A methods section reports what you did, in enough detail for a reader to judge the study and repeat it. Write it in the past tense and in a predictable order: design and setting, participants, interventions or exposures, outcomes, sample size, statistical analysis and ethics. Describe new or modified methods in full, and cite established ones.
Journals head it Methods, Materials and Methods, or Patients and Methods. The International Committee of Medical Journal Editors (ICMJE) states its purpose in the Recommendations on preparing a manuscript: "clarity about how and why a study was done in a particular way". Our overview of the IMRaD structure gives the Methods three sentences, and one design has its own guide: the Methods for a retrospective chart review. This guide covers any design.
The running order of a methods section, and how it changes by design
The order below follows the Methods items of the STROBE checklist, with ethics added at the end; CONSORT 2025 covers much the same ground for trials. Use the names as subheadings, renamed to fit your design. Some journals want the ethics statement first or in a separate declarations section.
| Subsection | Its job | Randomised trial | Cohort, case-control or cross-sectional | Laboratory or experimental |
|---|---|---|---|---|
| 1. Design and setting | How to weigh everything that follows | Design, allocation ratio, sites, dates | The named design, source population, setting, dates | In vitro or animal work; dates |
| 2. Participants or materials | Who or what was studied, and how they were found | Eligibility; recruitment | Eligibility; how cases, exposure groups or the sample were identified | Source of each cell line, specimen, animal and key reagent |
| 3. Intervention or exposure | What differed between the groups | Each arm: what was done, by whom, when; the comparator | How the exposure was defined, measured and timed | Each condition and its control |
| 4. Outcomes and variables | Each measurement, defined so it could be repeated | Primary outcome, then secondary outcomes and harms, each with a time point | Outcomes, exposures, confounders and effect modifiers (STROBE item 7) | Assays and instruments, with settings |
| 5. Bias | How the comparison was kept fair | Sequence generation, allocation concealment, blinding | Efforts to address bias, including confounding (STROBE items 9 and 12(a)) | Randomised sample order; blinded assessment |
| 6. Sample size | Why the study is this size | The calculation and its assumptions | A calculation, or why the available data fixed the number | Independent experiments and technical replicates |
| 7. Statistical analysis | Lets a reader judge and check the analysis | Analysis population; missing data (CONSORT 2025 item 21c) | Adjusted models; missing data (STROBE item 12(c)) | Unit of analysis; how replicates were combined |
| 8. Ethics and registration | Shows the study was approved | Committee, reference, consent, registry number | Committee and reference, or a documented waiver of review; consent, or why it was waived | Animal ethics approval, reported with ARRIVE; approval for human specimens |
Case reports, systematic reviews and qualitative studies follow their own checklists, listed by design on the EQUATOR Network.
Describe new methods in full and cite the standard ones
The ICMJE Recommendations set the level of detail: "Give references to established methods, including statistical methods"; "provide references and brief descriptions for methods that have been published but are not well-known"; and "describe new or substantially modified methods, give the reasons for using them, and evaluate their limitations." In practice:
- Established method: cite it and state the version, cut-off or setting you used.
- Published but not well known: cite it and summarise its essentials in a sentence or two.
- New or modified: describe it fully, with the reason and its limits, stating any change in the original's terms.
"As previously described" is acceptable only when three things travel with it: the citation, a one-sentence summary of what was done, and any change you made. If the cited paper itself says "as previously described", cite the paper that holds the full description.
Justify a choice only where a reader would expect a different one, such as a non-standard cut-off. One clause with a reason or a citation is enough. For surgery, see writing an operative technique section another surgeon could repeat.
Rewrite the protocol as a record of what was actually done
A protocol states what will happen; the methods section states what did happen. Changing "will be" to "were" is not enough, because a copied plan can report steps that never took place. Check each sentence against the study records, then write the fact.
| Protocol wording (a plan) | Methods wording (a record) |
|---|---|
| Participants will be randomised using a computer-generated sequence. | [Role], who had no part in recruitment, generated the sequence by computer using [method]; allocation was concealed by [method]. |
| Outcome assessors will be blinded where possible. | Outcome assessors did not know the allocation; [participants or clinicians] could not be blinded because [reason]. |
| Subgroup analyses may be performed. | One subgroup analysis, by [variable], was pre-specified in [protocol version and date]; the analysis by [variable] was added later and is reported as exploratory. |
For a trial, say where the protocol and statistical analysis plan can be read; CONSORT 2025 gives this its own item (item 3). Report changes made after the study began with dates and reasons, as set out in our guide to writing a research protocol, and drop the background the ethics committee needed.
