The discussion section of a research paper is where many otherwise solid manuscripts fall apart. Authors who handled their methods and results with confidence suddenly start summarizing instead of interpreting, hedging every claim into mush, or quietly overclaiming what the data can support. Reviewers notice, because the discussion is where they look to see whether you understand your own findings — not just what you found, but what it means and why anyone should adjust their thinking because of it.
The section is easier to write when you stop treating it as a free-form essay and start treating it as a sequence of moves. This guide walks through those moves in order, shows how to keep the discussion distinct from the results and the literature review, and flags the mistakes reviewers cite most often when they recommend rejection or major revisions.
What the Discussion Section of a Research Paper Is For
In the standard IMRaD arrangement — introduction, methods, results, and discussion — each section answers one question. The introduction answers why the study matters. The methods answer how you did it. The results answer what you found. The discussion answers the question readers actually came for: so what?
That division of labor explains the two most common failures. A discussion that re-narrates the results answers a question the paper has already answered. A discussion that wanders into grand claims answers a question the data never asked. The section earns its place by doing interpretive work: connecting findings to the problem posed in the introduction, to prior scholarship, and to consequences for practice or theory. If the architecture of the whole paper is still fuzzy, our overview of the IMRaD structure and why reviewers expect it is a useful companion to this guide.
The Five Moves of a Strong Discussion
Across disciplines, well-received discussions tend to make five moves in a predictable order. You can treat each as a paragraph or a cluster of paragraphs depending on the complexity of the study. Field conventions shift the emphasis — clinical journals expect prominent limitations, humanities articles blend interpretation throughout, and some social science outlets merge implications into a separate conclusion — but the underlying sequence travels remarkably well.
| Move | What It Does | Typical Share |
|---|---|---|
| 1. Restate the key finding | Opens with your main answer in plain language, tied to the research question | One short paragraph |
| 2. Interpret | Explains what the finding means and the mechanism or reasoning behind it | Largest share |
| 3. Situate in the literature | Shows where results confirm, complicate, or extend prior work | 1–2 paragraphs |
| 4. Acknowledge limitations | States honest boundaries on the claims and why they do not void them | One paragraph |
| 5. Implications and next steps | Spells out consequences for practice or theory and specific future research | 1–2 paragraphs |
Move 1: Lead With the Answer
Open the discussion by answering your research question directly, in one or two sentences a non-specialist could follow. Resist the urge to warm up with background — the introduction already did that. A clean opening move sounds like: the intervention improved completion rates, and the effect was concentrated among part-time students. Everything that follows hangs off that sentence.
Move 2: Interpret, Do Not Recap
Interpretation means proposing why the results look the way they do and what they reveal about the underlying phenomenon. Offer the most plausible explanation, then briefly weigh a rival explanation and say why you find it less convincing. Unexpected or null findings deserve the same treatment, not burial: an honest paragraph about a surprising result is often the most cited part of a paper, because it is where the field learns something.
Move 3: Put the Findings in Conversation
Return to the key studies from your introduction and say precisely how your results relate to them: consistent with, contrary to, or extending under new conditions. Two or three well-chosen comparisons beat a parade of citations. This move should read like a conversation among a handful of papers, not a second survey of the field — if you find yourself summarizing sources again, you have drifted back into the territory covered in our step-by-step guide to writing a literature review, and the material belongs there instead.
Move 4: Limitations That Build Credibility
State the genuine boundaries of the study — sample, design, measurement, context — and, for each, one sentence on why the core claim survives it. Skip the ritual confession of every conceivable weakness; reviewers read boilerplate limitations as a sign the author has not thought hard about which ones matter.
Move 5: End on Consequences, Not Clichés
Close with what should change because of this work: a practice recommendation, a theoretical refinement, or a policy consideration, scaled honestly to your evidence. Then name one or two specific next studies — the design you would run tomorrow — rather than the empty formula that more research is needed.
