Research Paper Structure
25 Research Paper Conclusion Examples by Methodology
Real-world conclusion examples for quantitative, qualitative, and mixed-methods research papers—with side-by-side comparisons and methodology-specific templates.
You've written the findings. The cursor is sitting under “Conclusion,” and every sentence sounds either too small or wildly overconfident.
A good research paper conclusion does one job: it closes the loop between your research question, your evidence, your limits, and the next useful step. The examples below are organized by methodology because a quantitative conclusion should not sound like an ethnography, and a mixed-methods conclusion has to do extra work that a single-method paper doesn't.
Use these as models for structure and language, not as claims about real studies. If you're comparing your own draft against them, an AI research workspace like Otio can help you check whether your conclusion restates the question, stays inside scope, and uses methodology-appropriate hedging.
Who this list is for
This list is for the student who understands the study but can't land the plane. Undergrads, master's students, PhD candidates, capstone writers, and researchers switching methods all run into the same problem: the conclusion has to sound final without pretending the study answered everything.
If you're writing Chapter 5, start with the examples closest to your design. A thesis conclusion usually has more room for limitations and recommendations than a journal article, but the core pattern is the same. The conclusion section of a research paper has to echo the question, synthesize the finding, and stop before it becomes a second discussion section.
Defense prep is another use case. When a committee asks, “So what did your study contribute?”, they're often asking whether your conclusion matches your design. A survey can estimate association. An ethnography can interpret meaning. A randomized trial can make a stronger causal claim, though even there the boundary conditions matter.
How we picked these 25 examples
I grouped the examples by methodology rather than discipline. That makes the patterns portable: a quantitative education paper and a quantitative health paper often close in similar ways, even when the subject matter has nothing in common.
The sources behind the guidance come from university writing centers, research-methods references, and open-access dissertation infrastructure. For actual dissertation models, Stanford Libraries points students to OATD, the Open Access Theses and Dissertations index, which is useful when you need to see how real committees accept conclusions in full-length projects.
A practical caveat: the example conclusions below are composite models. They mimic the structure, scope, and language of strong conclusions, but the study details are illustrative. Don't copy their statistics or pretend they came from your data. That's how a good template becomes academic misconduct in five minutes.
The selection criteria were simple:
Does the conclusion answer the research question without repeating the abstract?
Does it connect the finding back to literature, theory, practice, or policy?
Does it name limitations in a way that narrows the claim?
Does the final recommendation specify a design, population, mechanism, or setting?
Does the language fit the method?
If you're still shaping the whole paper, the conclusion will be easier after the methodology section of your research paper is clean. Muddy methods make for mushy endings.
What makes a strong research conclusion
A strong conclusion starts by returning to the research question in changed language. You don't need to write, “This study set out to…” every time. You do need to make clear what was asked and what your evidence now supports.
The San Jose State University Writing Center guide to research paper conclusions frames the conclusion as the place to synthesize rather than repeat. That's the line many drafts miss. A weak conclusion recaps section headings; a stronger one tells the reader what those sections now add up to.
The second move is literature connection. Your conclusion should say whether the findings support, challenge, refine, or extend prior work. Use restraint. “Extends” is safer than “overturns” unless your design truly earns the bigger verb.
Limitations belong here too. Not as an apology. As scope control.
Weak conclusion habit | Stronger move |
|---|---|
Repeats the abstract in new order | Synthesizes what the findings mean together |
Says “future research is needed” | Names a specific design or population |
Claims causality from correlation | Uses association language unless the design supports causality |
Hides limitations | States limits and explains how they shape interpretation |
Adds a new citation at the end | Works only with evidence already introduced |
The UNC Writing Center's guidance on conclusions warns against ending by merely restating the thesis. For research writing, the equivalent mistake is repeating the result without explaining its reach.
