This piece was originally posted as part of a discussion of preprints on the Auditory mailing list.

Preprints have been around and used extensively (particularly in maths and physics) for over 30 years at this point (arXiv was founded in 1991). Most major funders and journals recognise preprints, probably the majority of funders now have open access requirements that can be fulfilled with preprints, and a few are even mandating their use. It’s actually not much younger than the widespread use of peer review, which didn’t become a de facto standard until the 1960s-1970s (Nature didn’t use it until 1973 for example).

Preprinting papers is a huge advantage to authors, and the data is stark. Papers in biology which originally appeared as preprints get 36% more citations and the advantage is immediate and long lasting.

To make the argument clearer, let’s break it down into the different roles that preprints can have.

The first role is what preprints can do in the period following the publication of a paper in a journal. In this case, posting a preprint of a paper fulfills open access requirements and makes it possible for the whole world to read your paper, including the general public, and people at less wealthy universities and countries that cannot afford the journal subscription. I cannot see any coherent argument against this. It’s a disgrace that the public pays for science but is not able to access the results of the work they paid for, and it is only a hindrance to scientific progress to gate access to knowledge.

The second role is what preprints can do in the time between the journal accepting the paper and making it available. This is purely about speed of publication but I can’t see any reason why you wouldn’t want this speed? I just went to the most recent issue of JASA and looked at the first three papers as a rough sample, and this delay was 3 weeks, 3.5 weeks and 6.5 weeks. It’s not years, but might make the difference in someone’s job or grant application.

The third role is where some might disagree, the time period between publishing the preprint and journal acceptance. But I don’t really see any conflict here. If you don’t want to read preprints and prefer to wait, then just don’t read them. But they will have value for other readers (like me) who accept the limitations, and they have great value for the authors (36% more citations for example). For reference, for my sample of JASA papers above, the times from first submission to journal publication were 22 weeks, 27 weeks, and 38 weeks.

It has been argued that although peer review isn’t perfect, it does provide some degree of quality control. I agree, but I’m not sure how much quality control it really provides. A study of peer review at the BMJ with deliberate major and minor errors found that on average peer reviewers picked up on 2.6 to 3 of 9 major errors deliberately introduced. That’s some sort of quality control, but not enough to mean that you can uncritically read peer reviewed papers.

And on the other hand, there is also a downside to only reading peer reviewed work in that you are subject to editorial and reviewer biases. A PNAS study found that a paper submitted with a Nobel prize winner as author was recommended for acceptance by 20% of reviewers, but the very same paper with an unknown student as author was only recommended for acceptance 2% of the time.

More controversially perhaps, I think there is a potential fourth role for preprints that are never submitted to a journal. This is very common in maths, physics and computer science and works well there. I think it would work even better when combined with a post-publication peer review platform that made reviews open, prominently displayed with an at-a-glance summary, and easily accessible. But that’s a separate argument!