The Discussion is where a research paper stops reporting and starts arguing. The Results section establishes what was found; the Discussion has to explain what those findings mean, how they relate to what was already known, where the study’s conclusions run out, and why any of it matters. It is consistently one of the hardest sections to write well — not because the writing itself is technically difficult, but because it requires interpretive judgment the Methods and Results sections deliberately avoid. This guide focuses specifically on that interpretive work: how to structure a Discussion section, how to compare findings to prior literature without either ignoring it or over-crediting it, how to write an honest limitations paragraph, and the overclaiming pitfalls that reviewers flag most often. For sequencing the Discussion within the rest of the drafting process, and for register/tense conventions that apply across the whole manuscript, see CASRAI’s How to Write a Research Paper guide, which this page extends.
What the Discussion Section Is For
In the standard IMRaD (Introduction, Methods, Results, Discussion) structure used across most empirical scientific writing, each section has a distinct job, and mixing them is one of the most common structural weaknesses reviewers cite. Results reports what was observed, without interpretation. Discussion explains what those observations mean. A Discussion section that restates Results findings without adding interpretation is not doing its job; a Results section that starts interpreting findings mid-report is doing the Discussion’s job too early. Keeping the two separated — even when it feels repetitive to describe a finding again before interpreting it — is a convention worth following because it is what readers and reviewers expect. See CASRAI’s IMRaD structure entry for the full four-section breakdown.
A Discussion section generally needs to accomplish several distinct things, usually in something close to this order:
- Restate the main finding(s) briefly, without re-reporting statistics already given in Results.
- Interpret what the findings mean in relation to the study’s original question or hypothesis.
- Situate the findings against prior work — where they agree, where they diverge, and why.
- Acknowledge the study’s limitations honestly and specifically.
- State the implications — theoretical, practical, or both — without overreaching beyond what the data support.
- Suggest future directions, where genuinely warranted by an open question the study surfaced rather than as a stock closing sentence.
Not every paper needs all six in that exact sequence, and some journals combine Results and Discussion into one section, or use a structured Discussion with explicit sub-headings (a format increasingly common in clinical and health-sciences journals). Whatever the target journal’s convention, check its author guidelines before drafting, since it determines both structure and expected length.
Interpreting Results Without Overclaiming
The most common failure mode in a Discussion section is not omission but overreach: stating a conclusion with more certainty than the study design supports. A single study, especially an observational or modestly powered one, rarely “proves” anything on its own — it provides evidence consistent with, or inconsistent with, a hypothesis. Words like “prove,” “confirm,” or “establish” are frequently flagged by reviewers and editors as overclaiming unless the evidence genuinely warrants that strength of language (a well-powered randomized trial replicating an established effect is different from a single small pilot study).
A useful discipline while drafting is to trace every interpretive claim in the Discussion back to something actually shown in Results. If a sentence asserts something the data don’t directly support — a causal claim from correlational data, a claim about a population never sampled, a mechanism the study wasn’t designed to test — it belongs either in a hedged form (“these findings are consistent with…”, “one possible explanation is…”) or in the future-directions discussion, not stated as established fact. This same discipline is covered from the drafting-order side in CASRAI’s How to Write a Research Paper guide’s pitfalls section.
Distinguish explicitly between statistical significance and practical or clinical significance. A result can be statistically significant (unlikely to be due to chance, given the sample size) without being large enough to matter practically, and a large, meaningful effect can fail to reach statistical significance in an underpowered study. Conflating the two — treating “significant” as synonymous with “important” — is a specific, checkable overclaiming pattern reviewers watch for.
Non-significant or null results deserve real interpretation, not a one-line dismissal. A null result can still be informative: it can rule out an effect of a certain size, it can be inconsistent with a competing theory, or it can simply mean the study was underpowered to detect a real effect — and those are different claims that call for different Discussion text. Treating “we found no significant effect” as needing no further comment is a missed opportunity and, in some fields, a contributor to publication bias against null findings.
Comparing Findings to Prior Work
Situating a study’s findings against the existing literature is what turns an isolated result into a contribution to a field. This section of the Discussion should engage with prior work specifically, not generically — naming what earlier studies found and how the current findings agree, diverge, extend, or complicate that picture, rather than a vague gesture toward “previous research” already covered in the Introduction’s literature review.
A few patterns are worth applying deliberately:
- When findings agree with prior work, say so specifically, and note what the agreement adds — a different population, a different method, a larger sample, replication in a new context. Simple confirmation without added value is a weaker contribution than the Discussion should imply it is.
- When findings diverge from prior work, resist the temptation to either ignore the discrepancy or dismiss the earlier study outright. Consider genuine explanations: different populations, different measurement instruments, different effect sizes that are directionally consistent but not statistically distinguishable, or a genuine field-level disagreement that the current study doesn’t resolve on its own.
- Avoid citation-dumping — a string of citations with no explanation of what each one showed or how it relates to the current finding reads as padding, not synthesis, and reviewers notice the difference.
The Introduction’s literature review and the Discussion’s comparison-to-prior-work should not be identical text reused in a different place — the Introduction establishes what was known before the study; the Discussion revisits that same literature now that the study’s own findings exist, asking what has changed. If both sections read as interchangeable, one of them is not doing its distinct job.
