Mentoring Students Through Their First Peer Review

The first peer-review letter can crush a student or train a scientist, and how a mentor runs that round decides which one happens.

Guide · 12 min read · Directive Publications Editorial Team

Before submission

• Co-write the cover letter once together, then let the student draft next time
• Agree authorship and contributions early
• Run a lab “hostile read” of the draft

Co-writing means one screen, two chairs, and you narrating what each sentence is for as you type the first paragraph: this is the paper, this is why it fits this journal specifically, this is the corresponding author with an institutional email that will still work in two years. Then hand the keyboard over and let the student write the rest while you watch. The next paper, they draft and you edit. Two failure modes sit either side of this. If you write every cover letter yourself, a postdoc can leave the lab having never written one. If you refuse to touch it at all, the student's first letter is a reworded abstract, which earns no attention from the editor and teaches the student nothing about why.

Settle authorship at the point the analysis plan is fixed, not the week of submission. Put it in a shared file everyone can see: each name, what they will actually do in plain words — designed the assay, ran the fieldwork, wrote the statistics section — and who is corresponding author. If the target journal uses a contributor-role taxonomy such as CRediT, use its role names from the start so the submission form becomes a transcription rather than a negotiation. Revisit the file whenever the scope changes: a new dataset, someone leaving, a rotation student running three months of experiments. When nobody writes it down, the question surfaces the week of submission, and the person being asked to decide whether a departing colleague earned a place is a first-year who cannot say no to anyone.

A hostile read is a colleague reading the draft the way an unsympathetic reviewer would, before anyone outside the lab sees it. Make it structurally real rather than a favour. Give the reader the target journal's aims and scope, and ask for a short reviewer report in the journal's own format — a recommendation, three major points, a list of minor ones — rather than comments in the margins. Give them a week, and schedule it two weeks before the intended submission date, not two days, so there is time to act on what comes back. The best hostile reader is close enough to judge the field and far enough away not to know what you meant. Tell the student in advance that it is supposed to sting slightly, and that the entire point is for the first hard read to happen in a room where they are safe. If the report comes back saying the paper is nice and the typos need fixing, you did not get a hostile read. You got a friendly one, and the real version is now waiting at the journal.

Resources: For Authors · Suggesting Reviewers

When reviews arrive

1. Read the letter yourself first
2. Sit with the student and sort comments into easy / hard / disagree
3. Assign response-letter ownership to the student with your edit pass
4. Teach tone: respectful, specific, evidence-based

Open it alone. Your face is the student's first and most trusted piece of evidence about how bad the news is, and they read it more accurately than you think. Read once, do something else for an hour, then read again with the sorting in mind. Then, before the meeting, send one short message that names the outcome plainly: major revision, two reviewers, mostly method questions, roughly three weeks of work, meeting Thursday. A student who gets that message stops rereading the letter at midnight. The alternative — forwarding the decision email with no comment, or with a one-word groan attached — leaves a gap in which they construct the worst possible version, and by Thursday you are arguing with that version rather than with the review.

Sort mechanically, before anyone forms an opinion about the review as a whole. Number every comment the way the reviewers laid them out, R1.1, R1.2, R2.1, splitting any paragraph that bundles three requests into three numbered items, because you will answer them separately anyway. Then three piles. Easy is wording, a missing citation, an undefined abbreviation, an unlabelled axis. Hard is a robustness check, a reanalysis, a section that has to be rebuilt because two reviewers got lost in the same place. Disagree is usually one or two items. Do the whole sort out loud the first time and say why each comment lands where it does, because that judgement is the actual thing you are teaching. Two rules to hand over while you work: volume is not difficulty, and the editor's own letter outranks the reviewers' — if the editor says the statistics must be addressed and the rest is at your discretion, that is the priority order no matter how forcefully a reviewer wrote. On the next paper, let the student sort first and you check. You will usually move one or two items between piles, and that short conversation is the lesson.

Ownership means the student drafts every response and you edit. Edit in tracked changes, and attach a one-line reason to any substantive rewrite, so they see the rule rather than only the corrected sentence. A useful rule of thumb: if you are rewriting more than about a third of the letter, stop editing and go back to talking, because the problem is the brief and not the prose. If a deadline is genuinely too close and the student is stuck, say out loud what you are doing — you are drafting the response to R2.3 because time has run out and it needs the modelling detail, they take the rest, and the next one is done properly. Naming it as an exception is what stops it becoming the norm. The version to avoid is the silent rewrite, where you take the file, return a finished letter, the paper is accepted, and the student has learned only that revision is something that happens to them.

Teach tone as three tests the student can apply without you in the room. Would this response be comfortable to read aloud with the reviewer sitting across the table? Does it name a section and line number in the revised file? Does every disagreement rest on evidence rather than on assertion or on how standard something is in the field?

