Ten lessons from the funding agency open access policiesNow that we've seen the final version of the Research Councils UK (RCUK) OA policy, it's a good time to look back over the major policies and proposals and draw a few lessons. In this article I'll focus on the policies of the US National Institutes of Health (NIH), the Wellcome Trust, the RCUK, and the proposed policies in the CURES Act and Federal Research Public Access Act (FRPAA) now before Congress. In future articles I may look at more funder policies and more lessons.
SPARC Open Access Newsletter, issue #100
August 2, 2006
by Peter Suber
First, here's where to the find the policies themselves:
The NIH policy (took effect May 2, 2005)
My most recent SOAN article about it
The Wellcome Trust policy (took effect October 1, 2005)
My SOAN article about it
The RCUK policy (announced June 2006)
My SOAN article about it
The RCUK draft policy (released for comment June 2005)
My SOAN article about it
The CURES Act (introduced in the US Senate December 2005)
My SOAN article about it
The Federal Research Public Access Act (FRPAA) (introduced in the US Senate May 2006)
My SOAN article about it
* Lesson 1. The policy should require OA, not merely request it.
The Welcome Trust has proved that funder mandates work and the NIH has proved that mere requests, encouragements, and exhortations do not work. Requests are not enough to bring about deposit (as opposed to non-deposit) and not enough to bring about early deposit (as opposed to late deposit). There's no point in setting OA as a goal and then getting only 3.8% compliance from grantees, which is what the NIH has been getting. Funding agencies that want OA enough to set it as a goal now have empirical evidence to show that only a mandate is likely to help them reach their goal.
Authors are ready: 94% of them would comply with a mandate from their funder or employer (Swan and Brown, May 2005, p. 63). The NIH compliance rate could improve *ten-fold* without reaching *half* that rate.
The Wellcome Trust has a mandate and the CURES and FRPAA bills now before Congress would impose mandates. The NIH started with a request and is now moving toward a mandate. The Public Access Working Group and National Library of Medicine Board of Regents have recommended an NIH mandate and the House Appropriations Committee has decided to require one. In light of the NIH's documented failure to generate participation with a mere request and encouragement, it's hard to believe that any other funder would follow the NIH's false step and settle for a request. But one of the UK Research Councils (CCLRC) has done just that. Three other Research Councils have chosen mandates (BBSRC, ESRC, MRC) and the rest are still deliberating.
* Lesson 2. The mandate should apply to the final version of the author's manuscript, incorporating all changes introduced by the peer review process.
Funders can't mandate OA to the published edition unless they or their grantees hold copyright to the published edition, which is seldom the case. But they should give publishers the option to replace the author's version with the published version in the OA repository. For this purpose, the "published version" could be the very file used in the online edition of the journal or merely the final form of the text and pagination without the published look and feel.
The published version of the text is preferable for citing and quoting, but the version approved by peer review is all that scientists really need to learn about new results and start to apply, replicate, challenge, or build upon them.
Publishers who worry that mandating OA for the author's manuscript will aggravate the version control problem (the problem of having more than one version of an article in circulation) should take advantage of the funder's option to replace the author's version with the published version.
All the policies under review here except the RCUK policy apply to the author's final version of the manuscript and offer publishers the option to replace the author's version with the published version. The RCUK is silent on these questions, apparently leaving them up to the individual RCs. FRPAA lets publishers substitute the published edition for the author's edition but with an interesting qualification: agencies need only allow the substitution when the published edition advances the agency's "goals...for functionality and interoperability". In principle, an agency could decide that a raw or enhanced submission from authors is more useful than the published edition.
* Lesson 3. Without compromising their own interests, funders should help publishers.
Funders should not retreat from an OA mandate or accept longer embargoes than they want; that would compromise their interests. But there are other ways in which they can help publishers.
For example, funders should share traffic and download data with publishers, when they can. Sometimes they can't, for example, when they give authors a choice of repositories (which I also recommend below, Lesson 6). But in that case funders can at least recommend data sharing to repository maintainers.
Data sharing will help all who care to have accurate traffic data for an article --mostly publishers and authors. It will reduce publisher opposition to funder mandates and increase benefits for authors. The benefit will be even more critical for publishers who choose to replace the author's final manuscript with the published edition.
