Academia_screenshot_no may need to change the second step in its sign-up process given publisher concerns regarding posting of research articles without rights holder permission

The Chronicle of Higher Education reports that Elsevier has issued a sweeping series of Digital Millennium Copyright Act (DMCA) take down notices regarding Elsevier-published content to, a file-sharing network for researchers and other academics.

This has prompted a storm in the Twittersphere, a response from Elsevier, a number of commentaries on blogs and list-serves, and a truly bizarre article from CNET that casts as a “new school” “digital era” “publisher” and rival to Elsevier (who is couched as an “old school” “traditional company” – which just incidentally owns and operates a platform very similar to for its part is reportedly encouraging authors of affected papers to sign this Elsevier boycott petition despite the fact that their own terms of use prohibit the posting of content that infringes on the copyright or license of publishers such as Elsevier.

Is this a footnote or the end of a chapter in the annals of digital science publishing? While it is too soon to know to what extent this incident will have broader implications, it does bring some important and long-simmering issues to the fore regarding sharing on professional networks, most of which the Chronicle article manages to miss.

Fortunately, the industry’s many pundits rode to the Chronicle’s rescue in the article’s comment section, which is worth reading in its entirety, and which better frames many of the issues involved. One particular exchange in the comments gets to the heart of the matter, but before we get to that a bit of background is in order.

It perhaps goes without saying that authors who publish with Elsevier and most other publishers sign publication agreements that transfer copyright or an exclusive publication license to the publisher and specify clearly what authors can and cannot do with their paper after publication. Generally speaking, publishers tend to frown upon systemic distribution and commercial reuse as such activities undermine their business models. Most publishers explicitly permit one-to-one or private group sharing, which may include emailing a paper to a colleague, using a paper in the classroom or a conference presentation, or other similar uses. Given that this is academic publishing, most publishers’ policies delineate educational or other noncommercial uses from commercial uses, allowing many of the former and restricting the latter without explicit permissions.

One can argue that authors should not sign such publications agreements. One can further argue that all researchers should publish only in journals that use CC-BY agreements, allowing unfettered systemic distribution and even commercial reuse. At some point in the future such arguments may win the day, but the reality today is that most researchers do sign such agreements and are legally bound to honor them.

Despite the fact that it is a form of (noncommercial) systemic distribution, publishers have, for the most part, become increasingly accepting (to greater or lesser degrees and up to a point) of Green Open Access (Green OA), whereby authors post PDFs of their papers on personal websites, institutional archives, or central subject archives such as PubMed Central or the arXiv. In many cases, publishers explicitly grant such permission or else do not enforce copyright or exclusive license provisions. In some cases publishers allow the final article of record to be deposited in such archives; in other cases only the accepted manuscript can be deposited. In some cases a delay of up to 12 months is required before repositories can make articles publicly available (though there is at least one proposed work-around for that which creates a one-to-one distribution model for such repositories), in other cases such deposits can be made public immediately., however, is not an institutional repository. Nor is it a subject-based repository like PubMed Central. Nor is it a noncommercial pre-print server like the arXiv or the new bioRxiv. Nor, despite its “.edu” domain address, is it a not-for-profit academic initiative of any kind. is a venture capital-backed software company that seeks to derive revenue by selling analytics about the activities of its installed user base, much like Facebook, LinkedIn, Mendeley, and many others. Central to its success, however, is the sharing of papers and the metrics around that activity. Or at least one might suppose it is central given this activity is enshrined in the organization’s tag line: “Share Research.”, however, likely does not have the legal right to host much of the research that is being shared, both systemically and for its own commercial purposes, on its network. So how has managed to attract funding when its business model hinges on being able to share content that it may not have the right to share? Part of the answer lies in the “safe harbor” provision of the Digital Millennium Copyright Act that limits the liability of companies like (and YouTube, Facebook, etc.) that often host content under copyright to third parties (like publishers) provided the content was posted by their users and provided they respond to take down notices from rights holders.

