Responding to Peer Reviewers: A Point-by-Point Guide

My first response letter as a corresponding author ran to eleven pages. About four of them were me explaining to Reviewer 2 why Reviewer 2 was wrong.

The reviewer had questioned our sample size. I had a power calculation. I had citations. I had, I thought, an airtight case. So I wrote a long, technically correct and faintly sarcastic rebuttal, changed almost nothing in the manuscript, and sent it back feeling rather pleased with myself.

The paper was rejected. The editor's note was short and polite. The gist was that the authors hadn't engaged with the reviewer's concern, and that the concern was really about whether readers would trust the subgroup results. I had answered a question nobody asked, and I'd missed the one they did.

Twenty years and a few thousand decision letters later, I can tell you that this is the most common way good papers die at revision. It usually isn't bad science. It's a response letter that argues with the reviewer when it should be helping the editor decide.

Who you're actually writing for

Most advice on responding to reviewers says your job is to satisfy the reviewers. I think that's slightly off, and the difference matters.

Your real audience is the handling editor. Reviewers advise and editors decide. In many cases the revised manuscript goes back to the same reviewers, but the editor reads your letter first and often decides whether to send it out again at all. A busy editor wants to finish your letter thinking that every point was addressed, that the changes are easy to find, and that any disagreements are reasonable. You aren't trying to win the argument. You want to make the editor's decision easy.

Once you read the letter that way, a lot of tone problems sort themselves out.

Read the decision letter twice, a day apart

The first read is for your feelings. Do it, close the file, and don't reply to anyone. Not your coauthors, and definitely not the editor.

The second read, the next day, is for information. That's when I'd start with the editor's own paragraph, because it's the most important text in the whole email and people routinely skim it. Editors often tell you outright which comments are essential. Look for phrases like "in particular," "the main concern," or "please pay special attention to." Those are instructions. If the editor writes "Reviewer 1's point about confounding is critical," that comment is your revision, and everything else is supporting work.

Triage every comment before you write a word

Copy every reviewer comment into a working document, verbatim, and number them (R1.1, R1.2, R2.1 and so on). Split long comments that bundle several requests. Then sort each one. I use four bins:

The "Clarify" bin is where most people go wrong. When a reviewer misunderstands your method, the instinct is to explain it to them in the response letter and leave the manuscript alone. Don't. The reviewer is your first reader, and they've just told you the text doesn't work. Fix the text.

If you end up with more than two or three comments in the "decline" bin, look again. Sometimes you really have drawn a difficult reviewer. More often you're defending sentences you're attached to.

How to structure the response letter

There's no universal format, but the one that works best has four parts for every comment: the comment, your response, the change, and where to find it. Editors love it because they never have to hunt.

Open with a short note to the editor. Thank them once, sincerely and briefly. Summarize the major changes in a few sentences, especially anything the editor flagged. Then go through the comments in order.

Here's a response I'd send back to the authors:

R1.3: The authors should adjust for socioeconomic status.

Response: We thank the reviewer for this comment. We agree this is important and have revised the manuscript accordingly.

It doesn't tell me what changed, how, or where. "Revised accordingly" is the most overused phrase in revision letters, and it makes editors suspicious. Here's the version that gets through:

R1.3: The authors should adjust for socioeconomic status.

Response: We agree that socioeconomic status is a plausible confounder. We have added area-level deprivation index (quintiles) to all adjusted models. The main estimate changed slightly, from an HR of 1.42 (95% CI 1.18 to 1.71) to 1.37 (1.13 to 1.66), and our conclusions are unchanged. We now note in the limitations that individual-level income was unavailable.

Change: Methods, Statistical analysis, paragraph 2; Table 2; Discussion, Limitations, paragraph 3 (tracked version, lines 188 to 192 and 402 to 405). New text: "Models were additionally adjusted for area-level deprivation…"

Quote the new text. It costs you a few lines, and it means the reviewer can approve the change without opening the manuscript. That alone speeds up second-round reviews.

On thanks: say it once at the top and then stop. Forty instances of "We thank the reviewer for this insightful comment" read as filler, and reviewers notice. "We agree" or "Good catch" is warmer and shorter.

Pushing back without starting a fight

You're allowed to disagree. Editors expect it, and a letter that accepts every single request can look as if the authors don't believe in their own design. What matters is how you do it.

A few years ago I worked with Marta, a biostatistician on a pragmatic trial, whose reviewer insisted that the per-protocol analysis should be primary. Marta was right that it shouldn't be. The trial was registered with intention-to-treat as primary, and switching would have been outcome-switching, which is exactly what CONSORT and the other EQUATOR reporting guidelines exist to prevent. Her first draft said, more or less, "This is a basic misunderstanding of trial methodology."

Her final version went like this:

We understand the reviewer's interest in the effect among adherent participants. Because intention-to-treat was prespecified as the primary analysis in our registered protocol, changing it now would depart from the registration. We have therefore kept ITT as primary and expanded the per-protocol analysis, which now appears in the main text (Results, paragraph 4; Table 3) rather than the supplement. The two analyses agree in direction and magnitude.

