River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

Response to Reviewers Letter Template

Three documents and three sheets that split every numbered comment into the separate things it actually asks for.

Free download  ·  No account needed

Request Register

Reviewer 3, comment 4. One number, five requests.

RowWhat it asks forKindIn the comment-by-comment draftDisposition
3.4aCounts are pooled across transects, ignoring the nesting structureAnalysisAnsweredMade
3.4bUnclear whether the counts are overdispersed, and if so Poisson is wrongAnalysisMissing, second sentenceMade
3.4cFigure 3 is unreadable at print size and its legend disagrees with Figure 2StructureMissing, parentheticalMade
3.4dThe four-week aggregation window needs justifyingReportingMissing, final sentenceMade in part
3.4eReplace the aggregation with a continuous time-varying termAnalysisMissing, final clauseDeclined

The draft written comment by comment answered comment 4 well and answered one fifth of it. Across three reports it answered all 24 comments and 33 of the 41 things they asked for.

The pack is three documents and three sheets and one rule. The rule is that a numbered comment is not one request. Reviewers write in paragraphs, so comment 4 asks for a random intercept, then asks in its second sentence whether the counts are overdispersed, then mentions in a parenthetical that a figure is unreadable, then wonders about the aggregation window. That is four requests wearing one number, and a fifth arrives in the last clause.

So the unit here is the request. Every numbered comment is split into its separable asks and each one becomes a row. Each row then needs three fields a letter usually leaves vague: a disposition in three values rather than the word addressed, the change that satisfies it, and the section, page and line where that change now sits. Two joins run across the register after that. Where one edit answers more than one reviewer, it goes in the letter under each of them. Where two reviewers want opposite things, it goes to the editor.

Everything else in this category answers the reviewer's numbering, because the numbering is what is printed. In the worked example that draft answers all 24 comments and 33 of the 41 things they ask for, and five of the eight it drops come from the reviewer who recommended rejection. Open the pack in River and the split happens first, before a word of response. Or download the sheets and keep the register beside the grant reporting pack.

Every sheet in the pack

The Request Register, the two joins that run across it, and the letter written from all three.

Request Register

41 rows, one per request, against 24 numbered comments. The column that does the work is the last one: a response with no location is a response an editor has to take on trust.

RowThe requestDispositionChange madeLocation in the revised fileAlso answers
1.1aA decline in verge-nesting species is asserted without a sourceMadeCited to the 2024 monitoring report with the interval, not the direction aloneIntro p.3 L62-66
1.2aWas the 2019 season excluded, and if so whyMade2019 included in the primary analysis, exclusion criteria statedMethods p.8 L176-1902.3a, 3.3a
3.4aCounts pooled across transects, ignoring the nesting structureMadeRandom intercept added, the estimate falls from 1.34 to 1.19 and both are reportedMethods p.10 L228-241
3.4eReplace the aggregation with a continuous time-varying termDeclinedFixed four-week rota leaves no within-window variation to estimate fromLimitations p.19 L452-459
2.6bAt least characterise adjacent land use if you cannot model itMadeLand use within 250 m tabulated per siteSupp. Table S7

Row 1.2a is one edit that answers three reviewers. It appears in the letter three times, under each of them, rather than once with a pointer.

Conflict and Overlap Map

The register run against itself. Two reviewers asking for opposite things is the editor's decision, and an author who settles it quietly has told the losing reviewer nothing.

KindRowsOne asksThe other asksResolutionTo the editor
Conflict1.3a / 3.6aCut the sensitivity analyses for lengthAdd two moreTwo added, all six moved to supplementary, main text shorter than beforeYes
Conflict2.2a / 3.5aMove Table 2 to supplementaryExpand Table 2 with the per-taxon breakdownKept and expanded, the other request declined with the conflict namedYes
Overlap1.2a / 2.3a / 3.3aWhy was 2019 excludedSame question, three phrasingsOne change, cross-referenced under all threeNoted
Overlap1.4a / 2.7c / 3.7aThe abstract overstates the designCausal language is not earnedNarrowing the abstract and the discussion answers all threeNoted

Two conflicts and four shared changes out of 41 requests. Both conflicts went into the cover note, where they read as a decision explained rather than a problem discovered.

Disagreement Log

Six declines, each with a reason class and the evidence it rests on. The last column is the test: write the reviewer's best comeback, and if it comes easily the decline is not finished.

RowThe requestReason classWhat it rests onDone insteadTheir strongest reply
3.4eContinuous time-varying termNot supported by the dataSurvey visits were on a fixed four-week rotaTwo-week window reported as a sensitivity analysisIt would still absorb seasonal trend
2.5aLandscape floral resource covariateWould require new data collectionCosted at one season, three observersNamed as a confounder with the likely direction of biasA remotely sensed proxy could substitute
2.2aMove Table 2 to supplementaryOutside the question the paper asksReviewer 3 asks for the same table to be expandedKept and expanded, conflict handed to the editorThe paper is long and something has to move
3.7bReframe the analysis causallyContradicted by the literatureAssignment follows road class, which also predicts the outcomeCausal framing dropped rather than reframedThen the paper says less

Row 2.2a is the only decline resting on a scope boundary, which is the weakest class there is. It is also the one the cover note asks the editor to rule on rather than settling.

