How to Write a Conclusion for a Research Paper

A Conclusion is where a research paper stops adding new interpretive argument and delivers its takeaway. By the time a reader reaches it, they have already seen the study’s aim in the Introduction, the evidence in Results, and the interpretation — findings in context, limitations, implications — in the Discussion. The Conclusion’s job is narrower than any of those: restate what the study set out to do, state in one or two sentences what it actually found and why that matters, and stop. It is not a second Discussion, and it is not the Abstract repeated at the end of the paper. This guide covers what a Conclusion section needs to contain, how it differs from the Discussion and the Introduction it mirrors, a practical sentence-by-sentence template, and the mistakes that most often make a Conclusion read as padding rather than a payoff. For where the Conclusion sits in the whole manuscript and how it fits the drafting process, see CASRAI’s How to Write a Research Paper guide, which this page extends.

Not Every Journal Requires a Separate Conclusion Section

Before drafting, check whether the target journal wants a standalone Conclusion at all. The IMRaD structure (Introduction, Methods, Results, Discussion) is what the ICMJE recommends as the standard shape for reporting original research, and its own guidance does not call for a separate Conclusions section — ICMJE’s Discussion guidance says to “link the conclusions with the goals of the study” as part of the Discussion itself, not as a distinct fifth section.

PLOS ONE’s submission guidelines make the optionality explicit rather than implicit: “These sections may all be separate, or may be combined to create a mixed Results/Discussion section … or a mixed Discussion/Conclusions section (commonly labeled ‘Discussion’).” In practice you’ll see three structural patterns across journals:

• No separate Conclusion at all — the Discussion ends with a closing paragraph that does the Conclusion’s job, under the Discussion heading.
• A combined “Discussion and Conclusions” heading — one section, with the final paragraph(s) functioning as the conclusion.
• A short, separate Conclusion (or “Conclusions”) section — common in clinical, engineering, and many non-English-language-adjacent journal templates, and often explicitly required by the author instructions.

None of this is a style preference you get to pick — it’s set by the target journal or venue’s author instructions, the same way structured-versus-unstructured abstract format is (see CASRAI’s How to Write a Research Abstract guide for that parallel case). If the instructions are silent, a short separate Conclusion is rarely wrong; folding it into the Discussion is the more common default in basic-science and biomedical journals that follow ICMJE closely.

What a Conclusion Section Has to Do

Whether it’s a standalone section or the closing paragraph of the Discussion, a Conclusion earns its place by doing a small, specific set of jobs and nothing more:

• Restate the study’s aim or question, briefly — one sentence, not a re-run of the Introduction’s motivation.
• State what was found, in plain language, without re-reporting the statistics already given in Results. The reader has seen the numbers; the Conclusion states what they mean in one or two sentences.
• State why it matters — the contribution, in a form scaled to what the study actually supports, not to what would make the best headline.
• Close, ideally with a concrete implication or a specific next question, rather than a sentence that could be pasted onto any paper unchanged.

What it should not do is introduce a new interpretive argument. Comparing findings to prior literature, explaining mechanisms, and writing the limitations paragraph are Discussion work (see CASRAI’s Writing the Discussion Section of a Research Paper guide for that); a Conclusion that starts doing that work for the first time is really an unlabeled second Discussion, and reviewers tend to ask for it to be cut or merged.

Conclusion vs. Discussion: Where the Line Actually Is

The distinction is about function, not length. The Discussion interprets: it explains what the findings mean, situates them against prior work, and discloses limitations, doing genuinely new argumentative work with the results. The Conclusion summarizes the outcome of that argument — it doesn’t build a new one. A useful test while drafting: if a sentence in the Conclusion is doing something a reader hasn’t already seen argued in the Discussion (a new comparison to another study, a mechanism not previously raised, a limitation mentioned nowhere else), it’s Discussion content that’s ended up in the wrong section.

This is also why a thin, one-line Conclusion is often the right length rather than a failure of effort — if the Discussion has already done its job, the Conclusion may legitimately be two or three sentences. Padding it out to look substantial by re-explaining points the Discussion already made is a more common problem than a Conclusion being too short.

Conclusion vs. Abstract, and Conclusion vs. Introduction

Two other sections get confused with the Conclusion, for different reasons.

The Abstract’s “Conclusions” field (in a structured abstract) and the paper’s Conclusion section are not the same text reused twice, even though they say similar things. The abstract’s conclusion has to stand completely alone — it’s read and indexed independently of the rest of the paper, under a hard word-count constraint, by a reader who may never see the full text (see CASRAI’s How to Write a Research Abstract guide). The paper’s Conclusion is read by someone who has just finished the Discussion; it can be slightly more expansive, can reference specific results by name rather than restating them, and doesn’t need to be self-contained in the same way. Rewriting the abstract’s conclusion sentence-for-sentence as the paper’s Conclusion is a common shortcut that reads as lazy to a careful reviewer, precisely because the two are answering to different constraints.

The Introduction and the Conclusion are structural mirror images. CASRAI’s guide to structuring the introduction describes it as a funnel — broad field context, narrowing down to the specific gap and the study’s aim. The Conclusion runs that funnel in reverse: it starts from the specific finding and widens back out to why it matters beyond this one study. Purdue University’s Online Writing Lab describes the same general-to-specific-to-general shape in academic writing more broadly, summarizing a closing section’s job as completing the movement the opening began, rather than introducing a new direction at the end. The practical version of that rule: nothing that first appears in the Conclusion should be new to the reader — not a new citation, not a new result, not a claim of significance broader than anything argued in the Discussion.

