The decision letter says major revisions. Somewhere in your inbox are two reviews of very different quality, one thoughtful and one apparently written on a train, and between them thirty-one numbered comments. What you produce next is a document that most authors have never been taught to write, and it carries more weight in the second decision than the revised manuscript does, because it is what the editor reads first and, in many cases, most carefully.
The response letter is not a formality attached to the revision. It is the argument for why the revision is sufficient. Editors use it to decide whether to accept your paper directly, to return it to the same reviewers, or to conclude that the authors did not engage. Reviewers who receive it read almost nothing else before forming a view; they scan for their own comments and check whether each one was answered.
This page covers the document structure that editors expect, the numbering and formatting that make it easy to verify, the three kinds of response and when each is appropriate, how to disagree with a reviewer and be believed, what to do when two reviewers demand opposite things, the requests that are cheaper to satisfy than to fight, and a full worked example from comment to reply.
A complete revision package normally has four parts, and missing one of them creates avoidable friction at the editorial office.
The first is a short cover letter to the editor. Not the point-by-point document: a separate half-page that thanks the editor, states that the manuscript has been revised in response to all comments, summarizes the three or four most substantial changes in plain language, and confirms anything the journal requires such as authorship, data availability or competing interests. Editors handling forty revisions a week want the summary before the detail.
The second is the point-by-point response, which is the main document and the subject of the rest of this page. The third is a revised manuscript with changes marked, usually as tracked changes or highlighted text. The fourth is a clean copy of the revised manuscript with no markup. Most journals ask for both versions; supply both even when only one is requested, because reviewers differ in which they prefer to read.
If your revision includes new analyses, new supplementary files or updated data deposits, list them explicitly in the cover letter with their filenames. Files that arrive without being announced sometimes do not reach the reviewers.
The organizing principle is verifiability. A reviewer should be able to find their comment, read your answer, and check the manuscript, without searching. Everything below serves that.
Use line numbers in the manuscript, continuous rather than restarting on each page, and refer to them. Keep the response document as a single file, and give it a filename that identifies the manuscript. If the journal’s submission system strips formatting, use plain text markers such as REVIEWER COMMENT and RESPONSE in capitals rather than relying on colour, which frequently does not survive conversion.
Length is not a virtue in itself, but this document is often longer than the manuscript and that is normal. What matters is that each individual response is short enough to read in one pass. Three or four sentences for most comments, a paragraph for the substantial ones, and any long technical justification moved into an appendix to the response letter that the interested reviewer can consult.
Every comment gets one of three answers, and clarity about which you are giving is half of what makes a letter persuasive.
The great majority of comments should get this. State the agreement briefly, describe the change, quote it, give the location. Do not pad it. “We agree. The sentence has been revised to read: […] (p. 14, lines 302 to 305)” is a complete response and reviewers appreciate the economy.
One rule that saves enormous trouble: when the reviewer has misread something, treat it as evidence that the manuscript is unclear rather than as an error on their part. The response “as we already stated on page 9” is the single most damaging sentence available to you, because it is simultaneously correct and useless: if a specialist reader of your paper misunderstood it, other readers will too. The productive version is “We had not made this clear enough. We have rewritten the passage on p. 9 to state explicitly that …”.
Used when the concern is legitimate but the proposed remedy is not, or is impossible. Name the part you accept, make a real change in response to it, and explain specifically why the remainder is not feasible or not appropriate. Then offer a substitute: an alternative analysis, a limitation stated in the discussion, a supplementary table, a narrowed claim.
The substitute is what makes this work. A partial acceptance without an offer reads as a refusal with extra words.
Legitimate, and necessary sometimes, because reviewers are wrong with some regularity. It works when four elements are present and fails when any is missing. Acknowledge the concern in the reviewer’s own terms so it is clear you understood it. Give the reason for disagreeing, and make it a substantive reason with a citation, a re-analysis or a piece of evidence, not a preference. Show that you took the possibility seriously, ideally by reporting that you ran the suggested analysis and what it showed. And offer a concession anyway: the point added to the limitations, the alternative reported in a supplement, the claim softened.
