Last updated: September 10, 2026

Key Takeaways

• Include the software and version, such as R 4.x or SPSS 29, when the output depends on it.
• Used software? Name the program and its version, such as R 4.x, SPSS, Stata, Python, or GraphPad Prism, because version differences can change outputs.
• Include inclusion and exclusion criteria, and any relevant quantities, such as n=48 samples, 200 survey responses, or 3 replicates per condition.
• Missing outcome data should be handled with a stated rule — complete-case analysis, imputation, or exclusion — if responses are missing the outcome.

Materials and methods tells a reader exactly what you used, and exactly how you used it. In a materials methods — complete guide, that section needs to be specific enough for another qualified reader to judge the work and repeat it. Thin? The whole paper looks shaky. Precise? Trust goes up. See the NIH writing guidance on reproducibility and the EQUATOR Network for reporting standards. https://www.nih.gov/research-training/rigor-reproducibility

Who this guide is for, and what I assume you already know

Researchers, thesis writers, lab-report authors, and people drafting methods-heavy articles all fit here — anyone whose Materials and Methods section has to stand on its own. In a materials methods — complete guide, I am assuming you already know your study question, your sample or dataset, and the broad field you are working in. I am not assuming a polished template, a statistics degree, or a lab manager hovering over every line.

Two jobs, really. One, the section has to let another qualified person reproduce the work; two, it has to show a reviewer where the weak spots are. That is the standard in most scientific fields, whether you are writing for a journal that expects a formal Methods section or for a thesis committee that wants to know what you actually did. Clinical work adds another layer: you also need the relevant reporting framework for your design, such as CONSORT for randomized trials, STROBE for observational studies, PRISMA for systematic reviews, or ARRIVE for animal research. Those initials are not window dressing; they decide what belongs in the section. For details, consult the EQUATOR Network and the linked reporting checklists. https://www.equator-network.org/

This guide is not for someone trying to bury a messy protocol under foggy language. Nor is it for a paper where the method is inherited intact from a canonical standard and nothing changed except the sample. In that case, a short reference to the standard may be enough, as long as you state any deviations. “Standard procedures were followed” is not enough. Not even close.

One practical rule: if a detail could change the outcome, it belongs in Materials and Methods. That includes instrument model, reagent grade, software version, sampling frame, inclusion criteria, timing, temperature, and any statistical threshold such as a two-sided α of 0.05 or a 95% confidence interval. Can a reader tell what would happen on a second run? If not, the section is unfinished. For a broader overview, see the NCBI Bookshelf guidance on methods reporting. https://www.ncbi.nlm.nih.gov/books/

What belongs in Materials and Methods?

Materials and methods covers every choice that shaped the data, from the sample to the analysis plan. The real question is simple: could another competent person reproduce the work from your description alone? In a materials methods — complete guide, that is the yardstick.

I would break the section into four blocks: materials, participants or specimens, procedure, and analysis. Wet-lab papers use “materials” to mean chemicals, buffers, assay kits, cell lines, instruments, and software. Social science papers usually mean participants, questionnaires, interview guides, coding software, and the setting. Computational studies point to the dataset, preprocessing steps, software environment, and model settings. Different nouns. Same logic.

A strong Materials and Methods section does more than list tools. It explains where the materials came from, the conditions used for data collection, the order of steps, and the exact way results were processed. A spectrophotometer, for example, matters less by model number than by wavelength, cuvette path length, blanking procedure, and calibration routine. A survey? Item wording, response scale, and recruitment path matter more than a vague label like “online questionnaire.” For sample reporting, the NIH and CONSORT guidance are useful starting points. https://www.nih.gov/research-training/rigor-reproducibility

You also need to say what you did not do. Did you exclude samples below a threshold? Discard incomplete responses? Stop a run because of contamination or instrument drift? Those calls belong here because they shape interpretation. A generic article often hides exclusions in the Results or tucks them into a footnote. That trick makes the method impossible to audit.

Statistics belong here too, in many fields. If you used a t-test, state whether it was paired or independent and whether assumptions were checked. If you used linear regression, identify the outcome, predictors, and any transformations. If you used a mixed-effects model, say what was treated as fixed and random, and consult a statistician or other qualified professional if that structure is not straightforward. Software matters when output depends on it: include the name and version, such as R 4.x, SPSS, Stata, Python, or GraphPad Prism, because version differences can change outputs. The R documentation and APA Statistical Reporting guidance are useful references. https://cran.r-project.org/manuals.html

The mistake I see most often? Chasing smooth prose at the expense of precision. Methods does not need literary polish. It needs traceability. A reader should be able to move from sample source to final figure without filling gaps.

How do I write a Materials and Methods section step by step?