A Practical Template

For a short, standalone Conclusion (two to five sentences is typical where a journal doesn’t specify a length), a reliable sentence-by-sentence sequence:

1. Restate the aim in one sentence — what the study set out to determine.
2. State the main finding in plain language, without repeating exact statistics already in Results.
3. State the contribution — what this adds to the field, scaled to the evidence (a single moderately powered study rarely “resolves” a question; it usually narrows, supports, or complicates one).
4. Optionally, name the single most consequential caveat — only if the Discussion’s limitations paragraph hasn’t already covered it, and only the one that most changes how a reader should weigh the finding, not a repeated list.
5. Close with a specific implication or next step — a particular follow-up study, a specific practical use, or a named open question the results themselves raised. A sentence a reader could paste unchanged into a different paper is a sign it isn’t specific enough.

Worked Example

The example below is an illustrative composite written for this guide to demonstrate the structure — it is not a real study, and no data, institution, or finding described in it should be treated as real.

This study set out to determine whether a brief, pharmacist-led medication-reconciliation call within 48 hours of hospital discharge reduced 30-day medication-related readmissions among older adults on five or more chronic medications. Patients who received the call had a lower 30-day medication-related readmission rate than those who received usual discharge care, and the effect was concentrated among patients discharged on a new anticoagulant or insulin regimen. These findings suggest that a low-cost, brief pharmacist call targeted at patients with the highest-risk medication changes — rather than a universal call to every discharged patient — may be a more efficient way to deploy limited pharmacist time. Because the trial was conducted at a single health system, replication in settings with different discharge-planning workflows is needed before the targeted-call model is adopted more broadly.

Why this works: it restates the aim in one sentence, states the finding without repeating exact numbers already given in Results, states a contribution scaled to what a single-site study can actually support (“suggest,” “may be,” not “prove” or “should be adopted”), and closes with a specific next step (replication in different discharge-planning settings) rather than a generic call for “more research.”

Common Mistakes

• Introducing a new result, citation, or comparison for the first time. Anything appearing in the Conclusion for the first time belongs in Results or Discussion instead — the same structural error the Discussion guide flags for new findings appearing there instead of in Results.
• Copying the abstract’s conclusion verbatim. The two answer to different constraints (see above); reusing the exact sentences reads as an unfinished draft.
• Overclaiming beyond what the data support. Language like “proves,” “establishes,” or claims of generalizability to a population never studied are flagged by reviewers in the Conclusion just as often as in the Discussion — the Conclusion is not exempt from the same evidentiary discipline.
• A generic closing line with no specific content. “Further research is needed” or “these findings have important implications,” with nothing naming what research or which implications, does no real work. Purdue OWL’s general writing guidance makes the same point about closings outside the sciences: simplicity and specificity beat a dramatic, vague finish.
• Re-summarizing the entire Discussion. A Conclusion that restates every point the Discussion already made is doing the Discussion’s job twice, at length, rather than delivering a short takeaway.
• Skipping it when the journal requires one, or including one when the journal doesn’t want it. This is a checkable, mechanical requirement in the author instructions — verify it rather than assuming ICMJE’s IMRaD default applies to every target journal.

Frequently Asked Questions

Does every research paper need a separate Conclusion section?

No. ICMJE’s own Discussion guidance folds the conclusion into the end of the Discussion rather than requiring a separate section, and PLOS ONE’s author instructions explicitly say Results, Discussion, and Conclusions “may all be separate, or may be combined.” Whether a standalone Conclusion is required, optional, or not used at all is set by the target journal’s author instructions — check those before drafting.

How long should a Conclusion section be?

There’s no universal length rule. Where a journal treats it as a short standalone section, two to five sentences is typical; some clinical and engineering journals allow a longer Conclusion of a paragraph or two. Check the specific target journal’s author instructions rather than assuming a round number, the same caveat that applies to abstract word limits.

What’s the difference between the Conclusion and the Discussion?

The Discussion interprets the findings — comparing them to prior work, explaining possible mechanisms, and disclosing limitations. The Conclusion summarizes the outcome of that interpretation in a few sentences without introducing new interpretive argument. See CASRAI’s Writing the Discussion Section of a Research Paper guide for the Discussion side of that boundary in full.

Can the Conclusion cite new sources or introduce results not in the Results section?

No. Every result discussed anywhere in the paper, including the Conclusion, needs to have been reported first in Results; every comparison to prior literature belongs in the Discussion. A Conclusion that introduces either for the first time is a structural error, not a stylistic choice.

Is the Conclusion the same text as the Abstract’s conclusion sentence?

No, even though they say similar things. The Abstract’s conclusion has to stand alone under a strict word limit for a reader who may never see the full text. The paper’s Conclusion is read immediately after the Discussion, can reference specific results by name, and doesn’t need to be self-contained in the same way. Reusing identical sentences in both places is a common but avoidable shortcut.

Related CASRAI Resources

• Scholarly Writing pillar — the full manuscript-preparation and publishing-craft cluster this guide belongs to
• How to Write a Research Paper: Drafting, Style, and Revision — drafting order, register, and revision guidance for the whole manuscript
• Research Paper Format: Margins, Fonts, Spacing, and Style Rules
• How to Write a Research Abstract (With Worked Examples)
• Writing the Methods Section So It Can Be Reproduced
• Structuring the Introduction of a Research Paper: The Funnel from Context to Gap
• Writing the Results Section of a Research Paper
• Writing the Discussion Section of a Research Paper
• IMRaD structure