How to Respond to Reviewers When the Comment Is Hard

Before writing anything, put every comment into one of four buckets. The sorting takes an hour and it determines the shape of the whole revision, because the letter should spend its effort in proportion to what each comment threatens.

Free

typos, a missing reference, a mislabelled axis. Fix them, reply in one line, do not spend paragraphs being grateful.

Cheap and improving

a clarification, a limitation, a robustness check you can run this week. Do all of these, including the ones you think are unnecessary, because credit here buys latitude elsewhere.

Expensive but possible

a new analysis, a reworked framing, a fresh dataset. Decide explicitly whether the paper is worth it, and if you do it, say so prominently in the summary.

Threatening

a comment that, if accepted, unmakes the central claim. This is where your real effort goes, and it is the only category where a long response is justified.

This is the most common hard case and it has a standard resolution: name the contradiction, choose, and explain the reasoning so the editor can overrule you. What fails is quietly siding with one reviewer, because the other reads a revision that ignored their central point without saying why.

Quote both comments together in one place, so the editor sees the conflict in the form you saw it.

Look for the concern both comments share. A section one reviewer finds too long and another too compressed is usually a section that is not earning its length, and restructuring serves both.

State the trade-off in a sentence and hand the choice to the editor, who can see both reports and any third one you were not shown.

Cross-reference the answer from the other reviewer's numbered section, so neither has to hunt for it.

Do not resolve it by asserting that one reviewer misunderstood the paper unless you can point at the passage that misled them, in which case fix the passage.

A reviewer asks for an experiment that needs a licence you do not hold, a cohort that no longer exists, or two years of follow-up. Treating this as a disagreement is the mistake, because you do not disagree. Name the obstacle, answer the concern behind the request, and let the editor decide whether the gap is fatal.

Give the concrete obstacle

an ethics approval, a closed cohort, an instrument you no longer have access to. An unexplained refusal reads as unwillingness.

Answer what the request was protecting against. A reviewer asking for genetic confirmation of a mechanism is usually asking you to stop overclaiming, and weakening the claim addresses most of it.

Offer whatever independent evidence you do have, even if it is weaker than what was asked for.

Put the gap in the Limitations section of the paper, not only in the letter. That is what the reviewer checks on the second pass.

Invite the editor to rule against you if they consider the missing work essential. A clean decision now beats a second round that ends the same way.

Sometimes a request to add citations is a genuine gap in your reading, and sometimes it is coercive citation, where a reviewer pushes their own papers regardless of relevance. You usually cannot tell which, and you do not need to. Judge each suggested reference on whether it belongs, and let the pattern speak for itself.

Read the suggested papers. If they are relevant, cite them and say so plainly. Most such requests are legitimate.

If a reference does not fit, decline it the way you would decline any other comment: with a reason about the argument, never about the reviewer's motives.

Never suggest in writing that the reviewer is self-promoting. If you believe a request is coercive, raise it with the editor privately, and only when the pattern is clear, such as a long list of one group's work with no connection to your topic.

Add relevant work by others as well when a citation gap is real, so the reference list does not look like a concession to one reader.

Editors track this pattern across reviews. Reporting a specific concern once is reasonable; making it the theme of your letter is not.

You revised what was asked, and the report comes back with three concerns nobody raised before. This is legitimate in principle, because the reviewer is reading a changed paper, and exhausting in practice. The response is the same as the first round with one addition: note the sequence factually, once.

Answer the new comments properly. Refusing on the grounds that they are late is the fastest route to a rejection.

It is fair to note, without complaint, that a point concerns a passage unchanged since the original submission, and to ask the editor whether it should be addressed in this round.

Watch for a moving target

when a second round asks for something the first round explicitly said was not required, that is a question for the editor rather than for the reviewer.

Ask for an extension before the deadline rather than after it, if the new comments require real work. Most journals grant one.

If the paper has been through three rounds and the target is still moving, ask the editor directly whether there is a version of the paper they would accept. The answer, either way, is worth more than a fourth round.

Revision is where new problems get introduced: references added but not in the bibliography, a new table without a caption, an abstract that grew past the word limit, a figure that no longer matches its number in the text. These are exactly the errors that produce a technical check failure and add another week to a paper that has already waited months.

Check that every comment has a numbered reply, including the ones you answered in a word. A missing entry reads as an evasion.

Check every page and line reference against the clean file, since they all moved when you edited.

Check the reference list against the in-text citations in both directions, because new citations are the most common thing added during revision.

Check the abstract and any word or figure limits again, since a revised abstract frequently breaks them.

Run the revised manuscript through the manuscript checker for a full compliance pass before resubmission, which catches the mechanical problems that a careful reading of your own prose reliably misses.