The first invitation to review usually lands in your busiest week. Plenty of early-career researchers accept because it feels like recognition, leave it until the deadline, then send three lines, “solid paper, needs English editing”, with a minor revision tick. The editor reads that in ten seconds and quietly makes a note not to invite you again.
The opposite is also true. A careful, well-argued report from a junior researcher sticks in an editor’s memory, because good reviewers are always in short supply. Reviewing is a learnable skill, and nothing teaches you more about how your own papers will be read.
Accept or decline: three questions, five minutes
- Can I assess most of this paper? You do not need expertise in every aspect. If you are strong on the methods but new to the setting, accept and tell the editor which parts you can and cannot judge.
- Do I have a conflict of interest? A colleague in your department, a former supervisor, a direct competitor: decline or flag it.
- Can I meet the deadline? If not, decline the same day and suggest one or two suitable names. A fast no is far more helpful than a yes that runs a month late.
The manuscript is confidential. Do not share it, do not paste it into AI tools, and do not use its ideas in your own work. If you would like your supervisor to co-review as a learning exercise, ask the editor first; many journals allow it and some formally credit co-reviewers.
First pass: read like the editor
Read straight through without detailed notes and answer four questions:
- What is the research question, and is it worth asking?
- Can this design actually answer it?
- Do the conclusions go further than the data?
- If everything were fixed, what would readers of this journal gain?
You will usually have a feel for your recommendation by now. Do not commit yet, but write that feeling down in one sentence so you can compare it later.
Second pass: go where papers break
New reviewers spend most of their energy on the introduction and the prose. Experienced reviewers spend it where manuscripts are most fragile:
| Section | Question to ask | Common problem |
|---|---|---|
| Methods | How was the sample chosen, who was excluded, how were variables measured? | No exclusion criteria; scales without a source |
| Analysis | Does the analysis fit the data type and design? | Many unadjusted tests; nested structure ignored |
| Results | Do numbers in tables match the text? | Percentages not summing to 100; n changing without explanation |
| Discussion | Is interpretation beyond the evidence? | Causal language for cross-sectional data |
| Introduction | Is the gap real; are key works missing? | Literature that stops five years ago |
One small habit pays off: add things up. Check whether the sample size in the methods matches the degrees of freedom in the results tables. Mismatches like this surface quickly and often point to bigger issues.
A five-part report structure
- Summary in your own words (3–5 sentences). Authors see whether you understood the paper; editors use it to compare reviewers.
- Overall assessment: the genuine strengths and the single biggest concern, one or two sentences each.
- Major issues, numbered. For each: what you observed, why it matters, what the authors could do.
- Minor issues, numbered with page and line references. Scattered typos deserve one general line.
- Confidential comments to the editor: your recommendation and reasoning, plus any ethical concerns. Do not use this box to be harsher than in the report the authors will see.
One to two pages is plenty for a full research article. Length is not quality; twelve minor points never outweigh one well-argued major point.
Phrasing critique that lands
- Instead of “The analysis is wrong”: “Because pupils are nested in classes, I suggest a multilevel model or at least class-clustered standard errors; otherwise the standard errors are likely underestimated.”
- Instead of “The literature review is thin”: “The introduction would be stronger if it positioned the study against work using the same instrument in Asian settings.” If you suggest references, suggest other people’s.
- Instead of “Poor English”: “Several paragraphs in the results are hard to follow; a round of language editing would help readers.”
Final test: reread your report as if you were the author and this were your first paper. Any sentence that would make you shut the laptop without learning anything should be rewritten.
Your recommendation is one vote
The editor decides, not you. Make sure your recommendation matches your report: three unfixable design flaws followed by “minor revision” confuses everyone. A rough guide:
- Accept: rare in the first round.
- Minor revision: presentation problems only, no reanalysis.
- Major revision: extra analyses or rewritten sections, but a sound design.
- Reject: problems in design or data that no revision can solve.
What not to do
- Ask the authors to write the paper you would have written.
- Request citations to your own work. Many journals treat coercive citation as misconduct.
- Copy-edit sentence by sentence.
- Try to guess or contact the authors in an anonymised review.
- Let an AI tool draft the report; many publishers forbid it because the manuscript is confidential.
In practice: a two-week review schedule
- Day 1: read the abstract and accept or decline within 48 hours.
- Week 1: first pass, one-sentence impression.
- Week 2: second pass using the table above, notes keyed to line numbers.
- Let the draft rest for a day, reread as the author, soften anything gratuitously sharp.
- Submit a few days early and keep a private copy for your records.
Next step: if you have never been invited, make sure your online research profile lists your specialist topics, and tell your supervisor you would welcome co-reviewing. The first report you write carefully is what brings the second invitation.
Câu hỏi thường gặp
Can PhD students review journal articles?
Yes. Editors regularly invite late-stage doctoral students with strong methods training. If you can only assess part of the paper, say so clearly in your note to the editor.
Do peer reviewers get paid?
Most journals do not pay reviewers. Some offer fee discounts or database access, but the real benefits are the skill you build and the service record on your CV.
How long should a peer review report be?
For a full research article, one to two pages is usually enough. What matters is that the major issues are clearly argued and come with suggestions.
Can I ask a colleague to help with a review?
Only with the editor’s permission. Manuscripts are confidential; if you want your supervisor to co-review, ask first and name them in the submission.
What should I do if I suspect misconduct while reviewing?
Do not accuse the authors in the report. Describe what you observed factually in the confidential comments to the editor and let the journal follow its procedures.
