Most authors learn peer review one rejection at a time. Here is what actually happens to a manuscript between submission and a decision, and who is doing what at each step.
Peer review is the set of steps a journal runs between receiving your manuscript and telling you what happens next. It is not one event. It is a sequence of handoffs, and each handoff is where time and judgment get added.
The manuscript first reaches an editorial office, not a reviewer. A managing editor or editorial assistant checks that the file is complete: the right sections, a word count within range, disclosure forms attached, figures readable. Papers get returned at this stage for missing pieces, before any scientific reading happens.
A handling editor, usually one member of the editorial board with relevant expertise, is then assigned. That person reads the manuscript with one question in mind: is this worth sending out for review at all. A paper judged out of scope, too preliminary, or unlikely to clear review even at its best is desk rejected here, before a single outside reviewer sees it.
If the paper clears that check, the handling editor invites reviewers, usually two or three, sometimes more depending on the journal and the field. Each invited reviewer can accept, decline, or simply not respond. Editors often send more invitations than they need, expecting some fraction to decline, and wait for enough acceptances before review formally starts.
Once reviewers are in place, they read the manuscript independently and submit reports, usually within a window the journal sets. The handling editor reads all the reports, weighs them against their own reading of the paper, and writes a decision letter. That letter, not any single reviewer's report, is the outcome the author receives.
Authors frequently talk about being reviewed as though the reviewers decide the outcome. They do not. Reviewers advise. The editor decides. Conflating the two roles is the single most common misunderstanding about how peer review works, and it changes how you should read a decision letter once you see the distinction clearly.
A reviewer's job is narrow: read the manuscript in their area of expertise and report what they find. That includes flagging unclear methods, questionable claims, missing citations to relevant prior work, and anything that looks wrong. A reviewer is not deciding whether the paper gets published. They are handing the editor evidence.
The editor's job is broader and includes things no individual reviewer can see. The editor knows how the two or three reports compare to each other, how this manuscript compares to others in the same queue, what the journal's current acceptance rate and backlog look like, and whether the objections raised are the kind a revision can fix. An editor can accept a paper over the strong objection of one reviewer, or reject a paper that came back with two positive reports, if their own reading says otherwise.
This matters practically. A reviewer who says "reject" has not rejected your paper. They have written one input into a decision someone else makes. Authors who treat every reviewer comment as a verdict are reading the wrong document as final.
Most authors imagine review as an open-ended critique. In practice, journals hand reviewers a short form. It typically asks for a recommendation from a fixed list, a set of scored or checkbox ratings on things like originality and methodological soundness, and a free-text box for comments to the author and sometimes a separate confidential box for comments to the editor only.
That structure shapes what comes back. A reviewer working through a form with a comment box is not writing a peer-reviewed critique of your critique. They are filling in a box, often between other obligations, and the length and thoroughness of what lands in that box varies enormously by reviewer, by field, and by how much time they had that week.
A typical review form, structurally
Recommendation: accept / minor revision / major revision / reject. Originality: low / medium / high. Methodological soundness: low / medium / high. Comments to author: free text. Confidential comments to editor: free text, not shown to the author.
Knowing this form exists explains a lot of otherwise confusing feedback. A terse, three-sentence review is not necessarily a sign the reviewer didn't take the paper seriously. It may be exactly as much as the box got, from someone reading on a train between two other manuscripts they also agreed to review that month.
This is worth its own section because it reframes almost every hostile-seeming report an author receives. Reviewers are not employees of the journal. They are researchers in the same field, asked to donate hours of unpaid time to read someone else's manuscript, on top of their own research, teaching, grant writing, and everything else already on their calendar.
Nobody trains them for this. There is no standard course a researcher takes before their first review. Most people learn to review by having been reviewed, by watching a senior colleague's reports, or by trial and error over a few invitations. The result is enormous variance in what a "review" looks like from one person to the next, even within the same field and the same journal.