Methodology also changes the verbs. Quantitative conclusions often use “indicates,” “is associated with,” “predicts,” and “may suggest.” Qualitative conclusions use “emerged,” “participants described,” “the analysis revealed,” and “the theme suggests.” Mixed-methods conclusions have to integrate: the qualitative strand should explain, complicate, or contextualize the quantitative strand.
Quantitative research conclusions (8 examples)
Quantitative conclusions need numbers, but not number soup. Report the main estimate, include uncertainty when available, then explain what the estimate means in relation to the hypothesis or model.
The PMC practical guide to writing quantitative and qualitative research questions and hypotheses is useful here because it treats question type, design, and inference as linked. If your question asks about association, your conclusion should not smuggle in causal language. Easy mistake. Painful in a defense.
Example 1: Experimental psychology study
“The results indicate that cognitive load significantly impaired working memory retention, with medium-to-large effects across task conditions. These findings extend working-memory models by suggesting that dual-task interference persists even under favorable encoding conditions. Future studies should test whether metacognitive training reduces this interference.”
Why it works: it restates the hypothesis, reports direction and magnitude, ties the finding to theory, then proposes a mechanism-specific next step.
Example 2: Epidemiological cohort study
“This longitudinal analysis found a dose-response association between physical activity and cardiovascular mortality across the follow-up period. The magnitude of the association aligns with prior population-level estimates and supports current public-health guidance. Limitations include self-reported activity and attrition; future studies should use objective activity tracking in more diverse samples.”
Why it works: cohort language stays careful. It doesn't say exercise “proved” lower mortality risk. It also names the measurement weakness directly.
Example 3: Clinical trial conclusion
“Compared with standard care, the intervention group showed a clinically meaningful reduction in HbA1c that persisted at 12-month follow-up. The effect suggests practical value for diabetes management under trial conditions. Adherence declined during the final six months, so maintenance strategies should be tested before wider implementation.”
Why it works: clinical significance gets separated from statistical significance. The conclusion also notices the real-world snag: people stop following protocols.
Example 4: Economics or policy analysis
“Regression discontinuity analysis suggests that the policy reduced unemployment claims in the first year, with smaller effects in later periods. Results remained stable across alternative model specifications, though regional variation indicates that implementation context shaped outcomes. Policymakers should test targeted rollout in jurisdictions with similar labor-market conditions.”
Why it works: the conclusion reports the effect pattern, checks sensitivity, and avoids turning one policy setting into universal law.
Example 5: Survey-based social science
“Structural equation modeling indicates that social support mediates a substantial share of the relationship between workplace autonomy and job satisfaction. The model explains more variance than prior cross-sectional estimates, suggesting that autonomy may operate partly through perceived support. Longitudinal designs are needed to establish timing and reduce concern about reverse causality.”
Why it works: mediation claims can get slippery. This version keeps the causal chain provisional.
Example 6: Biomedical lab study
“CREB1 knockout reduced tumor growth in xenograft models and increased survival relative to controls. Mechanistic analysis suggests that CREB1 loss impaired angiogenesis while increasing apoptosis in tumor tissue. These preclinical findings justify early translational testing, though off-target effects and immunogenicity should be evaluated before human trials.”
Why it works: it connects outcome to mechanism. It also marks the boundary between animal models and clinical application.
Example 7: Multivariate education analysis
“After controlling for socioeconomic status and prior achievement, teacher experience and class size were associated with between-school differences in test scores. The effect of teacher experience was strongest in high-poverty schools, suggesting that development programs may have uneven benefits across contexts. Future evaluation should test whether targeted teacher support improves outcomes in under-resourced settings.”
Why it works: the conclusion names the interaction instead of flattening the result. That keeps the implication useful.
Example 8: Meta-analysis conclusion
“Across randomized trials, cognitive-behavioral therapy produced a large pooled effect for anxiety symptoms, with moderate heterogeneity across samples. Subgroup results suggest stronger effects in adult samples and when delivered by specialists. Publication bias may inflate the estimate, but conservative adjustment still indicates clinically meaningful benefit.”
