How to structure a research paper (IMRaD)

IMRaD stands for Introduction, Methods, Results and Discussion. It became the dominant structure for original research articles in medicine and the natural sciences during the twentieth century and is now standard across most empirical fields. It works because it matches the questions a reader asks in order: why did you do this, what did you do, what did you find, and what does it mean?

Use it for empirical research papers, lab reports and many thesis chapters that report a study. Do not force it onto essays, humanities papers, literature reviews or theoretical work, which have their own structures. And always check your target journal or assignment brief: many journals rename, merge or reorder sections (for example, combined "Results and Discussion", or Methods placed at the end).

The Introduction moves from broad to narrow, often described as a funnel. A common pattern has three moves:

1. Establish the territory. What is the topic and why does it matter? Keep this short and specific to your field's readers.
2. Identify the gap. What is not yet known, contested or untested? Support this with the most relevant literature, not an exhaustive review.
3. Fill the gap. State your aim, research question or hypotheses, and briefly how you addressed them.

End the Introduction with the aim or hypotheses stated plainly. A reader should be able to find them in a few seconds.

The standard for Methods is reproducibility: a competent researcher in your field should be able to repeat the study from what you wrote. Typical subsections:

• Design: the type of study (for example, randomised trial, cross-sectional survey, case study).
• Participants or materials: who or what was studied, how they were selected, inclusion and exclusion criteria, sample size and how it was decided.
• Procedure or measures: what was done, in what order, with which instruments, and how variables were defined.
• Analysis: the statistical or qualitative methods, the software and version, and how missing data and outliers were handled.
• Ethics: approval body and reference number, and consent procedures, where applicable.

Describe what you actually did, including deviations from the plan. If you pre-registered the study, say so and note any differences.

Results present findings in a logical order, usually following the order of your aims or hypotheses. Interpretation belongs in the Discussion.

• Start with descriptive information (sample characteristics, response rates) before inferential results.
• Report statistics completely: the test, the statistic, degrees of freedom, the exact p value, an effect size and, where possible, a confidence interval. Our guide to choosing a statistical test for two groups covers what to report for the most common comparison.
• Use figures and tables for detail, and use the text to point out what matters in them. Do not repeat every number from a table in the prose.
• Report results that did not support your hypotheses as clearly as those that did.

Every figure and table should be numbered, referred to in the text, and understandable with its caption alone. See how to make a publication-ready figure.

The Discussion runs the funnel in reverse, from your specific findings back out to the wider field:

1. Main finding. Answer the research question in a sentence or two.
2. Comparison. How do your results relate to previous studies? Where they differ, offer plausible explanations.
3. Limitations. Be honest and specific about what limits the conclusions (sample, design, measurement) and how much each matters.
4. Implications. What follows for theory, practice or future research?
5. Conclusion. A short closing paragraph, or a separate Conclusion section if the journal or brief asks for one.

Avoid claiming more than the data support. A correlational study shows an association, not a cause; a result in one sample does not automatically generalise to others.

Title#

Specific and informative: name the main variables, population or system, and ideally the design. Avoid jokes and question titles unless your field uses them.

Abstract#

A self-contained summary of the whole paper, usually 150–300 words depending on the journal. Many journals require a structured abstract with headings such as Background, Methods, Results and Conclusions. Include your main numerical results, not just "results are discussed".

References#

Cite in the required style, and check each reference against its source. See the citation style guides and how to check that a citation is real.

You do not have to write in reading order, and most experienced authors do not. A practical order:

1. Methods, while the details are fresh.
2. Results, starting from your figures and tables.
3. Discussion, once you know what the results show.
4. Introduction, now that you know exactly which gap your findings address.
5. Abstract and title, last, so they describe the paper you actually wrote.

Tense and voice#

A widely used convention: past tense for what you did and found ("Participants completed...", "Scores were higher..."), present tense for established knowledge and for what your results mean ("This suggests..."). Many journals now accept or prefer the active voice ("We measured..."); check recent papers in your target journal.

• An Introduction that is a literature review. Cite what the reader needs to understand the gap, not everything you read.
• Methods that rely on "standard procedures". Name the procedure and cite where it is described in full.
• Interpretation in the Results. "Scores were higher in group A" belongs in Results; "suggesting that feedback improves motivation" belongs in the Discussion.
• Results that appear for the first time in the Discussion. Every finding you discuss should be reported in the Results first.
• A Discussion that restates the Results. Say what they mean and how they relate to other work, rather than listing the numbers again.
• An abstract that does not match the paper. Rewrite it last, after the final edit, and check every number against the Results.

For many study designs there is an agreed checklist of what to report. The EQUATOR Network collects them. Three of the best known:

• CONSORT for randomised controlled trials,
• STROBE for observational studies in epidemiology,
• PRISMA 2020 for systematic reviews and meta-analyses.

Even if a journal does not require one, working through the relevant checklist is a quick way to catch missing details. For a printable section-by-section skeleton, use the IMRaD outline template; for examples by paper type, see research papers and lab reports.