Likewise, when publishers do not replace the author's final manuscript with the published edition, funders should do what they can to ensure that the OA copies include citations and links to the published versions. This helps authors and readers by showing that the article was peer-reviewed and where it was peer-reviewed. It helps publishers preserve and spread their brand.
The purpose of OA is to provide access for those who don't have it or can't afford it. If it serves its purpose, and a reader wants to move from the OA edition to the published non-OA edition, we have no reason to make that difficult and many reasons to make it easy (for example, on the principle that nearly anything that helps readers helps authors). The claim that OA is a superior form of access and distribution does not entail the claim that the OA edition of an article is superior to some other edition.
The new RCUK policy requires a link to the published edition "whenever possible". The NIH provides such links in practice, though the NIH policy doesn't require it. The other policies are silent on the question.
* Lesson 4. The policy should apply primarily to work published in peer-reviewed journals.
It should focus on work voluntarily disseminated by authors to the public and approved by independent peer review. It might also apply to work presented at conferences, but should not apply to work that authors choose not to publish at all (lab notes, phone records, preliminary data and analyses, drafts), work that by law may not be disseminated (classified research), work that is rejected by journals, or work that is only published in royalty-producing forms like books.
Authors who make patentable discoveries will not want to publish until they have the protection of a patent or patent application. But accommodation is built in to this rule: the OA policy only kicks in when authors decide to publish.
The focus on peer-reviewed journals sends two messages: that the funder supports peer review and that the funder will not conduct the peer review itself. This answers two publisher fears --perhaps superfluously, since both fears are groundless. One fear is that authors might want to bypass peer review, and the other is that funders (especially public funders) might become publishers. Funder policies can bring about OA without diminishing peer review or increasing government control of science.
The point of this provision is to protect certain work from forced OA, not to entrench the journal as the primary vehicle of dissemination. If journals undergo fundamental change or displacement, funders should review their wording of this provision.
The only research products not published by authors to which the policy might apply are data files. I support an OA mandate for data files, but because OA data issues differ substantially from OA literature issues, it should probably be the subject of a separate policy. The NIH, for example, separates its article-access policy from its data-access policy.
The RCUK policy applies to conference presentations as well as peer-reviewed articles. The NIH, FRPAA, and Wellcome Trust policies only apply to peer-reviewed articles. CURES applies to peer-reviewed articles but also to some kinds of data files.
* Lesson 5. The policy should allow grantees to use grant funds, or to apply for supplementary funds, to pay the publication fees at OA or OA-hybrid journals that charge fees.
The reason to provide these fees is to support fee-based OA journals, not just OA repositories. The reason to provide supplementary funding for the fees is to prevent article processing fees from competing with equipment, assistants, and supplies for precious and limited grant funds. That would be a disincentive to publish in OA journals, the opposite of what's intended.
This provision is one of several responses to publishers who object that OA archiving will kill subscriptions. Through this policy, funders are saying that they will support journals that shift to an alternative business model, which could be hybrid OA (author-choice OA) or full OA. Of course this does not presuppose that OA archiving really will kill subscriptions. More broadly, this policy supports a new generation of OA peer-review providers, which will be necessary if high-volume OA archiving does undercut subscriptions and which is desirable even if subscription-based journals thrive alongside OA archiving, as they now do in physics.
The policy should not require grantees to submit their work to OA journals. There aren't enough OA journals to meet the demand and, even at a future time when there might be enough, this would limit authors' freedom to publish in the journals of their choice.
I have no opinion on whether funder policies should allow grantees use grant funds to pay page charges and other fees at non-OA journals. But if funders end up paying any publication costs at non-OA journals, they have a right to put an OA condition on their largesse.
I like this variation on the theme: the policy might offer to pay OA publication fees, but only at journals that allow OA archiving within six months of publication. In practice this would only apply to hybrid journals, since full OA journals don't put embargoes on OA archiving. The Medical Research Council comes close to this position but doesn't quite put it this way.