Why it works: meta-analysis conclusions should mention heterogeneity and bias. If they don't, readers will ask.
For more help reading the numbers before writing the ending, see this guide on how to analyze a research paper. It pairs well with quantitative conclusion drafting because the conclusion can only be as precise as your interpretation of the results.
Qualitative research conclusions (7 examples)
Qualitative conclusions close around meaning, not estimates. The evidence is still empirical, but the claim usually concerns themes, processes, identities, discourse, or experience.
The trap is sounding vague. “Several themes emerged” doesn't tell the reader much. A strong qualitative conclusion names the themes, explains the relationship between them, and makes a bounded claim about transferability.
Example 1: Phenomenological study
“Participants' lived experience of chronic pain centered on loss of identity, encounters with medical systems, and efforts to reclaim agency through self-advocacy. These themes appeared across participant backgrounds, suggesting that they may represent recurring dimensions of chronic illness. Because the sample centered English-speaking insured participants, future research should examine how language access and insurance status shape the pain experience.”
Why it works: the conclusion names the phenomenon and the sample boundary in the same breath.
Example 2: Grounded theory study
“The core category ‘managed ambiguity’ explains how early-career academics navigate publishing pressure, uneven mentorship, and professional identity formation. Participants adapted through selective disclosure, strategic collaboration, boundary-setting, and reframing failure. The model extends communities-of-practice theory by showing how newcomers actively construct belonging under uncertain institutional norms.”
Why it works: grounded theory needs a core category. Without one, the conclusion reads like thematic analysis wearing the wrong coat.
Example 3: Case study
“The case of Organization X shows how decentralized decision-making, paired with transparent communication norms, supported rapid adaptation during crisis. Distributed authority and cross-functional trust interacted to make decisions faster without removing accountability. Although the findings are context-specific, the mechanisms may transfer to similarly hierarchical organizations facing disruption.”
Why it works: case study conclusions should not pretend one case equals a law. They can still identify mechanisms.
Example 4: Narrative analysis
“Participants' recovery narratives shifted from shame-centered accounts toward meaning-centered identities, often after relapse, mentorship, or spiritual experience. This arc aligns with redemption-sequence theory while showing how social witnessing can accelerate narrative reconstruction. Recovery interventions should create room for narrative work alongside behavioral support.”
Why it works: narrative conclusions should describe movement. Here, the movement is identity reconstruction.
Example 5: Ethnographic study
“Six months of observation in the emergency department revealed clinical decision-making as a distributed social process shaped by space, tools, routines, and relationships. Experienced nurses ‘read the room’ by integrating patient cues with team dynamics under time pressure. This perspective challenges individualistic models of clinical judgment and suggests that training should include collaborative sensemaking.”
Why it works: ethnography earns its keep by showing what surveys often miss: the lived choreography of practice.
Example 6: Discourse analysis
“Analysis of policy documents and stakeholder interviews showed that ‘resilience’ functioned as a contested term. Government actors used it to emphasize individual responsibility, while community organizations used it to argue for collective support. These competing framings shaped which interventions appeared fundable and whose account of harm became visible.”
Why it works: discourse analysis should connect language to power. This conclusion does that without overexplaining the method.
Example 7: Phenomenological focus group study
“Across focus groups with parents of children with autism, the central tension was accepting neurodiversity while advocating for support in a neurotypical world. Parents described moving between these positions rather than resolving them. Support programs should validate this ambivalence instead of pushing families toward a single preferred stance.”
Why it works: the conclusion preserves complexity. That's often the mark of good qualitative writing.
If you're still choosing a design, compare these examples with the main types of qualitative research methods. A phenomenological conclusion and a discourse-analysis conclusion should not sound interchangeable.
Mixed-methods research conclusions (5 examples)
Mixed-methods conclusions have a harder assignment: they must report integration, not sequence. If the quantitative and qualitative findings sit beside each other like two strangers on a bus, the conclusion isn't finished.
