If the Results section is where you report what happened, the Discussion is where you say what it means — and it is where many otherwise-solid papers stumble. The most common failure is simple: the authors just restate their results in words, adding no interpretation. A good discussion does something harder and more valuable. It takes your findings, places them in the wider conversation, and tells the reader what they should now think.
Interpret, don't repeat
The cardinal rule: the Discussion interprets; it does not re-report. Your reader already saw the numbers in the Results. Here, you explain what those numbers imply — whether they support your hypothesis, what mechanism might explain them, and how confident the reader should be. If a sentence in your Discussion could sit unchanged in your Results, it is in the wrong section.
A worked example: restating versus interpreting
The study below is invented to illustrate the structure; its numbers are not real data. Suppose a cohort study found fewer wound infections with a new dressing, with a risk ratio of 0.70 and a 95% confidence interval (CI) of 0.52 to 0.94.
| Restating (belongs in the Results) | Interpreting (belongs in the Discussion) |
|---|---|
| Infection occurred in fewer patients in the dressing group (risk ratio 0.70, 95% CI 0.52 to 0.94). | The lower infection rate is consistent with [reference]. Patients were not randomised, so the difference may partly reflect which patients received the dressing. |
The second version adds what the first cannot: a link to earlier work and an honest reason for caution.
A structure that works
Most strong discussions follow the same arc, from specific to broad:
- Lead with your key finding. Open the Discussion by stating, in plain words, the most important thing you found and whether it answered your question.
- Place it in the literature. How does it agree with, extend, or contradict previous work? Engage honestly with studies that disagree — do not cite only the ones that support you.
- Offer an explanation. Why might you have found this? Propose a mechanism or interpretation, and be clear about how speculative it is.
- State the limitations. Name the boundaries of what you can conclude (see writing a limitations section).
- Say what it means. The implications for practice, policy or future research — concrete, not "more research is needed".
Match your Discussion wording to your study design
The study design sets the ceiling on what the Discussion may claim. Words such as effective or superior need a comparison group, and a direct causal claim needs a randomised design. Our guide to choosing between a case report, case series and cohort study explains why.
| Design | Wording that fits | Wording to avoid |
|---|---|---|
| Case report | "This case shows that [finding] can occur" | "[treatment] is effective" |
| Case series | "In these patients, [procedure] was feasible" | "[procedure] caused the improvement" |
| Cross-sectional study | "[exposure] was associated with [outcome]" | "[exposure] leads to [outcome]" |
| Cohort study | "[exposure] was associated with a higher risk of [outcome] after adjustment" | "[exposure] causes [outcome]" |
| Randomised controlled trial | "[intervention] reduced [outcome] in this population" | "[intervention] works for all patients" |
Interpreting statistics in the Discussion without overstating them
Discuss the effect size and its confidence interval, not the p-value alone. The interval shows the range of effects your data are compatible with. That range is what a reader needs to judge clinical relevance.
- A non-significant result is not evidence of no effect. Say whether the interval is wide enough to include an effect that would matter.
- A p-value is not the probability that your hypothesis is true. Do not write that a result proves anything.
- Do not call a result that missed your threshold "almost significant". Report it as it is and discuss the interval.
- Do not add an observed power calculation to explain a null result. The confidence interval already carries that information.
Our guide to p-values, confidence intervals and effect sizes covers how to report them in the Results.
Common mistakes
- Overreaching. Claiming more than your data support is the fastest way to lose a reviewer. Match the strength of your claim to the strength of your evidence.
- Ignoring contrary evidence. If prior studies disagree with you, address them. Pretending they do not exist reads as either ignorance or spin.
- The mini-review. The Discussion is not a literature survey. Cite what is relevant to your finding, not everything in the field.
- New results. An analysis that appears for the first time in the Discussion. Describe it in the Methods and report it in the Results before you interpret it.
- Unlabelled speculation. A proposed mechanism written as if it were established. Mark it as a hypothesis and say what evidence would test it.
What reporting guidelines ask the Discussion to cover
Reporting guidelines set the minimum content for each section of a paper. Find the one for your design on the EQUATOR Network, then check your Discussion against it.
- STROBE (observational studies): key results, limitations, interpretation and generalisability.
- CONSORT (randomised trials): limitations, generalisability and an interpretation consistent with the results.
- CARE (case reports): strengths and limitations of the approach, the relevant literature, the rationale for your conclusions and the main take-away lessons.
A self-check before you submit the Discussion
- Does the first paragraph state the main finding and whether it answered the research question?
- Does every paragraph add interpretation rather than repeat a number?
- Have you discussed the studies that disagree with you, if any exist?
- Is every proposed mechanism marked as a hypothesis?
- Does every claim stay within what your design can support?
- Is each implication specific enough for a reader to act on or test?
End on a clear conclusion
Finish with a short paragraph that answers the question you started with — one or two sentences on what you found and why it matters, without hedging or new information. A reader who read only your first finding and your last paragraph should come away knowing exactly what your paper contributes.
To match the final claim to your study design, see how to write the conclusion of a research paper, which includes a wording ladder.
Updated 17 September 2026: We expanded this guide with a worked example, a design-by-wording table, statistics rules and a self-check list.