A practical structure for the Discussion section of a medical manuscript — from answering your research question to writing honest limitations.

Structuring a Discussion Section That Answers Your Own Question

Ask a peer reviewer what goes wrong most often in a Discussion section, and two answers come up constantly: it quietly turns into a second Results section, restating numbers instead of interpreting them, or it drifts into an unearned sales pitch for the study's importance. Neither serves the reader. The Discussion has one job the rest of the manuscript hasn't done yet — answering the question the Introduction posed, in light of what you actually found and what everyone else has already published.

Start With Your Research Question, Not a Restatement of Results

The single most useful edit to almost any Discussion section is moving the direct answer to the research question into the opening paragraph. Not "our study examined whether X was associated with Y and found several notable results" — a plain, direct answer: "In this cohort, X was associated with a [magnitude] increase in Y." Readers, and reviewers, should be able to read the first paragraph of your Discussion and know your answer without reading the Results section again.

This sounds obvious, but it's the step most manuscripts skip, usually because the Discussion is drafted by summarizing what was found rather than by returning to the question posed at the end of the Introduction and answering it directly.

The Funnel-Shaped Discussion: A Structure That Works

A useful way to think about manuscript structure, described by Cals and Kotz in their Journal of Clinical Epidemiology series on scientific writing, is that the Introduction is shaped like a funnel — starting broad, with context and background, and narrowing to the specific research question — while the Discussion is an inverted funnel, starting narrow, with the direct answer to that question, and gradually widening out to comparisons with other literature and broader implications. Read together, the Introduction and Discussion form an hourglass.

In practice, that gives a reliable sequence:

1. Principal findings — a direct, specific answer to the research question, in one or two sentences.
2. Interpretation — what the finding means, and, where relevant, a plausible mechanism, stated with appropriate caution about the strength of the design.
3. Comparison with existing literature — how this result agrees or disagrees with prior work, and why.
4. Strengths — what this study did well methodologically, stated briefly and specifically, not generically.
5. Limitations — specific, relevant weaknesses and their likely direction of effect on the result.
6. Implications — what should change, or what should be studied next, stated concretely.

Not every paper needs all six as separate subheadings — shorter papers often fold several into flowing paragraphs — but the sequence itself rarely needs to change.

Interpreting Findings Without Overreaching

The interpretation paragraph is where discussions most often overreach, and it's worth being deliberate about the language your study design actually supports. An observational study can describe an association; only a well-controlled randomized trial can reasonably support language implying causation. "X was associated with a reduction in Y" is defensible from a cohort study; "X reduced Y" implies a causal claim the design may not support.

The same discipline applies to statistical results discussed elsewhere on this blog: a wide confidence interval or a heterogeneous meta-analysis pooled estimate should be interpreted with the same caution in the Discussion that it was reported with in the Results — restating a result as more precise or more universal in the Discussion than the numbers actually showed is a common and avoidable inconsistency that reviewers notice.

It's equally important not to extrapolate beyond the population actually studied. A finding in adults aged 40–65 with moderate disease severity, at a single tertiary center, should be discussed as applying to that population — with generalizability addressed explicitly, not assumed.

Situating Your Study Within Existing Literature

Comparing your findings with prior work is not a literature-review recap; it's an argument. For each relevant prior study, the useful question is: does it agree or disagree with what you found, and why might that be? Differences in population, dose, follow-up duration, or outcome definition are usually more informative to explain than a vague "further research is needed to clarify this discrepancy."

Citing a study you disagree with deserves the same specificity as citing one you agree with. "In contrast to Smith et al., who reported no association, our larger sample and longer follow-up period may have provided greater power to detect a modest effect" gives the reader something concrete to evaluate; "these results differ from previous studies" does not.

Writing Limitations Honestly, Without Self-Sabotage

A generic limitations paragraph — "this study has several limitations, including its retrospective design and modest sample size" — with no discussion of what those limitations actually mean for the result, tells a reviewer the authors haven't thought carefully about their own study's weaknesses. A useful limitation statement names the specific issue, its likely direction of effect, and (where possible) what was done to mitigate it: "Outcome ascertainment relied on chart-documented diagnoses, which may have led to underestimation of true incidence if milder cases went unrecorded; sensitivity analysis restricting to laboratory-confirmed cases showed a consistent direction of effect."

This isn't about undermining your own work — reviewers and readers trust a paper more, not less, when its authors demonstrate they understand exactly where its weaknesses lie and how those weaknesses would bias the result, rather than listing limitations as a formality to satisfy a checklist.

Ending With Implications, Not Overclaiming

The closing paragraphs should state concretely what the finding means for practice, policy, or the next study — and resist the temptation to claim more than the data support. "These findings suggest that clinicians should consider X" is a stronger, more useful statement than "further studies are needed," provided the data actually support that suggestion; where they don't yet, saying so plainly, and specifying exactly what evidence would resolve the uncertainty, is more valuable to the field than a vague call for more research.

A well-built Discussion doesn't need to inflate its findings to be worth publishing. A modest, precisely stated, honestly limited answer to a well-framed research question is exactly what a rigorous journal is looking for — and it reads, to an experienced reviewer, as evidence that the authors understand their own work.

Frequently Asked Questions

What's the real difference between the Results and Discussion sections?

Results present findings objectively without interpretation, typically in the past tense with minimal editorializing. Discussion interprets what those findings mean, compares them with existing literature, and states their implications and limitations. A Discussion that mostly restates numbers is functioning as a second Results section.

How long should a Discussion section be?

There's no universal word count, but most well-structured Discussions run considerably shorter than the Introduction plus Methods combined, and follow the same "principal findings, interpretation, comparison, strengths, limitations, implications" sequence regardless of length. Padding with restated results or generic statements tends to weaken rather than strengthen a Discussion.

Do I need a separate Limitations subheading?

Journal conventions vary — some require a dedicated Limitations subsection, others expect limitations woven into the discussion narrative. Either way, each limitation should be specific to your study and paired with its likely direction of effect on the result, not a generic list.

How many references should I cite in the Discussion?

Enough to substantively compare your findings with the relevant prior literature — typically fewer, more purposefully chosen citations than in the Introduction, each tied to a specific point of agreement or disagreement rather than a broad background citation.

Should I mention my study's strengths?

Yes, briefly and specifically — a genuine methodological strength (a validated outcome measure, a low loss-to-follow-up rate, a pre-registered protocol) helps a reader calibrate how much weight to give the findings. Generic claims of strength ("this is the first study to...") are less persuasive than specific ones.

Can I speculate about mechanisms in the Discussion?

Yes, but the language should clearly signal that it is speculation grounded in biological or clinical plausibility, not a claim supported directly by your data — phrases like "one possible explanation is..." keep this distinction clear for the reader.

References

1. Cals JWL, Kotz D. Effective writing and publishing scientific papers, part VI: discussion. Journal of Clinical Epidemiology. 2013;66(10):1064.
2. International Committee of Medical Journal Editors (ICMJE). Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals.
3. von Elm E, Altman DG, Egger M, et al.; STROBE Initiative. The STROBE statement: guidelines for reporting observational studies. The Lancet. 2007.
4. Schulz KF, Altman DG, Moher D; CONSORT Group. CONSORT 2010 Statement. Annals of Internal Medicine. 2010.