The NIH is willing to pay article processing fees for its grantees, but this is separate from its public-access policy. It also pays about $30 million/year in page charges at subscription journals, about 10 times the entire cost of implementing its public-access policy. The Wellcome Trust makes supplementary funding for publication fees available to grantees. CURES and FRPAA do not let grantees use grant funds for publication fees. The RCUK generally declines to do so, leaving this decision to "authors' institutions", but one RC (the MRC) has offered to pay these fees. (The RCUK still uses false and misleading term "author pays" for journals that charge fees.)
* Lesson 6. The policy should let authors choose which OA repository to use, provided it meets certain conditions of OA, interoperability, and long-term preservation.
Distributed archiving has several advantages over central archiving: it helps local institutions, adds local incentives to funder mandates to increase author participation, adds robustness to preservation, and in the age of OAI interoperability doesn't detract from searching.
If the funder has reasons to require deposit in a certain repository (say, one controlled by the funder the way NIH controls PubMed Central), then that requirement should not be exclusive. Authors should be free to deposit in other repositories as well.
Funders with central repositories will naturally have a procedure to process submissions. But they should also have a procedure to harvest eprints deposited in other repositories. It may turn out that harvesting an already-submitted eprint is even less expensive than processing a raw submission.
The draft RCUK policy preferred institutional over central repositories, but the final policy leaves this decision to the separate RCs, just as FRPAA leaves this decision to the separate federal funding agencies. Of the three RCs that have already chosen to mandate OA, two (MRC and ESRC) require deposit in central repositories and one (BBSRC) requires deposit "in an appropriate e-print repository". The NIH and Wellcome Trust use central repositories (PubMed Central and the UK PubMed Central).
* Lesson 7. The policy should apply to articles that result from research funded in whole or in part by the funder's grant.
This rule makes it unnecessary to decide whether the funder funded a "sufficiently large" portion of the underlying research. The decision is clean: by accepting the funds, the author agrees to provide OA to all articles that arise from the research.
It follows that the policy applies to co-authored articles even if only one author was funded by this funder. This also makes the decision clean: we don't have to decide what quorum of authors is large enough to make the compliance decision and we don't have to let the funder's interest in OA be held hostage by unfunded or differently-funded co-authors.
Finally, in the case of public funding agencies, this rule makes clear that the policy applies equally to agency employees and agency grantees, two groups who in other respects might be subject to different rules. (For example, in the US research by agency employees is uncopyrightable while research by agency grantees is copyrightable.)
The June 2003 Sabo bill in the US (Public Access to Science Act) would have applied to works "substantially funded" by public money, leaving the precise threshold to be decided by individual funding agencies. The bill was not adopted and no other OA policy has followed this model.
The NIH policy, the Wellcome Trust policy, CURES, and FRPAA all apply to work funded in whole or in part by the relevant agency. The RCUK is silent on the question, in effect leaving it to the separate RCs.
* Lesson 8. The policy should adopt the dual deposit/release strategy.
This strategy rests on two critical distinctions: (1) the distinction between the metadata and full-text of an article, and (2) the distinction between the time of deposit in a repository and the time of OA release.
The strategy calls for immediate deposit of both full-text and metadata, and immediate OA release of the metadata. "Immediate" here means at the time the article is accepted for publication. But the full-text needn't be released to OA until after the author-requested embargo or the funder's deadline (e.g. six months), whichever comes first.
Immediate release of OA metadata makes the article searchable and does not threaten publishers who worry about the effects of early release of OA full-text. (In fact, most publishers provide immediate OA metadata themselves, and for the same reason.) When the time comes to release the deposited text to OA, someone only needs to flip a switch; the text is already in the repository and needn't be hunted down, deposited, scanned, rekeyed, or left out. Flipping the switch might even be done by the repository software, programmed at the time of deposit to create an embargo of a certain length.
This strategy is entirely independent of length of the embargo. When embargoes exist, they only apply to the OA release of full-text, not to the deposit of full-text or to the deposit or release of the metadata.
Immediate deposit gives the funders or repositories time to process submissions so that they can really be released at the end of the embargo period. (The NIH, for example, does extensive processing of submissions to put them in a common format and add links to NIH-supported OA databases.) If the processing doesn't start until the embargo period ends, because the text isn't on deposit until then, then the OA release is delayed even further.
