1. What is a preprint, and how is it different from a journal paper?
When scientists finish a study, they write it up as a paper and usually send it to a scientific journal. The journal asks other experts in the same field to read it and look for mistakes or parts that need more explanation. This is called peer review, and it can take months.
A preprint is a version of the paper shared publicly before it appears in a journal, so it may not yet have been peer reviewed. The big advantage is speed: anyone can read new results right away. The catch is that other experts have not yet vouched for it. The servers say so plainly. arXiv states that material is not peer-reviewed by arXiv [1], and bioRxiv says articles are not peer-reviewed, edited, or typeset before being posted [3].
Some checking does happen, but it is a different kind. arXiv moderators check that a submission fits the subject area and has scholarly value [1]. bioRxiv screens for offensive or non-scientific content and health risks, and checks for plagiarism [3]. None of this decides whether the science is correct. Many authors also send the same paper to a journal afterward, so many preprints are early versions of papers that are published later.
2. How did researchers share new papers before the internet?
Before the internet, a researcher who finished a paper might mail photocopies to a small number of colleagues. The physicist Paul Ginsparg, who later started arXiv, remembered that researchers lower down the ladder depended on the goodwill of people on what he called the "A-list" [2]. In his account, someone at a less well-known institution might not hear about new results for a long time, and he saw electronic sharing as a way to level that field [2].
So preprints were already a habit among scientists. What arXiv changed was how they were handed out: to everyone, automatically, instead of to a short mailing list.
3. How did arXiv start in 1991, and why was nothing ever deleted?
In June 1991, a chat at a physics conference in Colorado gave Ginsparg the idea of a central, automated store of papers that would send out the full text whenever someone asked [2]. The first paper arrived by email on August 14, 1991 [2].
It began as a modest tool for one small corner of particle physics, and Ginsparg expected about 100 submissions a year. Instead it received 400 in its first half year [2]. Papers were supposed to be deleted after three months, but "by popular demand" nothing was ever deleted [2]. arXiv is still going, and now covers physics, mathematics, computer science and several other fields [1].
One famous example comes from mathematics. In November 2002, Grigori Perelman posted a paper on arXiv that became the first of three papers that together led to the solution of the Poincaré conjecture, a puzzle about shapes that had stayed open for nearly a century [6]. It was a preprint, read by experts around the world.
4. Why did biology and medicine take longer to accept preprints?
Ginsparg notes that in physics the idea caught on quickly. In biology and medicine it was slower. He points to a long tradition of treating only a peer-reviewed journal publication as the proper way to claim a discovery first, and to worries about the public-health effects of unchecked results [2].
Even so, Cold Spring Harbor Laboratory launched bioRxiv for the life sciences in 2013 [3]. For health research there is medRxiv, which puts a clear warning up front: preprints are preliminary reports that have not been certified by peer review, they should not be relied on to guide clinical practice or health-related behavior, and they should not be reported in news media as established information [4].
What happens to preprints afterward? Two researchers, Richard Abdill and Ran Blekhman, studied all 37,648 preprints posted to bioRxiv in its first five years. Of those posted between 2013 and the end of 2016, 67.0% were later published in a journal [5]. That figure came from automatic matching, and in a manual check, 37.5% of the preprints that had been marked "unpublished" had in fact appeared in a journal [5]. So the true share may be higher than 67%, but the study could not measure it exactly.
5. How can you tell whether a study you read about is a preprint?
A preprint is not bad science, and it is not automatically right either. It is an early report that is still waiting for other experts to look at it. Three questions help.
Does it say it has not been peer reviewed? The preprint servers label their papers this way, and medRxiv adds a warning notice [4].
Has it been published in a journal since? bioRxiv usually adds a link to the published version automatically within several weeks of publication, though sometimes a match is not made [3]. The journal version may differ from the preprint, so note which one you are quoting.
What was the news story based on? A story about "a new study" may be describing a preprint. If you can find the original paper, look for the words "preprint" or "not peer-reviewed."
Preprints also have an upside: being shared early gives more experts a chance to read them, and that may help mistakes come to light sooner. The safe way to read one is as a progress report, not a final answer.
6. How can you trace a science story back to its source?
Next time you see a science story, try to find the paper behind it with an adult. Look for the researchers' names, the institution, or a link in the article. Can you tell whether it is a journal paper or a preprint? If the story does not say, that may be a reason to look for the paper yourself.
For a deeper look, Ginsparg wrote his own account of arXiv's first twenty years [2]. It is a clear, readable history written by the person who built it.