Evolving the correction of the literature : manuscript versioning , error amendment , and retract and 1 replace 2 3

8 Academic publishing is undergoing a highly transformative process, and many rules and value systems that 9 were in place for years are being challenged in unprecedented forms leading to the evolution of novel ways 10 of dealing with new pressures. One of the most important aspects of an integrated and valid academic 11 literature is the ability to screen publications for errors during peer review to weed out mistakes, fraud and 12 inconsistencies, such that the final published product represents a product that has value, intellectually, and 13 otherwise. It is difficult to claim the existence of perfect manuscripts. The level of errors that exist in a 14 manuscript will depend on the rigor of the research group, as well as the peer review that screened that paper. 15 When errors slip through into a final published paper, either through honest error or misconduct, and are not 16 detected during peer review and editorial screening, but are spotted during post-publication peer review, an 17 opportunity is created to set the record straight, and correct it. To date, the most common forms of correcting 18 the literature have been errata, corrigenda, expressions of concern, and retractions. Despite this range of 19 corrective measures, which represent artificially created corrals around pockets of imperfect literature, 20 Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 8 August 2017 doi:10.20944/preprints201708.0029.v1 © 2017 by the author(s). Distributed under a Creative Commons CC BY license. 2 certain cases do not quite fit this mold, and new suggested measures for correcting the literature have been 21 proposed, including manuscript versioning, amendments, partial retractions and retract and replace. A 22 discussion of the evolving correction of the literature is provided, as are perspectives of the risks and benefits 23 of such new measures to improve the literature. 24 25

2017b).However, this militarization of the publishing system, with endless checks and balances (Teixeira da Silva, 2016a) is fraying authors' rights (Al-Khatib and Teixeira da Silva, 2017a).Some of the most prominent changes include whistle-blowing, with PubPeer 3 leading the way, public shaming by Retraction Watch 4 to expose cases of misconduct, errors and retractions, and the expansion of the concept of journal clubs, including via online channels, to expose errors and fraud and expand the debate on these issues with the ultimate objective of correcting the literature (Teixeira da Silva et al., 2017).The effect of these changes are augmented by social media sites that allow error, and scandal, to reach an unlimited audience within a very short space of time, causing, in negative cases, devastating and irreparable consequences, and the fatal termination of legends, sometimes within months (Teixeira da Silva et al., 2016), tearing down decades of career-building developed on a publication base of sand.This entire evolving PPPR is not without its biases, and there are many vested interests in creating new publishing models that can render already published as well as newly published literature as free of error as possible.In the case of for-profit publishers, a clean image and error-free literature represents a product with increased market value, whether that be a pay-per-view or pay-per-download article, or a journal subscription.Similarly, for academic journals published by purely academic societies, such literature represents a journal and society with integrity, that can be trusted, thereby attracting new membership, or maintaining a regular stable of members.Therefore, except for unscholarly and/or "predatory" publishers, whose objectives are largely not academic, or whose objectives are overly commercial, the correction of the literature is in the interests of all 3 https://www.pubpeer.com/parties involved.
However, the increase in retractions, as is being documented to some extent at Retraction Watch, and observed across some of the most powerful and profitable mainstream publishers, indicates that failure in peer review and the current publishing model has been widespread (Teixeira da Silva and Dobránszki, 2015a; Teixeira da Silva, 2016b), causing reputational damage to authors, editors, journals and publishers alike.Reducing the risk of erroneous literature would involve measures that should be implemented long before the publication process during the development of an academic prior to the publication of their first paper, such as rigorous reproducibility testing, the valorization of work based on its content and not on the venue where it has been published, as well as a host of other measures and checks and balances that lie beyond the scope of this paper, but some of which were discussed more extensively recently (Teixeira da Silva and Shaughnessy, 2017).This paper assumes that such preventive measures were not taken, leading to the publication of erroneous literature.