In general, publisher insistence on embargoes and funder deference to publishers have blurred both of the critical distinctions in play here. An x-month embargo is needlessly interpreted to apply to both deposit and release, and to apply to both full-text and metadata.
It would be easier to live with fairly long embargoes if we had a mandate for immediate deposit of the full-text and metadata, immediate release of the metadata, and eventual assured release of the full-text. But my preference would be to put an upper limit of six months on embargoes and gradually, over a year or two, and with full notice to authors and publishers, and perhaps with some exceptions, reduce the upper limit.
The dual deposit/release strategy only works for repositories that can accept "dark" deposits --that is, deposits that are not OA. But as far as I know, all the major OAI-compliant archiving packages can do this. The strategy works best for repositories that give users a simple online button for requesting an email copy of a "dark" full-text eprint, and gives authors a simple button for complying with the request. So far, Eprints and DSpace, the two most-used archiving programs, offer this feature, and I believe that other packages are considering it. When users discover an article through its OA metadata, the email button enables easy distribution prior to the release of OA full-text.
So far no funder implements the dual deposit/release strategy in full. The new RCUK policy requires deposit of both metadata and full-text, and for metadata requires it "[w]herever possible...at or around the time of publication." But unfortunately it says nothing about the timing of the deposit or release of full-text. The Wellcome Trust distinguishes the timing of deposit and release for the full-text (deposit immediately upon acceptance, release six months after) but doesn't apparently release metadata before full-text; the NIH does the same but has a longer period of permissible delay before full-text release. CURES, and FRPAA put a deadline on the timing of release but say nothing about the timing of deposit or the deposit/release of metadata.
* Lesson 9. With or without sanctions, the mandate should be enforceable.
There are six *universities* with OA mandates and none of them uses sanctions to enforce the policy. On the contrary, all use some combination of mandatory language, encouragement, education, assistance, and incentives. These suffice for universities and might suffice for funding agencies. There are still too few funder policies in effect to see the results of different approaches.
The CURES Act is one of the only funder policies to specify a sanction. It says that grantees who do not comply with the policy, including the six month limit on embargoes, risk the loss of future funding. (The policy says that non-compliance "shall" be considered during grant compliance review and "may" result in "the loss of public funding".) There are at least three forms of this sanction. The mildest is to prefer compliant to non-compliant applicants in case of a tie. For example, if the funding agency has to choose between funding two researchers, A and B, when their projects are of equal scientific merit, when both have received grants in the past, when A complied with the OA policy on the previous grant and B didn't, then the funder will choose A over B. Indeed, funders who don't prefer A to B under these circumstances make their policies meaningless, whether they use requests or requirements. A stronger form of the same sanction is to deny B a new grant, period, even if not tied with an applicant like A. The strongest form of this sanction is to ask non-compliant grantees to repay all or part of their grant.
FRPAA and RCUK are silent on sanctions. The NIH policy, to emphasize that it is not a mandate, explicitly says that the agency "does not penalize" non-compliance with its request and that non-compliance "will not be a factor" in reviewing a grantee's work. These provisions deliberately retreat from enforceability and must explain part of the NIH's very low compliance rate.
* Lesson 10. The legal basis of the funder's dissemination of these texts should either be a government license (for public funders) or the funding contract with the grantee (for public or private funders) . The policy should not rely, directly or indirectly, on publisher consent.
There's a fundamental distinction between (1) rights that publishers might possess under copyright law and (2) the right of publishers to refuse to publish anything for any reason, or more broadly, to put conditions on what they will publish. Let's call the first publisher's copyright and the second publisher's discretion. Publisher's discretion does not derive from copyright law. But when publishers make demands, e.g. for an embargo of a certain length, it's not always clear whether they are speaking from their rights as copyright holders or their right of discretion. But it's very important for funders to become clear about this.
Funders have to live with publisher's discretion. It's not only a reality beyond their control. It's also a good thing that none of us should want to restrict. But it doesn't follow that funders have to defer to publisher decisions as if this deference were required by copyright law.
Publishers who dislike a funder OA policy can use their discretion to resist, but if the OA policy is well-written, they cannot use copyright to resist. They can use discretion to resist by refusing to publish work by authors bound by a funder's OA mandate. It's significant that so far, no publishers have taken this stance with regard to any funder OA policy.