The reviewer accepted it without comment. That letter worked because it named the concern behind the request, gave a reason grounded in something external (the registration) rather than in Marta's opinion, and still gave the reviewer something they could see.

That's the pattern I teach. Acknowledge the underlying concern, give your reason with evidence, and offer a concession. Some sentences I find myself suggesting again and again:

• "We considered this carefully and have chosen not to make this change, because…"

• "We agree this would strengthen the work, but it requires data we did not collect. We now state this explicitly as a limitation (line 410)."

• "This analysis is beyond the scope of the current paper, which focuses on…; we have added it to the future directions."

• "We may not have been clear. We have rewritten this paragraph to make the distinction explicit."

Write to the comment and not to the person. Never use "clearly," "obviously," or "as stated in the manuscript." Every one of those means "you didn't read carefully," and the reviewer will hear it that way.

When the request is inappropriate

Occasionally a reviewer asks you to cite five of their own papers, or asks for something that crosses an ethical line. Coercive citation is a recognized problem, and COPE's guidance for editors and reviewers treats it as such. You can cite work that is genuinely relevant. For the rest, decline politely in the response letter and, if it's egregious, mention it in a confidential note to the editor. Editors want to know about it.

When reviewers contradict each other

Dan, an industry scientist I mentored through his first academic submission, got two reviews that pulled in opposite directions. Reviewer 1 wanted the assay validation moved to the supplement. Reviewer 2 wanted it expanded in the main text. Dan, eager to please, tried to do both. He shortened the section and added a new paragraph elsewhere, and the result satisfied neither reviewer and confused the editor.

When reviewers conflict, pick the option that best serves the paper, then say so openly in both responses: "Reviewer 2 requested the opposite change (see R2.4). We have kept the validation in the main text but condensed it, and moved the full protocol to Supplementary Methods." Editors deal with conflicting reviews every week, and they'd much rather see you make a reasoned call than a contortion. If the conflict is fundamental, it's fine to email the editor before resubmitting and ask which direction they prefer. That email is short and saves a round.

Track every change, or the resubmission falls apart

This is the unglamorous part, and it's where I see the most avoidable errors. Line numbers in the response letter don't match the manuscript. A table was updated but the abstract still quotes the old number. A reviewer's comment was answered in the letter but the edit never made it into the file.

Keep a change log from day one. It can be a simple spreadsheet with one row per comment and these columns: comment ID, bin, owner (which coauthor), status, manuscript location, and a "knock-on effects" column. That last column is the one people skip. Changing one estimate can ripple into the abstract, the results text, a figure, the discussion and the conclusions. Write down every place a number lives.

Some practical habits that save resubmissions:

• Produce the tracked-changes version and the clean version from the same final file, at the end, never in parallel.

• Add line or page references to the response letter last, once the text is frozen.

• Re-check your reporting checklist. If you submitted a PRISMA or CONSORT checklist, its page numbers are now wrong. Update it.

• Search the manuscript for every number you changed, including in the abstract.

• If authorship or contributions changed during revision, every author must agree in writing, consistent with the ICMJE authorship criteria, and your CRediT contributor statement should be updated to match.

A pre-resubmission checklist

1. Every comment has a numbered response, including the ones that just say "minor typo."

2. Every response points to a specific location, and the location is correct.

3. The editor's flagged concerns are addressed first in the cover note.

4. All numbers match across the abstract, text, tables and figures.

5. The reporting checklist, data availability statement and funding statement are current.

6. A coauthor who didn't do the revision has read the letter cold, purely for tone.

7. You've resubmitted within the deadline, or asked for an extension before it passed.

Item six is my favorite and the most skipped. The person who did the revision can't hear their own defensiveness. Someone who wasn't in the trenches can.

Keeping the context together

Much of the pain in a revision comes from the fact that the pieces live in different places. The reviews sit in one email, the PDFs you need to cite to justify a decision are in another folder, the manuscript is in a shared document, and the change log is a spreadsheet that one coauthor forgot to update. When a reviewer says "the authors ignore recent work on X," you end up searching, reading, summarizing and then writing the new paragraph across four separate tools.

This is the sort of work where an AI-native research workspace like PACR earns its keep. It connects the search, reading, summarizing and writing steps, so the literature you find while answering a reviewer stays linked to the paragraph you write in response. A realistic workflow is to search for the literature a reviewer says you missed, summarize what's actually relevant, draft the new text next to it, and share it with coauthors without copy-pasting between apps. It won't decide whether to push back. That's still your judgment. It does cut down on the clerical work that causes so many mismatched line numbers.

If you want a sense of how authors and reviewers actually experience this process, there's a decent body of research on peer review itself; a PubMed search on author responses and revision is a good starting point and humbling reading for anyone who thinks their reviewer was uniquely unfair.

One last thing

A major revision decision is good news. I know it doesn't feel that way when you're reading criticism of something you've worked on for two years. But the editor is telling you the paper is publishable if you do the work, and most rejections never offer that.

The next time a decision letter lands, don't open the manuscript first. Open a blank document, paste in every comment, number them, and sort them into bins. That hour of triage will shape a better revision than a week of rewriting. And if you'd like your literature, notes and drafts in one place for it, you can start a PACR workspace before the next round arrives.