收到 comments 的第一件事
一：休息 24 小時。第一次讀 comments 大多會生氣。讓自己冷靜後再回頭——你會發現 80% 的批評是有道理的。
二：分類 comments。用 spreadsheet 把每位 reviewer 的每點分為：(a) 同意 + 容易改、(b) 同意 + 需新實驗/分析、(c) 不同意但能解釋、(d) 不能改 (技術上 / 範圍上)。
三：寫 point-by-point response.每點都要回，逐字引用 reviewer 原話 → 我們的回應 → 文中改了哪裡（頁碼/行號）。
1. Take 24 hours. First read = anger. Cool down, come back — you'll find 80% of critiques are reasonable.
2. Classify comments. Use a spreadsheet to bucket each reviewer comment: (a) agree + easy fix, (b) agree + new experiment/analysis needed, (c) disagree but explainable, (d) can't fix (technical / scope).
3. Write point-by-point response. Every point gets a reply: quote reviewer verbatim → our response → where we changed (page/line).
一、5 種決定信
| 決定 | 意涵 | 行動 |
|---|---|---|
| Accept | 恭喜，幾乎不用修改Congrats — minimal/no changes | 校稿、簽 copyright formProofread, sign copyright |
| Minor Revision | 基本接受，小修Basically accepted, small fixes | 2 週內回覆，95%+ 接受率Revise within 2 weeks; ~95% acceptance |
| Major Revision | 有機會但要做大量工作Real chance but significant work | 2-3 月內回覆，~50-70% 接受率Revise in 2-3 months; ~50-70% acceptance |
| Reject & Resubmit | 「重投」當新稿、舊評論可參考Resubmit as new paper; reuse comments | 值得重做嗎？做完當新案投回Worth redoing? If yes, treat as new submission |
| Reject | 不適合本刊；不是論文沒救Not for this journal; not unsalvageable | 參考意見後改投 next-tier 期刊，不要原稿丟去下一家Address comments, submit to next-tier; don't resubmit unchanged |
二、Response Letter 結構
總覽信
感謝 editor 與 reviewers 時間 + 1 段 highlights：「我們新增了 3 個實驗、增加了 1 個 cohort 驗證、重做了 statistical analysis。所有 comments 已逐點回應。」
Thank editor + reviewers + 1 paragraph of highlights: "We added 3 new experiments, included 1 validation cohort, redid the statistics. All comments addressed point-by-point."
Reviewer 1 逐點回應
每點 4 元素：(1) 引用 reviewer 原話 (italic) → (2) 我們的 response → (3) 文中改動 (含頁碼/行號) → (4) 必要時引新數據/圖。
Each point has 4 elements: (1) quote reviewer (italic) → (2) our response → (3) manuscript changes (page/line) → (4) cite new data/figures as needed.
Reviewer 2, 3...
每位 reviewer 獨立 section。即使 R2 跟 R1 同問題，也分別回應 (但可寫 "as detailed in our response to R1")。
Each reviewer in its own section. Even if R2 asks the same as R1, reply separately (you can write "as detailed in our response to R1").
改動 manuscript
附上 (a) clean version + (b) tracked-changes / coloured version。讓 editor 一眼看出修改範圍。
Attach (a) clean version + (b) tracked-changes / coloured version so editor sees changes at a glance.
三、回應對照
❌ 情緒化反駁
「Reviewer 顯然沒理解我們的方法。我們已經做了非常多分析，沒辦法再做更多了。這個批評不公平。」
"The reviewer obviously didn't understand our method. We have done extensive analysis already and cannot do more. This criticism is unfair."
✅ 禮貌 + 有依據
「Reviewer 1, comment 3: 'The clustering resolution of 0.6 seems arbitrary; please justify or test alternatives.'
Response: We thank the reviewer for this important point. We have now systematically tested resolutions from 0.2 to 1.2 in steps of 0.2 (new Supplementary Figure S4). Cluster identity stability (Adjusted Rand Index vs original 0.6) plateaued at resolution ≥0.6, supporting our original choice. We have added the rationale on p.7, lines 12–18 and the comparison plot as Supp Fig S4.」
"Reviewer 1, comment 3: 'The clustering resolution of 0.6 seems arbitrary; please justify or test alternatives.'
Response: We thank the reviewer for this important point. We systematically tested resolutions 0.2 to 1.2 in steps of 0.2 (new Supplementary Figure S4). Cluster identity stability (Adjusted Rand Index vs the original 0.6) plateaued at ≥0.6, supporting our choice. Rationale added on p.7, lines 12–18, with the comparison plot as Supp Fig S4."
❌ 不同意但無依據
「我們不同意 reviewer 的看法。我們認為原稿已經夠了。」
"We disagree with the reviewer. We believe the original manuscript is sufficient."
✅ 禮貌不同意 + 依據
「我們感謝 reviewer 的意見，但對此點我們有不同看法。原因：(1) 我們原文 p.5 已說明 SCTransform v2 是經過 benchmark 推薦的方法 (Choudhary & Satija, 2022)；(2) 我們在新 Supp Fig S5 提供 LogNormalize vs SCTransform 的並排比較，cluster 結構幾乎一致 (ARI=0.91)。我們希望這已充分回應 reviewer 的疑慮，但若 reviewer 認為仍需更換，我們也願意配合。」
"We thank the reviewer but respectfully disagree on this point. Reasons: (1) p.5 already cites Choudhary & Satija (2022) showing SCTransform v2 is the benchmarked recommendation; (2) we now provide a side-by-side LogNormalize vs SCTransform comparison in Supp Fig S5 — cluster structure is nearly identical (ARI=0.91). We hope this addresses the concern, but are happy to switch methods if the reviewer considers it necessary."
四、輪到你當 reviewer
大多數研究者在投稿 5-10 篇後會被邀請當 reviewer。好的 review 既能幫作者改善論文、也能訓練自己批判性思維。Reviewer 訓練資源：Publons Academy, ASAPbio Reviewer Trainee Programme, Sense About Science。
結構建議：
- Summary (1 段)：用自己話總結作者做了什麼（讓 editor / authors 確認你看懂）。
- Major points (5-10 條)：影響核心結論的批評，分點編號。
- Minor points (5-15 條)：typos, figure issues, citations。
- Confidential to editor：對 novelty / fit / scientific misconduct 的隱密意見。
Most researchers are invited as reviewers after 5-10 published papers. A good review both improves the paper and trains your critical thinking. Training: Publons Academy, ASAPbio Reviewer Trainee Programme, Sense About Science.
Recommended structure:
- Summary (1 paragraph): in your own words, what the authors did (lets editor + authors confirm you understood).
- Major points (5-10): critiques affecting core conclusions; numbered.
- Minor points (5-15): typos, figures, citations.
- Confidential to editor: private notes on novelty / fit / suspected misconduct.
五、收到回覆後的 7 步驟
🌳 從收到 comments 到 resubmission
📝 自我檢測
1. 收到 Major Revision 的最佳心態是？
1. Best mindset on receiving Major Revision?
2. Point-by-point response letter 中，每個 comment 的最佳回應結構是？
2. Best response structure per comment in a point-by-point letter?
3. 當你被邀請當 reviewer 時，最嚴重的倫理違規是？
3. As a reviewer, the most serious ethical violation is?