Funders may have mission-related reasons to defer to publishers, for example because they believe, correctly or incorrectly, that some OA policies might harm publishers and that harm to publishers would harm science. Funders may have political reasons to defer to publishers because publishers are a significant lobbying force. Or funders may want to protect their grantees from adverse publisher decisions under the power of publisher discretion. But these reasons to defer to publishers should never be confused with a copyright-based legal requirement to defer to publishers.
Grantees sign funding contracts before they sign copyright transfer agreements with publishers. Funders are upstream from publishers and have the right to insist that their contracts be enforced, with the consequence that a grantee's subsequent publishing contract is subject to the terms of the prior funding contract. The Wellcome Trust OA policy is the clearest and strongest of all of the funder mandates to date because it is the only one to recognize this simple truth. (The Wellcome Trust considers its grants to be closer to gifts than contracts, but that doesn't affect this analysis because the WT regards the terms of its grant to be enforceable.)
If a funder lets authors decide the length of the embargo, then in practice most publishers will take the decision from authors, using their power of discretion. So even funders who don't think they are deferring to publishers, but only to authors, are deferring to publishers.
If funders let the length of the embargo depend on author or publisher decisions, then most publishers will not only take the decision from authors but press for the maximum embargo. I've documented this in the case of the NIH policy.
http://dash.harvard.edu/bitstream/handle/1/3967550/suber_news86.html#nih (June 2005)
http://dash.harvard.edu/bitstream/handle/1/3967549/suber_news87.html#nih (July 2005)
A funder's OA policy must acknowledge publisher discretion, but must allow no *other* kind of publisher resistance to interfere with the OA policy. After funders decide what they want, taking their mission-related and political relations with publishers into account, they should require it. If they make the requirement depend on the consent of third parties, they simply invite third-party interference. Funders who make this mistake actually *give* publishers the power to resist the policy as a matter of right, not just as a matter of discretion. But even then, the right is created by the funder policy, not by copyright law.
The US has enacted a "government-purpose license" allowing federal agencies to disseminate the results of agency-funded research. The NIH acknowledged its existence but deliberately decided not to use it, relying instead on publisher consent. Both CURES and FRPAA fix this mistake, expressly rely on the government-purpose license, and make publisher dissent irrelevant. The Wellcome Trust bypasses publisher consent with the simple but sufficient device of holding grantees to their funding agreement.
If any major policy is worse than the NIH's on this score, it's the new RCUK policy. The 2005 RCUK draft mandated OA "subject to copyright or licensing arrangements". The final version of the policy restates this vague deference to publishers at least as vaguely and even more deferentially: "Full implementation of these requirements must be undertaken such that current copyright and licensing policies, for example embargo periods or provisions limiting the use of deposited content to non-commercial purposes, are respected by authors." Both formulations subordinate the funder's interests to the publisher's interests. Both suggest that funders are required by copyright law to roll over for publishers in this way, which is incorrect and harmful. If funders are really worried only about running afoul of copyright law, and not publisher's discretion, then they never would subordinate their own funding contract to *subsequently* signed publishing contracts. I say more about the RCUK's position on this in last month's newsletter.
In countries that lack a government-purpose license, activists or agencies should work with the legislature to create one. The gist of it is that government funding agencies should have a right to disseminate the results of agency-funded research, in print and online, without permission from any other party. This is a clean and easy way to make publisher consent unnecessary. Note, however, that the Wellcome Trust's contractual approach works just as well without any legislative action.
Finally, legislatures laying down OA policies for public funding agencies should understand that they can mandate OA without amending copyright law. There are many excellent reasons to amend copyright law (don't get me started), but it's not necessary for the limited goal of OA. Moreover, copyright reform --of the right kind-- has more opponents than OA, and turning to it simply to achieve OA is politically perverse and fatal. Copyright reform is still very desirable, but it is without doubt the most difficult and least propitious way to achieve OA. And fortunately for OA, we don't have to wait for it. We don't have to fight its well-funded opponents, live with its compromises, or endure its delays.
