How to Write the Conclusion of a Research Paper (With Example)

The conclusion of a research paper is the short final section where you state what your study found, what that finding means, and what should happen next. It answers the research question in plain words and then stops. Most conclusions work best at one to three paragraphs, and they should contain no new data and no new argument.

A good conclusion is short enough for a reviewer to read in under a minute, and it tells them what the paper adds and what it leaves open.

What a conclusion does and what it leaves out

A conclusion gives the reader the answer to your research question and the reason it matters. It is the last thing the reviewer sees, so it affects what they remember about the paper.

The UNC Writing Center describes the conclusion as your chance to have the last word on the subject. It lists what a good ending does: it shows how the main points fit together, returns to the themes of the introduction, and points to wider implications or further study. That advice is written for essays, but it applies to research papers too.

The same page lists what to leave out. Don't introduce a new idea, don't reword the thesis and call it a conclusion, and don't pull in quotations or statistics that belong in the body. Don't add a pile of leftover facts you couldn't place elsewhere.

In a research paper, I'd add three more items to that list:

• Don't report a new analysis. If a number first appears in the conclusion, the reviewer can't find where you got it.
• Don't claim more than the data supports. The ICMJE recommendations for manuscripts say to link the conclusions with the goals of the study and to avoid conclusions the data don't support.
• Don't apologize. Limitations belong in the discussion, and the conclusion can name the biggest one in a clause.

What goes in is short. You state the finding, the reason it matters to your field, and one realistic next step. Those three jobs usually fit in a paragraph each, and sometimes in one paragraph for a short paper.

Conclusion versus discussion

The discussion interprets your results and the conclusion states the answer. The discussion argues, compares and qualifies, and the conclusion gathers the outcome of that argument into a few sentences a reader can quote.

Writers mix these up because both sections look back at the results. The difference is the job each does for the reader. The discussion shows your reasoning, and the conclusion shows where the reasoning ended up.

If you still need to write the earlier section, our guide to the discussion section explains the structure reviewers expect. A good conclusion is easy to write once the discussion is finished, because each of its sentences is a compressed version of a discussion paragraph.

Does your journal want a conclusion at all?

Not every journal wants a separate conclusion, and you should check before you write one. The classic IMRaD structure, which stands for Introduction, Methods, Results and Discussion, has no section labeled Conclusion. The Discussion may absorb the concluding content, with or without a subheading. Some journals add a separate Conclusion at the same heading level as the Discussion.

PLOS ONE shows how flexible this is. Its submission guidelines list "Conclusions (optional)" among the middle sections and say the Results, Discussion and Conclusions may be separate or combined. The guidelines also say authors should avoid overstating their conclusions.

Here is how to find out what your target journal expects:

1. Read the author guidelines for a heading list. Many journals name every required section.
2. Open three recent papers from the journal and look at how each one ends. If none has a Conclusion heading, the last paragraph of the Discussion does that job.
3. Check the abstract format. Structured abstracts often end with a labeled Conclusions line, and the final section of the paper should agree with it.

When the journal has no conclusion section, write the last paragraph of your Discussion as if it were one. It should state the answer, the meaning and the next step in a few sentences. Readers who skip to the end of a paper will go there first.

Theses and dissertations are different. A thesis almost always has a conclusion chapter, and it is longer because it ties several chapters together. Our master's thesis guide covers how that chapter fits the whole project.

How long a conclusion should be

A conclusion is usually between 150 and 400 words in a journal article, and it can run to several pages in a thesis. These are my working ranges, so treat the journal's own instructions as the final word.

Length follows the number of claims you have. A paper with one finding needs one paragraph. A thesis with four research questions needs room to answer each one.

If the conclusion runs long, cut material instead of compressing sentences. Check the word count against the journal limit as you trim.

How to write the conclusion of a research paper in four steps

Most strong conclusions follow four steps in order. You can write each in one to three sentences, and the paragraph breaks go where the steps change.

1. Answer the question. Start with the finding, in the same terms your introduction used for the research question. If the introduction asked whether X affects Y, the first sentence says whether it does and by how much.
2. Say what it means. Tell the reader what changes because of the answer. This is the contribution: a theory that needs revising, a practice that should change, or a measurement that now has evidence behind it.
3. Name the boundary. State the most important limit on the claim, such as the sample, the setting or the design. One clause is enough, because the discussion already covered the rest.
4. Point to the next step. Give one specific study, policy or practice that follows from the work. "More research is needed" tells the reader nothing, and a named design or population tells them what to do.

The order matters. If the boundary comes first, the reader hears an apology before they hear the result. If the next step comes before the meaning, the paper ends on a task list and loses its point.

You can write the answer sentence before the paper is finished. I often draft it first. It works as a test, because if you can't state the answer in one sentence, the argument isn't settled yet.

A worked example from an invented study

The example below comes from a made-up study, written for this guide. It is not real research, and the numbers are invented to show the structure. The study tests whether weekly retrieval quizzes improve exam scores in a first-year statistics course.

Here is a weak first draft of the conclusion:

In conclusion, this study looked at retrieval quizzes and exam scores. Many interesting results were found. The quizzes may have had some effect on learning, although more research is needed. Students also reported liking the quizzes, and a new survey found that 78% of them preferred online formats. Further studies should look at other courses.

This draft has five problems. It opens with a stock phrase and then describes the study instead of stating the finding. "Many interesting results" says nothing. The hedging ("may have had some effect") is too heavy for an effect the study measured. The 78% figure is new, because it appears nowhere in the results. The last sentence is a vague call for more research.