Current models to correct erroneous or fraudulent literature
An ideal literature would be error-free.However, since this is not possible, and is also likely not realistic because authors make errors, because the peer review system is limited to screening by a very limited set of eyeswhen it is in fact conducted at allthat are not financially compensated for their professional activity (Teixeira da Silva and Katavić, 2016), and because productivity and growth are not always compatible with high quality, it requires astute readers and proactive PPPR participants to step forward to present these errors.
Even when evidence is presented, however, there is tremendous resistance to correcting the literature, even among member journals and publishers that claim to follow COPE (Committee on P ublication Ethics) editorial-and publishing-related ethical guidelines, which is reflective of an organization that has been serving the interests of the for-profit publishing industry but has now been met with challenges it had not anticipated, revealing COPE's limitations, weaknesses and lack of accountability (Teixeira da Silva, 2017c).
This imbalance between what requires correction and what has not been corrected, either because of a lack of transparency, a lapse in editorial or publisher responsibility, legal impasses, and a wealth of other factors, makes it clearly apparent that it is currently much easier to publish than to correct the literature (Teixeira da Silva, 2017d).
It is not helpful that retracted papers continue to be cited, indicating that the downstream processes within academic publishing that are meant to circumvent such situations, i.e., the citation of invalid literature, are failing, or are not robust enough to detect, and prevent the citation of, retracted literature (Teixeira da Silva and Bornemann-Cimenti, 2017;Teixeira da Silva and Dobránszki, 2017a, 2017b, 2017c).It is therefore not surprising to note that: a) the wording used to define four main categories of literature correction policies, namely retractions, expressions of concern (EoCs), errata and corrigenda in almost 90% of 15 leading science, technology and medicine (STM) publishers differ from the definitions provided by COPE; b) as much as 61% deviation in both the wording and meaning of policies between these STM publishers exists (Teixeira da Silva and Dobránszki, 2017a).It can therefore be concluded that as it currently stands, the correction of the literature is ineffective because: a) there is variation in the will, desire and effort to correct erroneous literature by authors, editors, journals and publishers, even among STM publishers that are COPE members; b) the policies and notices that had, to a large extent, been drafted by COPE and then applied by leading STM publishers, were incomplete and insufficiently robust to deal with many of the challenges related to the age of whistle-blowing (PubPeer) and public shaming (Retraction Watch) that started in 2010-2012, then strengthened and became mainstream in 2015-2017.What became evident was that any association with these sites, and with correcting the literature, was, to some extent, an automatic association with bad science, fraud, or misconduct, even if in fact plain or innocent errors had been made.
Close examination of the parties behind this name-and-shame trend and campaign behind exposing fraud, misconduct, errors and faults in the published literature will lead to a close link to the John Arnold "war on bad science" 5 .Thus, willfully, or by association, there is a negative stigma associated with correcting the literature.Proponents of the STM publishing industry, alarmed by this trend, and shaken by the reputational damage it has been causing in just a few years to their publishing models, but unable to effectively reign in fraud and misconduct, despite all of the new measures being put into place described in the section above, have scrambled to find effective solutions to soften the tone related to correction of the literature.Part of this realization has to do with controlling reputational damage, attempting to minimize the publishing crisis 6 , and making correction of the literature a politically correctand thus marketing-friendlyprocess, rather than an aggressive affront to all parties involved in the creation and publication of erroneous literature.It is within this highly volatile environment that the new models discussed next, emerged.