So why did and their funders think that publishers would not aggressively issue such take down notices? One possible explanation might be found in the exchange from the comments section of the Chronicle article mentioned above. The exchange takes place between a commenter named “N.W.J.”, Amanda French of George Mason University, and William Gunn (a.k.a. “Mr. Gunn”) of Mendeley. I have reposted the comments below:


William Gunn puts forth an argument that Mendeley has long used for hosting content they do not have explicit permission from rights holders to host: that an individual’s profile page on Mendeley (or by extension, is that individual’s “personal website” and therefore covered under the exemption that many publishers provide to authors in both copyright and exclusive licensing agreements, allowing authors to post PDFs of their work to their own personal or institutional (e.g. their laboratory or departmental) website. Since papers are loaded to both Mendeley and by authors, the argument is that the paper is “self-archived” to the author’s personal website.

Leaving aside the fact that many publishers permit only the author’s accepted manuscript (and not the final PDF) to be self-archived, were one to accept this argument, by logical extension this would mean any commercial site (, ResearchGate, Facebook, LinkedIn, Scribd, Google, etc.) that sets up a profile page would have the right to host any research loaded to the site by an author. The distinction between a profile page on an academic or professional network and a personal or institutional website does not strike me as difficult to make and if it were ever an open question, as Mr. Gunn asserts, the question seems to have just been answered by the legal department of his own employer.

Many publishers, including Elsevier, have let this argument slide (or at least have not aggressively issued take down notices) for many years now because academic networks, including Mendeley,, ResearchGate, and others were nascent and publishers wanted to see how they would grow and mature, whether viable business models would emerge, and whether the networks would ultimately have anything to offer publishers (user intelligence, licensing revenues, marketing channels, etc.).

With Elsevier’s acquisition of Mendeley earlier this year and recent escalation of take down notices to, we appear to have moved beyond the nascent phase (a more cynical observer might call it the “Napster phase”) of such networks and into a phase where they must be more rigorous with regard to content rights. This also means Elsevier’s own Mendeley will need to do same or face take down notices from the rest of the world’s STM publishers.

Of course this could just be a business tactic of Elsevier to create friction for competitors of Mendeley. Elsevier publishes a vast amount of content and I do not imagine they will issue take down notices to themselves if authors upload their own Elsevier-published papers to Mendeley. by contrast (and despite what CNET’s reporter seems to think) is not a publisher and therefore can’t retaliate in kind giving a potential advantage to Mendeley.

And while perhaps at a competitive disadvantage, this does not mean and ResearchGate will have to close up shop. First, are savvy enough that they could easily employ safeguards that enable the posting of PDFs where authors have systemic distribution rights (such as those published under CC-BY licenses or from publishers with liberal sharing policies). For those papers that are under copyright protection by a publisher, there are a number of options for sharing articles that do not involve uploading illicit PDFs.

Instead of hosting article PDF files themselves, such networks can instead link, via OpenURL and an integration with institutional libraries, to the publisher’s website, thereby providing seamless access to full-text content via the user’s institution. Indeed, and to its credit, Mendeley appears to already be doing this (though I have not seen evidence of their also scrubbing the site of previously posted PDFs nor of preventing users from posting PDFs they do not hold rights to). This may even lead to services and business models designed around institutional customers with workflow solutions tailored to departments, laboratories, classrooms, and other workgroups. Such solutions may actually be more lucrative than selling analytics about the activities of a self-selected user base to third parties. 

In cases where a paper is under third-party copyright protection, instead of hosting PDF files themselves, academic networks might provide links to institutional repositories. Authors can self-archive whichever version of the article (final or accepted manuscript) is stipulated in their publishing agreement to a bona fide personal website or to an institutional or subject matter repository and then provide links from the academic network directly to the archived PDF (from a user perspective, there is no difference as to whose server hosts the file). Granted in many cases these would be accepted manuscripts and not the final version of record – though the final version could also be linked to via the publisher’s website (DOIs should make this fairly easy to do).