Write it in the same order the work happened: define the material, define the sample, describe the setup, record the procedure, state the analysis, and note the controls and exclusions. Order matters because readers use it to check whether the study’s logic holds up. In a materials methods — complete guide, that sequence keeps the section readable and auditable.

1. Start with the study design and setting. Put the design in one sentence: randomized controlled trial, case-control study, cross-sectional survey, bench experiment, cohort analysis, or computational comparison. Add the setting and time window, such as “single-center outpatient clinic,” “January to June 2024,” or “batch analysis in a biosafety cabinet.” Check that the design matches the question. If you cannot state the design in one sentence, the problem is usually not the wording; it is the study itself.
2. Define the sample or materials precisely. Name the organism, dataset, specimen type, participant group, or material source. Include inclusion and exclusion criteria, and any relevant quantities, such as n=48 samples, 200 survey responses, or 3 replicates per condition. Make sure the reader can tell what entered the study and what was left out. “Various samples” or “available records” is a problem; that sort of phrase hides selection bias.
3. List the instruments, reagents, and software that could affect the result. Give model names, manufacturers, software versions, and key settings only when they matter. For example, state the assay kit, analyzer model, PCR cycle conditions, or R package version. Verify calibration or validation status if the tool required it. A plain instrument list with no operational detail is a dead end — nobody can reproduce the measurement from that.
4. Describe the procedure in chronological order. Break the process into steps exactly as it happened: preparation, measurement, treatment, observation, and storage. If the sequence is not obvious, consult a field-specific protocol or an experienced professional before you write it up. Include specific parameters such as temperature, duration, volume, concentration, speed, wavelength, or threshold. Each step should have an endpoint the reader can recognize. Skipping around in time, or hiding behind verbs like “processed,” muddies the trail.
5. State the controls, comparisons, and randomization if used. Explain what served as the control, how allocation was done, whether blinding was used, and how many replicates were run. Include the randomization method if there was one, such as computer-generated assignment or block randomization. Make the comparison fair and the control condition obvious. No baseline? Then the result is just procedural noise in a nice outfit.
6. Explain how data were recorded and cleaned. Say what was measured, how often, in what unit, and what happened to missing or outlier values. If you removed data, describe the rule, not just the fact. Verify that the cleaning rule was set before analysis, or state clearly if it was not. Silent deletion is a problem because it makes the analysis look cleaner than the study actually was.
7. Specify the statistical analysis or evaluation method. Name the test, model, or coding framework, and state the alpha level, confidence interval, correction for multiple testing, or thematic approach if relevant. Include the software and version, such as R 4.x or SPSS 29, when the output depends on it. Check that the analysis matches the design and the scale of the data. A test that ignores pairing, clustering, or non-normal distribution will bite you later.
8. Close with any deviations, ethics, or special conditions. Report protocol deviations, IRB or ethics approval if required in your field, consent procedures, biosafety level, or storage conditions. If the procedure changed midstream, consult a professional or the relevant reporting standard so you can describe the deviation clearly and honestly. Verify that nothing important was left unspoken, especially if the method changed midstream. Pretending the procedure was perfectly uniform when it was not? Reviewers notice fast.

A good test is simple: could someone competent in the field, but not in your project, follow each step? If they would have to guess the 37°C incubation time, the 0.22 µm filter, the 5-point Likert scale, or the exact exclusion rule, you have not written enough. And if you catch yourself typing “then the usual analysis was performed,” stop. Replace it with the actual test, model, or coding method.

What details do reviewers expect, and which ones can stay out?

Reviewers want the details that change the meaning of the result, not every object on your bench or desk. So include the things that affect validity, reproducibility, or interpretation; leave out the trivia that does nothing.

Here is the dividing line: if an item could be a source of variation, name it. If it could not, do not clutter the page with it. Printer paper brand? Irrelevant. Assay lot number? Maybe matters. A barcode scanner in a survey lab? Probably not. The phone model used by respondents? Could matter if the study depends on a mobile interface.

Laboratory papers usually need sample source, preparation method, reaction conditions, equipment model, and detection threshold. If you used a commercial kit, the exact product name matters because kits are not interchangeable, and small formulation changes can alter signal. If you prepared your own buffer, the recipe matters: concentrations, pH, solvent, and order of mixing. If the method depends on incubation, time and temperature belong in the text. A reviewer does not need a story about the reagent rack; they need to know whether the sample sat at 4°C for 10 minutes or at room temperature for 2 hours.

Human-subjects papers need recruitment source, eligibility criteria, consent process, survey instrument or interview protocol, and missing-data handling. If the sample comes from a clinic, the clinic type and date range matter. If you used a validated scale such as the PHQ-9 or a custom questionnaire, say which one and whether you modified it. If you altered the wording of a standard instrument, say so explicitly, because even small wording changes can break comparability.