Experimental publishing models, partial retractions and mega-corrections
5 https://www.wired.com/2017/01/john-arnold-waging-war-on-bad-science/One of the reasons for an associations with the Retraction Watch -PubPeer -LJAF alliance involves the lack of reproducibility associated with flawed literature.Although solving the reproducibility crisis is beyond the scope of this paper, it is worth noting that some measures in academic publishing evolved to try and deal with issues related to reproducibility, improved accountability, and openness of the publishing process.The first is PPPR publishing, in essence blogs, which would give free reign to any academic to launch their own "journal" that is independently controlled, although excessive self-published journals or blogs could potentially drive publishing into a chaotic state since there would be the lack of central regulatory bodies to oversee quality.Currently, a fairly centralized anonymous PPPR platform exists, PubPeer, but it is self-moderated, while a signed PPPR platform is available, PubMed Commons7 .Publons8 serves as a hybrid PPPR platform, either with anonymous or signed comments on published literature.To deal with the issue of reproducibility, and thus theoretically reduce the need to correct the literature, would also involve the need to publish negative results (Teixeira da Silva, 2015b), which are currently not given importance by most mainstream STM journals, either because they are not glitzy enough or due to journal page restrictions, even though negative results form an important aspect of the bulk of most biomedical research studies.
In cases where multiple corrections are made to a manuscript, in some cases more than a dozen, but not enough to merit a retraction because the main findings supposedly remain intact, the publisher may opt to issue a mega-correction 9 .However, the subjective nature of editorial independence and differences between individual characterization about what constitutes a large or a mega-correction, impedes the wide, standardized or effective use of this category of correction.This ambiguous form of correcting the literature also opens up amphibious, flexible and odd rules, such as the "acceptable" level of plagiarism before it should be considered for retraction 10 .This flexibility in the interpretation of rules to correct the literature could lead to ambiguities as to the best course of action or how best to correct the literature when issuing a partial retraction or a mega-correction.The issue of whether to retract an entire paper because one table or one figure has been tampered with lies at the heart of the debate of whether partial retractions 11 have merit, or not.Those in favor of partial retractions claim that it allows readers to ignore only specific faulty or erroneous parts of a paper, while recognizing the remaining parts as valid or reliable.In contrast, those who disagree with partial retractions will argue that if one part has been found to be fraudulent, or false, given the interlinking nature of all parts of a paper, that the entire paper, as a whole, should be retracted because none of it can be relied upon.At the end of September, 2016, after a 10-year experiment and 42 partial retractions, the U.S. National Library of Medicine, which runs MEDLINE/PubMed, announced that it would no longer identify this form of correction, treating them instead as an erratum 12 .It is likely impossible to reconcile both parties (for and against partial retractions), leading some academics and STM industry-related parties to rethink corrections of the literature and propose novel suggestions, as discussed next.

New and emergent models to corre ct the literature : outline and potential risks
The new models for correcting the literature that will be discussed in this paper relate primarily to That case, which was quickly dubbed "Pizzagate" by Wansink critics, was initially published as a PeerJ preprint (van der Zee et al., 2017a), and quickly spread to the analysis of a wide swathe of the Wansink literature, revealing considerably more errors 16,17 , some of which have begun to be corrected, including one retraction 18 .However, one of the original authors of the team that originally analyzed Wansink's papers discovered more errors in the correction of two papers that had issued errata 19 , raising doubts about the competence not only of the authors, but also of the journal's editors.This case fortifies the notion that correcting the literature is essential, but that a more flexible system of correction is required, especially if more errors are encountered in the original paper, or in the published erratum.
The second case involves the retraction of a retraction in a Wiley journal 20 .In this case, the paper (Namazi and Kulish, 2016) was retracted after Nanyang Technological University (NTU) in Singapore requested the journal to pull it, indicating that proper ethical approval had not been obtained from NTU.The first author refuted this claim, indicating that ethical approval had been obtained from a non-NTU institute.
The publisher retracted the retraction notice, but the case has yet to be completed.The retraction notice is pay-walled, which is against COPE policies for retractions (Wiley is a COPE member) (Teixeira da Silva, 2015c).These two cases highlight how correcting the literature under the current existent structure is clearly inefficient and is unable to deal efficiently with complex cases.Barbour et al. (2017) propose referring to all corrections and retractions as "amendments", thereby removing the negative stigma associated with correcting or retracting faulty, erroneous or fraudulent literature.A priori, it is important to note that several of those who are putting forward these new models have massive invested interests, financially and otherwise.Barbour, who is the former COPE Chair (2013-2017) , and colleagues including Theo Bloom of the BMJ, Jennifer Lin of Crossref, and Elizabeth Moylan of BioMed Central, all industry insiders claiming to represent COPE in this reprint, put forth a system in which amendments that are made would be classified as "insubstantial, substantial, and complete".
