How to Write the Conclusion of a Research Paper

The conclusion is often written last, in a hurry, the night before submission — and it shows. Editors and reviewers read it closely because it tells them what the authors believe the paper has actually established. A good conclusion answers the research question directly, says what is new, admits what the study cannot show, and points to sensible next steps. This guide explains each of those jobs, how long the section should be, what does not belong in it, and how weak and strong versions differ, with examples you can adapt.

What the Conclusion Section Must Do

A conclusion is not a summary of everything in the paper. It is the authors’ final, considered answer to the question the paper set out to address. Readers who jump from the abstract to the conclusion — and many do — should come away knowing what was asked, what was found, why it matters and how far the claim can be trusted.

In practice the section has five jobs:

1. Answer the research question in direct language, without re-reporting every statistic.
2. State the contribution — what this study adds that the literature did not already contain.
3. Mark the limits a reader must keep in mind when applying the findings.
4. Suggest next steps for research and, where justified, for practice or policy.
5. Close the argument so the paper ends on its main message rather than a side point.

If you are unsure what your conclusion should say, reread your introduction. Its final paragraph usually states the aim or research questions; the conclusion should respond to each of them, in the same order. For the overall architecture of a journal article, see our IMRaD writing guide.

Conclusion, Discussion and Abstract: How They Differ

Three parts of a paper talk about findings, and authors often blur them. The table separates their roles.

Some journals have no separate conclusion; the last paragraph of the discussion does the same job, sometimes under a sub-heading such as “Conclusions” or “Implications”. Others — common in engineering, management and the social sciences — expect a stand-alone section, occasionally titled “Conclusion and Recommendations”. Check the author instructions of your target journal before deciding; at JMVRA, for example, the expected sections are set out in the publication guidelines.

Whatever the format, do not paste sentences from the abstract. Reviewers notice repeated text and read it as padding.

Answering the Research Question Directly

Open the conclusion with the answer rather than building up to it. One sentence recalls the aim; the next states what the study found, in plain terms. Keep to the few numbers that carry the main message — detailed results belong in the results section.

If the study had several research questions or hypotheses, answer each one using the same labels as the introduction (RQ1, RQ2, H1…). Say clearly when a hypothesis was not supported; a null or mixed result is still an answer. Avoid both heavy hedging (“it may be possible that there could be some association”) and turning an association into a cause.

Match the strength of your verbs to the design: shows or demonstrates for well-controlled experiments; suggests or is associated with for observational and cross-sectional data; indicates or illustrates for qualitative work. How the question was framed in the first place matters too — our guide to research questions and hypotheses covers that step.

Stating Your Contribution Without Overclaiming

After the answer, explain why it matters. Reviewers want to know what is different now that this study exists. Contributions usually fall into one or more of these types:

• Empirical — new evidence from an under-studied population, region, sector or period.
• Theoretical — a model is extended, tested in a new setting, or a boundary condition is identified.
• Methodological — a new or adapted instrument, dataset, algorithm or analysis procedure.
• Practical — evidence that can inform teachers, clinicians, managers, engineers or policy makers.

Name the contribution specifically and link it to the gap identified in your literature review. “This study contributes to the literature” tells the reader nothing. “To our knowledge, this is the first test of the technology acceptance model among rural primary-school teachers, and it finds that facilitating conditions matter more than perceived usefulness in this group” tells them exactly what is new.

Treat words such as first, novel and breakthrough with care. Use “first” only after a documented search, and qualify it. Claims that results apply to “all organisations” or “every developing country” after a single-site study are a frequent trigger for major revision. If the gap was weakly argued at the start, no conclusion can repair it; strengthen the literature review instead.

Limitations: What to Say and Where

Every study has limits, and acknowledging them builds trust. Many journals expect the full limitations paragraph in the discussion; the conclusion then carries only the one or two limits that most affect how the findings should be read. If the journal has no separate discussion, place a compact limitations paragraph just before your suggestions for future research.

Useful limitations are specific and explain their consequence:

• Design — a cross-sectional survey cannot establish which variable came first.
• Sample — a convenience sample from two urban hospitals may not represent rural patients.
• Measurement — self-reported study hours tend to be over-estimated.
• Context and time — data gathered during a policy transition may not reflect the settled situation.