Computational papers call for the dataset version, preprocessing steps, feature selection, model type, train-test split, cross-validation scheme, and evaluation metric. If the model is a random forest, a support vector machine, or a neural network, say the parameters that influence output. If you set a random seed, include it. If you trained on one dataset and tested on another, name both. “The algorithm was optimized” says almost nothing.

What can stay out? Repeated brand names that do not affect the result, supplier addresses in ordinary cases, and any detail no one would need to repeat the study or audit the analysis. I would also cut over-explanation of well-known standards unless your readers are outside the field. A journal article does not need a miniature textbook inside the Methods section.

The common failure is stuffing one paragraph with every object on the bench while skipping the actual workflow. The reader does not need a warehouse inventory. They need the chain of cause and effect.

When is the standard method not enough?

The standard method falls short when the sample is unusual, the measurement is modified, or the analysis has to handle structure that the standard approach ignores. In those cases, say exactly what changed and why.

Small sample size is one common edge case. With only a handful of subjects or specimens, methods that assume large-sample stability may mislead. You may need exact tests, wider confidence intervals, or a cautious interpretation of null results. Six samples per group? I would expect the author to acknowledge limited power and avoid grand claims.

Mixed-source samples are another. If some observations come from one site and some from another, you may need to account for site effects. In a clinical or survey paper, that can mean stratification or mixed models; in a lab setting, batch correction may be the right answer. The reader needs to know whether the site or batch was balanced, because imbalance can look like a treatment effect.

Modified instruments and custom assays need extra care. If you changed a commercial protocol, state the modification and, if possible, its validation status. Even a small shift in incubation time or reagent volume can alter sensitivity. Nobody should have to guess whether your “adapted ELISA” still behaves like the original assay. If you validated the change against a reference method, put that comparison here.

Missing data is another trap. Standard methods often assume clean, complete records, but real data are not clean. If responses are missing the outcome, say how you handled them using a stated rule, such as complete-case analysis, imputation, or exclusion. If you dropped samples because the instrument failed, say when and how many. Silent omission is one of the fastest ways to lose a reviewer.

Protocol changes mid-study happen in lab work, field work, and analysis pipelines. A generic article pretends the plan was fixed from the start. A good one says what changed, why it changed, and whether the change affected only future samples or the full dataset. Honesty there is not a weakness; it is how readers judge whether the results still hold.

The rule I use is simple: if the standard method works cleanly, say so with the necessary detail. If it does not, explain the deviation before the reader has to ask.

What are the most common Materials and Methods mistakes?

The most common mistakes are missing detail, wrong order, hidden exclusions, sloppy analysis language, and pretending the method was cleaner than it was. Each one chips away at trust.

1. Using vague verbs instead of exact actions. Saying “samples were processed” or “data were analyzed” does not tell the reader what happened. The consequence is unreproducibility. The fix is to name the action: centrifuged at 3,000 × g for 10 minutes, coded in NVivo 14, or analyzed with a two-sample t-test.

2. Leaving out the decision points. If you filtered out low-quality samples, excluded incomplete surveys, or removed outliers, the reader needs the rule. The consequence is selection bias that looks accidental. The fix is to state the criterion, such as excluding observations with missing primary outcomes or samples below a predefined signal threshold.

3. Listing tools without settings. A model name alone is often not enough. A PCR machine, microscope, or regression package can behave very differently depending on settings or version. The consequence is that another person cannot match your output. The fix is to include the setting that matters: cycle numbers, magnification, thresholding rule, software version, or package name.

4. Writing the section in the same order as the results, not the work. Methods should follow the procedure, not the narrative arc. The consequence is confusion about what came before what. The fix is a chronological structure: materials, preparation, procedure, measurement, analysis.

5. Hiding a nonstandard modification inside a standard label. “We used the manufacturer’s protocol” is not enough if you changed the incubation time from 30 minutes to 45 or altered the solvent ratio. The consequence is that the method is not actually the standard method. The fix is to state the exact deviation.

6. Overexplaining the obvious and underexplaining the risky. You do not need three sentences on what a pipette is, but you do need the concentration, the temperature, and the control condition. The consequence is bloated text with missing evidence. The fix is to spend words where variation matters.

A bad Methods section often sounds fluent because it is vague. That is the trap. Fluency is not clarity. Clarity in this section comes from numbers, names, sequence, and exclusions. If a sentence could survive unchanged in five different studies, it probably is not specific enough for any of them.

How do I write a Methods section that passes scrutiny?

A Methods section passes scrutiny when every claim ties back to a real action, a real setting, or a real rule. The first test is whether the reader can reconstruct the path from raw material to final figure without guessing.

Start with the question a reviewer is always asking: did the method fit the question? Comparing two groups? Define the groups before analysis. Measuring change over time? State the time points. Doing qualitative coding? Describe the framework clearly enough that another coder could follow it. Miss those basics, and no amount of polish will save the page.

Admin