A reviewer working from a short form, with limited time, and no formal training, is going to produce a report shaped by all three constraints. A comment that reads as curt, or that jumps straight to a criticism without context, is frequently just what a busy person managed to type into a box before their next meeting. It is not, usually, a personal judgment about the author.
This does not mean every harsh review is accidental, and it does not mean you should ignore a comment because the reviewer might have been rushed. It means the tone of a review is weak evidence about its accuracy. A curt comment can still be the correct catch. A warm, detailed comment can still be wrong. Reading past the tone to the specific claim being made is the useful skill here, and it gets easier once you stop assuming the tone was chosen deliberately.
Consider a reviewer who writes: "Sample size is inadequate, reject." Read as a verdict, that sentence is discouraging and says nothing you can act on. Read as one line in a comment box, filled in by someone who has fifteen minutes before they need to leave for a seminar, it is an incomplete thought worth a direct question back to the editor: what power calculation, if any, would satisfy this concern. The editor, who has read all the reports and not just this one, is usually able to tell you whether that objection is central to the decision or one input among several.
A reviewer's job stops well short of rewriting your paper for you. They are not there to redesign your study, run your analysis for you, fix your prose, or produce a publishable manuscript out of a weaker one. Reviewers who cross into rewriting territory, sketching out the paper they would have written instead, are giving you something other than a review, and editors increasingly discount reports that read that way.
A reviewer is also not a fact-checker for every claim in your paper, an editor of your English, or a guarantee that publication means the results are correct. Peer review catches some errors. It is not a proof of correctness, and treating a published, peer reviewed paper as automatically true is a separate mistake from the ones covered here, but a common one.
Journals run review under different visibility rules, and the rule in force changes what a reviewer knows and what an author can infer about who is reviewing them.
Single blind is the most common arrangement in many fields. The reviewer knows the author's identity and institution. The author does not know who reviewed the paper. Critics of this model point out that knowing an author's name and affiliation can bias a reviewer, favorably or unfavorably, before they have read a word of the actual argument.
Double blind removes author identity from the manuscript sent to reviewers, at least in theory. In practice, double blinding is imperfect. A niche subfield, a distinctive dataset, a self-citation, or a preprint already posted under the author's name can make identity guessable even when it has been formally stripped from the file.
Open review covers a range of practices, from reviewer names being disclosed to the author, to full reports being published alongside the paper. The stated goal is accountability: a reviewer attaching their name to a report has more reason to write a careful one. Whether it changes reviewer behavior in practice is debated, and adoption varies widely by field and publisher.
Post-publication review moves some or all of the evaluation after the paper is already public, through open commentary platforms, letters to the editor, or journals that publish first and formally review after. This does not replace pre-publication review so much as add a second layer after the fact, and it means a published paper is not necessarily a finished conversation.
Which model a given journal uses is a policy choice made by that journal, and it is usually stated in the author guidelines. It is worth checking before you submit, because it changes what you should and should not put in the manuscript itself, particularly under double blind rules where self-identifying details need to be removed.
Editors choose from a small set of decision categories, and the exact wording differs by journal, but most fall into a similar shape.
Accept, sometimes with minor copyediting still to come, means the paper is going to publication largely as submitted. This is uncommon on a first submission in most fields.
Minor revision means the editor believes the paper is close, and the requested changes are things like clarifying a method, adding a citation, or correcting a table, rather than anything that could change the paper's conclusions.
Major revision means substantial concerns were raised, often about the analysis or the support for the paper's central claims, and the paper needs real rework before the editor can decide. A major revision is usually sent back out to the same reviewers for a second look, which is why it takes longer than a minor one.
Reject and resubmit, where journals use it, sits close to major revision in substance but different in bookkeeping: the paper is formally treated as a new submission rather than a revision of the old one, sometimes because the required changes are large enough that the journal wants a fresh review cycle rather than a patched one.