Avoid the bare line “the sample size was small” without saying small for what. If you calculated a target sample size in advance, say whether it was reached; if not, say what size of effect the study could reliably detect. A limitation is more convincing when paired with what you did to reduce it, for example: “although participation was voluntary, the sample’s age and gender profile closely matched district census figures.” Do not apologise; three to five well-chosen points are usually enough.

Future Research and Recommendations

“Further research is needed” may be the most common closing sentence in submitted manuscripts, and the least useful. Good suggestions follow from your limitations and describe the study that would settle the open question.

Recommendations for practice or policy are expected in many applied fields such as education, commerce, public health, agriculture and engineering management. Keep them tied to evidence:

• Each recommendation should trace back to a specific finding in your results.
• Name who should act (school heads, hospital administrators, a state agency) and what they should do.
• Match confidence to design strength: an exploratory study can recommend a pilot, not a nationwide rollout.

Recommendations the data cannot support — tax policy advice from a study that only measured consumer attitudes, for instance — are a red flag for reviewers and are often the first thing they ask you to cut during revision.

How Long Should a Conclusion Be?

There is no universal word count, but most journal-article conclusions fall between 150 and 500 words, roughly 5–10% of the main text. A short communication may need one paragraph; a long empirical paper with several research questions may need four. Thesis conclusions are far longer and must be cut sharply for a journal article.

A dependable structure for a stand-alone section:

1. Paragraph 1: the aim in one sentence, then the direct answer to each research question.
2. Paragraph 2: the contribution to knowledge and, where relevant, implications for practice.
3. Paragraph 3: the key limitations and specific directions for future research.
4. Final sentence: the take-home message, worded differently from the abstract.

Use the past tense for what the study did and found (“the survey showed”) and the present tense for what the findings mean now (“these results suggest that”). Sub-headings inside the conclusion are rarely needed unless the template asks for “Implications”, “Limitations” and “Future Research” separately.

What Not to Put in a Conclusion

• New results. Every finding mentioned must already appear in the results section; an extra analysis introduced here confuses reviewers.
• New literature. The conclusion is not the place to cite studies for the first time. Most conclusions need no citations at all.
• Detailed statistics. Leave p-values, confidence intervals and tables in the results; one or two headline figures are enough.
• Copied abstract sentences. Rewrite rather than paste.
• Overclaiming. Causal language from correlational data, or generalisation beyond the population sampled.
• Personal reflections. “This research taught me a great deal” belongs in a thesis preface, not a journal article.
• Decorative endings. Famous quotations and rhetorical questions rarely add information.
• Declarations. Funding, conflicts of interest and data access have their own sections; see our data availability statement templates.

Several of these problems appear on reviewers’ lists of the most common rejection reasons, especially overclaiming and conclusions that do not follow from the results.

Weak and Strong Conclusion Sentences Compared

The quickest way to improve a draft is to test each sentence against its job. The pairs below come from typical first drafts.

For qualitative studies the same logic applies, but the “answer” is a set of themes or an explanatory account rather than a number: “Teachers described adopting the platform in three stages — compliance, adaptation and ownership — with peer mentoring as the main driver of movement between them.” In double-blind review, which JMVRA and many other journals use, reviewers judge the conclusion only against the evidence in the manuscript, so every claim must be traceable to your results.

Quick Checklist

1. Answer each questionEvery RQ or hypothesis gets a direct reply
2. Name the contributionSpecific, linked to the stated research gap
3. Check verb strengthCausal words only where the design allows
4. Keep key limitsOne or two limits that change interpretation
5. Make future work concreteDescribe the design that would answer it
6. Remove extrasNo new data, citations or copied abstract text

Questions Researchers Often Ask

In Summary

A conclusion earns its place when it gives a clear answer to the question set out in the introduction, names a specific contribution, keeps only the limits that change how the findings should be read, and proposes concrete next steps. Keep it short, write it after the discussion, match your verbs to your design, and leave out new data, new citations and borrowed abstract sentences. Before you submit your manuscript, read the conclusion alongside your introduction and results; if all three tell the same story, reviewers are far more likely to accept the paper’s claims.

References and Sources

• American Psychological Association — APA Style
• ICMJE — Recommendations for the Conduct, Reporting, Editing and Publication of Scholarly Work
• EQUATOR Network — Reporting guidelines for health research