Tuesday, January 9, 2018

Why I love preprints

An increasing number of servers are becoming available for posting preprints. This allows authors to post versions of their papers before publication in a peer-reviewed journal. I think this is great. In fact, based on my experiences with preprints so far, if I didn’t need journal publications to get a job, I don’t think I would ever submit another paper to a journal again. Here, I describe the advantages of preprints, and address some concerns that I’ve heard from colleagues who are less enthusiastic about preprints.

The “How” of preprints
Preprints can be simply uploaded to a preprint server: for example, on psyarxiv.com, via osf.io, or even on researchgate. It’s easy. This covers the “how” part.

The “Why” of preprints
In an ideal world, a publication serves as a starting point for a conversation, or as a contribution to an already ongoing discussion. Preprints fulfil this purpose more effectively than journal publications. Their publication takes only a couple of minutes, while publication in a journal can take anywhere between a couple of months to a couple of years. With modern technology, preprints are findable for researchers. Preprints are often posted on social media websites, such as twitter, and are then circulated by others who are interested in the same topic, and critically discussed. With many preprint servers, preprints become listed on google scholar, which sends alerts to researchers who are following the authors. The preprint can also be linked to supplementary material, such as the data and analysis codes, thus facilitating open and reproducible science.

Preprints are advantageous to show an author’s productivity: If someone (especially an early career researcher) is unlucky in obtaining journal publications, they can demonstrate, on their CV, that they are productive, and potential employers can check the preprint to verify its quality and the match of research interests.

The author has a lot of flexibility in the decision of when to upload a preprint. The earlier a preprint is uploaded, the more possibilities the author has to receive feedback from colleagues and incorporate them in the text. The OSF website, which allows users to upload preprints, has a version control function. This means that an updated version of the file can be uploaded, while the older version is archived. Searches will lead to the most recent version, thus avoiding version confusion. At the same time, it is possible to track the development of the paper.

The “When” of preprints
In terms of timing, one option is to upload a preprint shortly after it has been accepted for publication at a journal. In line with many journals’ policies, this is a way to make your article openly accessible to everyone: while uploading the final, journal-formatted version is a violation of copyright, uploading the author’s version is generally allowed1.

Another option is to post a preprint at the same time as submitting the paper to a journal. This has an additional advantage: It allows the authors to receive more feedback. Readers who are interested in the topic may contact the author with corrections or suggestions. If this happens, the author can still make changes before the paper reaches its final, journal-published version. If, conversely, a mistake is noticed only after journal publication, the author either has to live with it, or issue an often stigmatising correction.

A final possibility is to upload a preprint that one does not want to publish. This could include preliminary work, or papers that have been rejected repeatedly by traditional journals. Preliminary work could be based on research directions which did not work out for whatever reason. This would inform other researchers who might be thinking of going in the same direction of potential issues with a given approach: this, in turn, would stop them from wasting their resources by doing the same thing only to find out, too, that it doesn’t work.

Uploading papers that have been repeatedly rejected is a more hairy issue. Here, it is important, for the authors, to consider why the paper has been rejected. Sometimes, papers really are fundamentally flawed. They could be p-hacked, contain fabricated data, or errors in the analyses; theory and interpretation could be based on non sequiturs or be presented in a biased way. Such papers have no place in the academic literature. But there are other issues that might make a paper unsuitable for publication in a traditional journal, but still useful for others to know about. For example, one might run an experiment on a theoretically or practically important association, and find that a one’s measure is unreliable. In such a scenario, a null result is difficult to interpret, but it is important that colleagues know about this, so they can avoid using this measure in their own work. Or, one might have run into practical obstacles in participant recruitment, and failed to get a sufficiently large sample size. Again, it is difficult to draw conclusions from such studies, but if the details of this experiment are publically available, this data can be included in meta-analysis. This can be critical for research questions which concern a special population that is difficult to recruit, and in fact may be the only way in which conducting such research is possible.

With traditional journals, one can also be simply unlucky with reviewers. The fact that luck is a huge component in journals’ decisions can be exemplified with a paper of mine, that was rejected as being “irritating” and “nonsense” from one journal, and accepted with minor revisions by another one. Alternatively, one may find it difficult to find a perfectly matching journal for a paper. I have another anecdote as an example of this: After one paper of mine was rejected by three different journals, I uploaded a preprint. A week later, I had received two emails from colleagues with suggestions about journals that could be interested in this specific paper, and two months later the paper was accepted by the first of these journals with minor revisions.  

The possibility of uploading unpublishable work is probably the most controversial point about preprints. Traditional journals are considered to give a paper a seal of approval: a guarantee of quality, as reflected by positive reports of expert reviewers. In contrast, anyone can publish anything as a preprint. If both preprints and journal articles are floating around on the web, it could be difficult, especially for people who are not experts in the field (including journalists, or people who are directly affected by the research, such as patients reading about a potential treatment), to determine which they can trust. This is indeed a concern – however, I maintain that it is an open empirical question whether or not the increase in preprint will exacerbate the spreading of misinformation.

The fact is that traditional journals’ peer review is not perfect. Hardly anyone would contest this: fundamentally flawed papers sometimes get published, and good, sound papers sometimes get repeatedly rejected. Thus, even papers published in traditional journals are a mixture of good and bad papers. In addition, there are the notorious predatory journals, that accept any paper against a fee, and publish it under the appearance of being peer reviewed. These may not fool persons who are experienced with academia, but journalists and consumers may find this confusing.

The point stands that the increase in preprints may increase the ratio of bad to good papers. But perhaps this calls for increased caution in trusting what we read: the probability that a given paper is bad is definitely above zero, regardless of whether it has been published as a preprint or in a traditional journal. Maybe, just maybe, the increase of preprints will lead to increased caution in evaluating papers based on their own merit, rather than the journal it was published in. Researchers would become more critical of the papers that they read, and post-publication peer review may increase in importance. And maybe, just maybe, an additional bonus will lie in the realisation that we as researchers need to become better at sharing our research with the general public in a way that provides a clear explanation of our work and doesn’t overhype our results.

I love preprints. They are easy, allow for fast publication of our work, and encourage openness and a dynamic approach to science, where publications reflect ongoing discussions in the scientific community. This is not to say that I hate traditional peer review. I like peer review: I have often received very helpful comments from which I have learned about statistics, theory building, and got a broader picture of the views held by colleagues outside of the lab. Such comments are fundamental for the development of high-quality science. 

But: Let’s have such conversations in public, rather than in anonymous email threads moderated by the editor, so that everyone can benefit. Emphasising the nature of science as an open dialogue may be the biggest advantage of preprints.

1 This differs from journal to journal. For specific journals’ policies on this issue, see here.