ATLAS.ti's guide on structuring a mixed-methods research paper emphasizes the need to bring methods and findings together in the paper's structure. The conclusion is where weak integration becomes obvious.
Example 1: Convergent mixed-methods study
“Survey results showed that most employees reported work-life conflict, with significant variation by gender. Interviews clarified the mechanism: women more often described invisible household labor, while men more often described blurred work boundaries. Together, the findings suggest that work-life conflict is unevenly distributed and that flexible scheduling alone may not address gendered expectations.”
Why it works: the qualitative strand explains the shape of the quantitative result. That's integration.
Example 2: Explanatory sequential mixed-methods study
“Initial survey data showed substantial unmet mental-health needs in primary care. Follow-up interviews with patients and providers identified stigma, short appointment times, and fragmented referrals as the main barriers. These qualitative findings explain why the service gap persists and point toward integrated-care trials that target patient, provider, and system constraints.”
Why it works: the second phase answers the “why” left open by the first phase.
Example 3: Exploratory sequential mixed-methods study
“Interviews with teachers first identified classroom relevance, peer collaboration, and administrative support as perceived drivers of professional-development quality. The survey then quantified these factors and found that collaboration and support mediated the relationship between relevance and student outcomes. The findings suggest that professional development should be designed as an institutional practice rather than a one-off training event.”
Why it works: the qualitative phase builds the instrument; the quantitative phase tests the pattern.
Example 4: Embedded mixed-methods study
“The trial showed a reduction in bullying incidents after the school-based intervention, but case studies explained why some schools improved more than others. In high-response schools, leaders modeled the intervention's values in staff meetings and discipline decisions. Implementation research should test whether leadership training strengthens the intervention effect.”
Why it works: embedded qualitative evidence explains variation inside the quantitative outcome.
Example 5: Convergent mixed-methods health outcomes study
“Patient outcome data showed reduced hospital readmissions after the telehealth intervention. Interviews with patients and clinicians suggested that the reduction occurred because patients caught problems earlier and clinicians gained better insight into home barriers. The combined evidence supports testing telehealth scale-up in similar populations, while equity concerns remain for patients without broadband access or digital confidence.”
Why it works: the conclusion integrates outcome, mechanism, and limitation. It doesn't just say both strands were positive.
For a broader overview of design choices, this guide to research methodology types is worth checking before you write the conclusion. Method choice determines how far the final claim can travel.
Experimental and action research conclusions (3 examples)
Experimental conclusions should talk about effects, controls, replication, and boundary conditions. Action research conclusions should talk about cycles, participation, local learning, and whether the change can survive after the study ends.
Example 1: Lab-based experimental design
“Across three experiments, growth-mindset language increased persistence on unsolvable puzzles and improved later problem-solving performance. The effect persisted after controlling for baseline motivation, but it weakened under high-stakes conditions. These findings support motivational theories of mindset while suggesting that brief primes may require reinforcement outside laboratory settings.”
Why it works: replication across experiments strengthens the claim. The boundary condition keeps it honest.
Example 2: Quasi-experimental interrupted time series
“After the emergency department introduced a new triage protocol, average wait times decreased and remained lower across the follow-up period. Patient satisfaction improved, but staff workload stress increased and turnover rose during the first year. Efficiency gains therefore came with a measurable human cost, and future protocol design should include frontline staff before implementation.”
Why it works: it reports intended and unintended outcomes. Many weak conclusions hide the second part because it complicates the story.
Example 3: Participatory action research
“Across three action-research cycles, community members helped design, test, and refine a neighborhood food-security initiative. Early cycles showed that access depended on cultural preference and trust, not proximity alone. The final model reached more eligible households after local leaders were trained, suggesting that sustainability depends on community ownership and policy support.”
Why it works: the conclusion follows the action-research cycle. It also centers local knowledge rather than treating the community as a data source.