Keep the tone level. Not deferential, not combative. Reviewers are reading for whether you engaged, and an author who disagrees calmly with a reason attached is read as competent, while one who disagrees defensively is read as difficult regardless of who is right.
Limit the number. Disagreeing with three comments out of thirty is unremarkable. Disagreeing with fifteen, however well argued, changes how the whole letter is read and usually produces a rejection at round two, because the editor concludes the paper cannot be brought to a publishable state through this process.
Four situations come up often enough to be worth planning for.
One wants the theoretical framework expanded, the other wants it cut in half. Do not silently follow one and hope the other does not notice, which is the most common approach and the one that produces an angry second review.
Name the conflict explicitly in your response to both comments, explain which course you took and why, and address the note to the editor. “Reviewer 1 asks us to expand the discussion of self-determination theory, while Reviewer 2 recommends shortening the theoretical section. We have taken a middle course, retaining the framework but reducing it from four paragraphs to two and moving the extended treatment to Supplement A. We would be glad to follow either recommendation more closely if the Editor prefers.” This hands the arbitration to the person whose job it is, and editors respond well to it.
Sometimes appropriate, sometimes a request for a citation that has no relevance to your paper. Read the suggested references. If they are genuinely relevant, cite them and say so without comment; this happens frequently and is not in itself improper. If they are not, cite the one or two that are closest and explain briefly why the others were not incorporated. Adding irrelevant citations to appease a reviewer degrades your paper and is visible to editors who look.
If the pattern is egregious, a large number of self-citations demanded with no substantive comments attached, you can raise it privately with the editor. This is a recognized problem that editors track.
The instinct is to argue. The cheaper path, in almost every case, is to run the analysis, report it in a supplement, and note in the response what it showed and why you have retained your original approach as primary. Running it costs an afternoon; refusing it costs a review cycle, and if the reviewer suspects you refused because the result would be inconvenient, it costs more than that.
Reserve outright refusal for analyses that are genuinely inappropriate rather than merely suboptimal, and when you refuse, explain the methodological reason rather than asserting one. If the request is for something like observed power, say precisely why it does not answer the question and offer the analysis that does.
A four-line review consisting of “the study is not novel and the writing is poor” gives you nothing to work with. Answer what can be answered: identify the novelty explicitly with a sentence in the introduction stating what the study adds, and have the manuscript language-edited, saying so in the response. Then, in the note to the editor, ask politely whether more specific guidance is available on which sections were unclear. Editors sometimes have the reviewer’s confidential comments and can direct you, and they are aware which of their reviewers write thin reports.
The comment, reproduced as it would appear in the response document:
Reviewer 1, comment 3: “The authors use a convenience sample of undergraduate students and then generalize to working adults throughout the discussion. Furthermore, the mediation analysis assumes a causal ordering that the cross-sectional design cannot establish. I am not convinced the conclusions are supported.”
A weak response, of a kind that appears constantly: “We thank the reviewer for this comment. We have added a limitation acknowledging the convenience sample and the cross-sectional design. We believe the conclusions are supported by the data.” It concedes nothing concrete, quotes nothing, and disputes the reviewer’s central point with an assertion. It will come back.
A response that closes the comment: “We agree with both points, and on rereading we found that the manuscript did generalize beyond what the sample supports. We have made three changes. First, the title, the final sentence of the Abstract and five sentences in the Discussion have been revised to refer to university students rather than to working adults; all five are listed with line numbers in the table at the end of this letter. Second, we have added the recruitment funnel to the Method (p. 7, lines 148 to 156): 412 students were invited through the departmental participant pool, 297 began the survey, and 264 provided complete data, and we now report how the sample compares with the faculty enrolment data on age, sex and year of study. Third, the mediation analysis is now presented as a test of whether the pattern of associations is consistent with the theoretical model, rather than as evidence of a mechanism. We have added the two alternative orderings to Supplement C, where they fit equivalently, and we state in the Results (p. 18, lines 391 to 397): “Because all variables were measured concurrently, the models depicted in Figure 2 and the two alternatives reported in Supplement C are statistically indistinguishable; the ordering was specified on theoretical grounds and should not be read as evidence of temporal precedence.” We have also removed the practice implication that assumed a causal effect, and replaced it with a statement of the prospective design that would be required to test it.”