Academic networks might work out licensing agreements or other in-kind arrangements with publishers. They might, for instance, trade analytics in exchange for rights to serve publisher content to those without institutional access. Alternately, academic networks could limit access to PDFs to those directly connected with the researcher or individuals who explicity request such content – thereby changing from a systemic to a one-to-one distribution model. 

Academic networks might further work out a deal with publishers that enables authors to “upgrade” their papers to Gold OA on the fly by paying an author publication fee (APC) at the point of sharing. The academic network might retain a commission for such transactions. Or, they might work out an integration with an article-rental service such as DeepDyve (or set up their own), which would provide another revenue source. 

Despite the criticism Elsevier has received from OA advocates, there may be more than one silver lining here for the OA movement. If et al. begin linking to institutional and subject repositories instead of hosting PDFs, this will likely increase rates of self-archiving if such archiving becomes a necessary preliminary step for some forms of sharing. Such archives are more stable than the servers of a start-up and are likewise more readily indexed. The inability to systemically distribute their published research papers with no restrictions may also prompt authors who value such rights to seek out Gold OA titles or journals with Gold OA options, thereby providing a market-based solution that respects intellectual property rights while providing authors with clear choices. If enough authors value systemic and unfettered distribution rights, this will in turn increase the prevalence and profile of titles that offer such rights. In other words, were more publishers to enforce their intellectual property rights with respect to academic networks, it may well lead, over the longer term and if the market wants it, to more uptake of both institutional self-archiving and Gold OA publication. 

Michael Clarke

Michael Clarke

Michael Clarke is the Managing Partner at Clarke & Esposito, a boutique consulting firm focused on strategic issues related to professional and academic publishing and information services.


34 Thoughts on "The End of an Era for and Other Academic Networks?"

In reading the comment from Elsevier (linked in the article) it would be helpful perhaps to hear the extent to which they tried a ‘softly softly’ approach. One hears the words “Take down Notice” and immediately tends to associate that with the very worst excesses of rights assertion. And frankly, not without good reason (take downs of NASA public domain video springs to mind). But here, it would appear that this was an act of last resort. The phrase “user friendly options for alignment” used by Elsevier could benefit perhaps from an expansion on what that meant… Anybody from Elsevier care to comment?

[Obvious disclaimer: I’m not “anybody from Elsevier” :-)]

I’ve got to admit I found the phrase “user friendly options for alignment” more than a little sinister. Surely when one party is the customer and the other party is a service provider, “alignment” can only mean the service provider providing the service the customer wants. Many (not all) of the problems of academic publishing come from the unhealthy shift of power from customer to provider.

Your last six paragraphs, while speculative, clearly illustrate the fluidity and complexity of the present publishing industry. This is what technological revolutions are like at the street level, where every action creates new possibilities. Confusion is the price of progress.

This is a very nice piece, Michael. I am particularly intrigued by your speculation at the end that Elsevier’s practice could in time make Green OA deposits more valuable. I suspect that the purchase of Mendeley was in part motivated to make Green OA irrelevant. We shall see which way the world turns.

I fear that Mendeley will be the next target for Elsevier…

to gain control over it: only deposit publisher’s pdfs from Elsevier if you paid the OA fee; or only have access to these papers if you subscribe (not for free) to the service

Removing an article from one open web source/site doesn’t remove all of them and I would argue that many users who want access to an article could care less about the url from where they get it. Perhaps the should but that’s another issue.

The Leonard article that was taken down last week and received a lot of attention remains available via several urls (easily discoverable from search engines) including ResearchGate. I’m not sure if they’re the final, published versions but some researchers might not care.

If/when these are taken down they would also need to be removed from the Internet Archive and other web archives.

Any researcher who had a link to the copy simply has to redirect the url to one of the other sites.