Here is the revised version:

Weekly retrieval quizzes raised final exam scores by 6.4 points on a 100-point scale (95% CI 3.1 to 9.7). The gains were largest among students who began below the class median. The result supports retrieval practice as a low-cost way to improve outcomes in large introductory courses. Individual feedback is hard to provide in those courses. The design was a single semester at one institution, so the effect needs testing in other subjects and student populations. A multi-site trial with a delayed retest, six months after the course, would show whether the gain lasts.

Here is what changed and why:

The revised paragraph is 94 words. It has no stock opener and doesn't copy the abstract. The reviewer gets a result, a reason and a next step.

Common mistakes and how to fix them

Most weak conclusions share a handful of problems. This table lists them with a repair for each.

The UNC Writing Center advises against opening with stock phrases such as "in summary" or "in closing." They add nothing, because the heading already tells the reader where they are.

One more problem is mismatch with the rest of the paper. The conclusion should answer the question from the introduction, and our guide to writing an introduction explains how to state that question so the ending has something to answer. It should also agree with the numbers in your results section and with the closing sentence of your abstract. Read the four pieces together before you submit and fix any claim that differs between them.

For a literature review, the conclusion has a different target. It names what the sources agree on, where they disagree, and which gap remains. Our post on writing a literature review covers how to build the synthesis that this ending summarizes.

Using AI to draft or tighten a conclusion responsibly

AI can help you tighten a conclusion you've already written, and it is a poor source for the claims in it. The reason is simple: the conclusion states what your study found, and only you know that. A chatbot given a vague prompt will produce a plausible ending that may include results you never reported.

Here is a workflow that keeps the content yours:

1. Write the four steps yourself in rough sentences, with your real numbers.
2. Use a proofreader to fix grammar, shorten long sentences and remove filler. ProofreaderPro.ai edits in your document and keeps the meaning of what you wrote.
3. If a sentence is clumsy, ask a paraphrasing tool for alternatives and choose one. The paraphrasing tool offers rewrites that you then check against your results.
4. Compare every number and claim in the final text with your results section.
5. Record the AI use in the disclosure the journal asks for.

Disclosure rules exist, and they are specific. The ICMJE recommendations say authors should disclose at submission whether they used AI-assisted technologies. They should describe the use in the cover letter and the submitted work, and review AI-generated content carefully. They also say AI tools shouldn't be listed as authors, because humans are responsible for the submitted material.

Publishers differ in what they treat as disclosable, with some exempting basic language editing. Check your journal's policy instead of assuming. Our AI disclosure statement guide gives wording you can adapt, and the per-publisher guide lists what individual publishers ask for.

Tracked changes help here. On Academic Plus, ProofreaderPro.ai exports edits as tracked changes in Word, so you and any co-author can see exactly which words the tool changed. That record makes disclosure easier to write.

If the ending still sounds flat after editing, the cause is usually in the content, not the grammar. A conclusion with a clear answer, a stated meaning and a named next step works well in plain sentences.

Checklist before you submit

Run through these points once the draft is finished.

• The journal's guidelines and recent papers confirm whether a separate conclusion is expected.
• The first sentence answers the research question in the terms the introduction used.
• Every number in the conclusion appears in the results.
• No new data, citations or arguments appear.
• The verbs match the study design ("suggests" for observational work, "shows" for controlled tests).
• One limitation is named in a clause, and the full list stays in the discussion.
• The next step is specific: a design, a population or a variable.
• The wording differs from the abstract, but the claims match it.
• The section doesn't open with "in summary" or a similar phrase.
• The length fits the paper type and the journal's word limit.
• Any AI assistance is disclosed as the journal requires.
• The section has had a full proofreading pass for grammar and consistency.

Frequently asked questions

How long should the conclusion of a research paper be?

Most journal articles need one or two paragraphs, roughly 150 to 300 words. A short report may need only a few sentences, and a thesis chapter can run several pages. Follow the journal's instructions if they give a limit, because those override any general range.

What is the difference between a discussion and a conclusion?

The discussion interprets the results and compares them with earlier work, and the conclusion states the answer and what it means. The discussion shows the reasoning and covers limitations in full. The conclusion gives the outcome in a few sentences and names only the most important limit.

Should the conclusion repeat the abstract?

No. The two should agree on the claims, but the wording should differ. The conclusion can mention the meaning and the next step in more detail than the abstract has room for. Copying the abstract wastes the section and makes the paper feel repetitive.

Can I include new information in the conclusion?

No. New data, new citations and new arguments belong in the results or discussion, where the reader can see how you got them. The UNC Writing Center also lists new ideas as something a conclusion should leave out.

Do all research papers need a conclusion section?

No. The IMRaD structure has no separate Conclusion, and journals such as PLOS ONE treat it as optional. Where there is no heading, the end of the Discussion does the job. Check your target journal's guidelines and a few recent papers.

Can I use AI to write my conclusion?

You can use AI to tighten and proofread a conclusion you wrote, and you should check every claim against your own results. The ICMJE says authors should disclose AI-assisted technology at submission and remain responsible for the content. Don't ask a chatbot to invent findings.

How do I end a conclusion without a cliché?

End on the contribution or the next step. A specific recommendation, such as a named study design or a policy change, gives the reader something to act on. A line about "the future" or "the importance of the topic" says nothing the paper hasn't already shown.

Should I mention limitations in the conclusion?

Name the single most important one in a clause. The discussion should cover the rest. If the limitation changes how far the result applies, say so, because a reviewer will trust the conclusion more when it states its own boundary.

AI Proofreader for research papers

Fix grammar, shorten wordy sentences and keep your meaning in every section of your paper, with a Word export of the edited text.