When a newspaper reports that a new study has found a link between, say, diet and memory, it often adds a reassuring phrase: the research was published in a peer-reviewed journal. The words are meant to signal quality, and to a large extent they do. But what exactly does peer review involve, and how much confidence should it inspire? The answer is more interesting, and more modest, than the phrase suggests.
The basic process is straightforward. A team of researchers writes up its work as a paper and submits it to a journal. An editor first decides whether it is worth considering at all; many papers are rejected at this stage. If it passes, the editor sends it to several experts in the same field, the 'peers', who scrutinise the methods, the data and the reasoning. They might ask whether the sample was large enough, whether an alternative explanation has been ruled out, or whether the conclusions go further than the evidence allows. The editor then decides whether to accept the paper, reject it or, most commonly, return it with a request for revisions. A paper may go back and forth several times before it finally appears.
In most fields, reviewers are anonymous, which allows them to criticise senior colleagues without fear of repercussions. In some journals, the authors' names are hidden from the reviewers too, in an attempt to reduce bias. Others have moved in the opposite direction, publishing reviewers' reports alongside the paper so that readers can follow the debate for themselves. Each approach has its advocates, and none has proved clearly superior.
Although scientific journals date back to the seventeenth century, when the Royal Society in London began publishing its Philosophical Transactions, peer review as we know it is surprisingly recent. For much of history, editors simply relied on their own judgement. Only in the twentieth century, as the number of papers grew and science became more specialised, did sending manuscripts to outside experts become standard practice.
Its strengths are real. Reviewers routinely catch errors, demand clearer explanations and push authors to temper exaggerated claims. Many researchers admit, sometimes grudgingly, that their papers improved as a result. 'The reviews of my first paper were brutal,' recalls Dr Tomasz Novak, an ecologist. 'At the time I was furious. Looking back, the reviewers were right about almost everything, and the paper was far better for it.'
Yet peer review has well-documented limitations. Reviewers are usually unpaid volunteers fitting the work around their own research, so the process can be slow and the quality uneven. They see only what the authors choose to show them and rarely check the raw data, which means that honest mistakes buried in the analysis can slip through, and deliberate fraud, though rare, is very hard to detect. Reviewers may also be unconsciously swayed by an author's reputation, or be more sceptical of results that challenge accepted ideas.
Critics argue that the system should therefore be seen as a first filter rather than a seal of approval. Increasingly, scientists share early versions of their work, known as preprints, before review, allowing others to comment immediately. Errors spotted after publication can lead to a correction or, in serious cases, a retraction, in which the journal formally withdraws the paper. Some see the rising number of retractions as a sign that science is in trouble; others argue that it demonstrates the system's capacity to correct itself.
Perhaps the fairest conclusion is that peer review is not a guarantee of truth but one stage in a much longer process. A single study, however carefully reviewed, rarely settles a question. It is only when results are repeated, challenged and refined by other researchers over time that scientific knowledge becomes genuinely reliable. Peer review, in other words, is where scrutiny begins, not where it ends.