The discussion section is where a paper earns its keep. The results section reports what you found; the discussion explains what it means. A strong discussion interprets findings against your research question, places them inside the existing literature, and tells the reader why any of it matters. A weak one restates the numbers in slower English.
This guide walks you through a 5-part structure, gives you templates you can adapt, and shows strong vs weak examples for quantitative and qualitative work.
Need help writing your discussion section?
ThesisAI helps you turn raw findings into a clear interpretation grounded in the literature, without overclaiming.
Start WritingWhat is the Discussion Section?
The discussion section is the part of a research paper where you interpret your results in light of the research question and the prior literature. It is the second half of the D in IMRaD (Introduction, Methods, Results, and Discussion), and it usually sits between the results and the conclusion.
A good discussion section answers four questions for the reader:
- What do the results mean for the research question?
- How do they relate to what was already known?
- What are the implications - theoretical, practical, or methodological?
- What are the limits of what these results can say?
Results vs Discussion: Where to Draw the Line
A common mistake is to repeat the results in prose form. The line is simple:
- Results = what you found (numbers, tables, themes), reported neutrally.
- Discussion = what it means, why it matters, and how it fits the wider conversation.
If a sentence could appear in either section, it usually belongs in results.
The 5-Part Structure of a Strong Discussion
Most well-written discussions follow some version of this 5-part shape. You can rearrange, but you cannot skip.
1. Restate the Key Finding (Briefly)
Open by recalling your main result in one or two sentences. Not all of them - just the one that answers the research question. This anchors the reader before you interpret.
2. Interpret What the Finding Means
Explain the finding in conceptual terms. Why did you get this result? What mechanism or pattern does it suggest? Avoid simply restating the number.
3. Compare with the Existing Literature
Place your finding next to prior work. Does it confirm, extend, qualify, or contradict earlier studies? Be specific - name studies, not just "previous research".
4. Discuss Implications
Spell out what this result means beyond the paper: for theory, for practice, for methodology, or for policy. This is where you make the "so what" explicit.
5. Acknowledge Limitations and Future Directions
Close by naming the constraints that affect interpretation and pointing to the next study. Many papers split this off as its own section - either is fine, but it must appear somewhere.
Discussion Templates (Copy and Adapt)
Template 1: One Key Finding (Single Paragraph)
Fill-in-the-blanks template
The central finding of this study is that [restate finding in plain language]. This suggests that [interpretation / mechanism]. The result is consistent with [prior study], which reported [related finding], and extends that work by [what is new here]. It implies that [theoretical or practical implication]. These conclusions should be read in light of [main limitation], and future research should [specific next step].
Template 2: Multi-Finding Discussion (3 paragraphs)
Fill-in-the-blanks template
Para 1: The study set out to test whether [research question]. Three findings stand out. First, [finding 1]. Second, [finding 2]. Third, [finding 3].
Para 2: Taken together, these findings suggest that [overall interpretation]. The pattern aligns with [theory or prior work] but qualifies it in [specific way]. In particular, [the most novel point].
Para 3: For practice, this means [practical implication]. For theory, it implies [theoretical implication]. These conclusions are constrained by [limitation], and future work should [next study].
Full Discussion Examples (Strong vs Weak)
Example 1: Quantitative - Sleep and Memory Consolidation
Weak version:
The results showed that the sleep group performed better than the wake group on the memory test (p < 0.05). This is in line with previous research. The findings are important and suggest that sleep is good for memory. More research is needed.
Why it's weak: Restates the result without interpretation, vague reference to "previous research", no mechanism, no implication, and a generic close.
Strong version:
Participants in the sleep condition recalled 23% more word pairs at 24 hours than those who remained awake, a difference consistent with a consolidation account of sleep-dependent memory. The effect was larger for emotionally arousing items than neutral ones, suggesting that sleep selectively strengthens memories tagged as salient during encoding rather than rehearsing all items uniformly. This pattern extends Diekelmann and Born's (2010) review by showing that the emotional gating effect holds in a within-subjects design with controlled encoding exposure, and it complicates the alternative interference account, under which the sleep advantage should not vary by item type. Practically, the result suggests that revising emotionally charged material before sleep may yield disproportionate retention gains, though the effect size in our sample (d = 0.42) is modest enough that we would not yet recommend changes to study advice. Replication with longer retention intervals and an active wake-control condition (e.g., low-load cognitive task) is needed before stronger claims can be made.
Why it works: Restates the finding once, then interprets it; ties to a specific theory and a named prior study; explains the novel contribution; gives a calibrated practical implication; closes with a concrete follow-up design.
Example 2: Qualitative - First-Generation Students and University Belonging
Weak version:
First-generation students said they felt out of place at university. Many felt like impostors. This shows that universities should do more to support first-generation students. The findings are important for higher education.
Why it's weak: Restates a theme without interpretation, no engagement with literature, vague implication, and no acknowledgement of the bounded nature of qualitative claims.