Although the idea that a paper can be updated real time is not a novel ideathe concept already exists for books and preprintsit is the reasoning behind this new proposal that is alarming and of concern: themselves have failed to correct their own paper.bioRxiv, which is aware of these issues, has also done nothing to encourage the authors to address these errors and concerns, casting concerns about this preprint server.In fact, several serious concerns about bioRxiv already exist 22 .
In some cases, "pervasive" errors may change the final conclusions, and in such cases Heckers et al. (2015) suggest retracting and replacing the paper with a new version only if the authors are able to address those errors following reanalysis of their study.This model is a reasonable and realistic option for authors for the following reasons and provided that the following cautionary measures are followed: a) it allows authors to correct the literature, and set the record straight; b) it allows the journal and publisher to offer a new and fair opportunity to correct the literature by allowing for a fresh peer review; c) it assigns responsibility for error, both on the authors for creating the errors, as well as the journal and publisher for not detecting them during peer review and editorial quality control; d) the model is only good if the flawed "old" version remains, with a "retracted" stamped across it to indicate that it should not be used, and not simply replaced because replacing would erase a historic version of record.Point d) is extremely important and any editor or journal that attempts to use the "retract and replace" policy to hide misconduct by substituting the original study's files (html, PDF, etc.) with the "latest, correct" version, would be in fact committing publishing misconduct, and possibly fraud, by deliberately concealing the erroneous version of the paper and hiding its true publishing history.Such an action would be as unethical as silent retractions, where publishers hide the existence of a retracted paper, to save their own image, thereby depriving the public of knowledge of serious 22 https://medium.com/@OmnesRes/why-biorxiv-cant-be-the-central-service-24530d119c89 error, misconduct, and negligent editorial processing (Teixeira da Silva, 2016c).Using a specific case 23 to illustrate their model, the Heckers et al. (2015) suggestion is an excellent model that could serve the academic community well.It certainly would involve a considerable amount of additional work and investment by authors, the journal and publisher, as distinct DOIs (digital object identifiers) and PubMed entries are required, but ultimately with the correction of the literature in mind.Other papers have been "retracted and replaced" in JAMA Network journals as this new model of correcting the literature begins to take hold 24 .Two specifics not covered by Heckers et al. (2015) are: a) in such cases, where an article processing charge (APC) is levied for the "original" paper in an open access model, is this APC refunded, or, alternatively, is no new APC levied for the "retracted and replaced" paper?b) Will the authors and journal/publisher responsibly inform the authors and editors of any papers that have cited the "retracted and replaced" paper to correct their citations to reflect the new version (see link to discussion on downstream literature below)?
A similar concept to "retract and replace" introduced by JAMA Network journals was also instituted in 2015 by The Lancet and The Lancet Respiratory Medicine as "retract and republication" (The Lancet, 2015; Cagney et al., 2016).