Until the issues discussed in the article are resolved (perhaps as suggested by the author of this blog post) this entire situation will grow larger and more heated.

Does anybody know how alternative copies of articles affects Impact Factor? Does Thompson Reuters account for multiple versions on different sites?

How, for example, should one cite a paper one reads on Does the cite point to the official copy, e.g. by using the DOI, or do you point the reader to the entry? As more and more cites become links, this is an important question.

The citation should always remain to the actual publisher version. Citations should remain unaffected as there is only one registered DOI for an article. ISI requires the correct first author, year, and/or page number and DOI.

As long as the uploaded version, either the version of record or the author’s accepted manuscript, contains the proper citation information for the version of record in the journal, it shouldn’t have an effect on the Impact Factor. If there are multiple version out with different or incomplete information, then it may reduce the number of citations attributed to the official count of Thomson-Reuters.

Having correct citation information is a big assumption for the alternative version. If it’s the author’s final draft, they don’t know the DOI, the page number, or even the volume/issue at the time its submitted. In a sense I’d be better off letting them post the PDF from the copy of record.

Each misdirected cite costs my journal about 0.004 off its IF.

I think you’re overstating the difficulty here. The author submits their paper to the journal, it gets accepted and published. They then deposit the accepted manuscript version of their paper in a repository, and include a link to the final version of the paper along with the full reference. Most publishers request this in their term for reuse of the manuscript in this manner, and many repositories similarly appreciate the inclusion of such information.

Something that’s not mentioned in your piece is the air of deception that surrounds many of the activities of these social media startup sites. I know I receive several emails a week from ResearchGate, some telling me that someone I don’t know from a university where I was 15 years ago published a paper in a field unrelated to mine, but all containing an exhortation to “add your publications…And introduce your research to a new audience.” This is usually accompanied by suggestions that all of my colleagues and coauthors are uploading their papers, which provides an air of legitimacy.

As you note, (along with ResearchGate and Mendeley) explicitly state in their terms of service that by uploading an article, you are vouching that you have full legal rights to do so, and that you absolve the host site from any legal responsibility. These terms though, are generally hidden several clicks away from the act of uploading the article. ResearchGate has a small and vague “View conditions for uploading files” link (does this refer to file sizes and formats or legal requirements?) that most are not going to click on. seems to have no such notice during the upload process.

Clearly it’s in the best interests of these networks to do the minimum necessary to meet legal requirements, while encouraging as many uploads as possible. They have to carefully toe the line to avoid explicitly encouraging infringement (, but it seems from’s campaign here, it’s being done implicitly with a wink and a nudge.

In the age of the 100-plus page EULA, it’s rare for a user of any online service to actually read the entire set of terms and conditions required. Researchers may wish to exercise caution though, to avoid potential legal jeopardy. While I can’t see any scholarly publisher following the RIAA’s example and suing authors directly, it would seem a concern that a researcher exposes himself to for seemingly little gain. As you note, there are many pathways to offering access to one’s work that don’t require the same legal risk.

When you store a personal PDF in Mendeley, there is a tiny little check box in the bottom corner for opting out of putting that paper in their accessible collection. It is, of course, automatically unchecked. So users can very easily load content without even knowing it.

Angela, that check box is for syncing documents to other computers you use Mendeley from. Posting your own PDFs to your profile is different and we give an excess of information at the point of posting.

I wonder how closely’s lawyers have studied the Grokster case? I wouldn’t be at all surprised if Grokster’s site had a similar disclaimer notice about the liability for posting content not owned by the poster. Yet that tactic didn’t save Grokster from being shut down by a unanimous ruling of the Supreme Court, which found Grokster to have been engaged in the active encouragement of infringing behavior.

Authors usually want their work to be discovered and appreciated which is what these services imply will be enhanced by posting. Assuming that impact metrics are a part of this desire, driving traffic to the location of the publication of record would be an important goal. Posting an abstract, précis or summary of the paper with a link to the publication of record might produce a win/win type compromise. Of course that wouldn’t obviate squabbling over how much is “given away.”

