Science journals: where they came from and how to read one

On 3 June 2020 the authors of a paper in The Lancet asked for it to be retracted. The paper, published that May, had reported that hydroxychloroquine did not help patients with COVID-19. The data came from a company called Surgisphere, which would not give the authors or an independent reviewer access to the underlying records, and the journal agreed to withdraw it. The episode was quick, public and awkward, and it shows plainly what a science journal is and what it is not. A journal is a promise that someone has looked. It is not a guarantee.

Two titles in 1665, and an argument

The first journals appeared within months of each other. The Journal des sçavans was printed in Paris on 5 January 1665, edited by Denis de Sallo, and covered book reviews, obituaries, reports of discoveries and court proceedings. Philosophical Transactions of the Royal Society started in March 1665 in London. As the general entry on scientific journals puts it, Philosophical Transactions is usually called the first scientific journal, and the French title the first academic one. Which of the two is the answer depends on what you count, and both camps have a case.

Both belong to the same first wave of periodicals that the page on the oldest magazines follows, and the line between a learned journal and a magazine was blurry for a long time. Neither looked like a modern journal. There were no referees in the modern sense and no standard format for a paper; an editor decided. The idea that other scientists should check a paper before printing grew slowly. The scientific journal entry dates a peer review system only to the 1970s in its modern form, although the Medical Society of Edinburgh established Medical Essays and Observations in 1733, which the same entry calls the first fully peer-reviewed journal. And in 1953 Nature printed Watson and Crick’s paper on the structure of DNA without sending it to referees; its later editor John Maddox said the correctness was self-evident.

The titles that lasted

Several journals and magazines founded in the nineteenth and twentieth centuries are still the ones people mean when they say “a science journal”.

Nature is the one with the best story. It was started on 4 November 1869 by Norman Lockyer, and its title comes from a line of Wordsworth about the mind that builds on the solid ground of nature. The long entry on Nature records that it is now published by Springer Nature. Science had a harder start. Founded in 1880 by the journalist John Michels with money from Thomas Edison, it stopped in March 1882, and Alexander Graham Bell and Gardiner Greene Hubbard revived it the next year. The American Association for the Advancement of Science took it over in 1900.

The Lancet is older than either. Thomas Wakley named it after the surgical instrument in 1823, and it has been owned by Elsevier since 1991.

What changed with preprints

Physics got there first. In 1991 Paul Ginsparg, working at Los Alamos, set up a server so that physicists could share their papers by email before the journals printed them. arXiv was run by Cornell University Library from 2011, announced in 2026 that it would become an independent nonprofit, and is explicitly not peer reviewed, though submissions are moderated. Medicine followed in June 2019, when Cold Spring Harbor Laboratory, Yale and the BMJ launched medRxiv.

The timing mattered. When COVID-19 spread in 2020, researchers posted early results on the servers instead of waiting months for a journal slot. That helped in speed and hurt in certainty: papers that had not been reviewed were reported as findings. If you read about a new study in a newspaper, check whether it is a preprint. The page it appears on will say so.

Impact factor and its critics

Journals are ranked by a number. Eugene Garfield, who founded the Institute for Scientific Information, devised the impact factor to help librarians choose titles, and it has been calculated since 1975. It divides the citations to a journal’s articles from the previous two years by the number of articles it published in that time. Clarivate now publishes it.

The trouble is how the number is used. A journal with a high impact factor may owe it to a handful of widely cited papers, while the typical paper in it is cited rarely. In 2013 the San Francisco Declaration on Research Assessment, started by the American Society for Cell Biology, asked institutions to stop judging researchers by the journal title alone. Whether the habit has changed is another matter, and the journal’s name still carries weight in many places.

Open access

The usual complaint about journals is cost: a university library might pay a large sum to read papers its own staff wrote. Open access publishing is the answer to that, and the main push in Europe came in September 2018, when a group of national funders from twelve countries announced Plan S. The cOAlition S site explains the rules, which require publicly funded research to be free to read at once, with no embargo. David Sweeney of UK Research and Innovation co-led the task force that wrote the implementation guidance, issued in November 2018.

Not everything is solved by it. Someone still has to pay, and where the author or funder pays, a fee can discourage poorer researchers from submitting. Science, still a subscription title, makes its research articles free a year after publication.

Journal or magazine

The word journal is used loosely, and the guide to types of journals sorts the families: learned society journals, commercial titles, trade periodicals and the rest. For this page the working distinction is who writes and who reads. In a journal such as Nature the authors are the researchers and the readers are other researchers. In a magazine such as New Scientist, which published its first issue on 22 November 1956 under founders Tom Margerison, Max Raison and Nicholas Harrison, the writers are journalists, and the readers are anyone curious. Both are useful. They answer different questions.

The two overlap in the middle. Nature and Science carry news and comment pages alongside papers, and they are read by many outside their fields. For the wider family of titles, the guides to magazines cover how each kind is made and sold.

Fields differ

A journal in mathematics is not run like one in medicine. Mathematicians post to arXiv before they submit anywhere, and a paper may sit in the archive for years while a referee checks the proof line by line. Clinical medicine moves in a different direction: trials are registered, results are reported by a set of rules, and an error can reach patients. That is part of why citation counts cannot be compared across fields, as the critics of the impact factor point out. A top journal in one subject may be cited far less than an average one in another.

Engineering and chemistry have their own sets of society journals and specialised titles, many of them small. Small does not mean weak: a narrow journal read by every specialist in its area may matter more to those people than Nature does.

When journals meet politics

The Surgisphere case landed in an argument that was already loud. The drug in question had become a political subject, and a paper that said it did not work was read by many people as a verdict on the people who had promoted it, not as a single study with a dataset. When the data failed to hold up, the retraction was read the same way.

The lesson for a reader is dull but useful. One study, in one journal, in one month, is a data point. A conclusion is what accumulates after others have checked it, repeated it and failed to knock it down. Journalists, and readers, who treat a single paper as a ruling end up having to take it back.

Reading one without a degree

Start with the abstract, which is a summary written by the authors. Then read the methods: how many people or samples, what was compared with what, and whether the study was set up before the data came in. A reader who goes straight to the conclusion misses the part where most claims succeed or fail.

Look at the funding and the declared conflicts of interest, near the end. They are short and routinely skipped, which is a pity, because a paper on a drug paid for by the company that makes it deserves a different level of caution from one paid for by a public body. Look, too, at whether the journal has issued corrections. In the case of the Lancet and the MMR vaccine, the 1998 paper by Andrew Wakefield was fully retracted on 2 February 2010, after most of its co-authors had already withdrawn their interpretation in 2004. The Lancet’s entry tells both stories, Wakefield and Surgisphere, in the same place. Twelve years is a long time for a wrong paper to stay in the record without a mark.

If the subject is general interest rather than research, the differences between periodicals are set out in the guide to types of journals. For a magazine that translates science into photography and maps, see National Geographic Magazine, and for the broader family of titles, the magazine guides.

The Surgisphere paper came down within days. The Wakefield paper took twelve years. Nothing in the format of a journal tells you which kind of problem you are looking at.