There are some cases in which errors exist in a paper as a result of honest error, estimated to be about 20% of the retracted literature (Fang et al., 2012).In such cases, when errors are minor, a correction will suffice, but in cases where key or core conclusions are altered, such papers still need to be retracted.It is then incumbent upon authors, if they are still actively researching, to reanalyze their data and resubmit and republish their work.However, honest errors that are met with a retraction are still met with negative stigmatization associated with the retraction, in most cases by other researchers who might not know the background of the case, or who might not be sympathetic to the notion of "honest error".One classic case is a paper by Mann et al. (2012), which was retracted from PLOS Computational Biology, and then republished, after a fresh round of peer review, in the same journal in 2013 (Mann et al., 2013).One criticism of this case is that the article processing fees had to be paid twice, a business model that is unsustainable because it taxes and exploits the authors, or their institutes, both intellectually, and financially (Al-Khatib and Teixeira da Silva, 2017b).A negative chain reaction to retractions may be annulled when authors retract of their own accord (Lu et al., 2013).The concept of authors seeking to retract their own erroneous papers, especially those whose conclusions may be irrevocably changed, prompted the science watchdog Retraction Watch to term the action as "doing the right thing" 25 , although Retraction Watch displays extreme subjectivity when deciding which cases qualify for this category.To eliminate this bias, which can still be detrimental to those who are shamed by Retraction Watch for not doing the right thing, "self-retractions" (Pulverer, 2015;Fanelli, 2016;Hosseini et al., 2017) could be considered a new category of retraction that has a positive connotation, lauding the authors for seeking to correct the literature out of their own initiative.The problem here is who exactly would be in an unbiased position to ascertain whether authors were "doing the right thing"?Making an erroneous distinction between those that do the right thing from those that do not could amount to libel, or slander, and thus finding a set of individuals who would be competent and unbiased enough to make such an assessment would be complicated.The current five categories suggested by Fanelli et al. at the 5 th WCRIthese categories are likely to change as their paper is currently under reviewto replace retractions where misconduct has not been shown include: a) withdrawal ("this is a peer-reviewed paper in which the authors retract one or more of their previous publications based on presenting detailed new evidence, data, methodologies, results or theoretical arguments that invalidate previously published claims"); b) retired ("a guideline or recommendation article is retired when its content is deemed outdated and its authors are unable to update them"); c) cancelled ("this is a full retraction of a paper due to an editorial, production or publishing mistake.It is, in essence, the retraction equivalent of an erratum"); d) self-retraction ("a short retraction notice signed by all co-authors of the original paper and issued if and only if the co-authors make a joint and unsolicited request of retraction to the journal"); e) removal ("under exceptional circumstances a publication may be entirely removed from the public record if its content presents a serious and substantial risk for society, individuals or the environmental").As these suggestions currently stand, there are several potential problems.Category a (withdrawal) sounds very much like the JAMA-suggested retract and replace policy, so this new category would be redundant.Similarly redundant would be the "cancelled" category because an erratum already fulfills this function perfectly well, so renaming it would add a layer of confusion for academics.Finally, category e ("removal") is exceptionally vague, and could be ripe for abuse by interest groups seeking to retract work by competing groups.These suggestions must thus be viewed with extreme caution.More importantly, who will be implementing these new categories?The groups or publishers that represent these new categories must be carefully analyzed to ensure that there are no hidden conflicts of interest.It will be interesting to see how many COPE member journals and publishers will adopt these new categories to replace the old COPE-enshrined errata, corrigenda, and retractions, and how they will re-characterize older retractions issued before these new categories are implemented.Once again, academics will be part of an experimental system, serving as guinea-pigs for lobbying and interest groups that are starting to plague academic publishing and gradually restricting authors' rights (Al-Khatib and Teixeira da Silva, 2017a).

Other polemic issues and unresolved aspects related to the correction of the literature
Several issues need to be debated urgently before this whole suite of new corrective measures starts to become implemented.The first issue that needs to be thoroughly considered is how to correct the downstream literature that cites literature that has been corrected and/or retracted (Teixeira da Silva, 2015d, 2015e).The corrective measures currently in place deal only with the top layer, but not the underlying layer of literature.Whose responsibility is it to correct any literature that has cited corrected or retracted literature?
The responsibility should lie squarely on the shoulders of authors, editors and publishers to issue a notice to readers of papers that have cited erroneous, corrected or retracted papers not to rely on, or cite, such "tainted" literature.In the case that editors of old papers are deceased or cannot be contacted, or are unwilling to participate in the corrective process, or in the case of deceased academics, publishers must be held responsible for correcting the downstream literature, even if in literature published by other publishers (Teixeira da Silva and Dobránszki, 2015b).Currently, this is not happening.