The second response is longer, and every additional sentence is a verifiable fact. It concedes the substance immediately, which costs nothing since the reviewer was right, and it spends its length on evidence rather than on defense. That is the shape to aim for.
Most journals give between 30 and 90 days for major revisions and considerably less for minor ones. Extensions are granted almost automatically if you request them before the deadline and give a reason; requesting one afterwards, when the system has already sent the reminder or withdrawn the manuscript, is harder. If new data collection or a re-analysis is involved, ask early and ask for more time than you think you need.
The errors that reliably cause trouble, in order of frequency: leaving a comment unanswered; answering in generalities without page and line numbers; making a cosmetic change and describing it as a substantive one, which reviewers detect immediately by opening the manuscript; being defensive in tone; changing the manuscript without saying where; disagreeing with too many comments; and, occasionally fatal, making changes in the response letter that are not actually present in the revised manuscript, which destroys trust for the remainder of the process.
One habit is worth adopting from the first round: keep a table at the end of the response listing every change with its location. It takes twenty minutes, it makes the editor’s verification trivial, and it forces you to check that every change you claimed is really in the file.
Finally, on rejection after revision. It happens, and the response document you wrote is not wasted: it is a detailed record of the objections your paper attracts, and the manuscript you send to the next journal should already contain the answers. Authors who transfer the improved manuscript and discard the reviewer knowledge tend to receive the same comments again from different people.
As long as the comments require, which is frequently longer than the manuscript itself. What matters is that individual responses are short. Three or four sentences for a routine comment, a paragraph for a substantial one, and any extended technical argument moved to an appendix of the response letter. A letter that is long because it quotes the revised text is fine; one that is long because it argues is not.
Yes, including trivial ones and ones already covered elsewhere. Skipping a comment is the most common cause of a second revision round, because the reviewer searching for their own point and not finding it concludes they were ignored. If a comment duplicates another, answer briefly and cross-reference the fuller response.
Yes, and sometimes you must. A disagreement is persuasive when it acknowledges the concern in the reviewer’s own terms, gives a substantive reason with a citation or a re-analysis rather than a preference, shows you took the suggestion seriously enough to test it, and offers a concession such as a limitation or a supplementary analysis. Keep the number of disagreements small; a letter that disputes half the comments reads as unwillingness to revise.
Name the conflict openly in your response to both comments, explain the course you chose and why, and address the point to the editor, offering to follow either recommendation if they prefer. Editors expect to arbitrate and respond well to being asked. The failure mode is quietly satisfying one reviewer and hoping the other does not notice, which usually produces a sharper second review.
Both, unless the journal explicitly forbids one. Reviewers differ in preference, and editors often want the clean copy for readability and the marked copy for verification. Use continuous line numbers in the clean version, since that is what your response letter will reference, and check that the numbers in the letter match the file you actually upload.
Usually yes, and report it in a supplement while explaining why you retain your original approach as primary. Running it costs an afternoon; refusing it costs a review cycle and invites the suspicion that the result would have been inconvenient. Reserve outright refusal for analyses that are genuinely invalid, and when you refuse, give the methodological reason and offer the analysis that does answer the underlying concern.
Answer what can be answered, then ask. If the criticism is that the work lacks novelty, add an explicit statement of what the study contributes; if it is that the writing is unclear, have the manuscript language-edited and say so. Then include a courteous note to the editor asking whether more specific guidance is available. Editors can sometimes point you to the sections in question, and they are aware which reviewers submit thin reports.