How to Write a Methods Section That Survives Review and Supports Reproducibility

Recent Nature Portfolio editorials and journal instructions are unusually clear about what goes wrong in methods reporting.

Nature Climate Change wrote in 2025 that clear methods reporting is essential for reliable and reproducible science and can also prevent an extended review process. It also noted that its Methods section appears at the end of the manuscript, is published online only, and allows roughly 3,000 words, while still expecting a brief methods description and discussion of caveats in the main text.

Scientific Reports gives concrete article-level limits that show how journals think structurally:

1. main text ideally 4,500 words, excluding Abstract, Methods, References, and figure legends
2. abstract no more than 200 words
3. title no more than 20 words
4. ethics and informed-consent statements should appear in the Methods section where relevant

Nature Cell Biology and other Nature titles go further by emphasizing:

• a mandatory Data Availability space
• a Statistics and Reproducibility section
• reporting summaries for many article types
• code details sufficient for others to follow the conclusions

Nature Methods has also argued recently for logically organized methods, no artificial fear of detail, and step-by-step protocol support when needed.

These are not cosmetic preferences. They show what reviewers and editors increasingly treat as baseline.

The methods section has three jobs at once.

1. Let reviewers assess validity

Reviewers are testing whether the result can be believed.

2. Let readers understand what the result means

Interpretation depends on design choices, exclusions, preprocessing, and endpoint definitions.

3. Support reproducibility or at least transparent reusability

Even when exact replication is unrealistic, the logic and procedure must still be inspectable.

When authors treat methods as a compressed ritual paragraph, all three jobs suffer.

Use the order below unless the target journal requires something else.

1. Study design

Start with the design logic:

• randomized trial, cohort, case-control, cross-sectional, qualitative, computational benchmark, etc.
• prospective or retrospective
• primary and secondary endpoints where relevant

If you bury the design until late in the section, reviewers have to reverse-engineer the paper.

2. Participants, samples, datasets, or materials

Be concrete about:

• who or what entered the study
• inclusion and exclusion criteria
• recruitment or sampling source
• dates or collection period when relevant
• final analytic sample size and attrition

3. Experimental or analytic procedures

This is where underreporting often becomes fatal.

State:

• what was done
• in what order
• using what platform, instrument, reagent, model, or software
• under what settings or thresholds

If a method is standard, you may cite the original source, but do not assume citation replaces explanation.

4. Outcome definitions

Many manuscripts fail because the methods never define the exact outcome the Results later analyze.

Specify:

• how variables were operationalized
• any transformations or derived measures
• primary versus exploratory outcomes

5. Sample size justification

This is one of the most common reasons for methods-section rejection: no power analysis or sample size justification. A peer-reviewed guide in the Indian Journal of Anaesthesia states that "not performing power analysis for sample size calculation is usually considered a good reason for article rejection."

For clinical trials: report the power calculation, the assumed effect size, the alpha level, and the resulting sample size requirement. For observational studies: justify why the available sample size is sufficient. For computational studies: explain why the dataset size supports the claims. Even when exact power analysis isn't applicable (qualitative research, case studies), state why the sample or dataset is appropriate for the conclusions drawn.

6. Statistical analysis

This should not be a boilerplate afterthought.

At minimum, spell out:

• which tests or models were used
• what assumptions mattered
• whether tests were one- or two-sided
• how multiple comparisons were handled
• how missing data were handled
• what software and version were used

If the paper uses modeling, machine learning, or complex preprocessing, the statistical and computational workflow may need its own subheadings.

6. Ethics, data, and code statements

Where relevant, include:

• IRB or ethics approval
• informed consent
• preregistration
• data availability
• code availability

Reviewers increasingly see these as part of methods quality, not administrative extras.

Use this as a pre-submission table.

1. Hiding critical details in the supplement by default

Supplementary methods can be useful, but if the main logic of the study depends on details that appear nowhere prominent, reviewers may suspect the manuscript is trying to protect a weak design behind opacity.

2. Writing as if all readers already know the field's conventions

Nature explicitly asks authors to write so that significance and background are understandable outside the narrow specialty. The same principle applies to methods. Reviewers from adjacent areas still need a workable explanation.

3. Reporting statistics too vaguely

Saying "data were analyzed using standard methods" is not reporting. It is evasion.

4. Omitting negative procedural detail

If participants were excluded, data were filtered, outliers were removed, or preprocessing thresholds were applied, say so clearly. Hidden curation choices are exactly what reviewers worry about.

5. Reusing old methods text without adapting it to the current paper

This creates both clarity and ethics problems. If the workflow changed, the old text misleads. If it did not change, you still need to make the present paper self-contained enough to read.

Enough detail means a knowledgeable reviewer could answer these questions:

• could this study be repeated in principle
• could I explain the pipeline to a colleague without guessing missing steps
• do I understand the inferential boundaries of the result

That is the standard. Not "did we mention the platform name."

This deserves special emphasis because many methods sections collapse here.

Use a template like:

1. define the analysis population
2. define the primary outcome
3. state the main model or test
4. specify adjustments, covariates, or stratification
5. describe missing-data handling
6. describe multiplicity handling if relevant
7. give software and version

If the paper involves machine learning, also report:

• train/validation/test logic
• tuning procedure
• evaluation metrics
• leakage prevention
• computational environment when material

Do not write the methods section last-minute from memory.

A better workflow:

1. export the actual protocol, notebook, script comments, or lab record
2. build the section from the real workflow
3. remove nonessential clutter without removing decision-relevant detail
4. ask a colleague outside the immediate project to read only the Methods and explain the study back to you

If they cannot, the section is still too implicit.

Authors often assume methods only matter after review. That is wrong.

A weak methods section can trigger desk rejection because it makes the manuscript feel immature, undercontrolled, or difficult to assess. Even before full review, editors use the section as a proxy for rigor and seriousness.

That is why this section pairs well with how to get published in a top journal, how to write an academic abstract, and statistical review common red flags.

Before submission, a final manuscript readiness check is a good way to catch unclear procedures, unsupported claims, and figure-method mismatches that the authors no longer see.

A methods section survives review when it makes the study legible, not when it sounds sophisticated.

Give reviewers the design, the procedures, the definitions, the analysis logic, and the oversight details they need to test trust. Anything less is an invitation to revision.