How does journal peer review actually work?

The process is more human and less systematic than most first-time authors expect. An editor, often a working academic doing this alongside a full job, reads the submission and decides whether it is worth reviewers' time. If it is, they start inviting people, and this is where the calendar goes.

Reviewer invitation is the bottleneck. Editors invite specialists who are busy, unpaid, and already reviewing for three other journals. Declines are common, some invitations go unanswered, and each round of invitations costs a week or two. When a paper sits at "under review" for months, the usual explanation is that reviewers were still being recruited rather than that someone is reading very slowly.

The division of labour is formalised more than authors realise. "Editor" covers editors-in-chief, managing editors, and associate editors, and reviewers are typically asked to recommend an outcome, assess the content, flag ethical concerns, and comment generally (Tennant and Ross-Hellauer, 2020, p. 1). Splitting the work that way spreads editorial risk across several people, and the same split weakens accountability, because no single person owns the decision (Tennant and Ross-Hellauer, 2020, pp. 4-5). This runs across a research enterprise that invests around $2 trillion a year and produces more than 3 million peer-reviewed articles.

What a reviewer is actually expected to check is often undefined. One recommended routine is three passes over the manuscript, first structure, contribution, and scope, then methodology, data integrity, and interpretation, then a coherence check on the feedback itself (Rahman and DeBlasio, 2026, p. 3). In practice reviewers may not know whether they are supposed to test your code or whether editorial staff will, and they cannot rerun your experiments once the data are collected (Tennant and Ross-Hellauer, 2020, p. 4). A few venues fix this by rule, and the Journal of Open Research Software and the Journal of Open Source Software both require code review.

Reviews themselves vary in depth from a paragraph to four pages. The editor reads them together, notices where they disagree, and makes a decision that need not follow the majority. An editor can accept a paper both reviewers were lukewarm about, or reject one both praised, if the journal's priorities point that way.

Knowing this changes how you read a decision. Contradictory reviews are normal and are not a sign that the process failed. What the editor tells you in the decision letter is the instruction that matters.

It also helps to hold the process's authority at the right level. Comparisons of arXiv and bioRxiv preprints with their final published versions found very few overall changes and similar reporting quality, which makes the value peer review adds hard to measure at scale (Tennant and Ross-Hellauer, 2020, p. 6). Arguments about anonymity are unresolved in the same way. Blinding may shield early-career and under-represented researchers from reprisals, and it may also license careless reviewing and hide structural bias, and one psychology survey called fears of reprisal greatly exaggerated while the authors warn that finding may not carry to other fields (Tennant and Ross-Hellauer, 2020, p. 5). Read a decision letter as one editor's judgement informed by reports, not as a reproducible verdict.

How long does peer review take?

Typically two to four months to a first decision, with wide variation by field. Most of that time is spent finding reviewers who agree, not reading. A paper under review for five months is usually a paper whose fourth or fifth invited reviewer has just accepted.

Many journals publish median times from submission to first decision, and those figures are worth checking before you submit if timing matters to you.

A polite enquiry after the journal's stated typical time has passed is normal and welcome. One short email to the editorial office asking for a status update, not more often than monthly.

What do the decision terms mean?

Accept is rare on a first decision. Minor revision means the reviewers are satisfied in principle. Major revision means substantive work is needed and acceptance is not guaranteed. Reject and resubmit means a new submission with a new review round. Reject is final for that journal.

Major revision is a good outcome despite how it reads. The editor has invested in the paper and is inviting you to earn acceptance, and a well-handled major revision is accepted more often than not.

The reports you receive will be uneven, and the good ones share a shape. Reviewing guidance asks for a point-by-point critique that says how important each problem is to the publication decision, avoids pushing the reviewer's own preferred hypothesis onto your paper, and commits to a definite rating, because fence-sitting only prolongs the process (Bourne and Korngreen, 2006, p. 974). Reviewers are also told to name weaknesses, propose ways to fix them, and state the strengths (Rahman and DeBlasio, 2026, pp. 5-7). When a report gives you none of that, ask the editor which points are binding for the revision.

Reject and resubmit is genuinely ambiguous and worth clarifying. Some editors use it to reset a review clock on a paper they want, and others use it as a soft rejection. If the letter is encouraging, ask the editor directly whether a resubmission would be welcome.

One detail worth knowing: the reviewers who see your revision are often the same people who wrote the first reports. Writing the response letter as though they will read it, because they will, changes how you phrase a disagreement.