It’s worth noting that Elsevier (well, actually a service they hired) sent take down notices for published PDFs on authors’ site hosted by their university. Yes, they are enforcing both their legal rights and their posted policy. But it isn’t just academic networks receiving these.

Fascinating, Dan. Do you have any data on this? Could we be seeing publisher push-back, something that should not surprise the OA folks. Industries do not usually roll over when activists try to roll them.

Many publishers do this as many publishers do not allow authors to post the final PDF, the one the publisher copyedited, tagged, added links to and made more readable. Many times, authors are allowed to post accepted manuscripts, sometimes with an embargo. This is usually detailed in whatever copyright transfer form an author signed. Authors usually do not have the right to post the publisher version of the paper on the open internet.

(Please see my response to a comment above.) Accepted drafts do not contain the necessary citation information. After publication, the author would need to look up the copy of record to get the DOI, page numbers, etc. Would an author really have the skill and interest to do this?

How would an author ever cite their own papers, or even include them on their CVs or grant renewals if they lack the skill to properly figure out the correct citation? Adding the citation to your CV, website, reference list, etc. is one of the first things an author does when a paper is published. It’s trivial to do so. You can’t be serious here. Are you suggesting that scientists can’t figure out how to cite an article? How does any article ever get cited then?

I’m not suggesting scientists don’t know how to cite. However, they don’t have the necessary information (DOI, pages, etc.) until the copy of record goes online. I am suggesting they may not have the time or inclination to add this info to the alternative copy, especially if they already have posted it elsewhere.

I know counting cites for individual authors is a dodgy business, but, from the author’s standpoint, it matters little whether the cite is to the official or the alternate copy. It matters a lot to the journal. Misdirecting a cite, thus lowering the journal’s IF, effectively steals credit from all the authors who publish with that journal.

Again, I’m not sure it’s all that difficult. Really, this is only a problem for an author who uploads his/her paper to a repository before publication occurs, perhaps more an issue for preprint servers than for repositories. Not sure that most authors are in such a rush, and given the relatively short time these days between acceptance and publish-ahead-of-print for most journals, you’re talking about a matter of weeks.
To quote OUP’s policy on the matter for preprints:

…provided that where possible they acknowledge that the article has been accepted for publication as follows:
This article has been accepted for publication in [Journal Title] ©: [year] [owner as specified on the article] Published by Oxford University Press [on behalf of xxxxxx]. All rights reserved.

And postprint reuse:

When uploading an accepted manuscript to a repository, authors should include a credit line (see below) and a link to the final published version of the article. This will guarantee that the definitive version is readily available to those accessing your article from public repositories, and means that your article is more likely to be cited correctly.

Citation counts are increasingly important to authors. There’s movement away from journal based metrics to article based metrics. Citation is still an important part of altmetrics as well. I know many academics who include citation counts of their articles as part of their review process. Since there’s no unified way for an author to track citations to repository versions of an article, it’s in their best interest to direct all citations to a trackable source.

OMG, I have never met an author that did not know how to cite their own paper.

Again, I’m not saying authors don’t know how to cite themselves. In theory, should have zero cites if everybody who reads and cites the papers there uses the correct form for the copy of record. But, has anybody actually followed up to see this is the case?

In a sense, if you read (and perhaps quote) a paper on, then citing back to is more correct, since the unofficial copy — i.e. the one you actually read — is NOT necessarily identical to the copy of record. Multiple copies, unless carefully synchronized, always present a citation risk.

If an author posts their accepted manuscripts, we require that they wait for publication (8 weeks or so after acceptance), that they include the proper citation, and that they provide a link to the version of record on our journal site. I don’t have time to police that so we are trusting that this happens. We do ask websites to remove the final PDF version of our papers.

Comments are closed.