Example 4: Design-based research
“Iterative testing of the digital tutoring module showed that student engagement improved when feedback was immediate and tied to visible progress markers. Classroom observations indicated that teachers adapted the tool most successfully when it fit existing routines. Future design cycles should test whether these features hold across schools with different technology access.”
Why it works: design-based research conclusions should feed the next design cycle. This one does.
Example 5: Program evaluation
“The after-school mentoring program was associated with improved attendance and stronger student-reported belonging, but academic gains were smaller and varied by site. Interviews suggest that mentor consistency shaped student engagement more than program length. Future evaluation should compare mentor-training models before expanding the program.”
Why it works: evaluation conclusions need to be useful to decision-makers. This one tells them what to test next.
Common pitfalls to avoid in your conclusion
The most common conclusion failure is new material. A citation you just discovered at midnight belongs in the literature review or discussion, not the final paragraph. If it appears only in the conclusion, the reader wonders what else arrived late.
The Purdue OWL guidance on conclusions warns writers not to introduce new evidence at the end. Research papers have the same rule, with higher stakes. New evidence changes the argument after the reader has stopped evaluating it.
Over-claiming is the second failure. “This proves” is rarely your friend. Use “suggests,” “indicates,” “is associated with,” or “may help explain” when the design calls for restraint.
Under-claiming is quieter but just as damaging. Some writers become so cautious that the conclusion says almost nothing. If your study found a pattern, name it. Then bound it.
Watch for these specific mistakes:
Introducing new data: save late findings for revision, not the final paragraph.
Repeating the abstract: the abstract previews; the conclusion synthesizes.
Ignoring limitations: committees notice. So do peer reviewers.
Writing vague recommendations: “more research is needed” is a placeholder, not a plan.
Forgetting the research question: the ending should answer the opening.
Harvard's writing guidance on effective conclusions frames the ending as the place where the reader understands why the argument mattered. In research writing, “mattered” usually means contribution, implication, or next inquiry. Pick the one your evidence can support.
How to use this list and next steps
Pick two or three examples from your methodology and read them aloud. You'll hear the verbs before you can explain the rule. Quantitative examples tend to move from estimate to implication; qualitative ones move from theme to interpretation.
Then mark your own draft against five checks:
Does the first sentence return to the research question?
Does the middle synthesize findings rather than list them?
Are limitations specific enough to guide interpretation?
Does the recommendation name a design, setting, population, or mechanism?
Does the language match the method?
If you're writing a mixed-methods paper, draft the joint insight before polishing sentences. One useful test: remove the qualitative finding. If the conclusion still says the same thing, you haven't integrated the methods yet.
I've watched students spend an hour swapping “shows” for “suggests” while leaving the main problem untouched: the conclusion never answers the original question. Start with the claim. Polish later.
If you want help stress-testing the draft, use Otio's AI chat with your paper and source documents to compare your conclusion against the examples above, check whether the scope matches your methodology, and tighten the next-step recommendation.
FAQ
Q: How long should a research conclusion be?
A: For most journal articles, one to three paragraphs is enough. For a thesis or dissertation chapter, the conclusion may run several pages because it often includes implications, recommendations, and limitations.
Q: Should I introduce new citations in the conclusion?
A: No. If a citation is necessary, it belongs earlier in the paper where it can be discussed and integrated. The conclusion should synthesize what you've already presented.
Q: How do I write a conclusion for a mixed-methods study?
A: State what each strand found, then explain the joint insight. The strongest mixed-methods conclusions show how one strand explains, complicates, or strengthens the other.
Q: What's the difference between a discussion and a conclusion?
A: The discussion interprets findings in depth and compares them with prior literature. The conclusion closes the paper by restating the central answer, naming limits, and pointing to the next useful step.
Q: How do I avoid over-claiming in my conclusion?
A: Match the verb to the design. Use “suggests” or “is associated with” for correlational work, reserve causal language for designs that can support it, and state limitations plainly.
