A lab on another continent wants to replicate your experiment. Their first email has seven questions: was the stated concentration final or stock, what temperature were samples incubated at, which samples were excluded and why, which software version did you use, was the test one-sided or two-sided. Your Methods section runs to 900 words, and it answers none of them.
This happens far more often than it should. Methods sections tend to be written last, in a hurry, from memory. This guide covers how to write a Methods section that a competent stranger in your field could use both to repeat your study and to judge whether you did it well.
Two different tests: can it be repeated, and can it be judged?
Your Methods section has three readers. Reviewers read it to decide whether the results deserve belief. Future readers use it to replicate or adapt your approach. And you, two years from now, will need it to remember what you did.
That gives two criteria to satisfy at once:
- Repeatable: enough parameters, materials and sequence for someone else to walk the same path.
- Assessable: enough information about the decisions that could bias the results — how units were selected and allocated, who knew which group was which, what data were dropped.
Many papers describe equipment in loving detail but never say whether the person scoring outcomes was blind to group. Reviewers care more about the second test.
Subsection order: follow the path your data took
No single order suits every discipline, but this sequence works for most empirical and survey research:
- Overall design — one or two sentences: study type, number of arms, follow-up period.
- Setting and participants — where, when, who or what, with inclusion and exclusion criteria.
- Materials and measures — instruments, reagents, scales, and where they came from.
- Procedure — in the chronological order participants or samples experienced it.
- Variables — how each primary and secondary outcome and each covariate was defined.
- Data analysis — from cleaning to the main model to sensitivity analyses.
- Ethics — approval, consent, data protection; some journals want this first.
How much detail each study type needs
| Study type | Must include | Commonly missing |
|---|---|---|
| Laboratory experiment | Sources of materials (supplier, catalogue number, cell line), conditions, number of replicates | Whether replicates were biological or technical; whether outcome reading was blinded |
| Intervention trial | Eligibility, randomisation method, allocation concealment, pre-specified primary outcome | Who generated the sequence and who enrolled participants — and whether they were the same person |
| Survey | Sampling frame, recruitment route, fieldwork dates, scale sources, translation process | How the response rate was calculated; how many people opened but abandoned the form |
| Qualitative | Participant selection, interview guide, duration, recording and transcription, coding approach | Researcher positionality, who coded, when and why recruitment stopped |
| Secondary data analysis | Data source, access date, filtering criteria, software version | Analysis code, random seeds, cleaning steps |
For every scale, report four things: name and source, number of items, one example item, and reliability in your own sample. “We used the validated X scale” on its own is the first sentence a reviewer will flag.
Describing the analysis: naming the test is not enough
“Data were analysed in SPSS using regression” tells the reader almost nothing. Compare:
“We estimated the effect of the intervention on anxiety scores at 8 weeks using linear regression adjusted for baseline anxiety, age and gender. Normality of residuals was checked with Q-Q plots. Tests were two-sided at α = 0.05; the three secondary outcomes were adjusted using the Holm method. Analyses were run in R 4.3 with the lme4 package.”
The second version tells the reader the outcome, the covariates, how assumptions were checked, the sidedness, how multiplicity was handled and what software was used. Add one sentence on missing data and one on sensitivity analyses, and a reviewer has nothing left to ask.
Tense, voice and the evasive passive
- Past tense for what you did: Participants completed…, Samples were incubated…
- Present tense for things that remain true: describing an instrument (The scale consists of 10 items), software, or signposting (Table 2 shows…).
- “We” is accepted by most journals today and usually produces shorter sentences. Check your target journal’s author guidelines.
Watch for passives that hide the actor and the rule: “Outliers were removed.” By whom, on what criterion, how many, and before or after looking at the outcome? Answer all four in the sentence.
What reviewers flag most often
- A sample size that appears from nowhere, with no justification or calculation.
- “Random sampling” that was in fact handing questionnaires to the author’s own class.
- No account of how many participants or samples were excluded, at which stage, and why.
- “As described previously [12]”, where [12] cites [8], and both were modified.
- An analysis that appears in Results but was never mentioned in Methods.
- No statement on whether outcome assessors knew group allocation.
When citing an established method, summarise its core in a sentence or two and state every modification.
Use reporting checklists as a self-audit
Reporting guidelines are collected by the EQUATOR Network. The ones you are most likely to meet:
- CONSORT for randomised controlled trials.
- STROBE for observational studies (cohort, case-control, cross-sectional).
- COREQ or SRQR for qualitative research.
- PRISMA for systematic reviews, ARRIVE for animal research, STARD for diagnostic accuracy studies.
Many journals require a completed checklist at submission, with page or line numbers for each item. Open it when you start drafting and treat it as an outline. An item you cannot report usually points to something you didn’t do or didn’t record.
How long, and what goes in supplementary material
Word limits tempt authors to cut Methods first. A better approach is layering:
- In the paper: everything needed to judge the credibility of the main result.
- Supplementary files: full questionnaires, interview guides, detailed parameters, secondary analyses.
- A public repository: analysis code, step-by-step protocols, and data where sharing is permitted.
Never push exclusion criteria or the definition of the primary outcome into an appendix. Reviewers need to see those immediately.
A working process
- Write while you work: keep a log of dates, parameters and every deviation from plan, then copy from it.
- Build the skeleton using the subsection order above, merged with the relevant reporting checklist.
- Fill it in, favouring numbers and decisions over adjectives.
- Cross-check against Results: every number and every analysis in Results must trace back to Methods.
- Test it on a stranger: give it to a colleague in your field and ask, “Could you repeat this? What’s missing?” Every question they ask is a question a reviewer would have asked.
Your next step: open your current Methods draft, find the reporting checklist that matches your design, and mark each item as present or absent. The empty boxes are this week’s revision list.
Câu hỏi thường gặp
How long should the Methods section of a research paper be?
There is no fixed length; it often takes up somewhere between a fifth and a quarter of the paper. The real test is whether a peer could repeat the study and assess the main result. Full questionnaires, protocols and code can go into supplementary files.
What tense should I use in the Methods section?
Use the past tense for what you did, such as participants completed or samples were incubated. Use the present tense for things that remain true, such as describing an instrument or pointing to a table.
Can I use “we” in a Methods section?
Most journals now accept “we”, and it usually makes sentences shorter and clearer about who did what. Some fields and journals still prefer the passive, so check the author guidelines of your target journal.
Which reporting guideline applies to my study?
Randomised trials use CONSORT, observational studies use STROBE, qualitative studies use COREQ or SRQR, and systematic reviews use PRISMA. The EQUATOR Network hosts these guidelines and a tool to help you pick the right one.
Do I need to describe a method if I cite a previous paper?
Yes, briefly. Summarise the core of the method in a sentence or two and state every modification you made. “As described previously” on its own forces readers to chase references and often triggers a reviewer request.
