edition of the work exists, then any number of non-open-access editions may also exist. These may be identical to the open-access edition or contain enhancements omitted from the open-access edition.
- When both open-access and non-open-access versions of the work exist, then the full and final open-access version should either be made public first or be made public simultaneously with the non-open-access version.
- More than one open-access version might exist, e.g. versions of the work at earlier stages of its evolution.
- Open-access works are easy to revise and periodic updating is one way to take advantage of the internet. However, when the full and final version of the open-access edition reflects the content of a non-open-access publication, then that version should not be altered except when the revisions are explicitly labelled as subsequent emendations.
- This condition applies even when this grant is only a part of the funding for a given research project.
Annotation. This term is only necessary in case it was not already read into Term 1. That term and this term together make clear that the open-access condition applies whether or not this grant is the grantee's only source of funding, whether or not this grant covers all the grantee's research expenses, and whether or not the grantee has funding beyond this grant.
- This condition applies regardless of who holds the copyright to the work.
- Grantees who transfer the copyright on works covered by this agreement to publishers should reserve the right to deposit the postprint (Term 4) in an open-access archive (Term 3.1). If they fail to reserve the right, they are still bound by the open-access condition.
Annotation. Because funders are upstream from publishers, and because grantees sign funding contracts before they sign copyright transfer agreements, the terms of this agreement will limit the rights that authors can transfer to publishers. Telling publishers that this open-access condition applies, or explicitly reserving the right to deposit the postprint in an open-access archive, are courtesies to publishers who might not otherwise know that the author has already made this agreement with his or her funding agency. But even authors who fail to take these steps cannot legally transfer to publishers the right to demand that authors refrain from providing open access to their articles. Publishers who wish to make this demand must base it on their optional article-submission policy, not on a right they supposedly hold under the copyright transfer agreement.
Here is language that the University of Kansas recommends to its faculty when signing copyright transfer agreements: "Notwithstanding the above language, I reserve the right to use this manuscript in my teaching and research, for my colleagues at [this university] to use this manuscript in their teaching and research, and I also reserve the right to place an electronic copy of this manuscript on a publicly accessible web site." To adapt the Kansas statement to the purposes of this policy, the final clause should say something like "...place an electronic copy of this manuscript in an open-access and OAI-compliant archive or repository". Foundations might want to propose specific language of their own.
- This condition applies to all research reports, articles, theses, dissertations, or publications directly based in whole or part on the research funded by this grant (see Term 1), regardless of how many there may be and regardless of whether their authors are among the grantees for this grant.
Annotation. One purpose of this term is to deal with grantees who produce many articles from the funded research, not just one. Another is to plug the loophole by which A could apply for the grant and B write up the results of the research, free of the constraints of the open-access requirement.
- The applicant agrees to repay the grant if this condition is violated.
Annotation. The most interesting violation of this policy is to publish a work in a conventional or toll-access form but not also in an open-access form. One remedy might be open-access dissemination at the time the violation is detected, no matter how late in the game. Foundations should consider whether to offer this more lenient remedy, but here are two reasons not to do so. First, one of the chief benefits of open access is accelerating communication and research, and this purpose is defeated when the open-access version is delayed. Second, the lenient option might encourage some authors to wait until their breach or delay is detected before making their work openly accessible, or encourage some publishers to encourage their authors to delay.
If compelling grantees to repay the grant is too strong, and compelling late open-access dissemination is too weak, then foundations might consider some intermediate options. For example, the foundation could reserve some additional "incentive funds" to be released only when the grantee has provided open access to works based on previous funds. Or the foundation could simply make non-complying grantees ineligible for future grants. I thank Jim Till for these ideas.
There are lesser violations, for example putting the open-access version online shortly after the toll-access version is published rather than before or simultaneously (Term 5); providing open-access to the preprint but not the postprint (Term 4); or providing open access to the postprint at a personal home page rather than in an OAI-compliant archive (Term 2.3). Foundations should consider elaborating the policy to provide lesser penalties for lesser violations.
One possible mechanism for monitoring compliance is for an author and grantee to create a standardized CV (example) linking to all open-access works based on the funded research. These CV's would be machine-readable, giving foundations an efficient way to count and compare open-access publications. (This idea is spelled out at greater length, for a slightly different purpose, in an Ariadne article from April 30, 2003, by Stevan Harnad, Les Carr, Tim Brody, and Charles Oppenheim.)
Afterword. More and more scientists and scientific organizations are talking about open access to data, exemplified e.g. by GenBank, on roughly the same terms as open access to articles that interpret and analyze data. Open access to data allows the collaborative building of datasets. It facilitates the process of verifying and replicating results. It creates the opportunity for other scientists to examine the data for themselves, draw their own conclusions, and challenge the interpretation of a published author. It gives scientists at less affluent institutions the same chance to build on the data as scientists at more affluent institutions. It creates a means to share data even from research that does not result in published reports or articles. In fields like paleoanthropology it gives scientists worldwide access to rare specimens that would otherwise require expensive travel and protocols for sharing fragile objects. There are similar benefits for data generated by rare or expensive equipment. Finally, it allows publishing scientists to document their claims with far more supporting data than journals can normally accommodate.
However, there are no standard formats for putting data online in useful ways. See e.g. the example of microarrays. Consequently, if foundations required grantees to provide open access to their data, compliance might be expensive and the result might not yet be useful to interested scientists. I recommend that foundations monitor progress on this front and be prepared to add a requirement for open access to data when the standards and technology have evolved further. In the meantime, foundations should consider whether to encourage open access to data without requiring it, or whether to be silent on the subject.
Further reading on open access to data:
John Whitfield, Phones Download DNA, Nature, February 26, 2003
Christine Soares, Building an 'EvoBank', The Scientist, February 24, 2003
Peg Brickley, Free Access Costs Money, The Scientist, February 21, 2003
Committee on Responsibilities of Authorship in the Biological Sciences, National Research Council, Sharing Publication-Related Data and Materials: Responsibilities of Authorship in the Life Sciences, National Academies Press, 2003
National Oceanic and Atmospheric Administration, Report to the Committee on Environmental and Natural Resources [on the creation of NEDI, the National Environmental Data Index]
[PS: This list of recommended reading is very much out of date. Unfortunately I don't have time to update it right now. Note added 9/23/04.]
Notes on the successive drafts:
Return to the Newsletter
Return to the Blog
Research Professor of Philosophy, Earlham College
Open Access Project Director, Public Knowledge
Senior Researcher, SPARC
This document is in the public domain.