Response Letter

Written in the reviewer's numbering, because that is what they will read it in. Underneath it, each part answered separately and labelled, so a reviewer can see their second sentence was read.

ElementWhat goes in it
OpeningThanks once. Then the count: 24 comments, 41 requests, 30 made, 5 made in part, 6 declined.
Summary of changesFour to six substantive ones, so a reader who stops here knows what happened to the paper.
The change nobody asked forA transcription error found while rebuilding the analysis, disclosed in the opening rather than left to be noticed.
Response 4.1, madeDisposition, then the change, then the location. Never the other way round.
Response 4.2 to 4.5The four requests a single response under comment 4 would have swallowed, each answered under its own label.
Response 4.5, declinedReason class, the evidence, and the smaller change made instead. Never a bare scope assertion.

Reviewer 3 recommended rejection, and their eight comments hold 16 requests, more than either colleague. Those are the rows that decide the paper, so they get answered at the same depth as everything else.

What's in the pack

01

Request Register sheet

One row per request, numbered reviewer, comment, part, with the clause quoted verbatim rather than summarised. A summary is already an interpretation, and the interpretation is what quietly narrows a request into something easier to answer.

02

Conflict and Overlap Map sheet

The register run against itself. Where one change answers two reports, and where satisfying one report means refusing another. In the worked example that is four shared changes and two real conflicts.

03

Disagreement Log sheet

Every decline with a reason class, the evidence behind it, and the reviewer's strongest reply written out. If that reply comes easily, the decline is not finished and it will be back.

04

Response Letter

In the reviewer's numbering, with each part of a multi-part comment answered separately and labelled. Disposition first, then the change, then the location, because the reader is scanning for the disposition and it gets missed under two sentences of thanks.

05

Editor Cover Note

Under a page, and a different document from the letter. What changed, the count, the conflicts that need adjudicating, and the declines in one line each. Nothing about any reviewer.

06

Revised Manuscript Note

What is in the package, how much of the file moved, and any change nobody asked for. An unexplained difference between two versions costs more attention than the correction it hides. If the decision goes the other way and the paper moves journals, the reformatting is a separate job.

07

The pre-resubmission check

Six counts before anything is uploaded: requests, made, made in part, declined, missing locations, missing responses. Line numbers move when text is added above them, so every location gets opened and confirmed in the current file.

08

A numbered comment is not one request

The standing space rule every prompt reads first. Anything a reviewer could later say was not addressed becomes its own row, including a bare question, because an unanswered question reads as an evasion.

How to use it

  1. 1

    Open in River, or download it

    Open the pack in River and the agent splits the reports with you before drafting anything, or download the three CSV sheets and three Word documents instantly, filled in with the worked example.

  2. 2

    Split before you draft

    Read each comment twice: once for what it is worried about, once for how many separate things it asks. Those readings disagree more often than not, and the second one is what goes in the register.

  3. 3

    Give every row a change and a location

    Section, page and line in the revised file rather than the original. Leave it blank rather than guess: a blank shows up in the check, a wrong line number does not, and it is the worse of the two.

  4. 4

    Argue the declines, then hand up the conflicts

    Six declines in the worked example, each with a reason class and evidence, the way an evidence register grades a claim. The two conflicts go to the editor in the cover note.

Frequently asked questions

Is this template free?

Yes. Download the whole pack as CSV sheets and Word documents with no signup and no credit card. Edit with AI is a separate, optional path that has the agent split your own decision letter and open the register. The template library holds the rest of the research packs.

What format are the downloaded files?

CSV for the three sheets and Word documents for the response letter, the manuscript note and the editor cover note, in one zip. They open in Excel, Numbers, Sheets, Word, Pages and Google Docs with nothing to convert. Inside River the same content opens as live Docs and Sheets.

Why split comments up when the journal numbered them?

The letter still uses the journal's numbering, and it has to. The split changes what goes underneath each number. Journals ask you to address the specific points each reviewer made, and the specific points are not the same object as the numbered paragraphs they arrived in.

Do I really need a line number for every change?

Yes, and it is the cheapest thing on the list. A response without one asks the reviewer to search for the change, and searching is what turns into a query and then into a second round. Give the section, page and line in the tracked-changes file rather than the original.

Is it safe to decline a reviewer's request?

Routinely, if it is argued. Six of the 41 requests in the worked example are declined. What is not safe is a decline resting on the sentence that it is beyond the scope of the present study, with nothing after it. Name a reason class, show the evidence, and concede something smaller.

What if two reviewers want opposite things?

That belongs to the editor, and it goes in the cover note rather than being settled quietly in the letter. Reviewers often see each other's reports, and may be asked to comment on the criticisms of other reviewers, so a side picked in silence tends not to stay silent.

Does River write the responses?

It writes the drafts and holds the record. What to concede, what to defend and what the data will support are judgements belonging to whoever signs the letter. The split, the locations, the cross-references and the check are the mechanical part, and that is where the round gets lost.

Find out how many things your reviewers actually asked for

Send the decision letter and the manuscript. The split comes back before any draft response does, with the count both ways.

Edit with AI