Strong version:
Across the eighteen interviews, participants described belonging not as a fixed feeling but as something they had to actively perform - watching what they said, monitoring how they dressed, withholding details about family work. This active labour of fitting in offers a useful refinement of Lehmann's (2014) notion of dispositional adaptation: where Lehmann frames adaptation as gradual habitus shift, our participants described it as an ongoing daily cost, often persisting into their final year. The implication for student-support practice is that interventions focused on the first-term transition may miss the more sustained labour reported here; longer-running peer-mentoring relationships, extending through the degree, are likely to map more closely onto the pattern participants described. As a qualitative account based on one Russell Group institution, the themes are not offered as statistically generalisable; rather, they are transferable accounts to be tested against other institutional contexts, particularly post-1992 universities and STEM disciplines, where the cultural distance from working-class home environments may differ.
Why it works: Names the central theme and reframes it conceptually; engages directly with a specific prior study; draws an actionable practical implication; ends with calibrated scope claims that acknowledge transferability rather than overclaiming generalisability.
Common Discussion Mistakes (And How to Fix Them)
Mistake 1: Repeating the Results Section in Prose
Problem: "Group A scored 14.2 and Group B scored 11.8, a difference that was statistically significant." That belongs in results.
Fix: State the finding once at the start of the relevant paragraph, then move immediately to interpretation: what mechanism explains it, what theory it speaks to, what implication follows.
Mistake 2: Vague Engagement with the Literature
Problem: "This is consistent with previous research." Which research? In what way?
Fix: Name the study, name the specific finding it reported, and explain precisely how yours confirms, extends, or contradicts it.
Mistake 3: Overclaiming Causality from Correlational Data
Problem: "These results show that X causes Y" when the design was cross-sectional.
Fix: Use language matched to the design. "Was associated with", "predicts" (statistically), "co-varied with" - not "caused" - unless you ran an experiment with random assignment.
Mistake 4: Introducing New Results in the Discussion
Problem: A new analysis or new number appears for the first time in the discussion.
Fix: Every number you cite in the discussion must already appear in the results. If something is worth interpreting, it is worth reporting properly first.
Mistake 5: Missing the "So What"
Problem: The discussion explains the results in detail but never says why a reader outside the immediate sub-field should care.
Fix: Write one sentence per finding that completes "This matters because...". If you cannot finish that sentence, the finding may not belong in the discussion at all.
Mistake 6: A Discussion That Reads Like a Defence
Problem: Every limitation is followed by an apologetic rescue ("but the study is still important").
Fix: Acknowledge limitations clearly, then state the bounded contribution. Defensive language signals insecurity; calibrated honesty signals maturity.
How to Write the Discussion: A 6-Step Process
Step 1: List Your Findings in Order of Importance
Write each main finding on a single line. Rank them. The discussion should foreground the one that most directly answers the research question.
Step 2: For Each Finding, Draft an Interpretation Sentence
One sentence that says, in plain language, what the finding means. Not the number again - the meaning.
Step 3: Pull the 2 or 3 Most Relevant Prior Studies
For each finding, identify the small set of studies your reader would expect you to engage with. Note whether your result agrees, extends, qualifies, or contradicts each.
Step 4: Write the Implication for Each Finding
One sentence per finding completing "This matters because...". If implications cluster, group them in a single implications paragraph.
Step 5: Add Limitations and Future Directions
Either as a subsection here or as its own section after the discussion. Pair each limitation with a concrete next-study suggestion.
Step 6: Read the Discussion Without the Results Section
Does it stand on its own as an argument? If it reads like a recap of numbers, rewrite it as interpretation.
Discussion Section in Different Paper Types
The five-part structure adapts to context:
- Empirical journal article: One discussion section, ~15-25% of the paper, with limitations as a subsection.
- Master's or doctoral thesis: A longer, often standalone discussion chapter, ~10-20% of the thesis, with limitations as its own section or subsection.
- Systematic review: Discussion focuses on the strength of the evidence, gaps in the literature, and implications for practice and future research, rather than a single empirical finding.
- Short conference paper: Compressed discussion of 1-2 paragraphs - one paragraph for interpretation and prior work, one for implications and limits.
FAQs About Writing a Discussion Section
How long should a discussion section be?
Roughly 20-30% of the paper for journal articles, and one chapter in a thesis. If it is shorter than the results, you have probably under-interpreted.
Should the discussion include references?
Yes. The discussion is where you place your findings inside the existing literature, so it tends to be the second-most-cited section after the introduction.
Can the discussion include figures or tables?
Usually not new ones - new data belongs in results. A summary figure tying findings to a model is occasionally acceptable, but most discussions are prose.
How is the discussion different from the conclusion?
The discussion interprets the findings in depth and engages with the literature. The conclusion is a short summary of the contribution and its implications, usually one to three paragraphs.
Should I use first person in the discussion?
Increasingly accepted in most fields. "We interpret these findings as suggesting..." is fine. Check your target journal's style guide.
What if my findings contradict the prior literature?
Say so directly. Explain plausible reasons (sample, measure, design, context), and treat the contradiction as the most interesting feature of the paper rather than a problem to manage.
A clear, well-shaped discussion is what turns a set of results into a contribution. Treat it as the section a careful reader will spend the most time on - because they will.