"Preprints
researchers continue to refine or extend the work, removing the emotive climate particularly associated with retractions and corrections to published work."Barbouret al.  recognize that the COPE Guidelines for retractions, which have been used and implemented by thousands of COPE member journals and publishers, are imperfect: "their consistent implementation has proved more difficult as publishing has evolved."In other words, until July 2017 and possibly beyond, academics have been exposed to an imperfect system that has been imposed by this ethics organization, despite COPE's knowledge of the flaws and imperfections of this system.Errata and corrigenda have always been used to distinguish edits made or suggested by the authors versus those suggested by the editor or publisher, an extremely important distinction because the historical record of a published manuscript should indicate the entity responsible for fault or error.Despite this important distinction,Barbour et al. refer  to the difference between errata and corrigenda as "a now meaningless and poorly understood distinction", i.e., COPE is partaking in ideological flip-flopping, claiming the importance of such concepts for years, imposing it on global academia, and then suddenly changing the tune regarding their importance.Such volatile pos itioning by highly powerful and influential groups such as COPE damages the integrity of the publishing system that academics are expected to respect and abide by.Although social media such as Twitter and blogs can certainly raise awareness and discussion about a published paper(Yeo et al., 2016), in cases where valid criticisms merit corrections, but where corrections are not made, or are not permitted by the journal because of editorial policy (e.g., letters to the editor are not published), then indeed an amendment policy would be required.As a result of vested interests by COPE to continue show-casing new guidelines, even if they contradict older ones, is the biased promotion of new models.TheBarbour et al. (2017) preprint was met almost immediately with ridicule by a prominent science watchdog-cum-journalist, Leonid Schneider (Teixeira da Silva, 2016d), who claimed that the abolishment of retractions by COPE would allow fraudulent academics to hide their misconduct, or to subtly replace fraudulent data with valid data without the negative stigma associated with a retraction 21 , thus escaping the possibility of being punished, academically and/or criminally.Incidentally, Klaas van Dijk, a Dutch ornithologist, identified dozens if not hundreds of errors, concerns or possible problems with theBarbour et al. (2017) preprint, calling publicly for its retraction in the first quarter of 2017, precisely when the Schneider blog was published.Although theBarbour et al. preprint  is already in version 4, the authors have apparently not yet addressed most of these concerns, i.e., while they are advocating that academics make changes to their own manuscripts using their "amendment" policy, they 21 https://forbetterscience.com/2017/03/29/cope-the-publishers-trojan-horse-calls-to-abolish-retractions/

8 August 2017 doi:10.20944/preprints201708.0029.v1 al
van der Zee et al. (2017b)bour et al., 2017),Hosseini et al. (2017), and ideas put forward recently at the 5 th World Conference on Research Integrity by Fanelli et al13.Although, in general, preprints should not be cited as they represent incompletely scrutinized documents, in this case, they serve the purpose of advancing the discussion related to how the literature could be more effectively corrected.A brief note is made on preprints as a prelude to the discussion of these new models to correct the literature, and two cases are discussed that set the stage for reform.Preprints represent one tool that is now being aggressively marketed by ASAPbio 14 (see criticisms of ASAPbio here 15 ) as a solution to the replication crisis and to speed up the publication of findings, within the wider context of open science, by presenting new or contradictory results, in a crude state, open to public review, screening and criticism, before such work enters the mainstream peer review path.However, as indicated in the next section, there are vested interests and inherent biases in and among the pro-preprint proponents, causing a distortion of the need and importance of preprints, and a realistic potential for their abuse (Teixeira da Silva, 2017e).Two clear cases demonstrate the need for reform in the correction of the literature.The first case relates to Brian Wansink at Cornell University (USA).As part of a wider examination of errors within the psychology literature, which reportedly contains as much as 50% erroneous statistical errors(Nuijten et al., 2016),van der Zee et al. (2017b)closely examined what they had perceived to be a large amount (150 in total) of statistical errors and inconsistencies in four papers emerging from the Cornell Food and Brand Lab.