The NIH uses publisher consent instead of a government-purpose license, even though it has such a license at its disposal. CURES and FRPAA use government-purpose licenses rather than funding contracts or publisher consent. CURES and FRPAA also make clear that they do not modify US copyright law. These provisions are lessons learned from the Sabo bill, which would have amended copyright law to put publicly-funded research into the public domain. The Sabo will was defeated in part because it aroused a well-funded opposition and in part because most friends of OA were not willing to support it in that form. The NIH and RCUK policies are the products of agencies, not legislatures, and lack the power to amend copyright law, although they needlessly defer to publisher decisions as if copyright law required it. The Wellcome Trust is a private funder and also lacks the power to amend copyright law; but instead of deferring to publishers under a mistaken interpretation of copyright law, the Wellcome Trust understands that a clear and enforceable funding contract can bypass all copyright problems.
* Postscript. In addition to the five policies covered above, here's my snapshot of other OA policies for public funding agencies, organized by nation.
Canada: Canadian Institutes of Health Research (CIHR)
--Developing a policy, will be a mandate
--April 2006 solicitation of comments
--June 2006 progress report
Canada: Canadian Breast Cancer Research Alliance (CBCRA)
--Not a mandate yet but may become one; see below in the Top Stories section
Canada: International Development Research Centre (IDRC)
--At least encouragement; not clear whether it's also a mandate
--December 2005 statement
Canada: Social Science and Humanities Research Council (SSHRC)
--Encouragement, not a mandate
--October 2004 statement
--April 2006 statement
--Report with OA policy recommendations (dated January 2006 but not apparently released until March 2006)
--My SOAN article on it
Finland: Finnish Ministry of Education
--Encouragement, not a mandate
--March 18, 2005 statement (English-language version)
France: Centre National de la Recherche Scientifique (CNRS)
--Encouragement, not a mandate
--June 21, 2006 letter asking CNRS unit heads to invite their researchers to self-archive
France: Institut national de la recherche agronomique (INRA)
--Mandate (all department research "is to be self-archived")
--April 3, 2004 statement
France: Institut national de recherche en informatique ete en automatique (INRIA)
--Encouragement, not a mandate
--May 30, 2005 statement
Germany: Bundesrat or Parliament
--A bill to allow author self-archiving of publicly-funded research six months after publication regardless of any agreement the author might have signed with the publisher, introduced May 2006
--My SOAN article about it
Germany: Deutsche Forschungsgemeinschaft (DFG)
--In between a request and requirement; grantees "should" (not "must") deposit work in an OA repository
--January 2006 policy
--My SOAN story on it
South Africa: Academy of Science of South Africa (ASSA)
--March 2006 report recommending both green and gold OA, but not a mandate
Sweden: the National Co-ordination and Development program (BIBSAM) of the National Library of Sweden
--BIBSAM coordinates a national OA initiative to promote both green and gold OA in Sweden (not a mandate), launched May 2006
--Parliament recommended an OA condition to publicly-funded research, December 2005
To supplement this list of policies and proposals at public funding agencies, see the list of funder policies from JULIET
...and the OA policies at universities and research centers from ROARMAP.
Read this issue online
SOAN is published and sponsored by the Scholarly Publishing and Academic Resources Coalition (SPARC).
Additional support is provided by Data Conversion Laboratory (DCL), experts in converting research documents to XML.
This is the SPARC Open Access Newsletter (ISSN 1546-7821), written by Peter Suber and published by SPARC. The views I express in this newsletter are my own and do not necessarily reflect those of SPARC or other sponsors.
To unsubscribe, send any message (from the subscribed address) to <SPARC-OANewsfirstname.lastname@example.org>.
Please feel free to forward any issue of the newsletter to interested colleagues. If you are reading a forwarded copy, see the instructions for subscribing at either of the next two sites below.
SPARC home page for the Open Access Newsletter and Open Access Forum
Peter Suber's page of related information, including the newsletter editorial position
Newsletter, archived back issues
Forum, archived postings
Conferences Related to the Open Access Movement
Timeline of the Open Access Movement
Open Access Overview
Open Access News blog
SOAN is licensed under a Creative Commons Attribution 3.0 United States License.
Return to the Newsletter archive