Writing the Limitations Paragraph
Every study has limitations, and naming them specifically is a marker of a credible manuscript, not an admission of weak work. Reviewers and readers both know no single study is definitive; what damages credibility is a limitations paragraph that is vague, perfunctory, or absent, not the existence of real limitations.
An effective limitations discussion is specific about what kind of limitation is being disclosed and what it means for interpreting the findings, for example:
- Sample — size, composition, recruitment method, and what population the findings can and cannot be generalized to.
- Design — cross-sectional versus longitudinal, observational versus experimental, and what causal or temporal claims the design does or doesn’t support.
- Measurement — instrument reliability/validity, self-report versus objective measures, potential measurement error.
- Analysis — statistical power, missing data handling, assumptions the chosen model depends on.
- Scope — what the study didn’t attempt to measure or control for, distinct from what it measured imperfectly.
A limitation is more useful to the reader when it’s paired with its consequence for interpretation, rather than listed as a bare fact. “The sample was drawn from a single institution” is a weaker limitations sentence than “the sample was drawn from a single institution, so findings may not generalize to settings with different patient demographics or care models.” The second version tells the reader exactly how much weight to put on the finding; the first just names a fact and leaves the reader to work out its implications alone.
Limitations belong in the Discussion, not buried in Methods (where they can read as an oversight rather than a disclosed constraint) and not deferred entirely to a peer reviewer to point out. Some journals require a dedicated “Limitations” sub-heading; even where that’s not required, giving limitations their own paragraph rather than scattering single sentences throughout the Discussion makes them easier for a reader to find and weigh.
Stating Implications and Future Directions
The implications paragraph answers “so what” — what changes, for theory, practice, or both, given what the study found, bounded by what the study can actually support. Implications should be scaled to the evidence: a single study’s implications for immediate practice change are usually more modest than its implications for how a field’s understanding of a question should shift, and conflating “worth further investigation” with “should change current practice” is a common overclaim.
Future-directions text is most useful when it names a specific, motivated next question the current study’s own findings or limitations surfaced — a confound that couldn’t be controlled for, a population not yet studied, a mechanism the design couldn’t test directly — rather than a generic closing line (“further research is needed”) that could be appended to any paper regardless of its findings. A future-directions sentence that a reader could paste into a different paper without noticing is a sign the sentence isn’t doing real interpretive work.
Common Pitfalls
- Restating Results instead of interpreting them. Repeating numbers already reported, without explaining what they mean, is the single most common Discussion weakness reviewers cite.
- Introducing a new finding in the Discussion. Any result mentioned for the first time in the Discussion, with no corresponding entry in Results, is a structural error — every finding discussed needs to have been reported first.
- A one-sentence, generic limitations paragraph. “Further research with a larger sample is needed” as a stand-alone limitations statement, unconnected to what was actually limited about the specific study, reads as boilerplate rather than genuine disclosure.
- Overclaiming causation from correlational or observational data. Language implying a causal mechanism when the design can only establish association.
- Ignoring findings that don’t fit the expected story. Selectively discussing only the results that support the paper’s narrative, while leaving contrary or null findings unaddressed, undermines credibility and is the kind of gap peer reviewers are specifically trained to look for.
- Padding with citations instead of synthesizing them. A list of loosely related citations with no stated relationship to the current findings.
Frequently Asked Questions
How long should a Discussion section be?
There is no universal length rule — it varies by journal, field, and study complexity. As a general orientation, many empirical papers give the Discussion roughly similar space to the Introduction, though some journals cap it explicitly in their author guidelines (particularly clinical and biomedical journals with strict word limits). Check the target journal’s instructions for authors rather than relying on a generic rule of thumb.
What is the difference between the Discussion and the Conclusion?
Where both exist as separate sections, the Discussion does the interpretive work described in this guide — interpreting results, comparing to prior work, and disclosing limitations — while a separate Conclusion (when a journal requires one) is typically a short, focused summary of the paper’s main takeaway and its broader significance, without introducing new interpretive arguments. Many journals combine the two into a single “Discussion” or “Discussion and Conclusion” section; check the target journal’s structure requirements.
Should limitations go in the Discussion or a separate section?
Convention and journal requirements vary. Most commonly, limitations are addressed within the Discussion, either in a dedicated paragraph or under an explicit “Limitations” sub-heading; some journals require it as a distinct sub-section, and a smaller number ask for it directly before the Conclusion. Whatever the placement, limitations should be specific to the study, not a generic disclaimer.
Is it acceptable to discuss a null or non-significant result at length?
Yes, and it is often necessary. A null result can meaningfully inform a field — ruling out an effect of a given size, contradicting a competing theory, or flagging that the study may have been underpowered — and the Discussion should say which of those interpretations applies, rather than treating a non-significant finding as not worth discussing.
How many citations should the Discussion include compared to the Introduction?
There’s no fixed ratio. What matters is that every citation used in the Discussion is doing comparative work — showing agreement, divergence, or extension relative to the current findings — rather than being reused from the Introduction’s literature review without added interpretive purpose.
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
- IMRaD structure
- How to Write a Research Paper in APA Style
- Citation Styles and Referencing Formats
- ICMJE Authorship Criteria
- What Is a Dissertation? Structure, Defense, and Deposit