Reject means the paper is not going forward at this journal, whether because it does not fit the journal's scope, the objections raised are not fixable within the current design, or the paper simply did not clear the bar relative to what else the journal has in its queue that cycle.
The same set of reviewer comments can lead to different categories at different journals. An objection that triggers major revision at a selective journal might be waved through as a minor note elsewhere. The category is the editor's judgment about this paper at this journal, not a fixed grade the manuscript carries with it everywhere.
The time between submission and decision is not one wait. It is several sequential waits stacked on top of each other, and seeing them separately explains most of what feels like delay.
First, the manuscript sits until an editor is assigned and reads it far enough to decide whether to send it out at all. Second, once the editor decides to proceed, invitations go out to potential reviewers, and the editor waits for enough of them to accept. Declines and non-responses are common, and each one restarts part of this step with a new invitation.
Third, once reviewers have accepted, they are given a window to submit their report, and that window competes with everything else on a volunteer reviewer's schedule. A reviewer who accepted in good faith can still run past the deadline, and editors typically send reminders before escalating to finding a replacement reviewer, which restarts the invitation step for that one slot.
Fourth, once all reports are in, the editor has to read them, weigh them, and write a decision letter, which competes with the editor's own research and their queue of other manuscripts. Practice on how long each of these steps takes varies enormously by field, by journal, and by time of year, and stating a specific number here would be more precise than the process actually is. What is consistent is the shape: several independent waits in sequence, each one able to stall on its own, rather than one continuous process.
None of this defends slow review as ideal. It describes the mechanism, because "review is slow" and "the editor is sitting on my paper" are different claims, and the second one is usually wrong even when the first one is true.
A guide that only defends peer review is not being honest about it. The system has real, well documented limits, and knowing them is part of understanding how it works, not a side issue.
Reviewer variance is large. The same manuscript sent to different reviewers can and does receive very different recommendations, because two competent reviewers can reasonably weigh the same limitations differently. There is no single ground truth score a manuscript is converging toward. Two independent review cycles for the same paper are not guaranteed to land in the same place.
Fit with a reviewer's own methodological stance can also shade a recommendation in ways hard to separate from the paper's actual quality, even when the reviewer is acting in good faith.
Review also does not catch everything it is sometimes assumed to catch. Fabricated data, subtle statistical errors, and image manipulation have all passed through peer review and into publication, because reviewers are reading a manuscript, not auditing raw data or rerunning the analysis themselves. Peer review is a check on argument and presentation more than it is a forensic audit, and treating it as the latter overstates what the process actually does.
None of this means the system should be discarded. It means the output of peer review, an accept decision, is evidence that a paper cleared a particular, imperfect filter, not a guarantee that everything in it is correct.
Understanding the process changes how you should prepare a manuscript before it goes anywhere near a reviewer. If a reviewer has a short form, limited time, and no training, the manuscript that survives that process best is one that makes their job as easy as possible: a clearly stated method, numbers that match between the text and the tables, and a Discussion that claims no more than the data support.
A paper that forces a rushed volunteer reviewer to hunt for the sample size, reconcile a table against the text, or guess at what statistical test was actually run is a paper that invites the kind of terse, frustrated comment described earlier in this guide, not because the reviewer is being unfair, but because friction in a fifteen-minute reading window turns into friction in the report that comes back.
What to check across methodology, statistics, writing, and formatting before you submit.
The reasons editors decline manuscripts before review, and how to avoid each one.
The errors reviewers catch most often, from misused tests to overstated effects.
Structure the response letter, disagree well, and handle contradictory reviewers.
The Reviewer 2 Generator reads an abstract free. The free assessment reads a whole manuscript and tells you where it stands. For a full review, PeerGenius runs a panel of specialist reviewers over the paper and returns their objections before an editor sees it.
PeerGenius is built by Neon Peach, LLC, a product studio.
Run the panel over your manuscript and see the objections a rushed reviewer would raise.
Get free assessment