What Is a Preprint?

A preprint is a complete research paper that its authors post publicly, free for anyone to read, before it has been through peer review, the step where a journal asks other scientists to check the work. It is finished work, with methods, results and conclusions. It is simply that nobody independent has signed it off yet.

The article here on peer review mentions preprints in passing, as the unfiltered alternative to a journal. This one looks at them properly: why they exist, where they live and what those sites actually check, what the rush of 2020 taught everyone, and what to do in practice when a news story turns out to be about one.

Why would anyone publish before review?

Imagine finishing a book and then waiting a year while a publisher decides what to do with it. All that time, nobody can read it, argue with it or build on it. For a novelist that is frustrating. For a scientist whose findings could shape other people's experiments, or in an emergency shape public health decisions, the delay has a real cost.

Preprints remove the wait. Researchers post the paper to a public website, it appears within a few days, and anyone can read it. Most submit the same paper to a journal at the same time, so peer review still happens. It just happens alongside the public version instead of in front of it.

There are other reasons too. Posting a preprint puts a public date on an idea, which matters when two groups are racing towards the same result. It invites comments from far more people than the two or three reviewers a journal would choose. And it makes the work free to read, which many journal articles still are not.

Where do preprints live?

On preprint servers: websites that host papers and give each one a permanent address. Physicists have used them for decades. In biology and medicine, the two best known are sister sites.

bioRxiv, pronounced bio-archive, was set up in 2013 by Cold Spring Harbor Laboratory, a research institute in New York, and is now run by a non-profit organisation called openRxiv. It describes itself as a free online archive for unpublished preprints in the life sciences, and states plainly that articles are not peer-reviewed, edited or typeset before being posted 1.

medRxiv, its counterpart for medical and health research, launched in 2019, founded by Cold Spring Harbor Laboratory together with Yale University and the medical publisher BMJ 2. Because medical findings can change how people behave, it carries a stronger warning. Preprints, it says, are preliminary reports of work that have not been certified by peer review; they should not be relied on to guide clinical practice or health-related behaviour, and should not be reported in news media as established information 2.

Both servers do screen what arrives. Papers are checked for offensive or non-scientific content and for plagiarism, and medRxiv also looks for material that might pose a health risk 12. That screening is a basic filter, like a doorman checking that visitors are dressed, not a judge assessing their argument. In bioRxiv's own words, no endorsement of an article's methods, assumptions, conclusions or scientific quality is implied by its appearance there 1.

What did 2020 teach everyone?

Before 2020, most people outside science had never heard the word. Then a new virus arrived, the world needed answers in weeks rather than years, and preprints went from a specialist habit to front-page news.

The scale was extraordinary. One study counted more than 125,000 scientific articles about COVID-19 released in the first ten months of 2020, of which more than 30,000 were hosted on preprint servers. The attention changed too: 28.7 per cent of COVID-19 preprints were featured in at least one news article, compared with 1.0 per cent of other preprints 3. For the first time, unreviewed science was routinely reaching the public directly.

So did it work? Mostly, yes. A follow-up study compared early pandemic preprints with the versions later published in journals. The figures and tables barely changed. The conclusions in the summary changed in a clear way for 17.2 per cent of the COVID-19 papers, against 7.2 per cent of other papers, and most of those changes did not alter the paper's main message 4. Peer review tightened things. It rarely turned them upside down.

But the exceptions were loud. At the end of January 2020, a preprint claimed that the new coronavirus carried short stretches of sequence with an uncanny similarity to HIV. Scientists around the world pointed out the problems in public almost at once; among other things, sequences that short can be found in a huge range of unrelated organisms. On 2 February the authors withdrew it, saying they intended to revise it in response to comments from the research community on their technical approach and their interpretation of the results 5.

The open, public checking had worked remarkably quickly, within days rather than the months a journal would take. The claim, however, had already escaped, and it went on circulating long after the withdrawal notice appeared. That is the 2020 lesson in two halves. Preprints made good science available sooner than ever, and most of it held up. They also let weak claims reach millions of people before anyone had checked them, and a correction never travels as far as the original.

The speed also had a quieter benefit that rarely made the news. Researchers could see what other groups were finding almost as it happened, spot mistakes early, and avoid repeating experiments that had already failed somewhere else. Much of the real value of preprints in 2020 was of that kind: scientists talking to scientists, faster than journals could carry the conversation.

What happens to a preprint afterwards?

A preprint is not a dead end. It is the first public stop on a journey, and the server keeps a record of the whole trip.

Most preprints are submitted to a journal, either before posting or soon after. If the journal's reviewers ask for changes, the authors often post a revised version to the server as well, and the earlier versions stay visible. When the paper is finally published, the server usually adds a link to the journal version at the top of the page, so a reader arriving at the preprint can see at a glance that a reviewed version exists.

Some preprints never reach a journal at all. That can mean the work was rejected, that the authors moved on, or simply that they never sent it anywhere. A preprint still unpublished several years later is not proof of a problem, but it is a reason for caution, because no independent specialist has ever put their judgement about it on the record.

And a few are withdrawn. On the big servers, withdrawal does not erase a paper. The page stays up, with a notice saying the authors have taken it back and, usually, why. That openness is deliberate. Someone who meets the claim quoted elsewhere can still find the original and see that its own authors no longer stand behind it.

How should you treat a preprint?

Not with suspicion, and not with trust. Treat it as a report from the field: useful, early, and not yet checked by anyone independent. Six habits help.

1. Notice that it is one. News stories do not always say so. The name of a preprint server, or the word preprint near the title, is the clue.
2. Check the version. Preprints can be updated, and every version is kept. The one being quoted may not be the latest.
3. Look for a published version. Servers usually add a link once the paper appears in a journal. If years have passed with no journal version, that is worth knowing.
4. Look for a withdrawal notice. Withdrawn preprints stay online with a notice explaining why, and some people carry on sharing the original regardless.
5. Ask the usual questions. How many participants? Compared with what? Who ran it, and who paid? Design matters just as much before review as after.
6. Hold it loosely. One preprint should nudge your view, not overturn it.

Preprints and the peptide literature

Research on peptides turns up on preprint servers like any other biology: a new structure, an early result in cells, a first analysis of a trial. That is healthy. It is how researchers share work quickly with one another and get it criticised early.

It also means an exciting claim about a compound can appear online, get quoted and requoted, and harden into a fact in people's minds while the paper itself is still waiting for its first reviewer. When you trace a claim back and find a preprint at the bottom, the right conclusion is not that it is false. It is that it is early, and that the checking has not happened yet.

Then it pays to look again in a few months. A preprint usually goes one of three ways: into a journal, into a revised version, or into a withdrawal. Knowing which of those happened is often the single most useful fact you can find about a claim, and it is usually one click away from the original page.