Keeping the Discussion Separate From the Results
Writers in fields that split results and discussion into separate sections often struggle with the boundary. A workable rule: the results section reports observations with numbers, tables, and neutral verbs like showed and increased; the discussion trades in meaning, using verbs like suggests, indicates, and challenges. If a sentence contains a statistic and no interpretation, it probably belongs in results. If it contains a claim about why or so what, it belongs in the discussion. Journals that combine the two sections still expect you to alternate deliberately between reporting and interpreting rather than blending them into porridge.
A Quick Before-and-After Example
Before: “As shown in Table 3, students in the flipped section scored higher on the final exam than students in the lecture section. Participation was also higher in the flipped section, as shown in Table 4. These results demonstrate that flipped classrooms are superior for student learning.”
After: “The flipped section’s advantage on the final exam suggests that moving practice into class time, where students could get immediate help, mattered more than the video lectures themselves. This reading is consistent with the higher participation we observed and helps explain why prior flipped-classroom studies that kept passive in-class time have reported weaker effects. Because sections were not randomly assigned, instructor enthusiasm remains a rival explanation, though both sections shared materials and exams.”
The first version repeats the tables and then leaps to a claim the design cannot carry. The second interprets, proposes a mechanism, connects to prior work, and folds in a limitation — four moves in three sentences, with no numbers restated.
Common Mistakes Reviewers Flag
- Overclaiming. Sliding from association to causation, or from one sample to humanity at large. Match the size of the claim to the strength of the design.
- Introducing new results. Every number discussed must already appear in the results section. New findings surfacing in the discussion make reviewers wonder what else moved around.
- Ignoring inconvenient findings. Reviewers read the results first. A discussion that skips the awkward table looks evasive rather than efficient.
- Hedging into meaninglessness. Triple-stacked qualifiers protect you from criticism by ensuring you said nothing. One honest qualifier per claim is usually enough.
- The mirror-image discussion. Walking through every result in the same order at the same length, which is a summary wearing a costume. Organize by importance instead.
A Drafting Process That Makes It Easier
Draft the discussion immediately after finalizing your results tables, while your reasons for every analytical choice are fresh. Start by writing the single sentence version of your answer, then expand each move in order. Read the introduction and discussion back to back before submitting: every question raised early should be answered late, and no promise should be left dangling. For sentence-level sharpening — cutting throat-clearing, taming nominalizations, controlling paragraph focus — the techniques in our collection of academic writing tips for clearer research papers apply to this section more than any other.
One more habit separates experienced authors: they test the discussion on an audience before submission. Presenting the paper at a conference forces you to compress your interpretation into its clearest form, and the questions you field are a preview of peer review. NIVA’s virtual conferences publish accepted work in official conference proceedings with an ISBN, and strong papers can develop toward NIVA’s peer-reviewed journals — a low-cost path for testing and strengthening a manuscript before it faces a journal editor.
Frequently Asked Questions
Five elements in order: a plain-language restatement of the key finding, interpretation of what it means, comparison with prior studies, honest limitations, and implications with specific future research directions. Each element can be a paragraph or more depending on the complexity of the study.
A common benchmark is roughly a quarter to a third of the manuscript, often four to eight paragraphs in a standard empirical article. Follow the norms of your target journal by examining recent published articles, and let the complexity of your findings, not padding, determine the length.
The results section reports what you found, using numbers, tables, and neutral description. The discussion interprets those findings: what they mean, how they relate to prior research, what limits them, and why they matter. Report in results, interpret in discussion, and never introduce new findings in the discussion.
Name the two or three limitations that genuinely constrain your claims, then explain briefly why your central conclusion holds despite each one. Selective, well-reasoned limitations signal analytical maturity and build reviewer trust, while an exhaustive list of every possible weakness reads as boilerplate and invites doubt.
Your discussion deserves a live audience before it meets a journal editor. Submit your work to NIVA’s next virtual conference and present from anywhere — faculty register for $150, students for $70, membership is free, and accepted papers appear in the ISBN proceedings.