Write the statistical analysis paragraph from fixed slots
The ICMJE asks for statistical methods in "enough detail to enable a knowledgeable reader with access to the original data to judge its appropriateness for the study". It also asks for software versions, and for pre-specified analyses to be distinguished from exploratory ones.
Statistical analysis. Continuous variables were summarised as mean and standard deviation (SD) or, when skewed, median and interquartile range; categorical variables as number and percentage. The primary outcome, [outcome], was compared between [groups] using [test], and the effect was estimated as [effect measure] with a 95% confidence interval. The adjusted analysis used [model] with [covariates], chosen before analysis because [reason]. Missing data were handled by [method and its assumption]. All tests were two-sided at a significance level of [0.05]. Multiple comparisons were handled by [correction, or by treating secondary outcomes as exploratory]. Analyses of [outcomes] were pre-specified in [protocol or analysis plan, version and date]; [analysis] was exploratory. Analyses used [software] version [number].
Delete any slot that does not apply. Choose adjustment variables from knowledge of what causes both the exposure and the outcome, not from p-values in univariable or baseline comparisons. How to pick each test is covered in our decision guide to statistical tests.
A methods section template to copy and fill in
Keep the subheadings that fit your design and replace every bracketed slot.
Study design and setting. This was a [design] at [number and type of centres] in [country] between [start date] and [end date], reported following [reporting guideline].
Participants. [Participants, records or specimens] were eligible if [criteria] and excluded if [criteria]. They were identified through [route].
Intervention or exposure. [What was done or measured, by whom and when.] The comparison group [received or had] [comparator].
Outcomes. The primary outcome was [outcome], defined as [definition], measured with [method] at [time point] by [role]. Secondary outcomes were [outcomes, each defined and timed].
Bias. [Randomisation, concealment and blinding; or confounders and how they were handled; or blinded assessment and replicates.]
Sample size. [The calculation and the sources of its assumptions, or why the data available fixed the number.]
Statistical analysis. [Summary statistics; the test and effect measure for the primary outcome; adjustment; missing data; sidedness and significance level; multiple comparisons; pre-specified and exploratory analyses; software and version.]
Ethics and registration. [Committee] approved the study (reference [number]). [Type of consent, or why it was waived.] [Registry and number, for a trial.]
One methods paragraph, rewritten from vague to reproducible
The study below is invented to illustrate the structure; its numbers are not real data.
Weak: Patients admitted to the medical ward were included. Hearing was assessed and patients were followed up for delirium. Data were analysed using standard statistical software. Ethical approval was obtained.
Strong: This prospective cohort study included consecutive patients aged 70 years or older admitted to the medical wards of one teaching hospital from 1 March to 31 August 2025, excluding those with delirium on admission. Hearing impairment was defined as a failed whispered-voice test in either ear, tested at admission by a trained nurse. The primary outcome was delirium within 7 days, assessed daily with the Confusion Assessment Method by assessors unaware of the hearing result. The association was estimated by logistic regression adjusted for age, chosen in advance as a likely common cause of hearing impairment and delirium. Analyses used R version 4.4.1. The hospital ethics committee approved the study (reference 2025/041), and written consent came from each patient or their representative.
Each vague phrase became a definition, a date or a named method, and the paragraph contains no result.
Four methods section faults and the fix for each
- Findings in the methods. Information obtained during the study, such as event counts or baseline comparisons, belongs in the Results.
- Argument in the methods. Claims that a technique is superior belong in the Introduction or Discussion.
- Outcomes named but never defined. "Complications were recorded" needs a definition, a grading system and a time window.
- Analyses that go missing. Each analysis in the Methods should appear in the Results, and each result should trace back to a described analysis.
At Directive Publications, the author guidelines ask you to match your manuscript to the relevant EQUATOR checklist; its Methods items make a good final pass. If you spot an error in this guide, email the editorial office as set out in how to report a problem to us.