Tone, in practice

Take a common shape of comment: a reviewer says the sample is too small to support the conclusion. A first draft written in the wrong hour comes out as, "The reviewer has clearly not understood our design. This sample size is standard in the field and the criticism is unfounded."

Three things are wrong with it, and they are worth naming separately rather than calling the whole thing rude. It blames the reader, which makes every later point sound defensive. It appeals to convention instead of evidence. And it changes nothing in the manuscript, so the editor sees a round that produced an argument and no revision.

The version that works has three parts — what you understood the reviewer to be asking, why you are not doing exactly that, and what you did instead: "We did not state our sampling rationale clearly, and have added it to Section 2.1 (lines 96-104), including the calculation the design was based on. We agree that the sample limits what can be claimed, and have narrowed the conclusion in Section 4 (lines 402-407) from a general statement to one about this population. A larger multi-site study is the obvious next step, and we now say so."

Have the student write both versions once, deliberately. Writing the letter they cannot send and then rewriting it teaches the difference far better than being told to be polite, and it makes the venting draft a legitimate private document rather than something to feel ashamed of.

What to model

• No blaming reviewers in the lab meeting
• No ghostwriting the entire response while the student watches silently
• Credit the student for revision craft in the next manuscript

The lab meeting where a review is read aloud in a mocking voice is memorable, and it teaches two wrong things at once: that criticism is illegitimate, and that this is how you talk about work you have been asked to judge. Students can carry that into their own reviewing. Say what each comment gets right first, including the badly written ones. Where a review really is unfair — a reviewer demanding their own citations, or plainly reviewing a different paper — name the specific problem, say what you do about it, which is one factual sentence to the editor, and then answer the comments that are answerable. Precision is the thing worth modelling. Contempt is not.

Ghostwriting is easy to fall into because it is faster, and the cost is invisible for exactly one paper. The alternative is to work in the margins rather than in the text, and to run at least one editing session with the student driving the keyboard while you talk, so they watch the decisions being made instead of receiving the output.

Credit has to be given where it counts, because revision craft leaves no trace. The published paper looks identical whether the student wrote the response letter or watched you write it. So name it in the group meeting: the response to Reviewer 2 is why this went through in one round, and here is who wrote it. Put it in the contributions statement of the next manuscript. Write it into reference letters in concrete terms rather than as diligence. Hand over the corresponding-author role once they have run a round, which is the form of credit that actually shows up on the record.

Integrity boundaries

Mentors should prevent:

• Gift authorship
• Conflicted reviewer suggestions
• Pressure to hide inconvenient data

Gift authorship is adding someone who did not contribute: the head of department, the person whose grant funded the work but who never saw the data, a collaborator added so a slot can be traded back later. The workable test is whether that person could defend the paper if a reader wrote to them with a hard question about the methods. If not, they belong in the acknowledgements. This is a mentor's boundary precisely because the student cannot hold it — a first-year cannot tell someone who will sit on their committee that they have not earned a place. If the expectation arrives from someone senior to you, make the criteria the subject rather than the person: point at the journal's authorship criteria and the contributions statement everyone has to sign, and ask which role they would like recorded. Many of these requests do not survive being written down.

Conflicted reviewer suggestions include recent co-authors, anyone at your own institution, your former supervisor, your student's committee members, and anyone you have already discussed the manuscript with. Many journals check these, and a list of friends is not a shortcut. The mild outcome is that the whole list is discarded and the editor thinks less of the submission; the bad outcome is an integrity query attached to a student's first paper. Teach the positive version instead: suggest people by expertise, with one line each saying why, ideally researchers whose work you cite and have never worked alongside. Teach the exclusion request as a separate and entirely legitimate tool — you may ask that a particular person not review, with a brief factual reason, and that is a normal thing to put in a cover letter.

Pressure to hide inconvenient data almost never arrives as an instruction to hide anything. It arrives as a suggestion that the replicate did not work so leave it out, or that the pre-registered outcome is confusing so lead with the secondary one, or that the outlier is obviously wrong so drop it. Give the student one rule that covers all three: anything that shaped the analysis appears in the paper or its supplement. An excluded outlier is reported together with the exclusion rule that was set in advance and, where it matters, the result with and without it. A failed replicate is described. A pre-registered primary outcome is reported first even when the secondary one is more interesting. If the pressure is coming from a co-author, or from someone senior to you, put your position in an email rather than settling it in a corridor, and take it to the research integrity office before the paper is submitted rather than after it is published. A correction notice attached to a first paper costs the student far more than a difficult conversation costs you now.

See: Authorship guidance · Peer Review Process

Bottom line

The measure of a first round is not whether the paper was accepted. It is whether the student could run the second one with you reading only the final draft. If they can sort a letter, hold a disagreement without blaming anyone, and tell you which comment is the expensive one, the round did its job even if the decision was a rejection.

Peer review is a curriculum. Treat the first round as teaching, not only triage. More at Directive Academy.