The discussion section explains what your results mean, and it moves from narrow to broad — the reverse of your introduction. Write it in six moves: restate your key finding in one sentence, interpret it, compare it with previous studies, acknowledge your limitations, state the implications, and point to future research. Do not repeat every result, and do not introduce new data.
This is the section where papers are won and lost. Reviewers rarely reject a manuscript because the methods were dull; they reject it because the authors could not explain why their findings matter, overstated what the data showed, or ignored a limitation the reviewer spotted immediately. This guide covers the structure that works, worked examples of weak versus strong writing, and how to handle limitations without undermining your own paper.
What Is the Discussion Section?
The discussion section is where you interpret your findings and place them in the context of existing knowledge. Your Results section reports what happened. Your Discussion argues what it means, why it matters, and how confident readers should be.
It is the most argumentative part of a scientific paper — and the part where your judgment as a researcher is most visible. Two authors can report identical results and produce discussions of completely different quality.
Discussion vs. Results vs. Conclusion
Confusing these three is the most common structural problem in submitted manuscripts.
| Section | Question it answers | Contains data? | Contains interpretation? |
|---|---|---|---|
| Results | What did I find? | Yes — full reporting | No |
| Discussion | What does it mean? | Only brief restatement | Yes — this is its purpose |
| Conclusion | So what, in one breath? | No | Yes, but compressed |
A useful test: if a sentence contains a p-value and nothing else, it belongs in Results. If it contains the word because, suggests, or consistent with, it belongs in the Discussion.
The Shape of a Good Discussion: An Inverted Funnel
Your introduction narrows from broad context down to your specific research question. Your discussion does the opposite — it opens with your specific finding and widens outward to what it means for the field.
Introduction: broad context → gap → your aim
Discussion: your finding → interpretation → broad implicationsPapers that ignore this shape tend to either open with a wall of background (which the reader already read) or end abruptly without saying why anyone should care.
The Six Moves of a Discussion Section
| Move | Purpose | Approx. share |
|---|---|---|
| 1. Restate the key finding | Orient the reader | 5–10% |
| 2. Interpret it | Explain the mechanism or meaning | 20–25% |
| 3. Compare with prior work | Situate it in the literature | 25–30% |
| 4. Acknowledge limitations | Establish credibility | 15–20% |
| 5. State implications | Answer “so what?” | 10–15% |
| 6. Suggest future research | Point forward | 5–10% |
These are moves, not mandatory subheadings. Some journals want explicit headings; most want the logic without them.
Step-by-Step: Writing Each Move
Move 1 — Restate your key finding (one sentence, no data dump)
Open with your principal result in plain language. One finding, not all of them. Resist repeating numbers you already reported.
Weak: “As shown in Table 3, adherence was 78.2% in the intervention group and 61.4% in the control group (p < 0.001), and systolic blood pressure decreased by 8.3 mmHg versus 2.1 mmHg.” Strong: “Pharmacist-led follow-up produced a substantial improvement in medication adherence, accompanied by a clinically meaningful reduction in systolic blood pressure.”
The weak version is a Results paragraph in the wrong place. The strong version tells the reader what to take away.
Move 2 — Interpret the finding
Explain why you observed what you observed. Propose a mechanism, but signal your confidence level honestly.
“The effect likely reflects more frequent contact and earlier identification of side effects, both of which are established drivers of discontinuation. Because the trial was not designed to isolate these mechanisms, this interpretation remains provisional.”
That final sentence costs you nothing and buys you significant credibility with reviewers.
Move 3 — Compare with previous studies
This is where a discussion earns its length. Do not list studies; synthesize them. Say where you agree, where you disagree, and — critically — offer an explanation for disagreement.
Weak (citation dumping): “Similar results were reported by several previous studies.” Strong: “Our adherence effect exceeds the 8–10 percentage point improvements reported in earlier pharmacist-intervention trials. The difference may reflect our 12-month follow-up period, which was roughly twice that of most prior studies, allowing the intervention more time to change routine behaviour.”
Disagreement handled well is a strength, not a weakness. Reviewers are far more suspicious of authors who claim every prior study agrees with them.
Move 4 — Acknowledge limitations (without gutting your paper)
Every study has limitations. Naming them yourself is not a confession — it is evidence that you understand your own design. Reviewers will find them regardless; far better that you found them first.
The technique that matters: state the limitation, then state its likely direction or magnitude of effect.
Weak: “This study has several limitations, including sample size and generalizability.” Strong: “Three limitations should be noted. First, pill counts may overestimate adherence, though this bias would apply equally to both arms and is unlikely to explain the between-group difference. Second, all six clinics were urban, so findings may not generalize to rural settings with different pharmacy access. Third, blinding of participants was not feasible, raising the possibility of a Hawthorne effect.”
Notice what the strong version does: it does not merely admit weakness, it reasons about whether each weakness could plausibly explain the result. That is the difference between an author who lists limitations and one who understands them.
Avoid the opposite failure too — a limitations paragraph so sweeping that it invites the reviewer to reject the paper on your behalf. Be honest and proportionate.
Move 5 — State the implications
Answer “so what?” for the people who might act on your work: clinicians, policymakers, engineers, or other researchers. Match the strength of the claim to the strength of the design.
“If replicated in rural and mixed-payer settings, pharmacist-led follow-up could offer a low-cost addition to hypertension management in primary care.”
The conditional framing (“if replicated”) is doing real work here. A single trial supports a proposal, not a mandate.
Move 6 — Point to future research
Be specific. “Further research is needed” is filler; name the study you would run next.
“A multi-site trial including rural clinics, with electronic adherence monitoring rather than pill counts, would address the two principal limitations of this study.”
Calibrating Your Claims
Overclaiming is the fastest way to lose a reviewer’s trust. Match your verb to your evidence.
| Study design | Appropriate language | Avoid |
|---|---|---|
| Observational / correlational | “was associated with”, “correlated with” | “caused”, “led to” |
| Single RCT | “improved”, “reduced” (in this setting) | “proves”, “establishes” |
| Pilot / small sample | “suggests”, “provides preliminary evidence” | “demonstrates conclusively” |
| Meta-analysis / replicated | “demonstrates”, “provides strong evidence” | “definitively settles” |
A reviewer who sees “proves” attached to a single trial will read the rest of your paper more skeptically. Precise hedging is not weak writing — it is accurate writing.
How Long Should the Discussion Be?
Typically 800–1,500 words, or roughly 25–30% of the manuscript body. It is usually the longest section after Methods.
| Signal | What it usually means |
|---|---|
| Discussion shorter than Results | Under-interpreted — you reported but did not explain |
| Discussion longer than Methods + Results combined | Over-speculating, or repeating results |
| Half the discussion is background | Introduction material in the wrong section |
Discussion Section Checklist
- Opens with the key finding, not a background recap
- Interprets rather than repeats results
- Compares with prior work and explains disagreements
- Limitations stated with their likely direction of effect
- No new data introduced
- Claims proportionate to the study design
- Implications are concrete and appropriately conditional
- Future research is specific, not generic
- Every citation earns its place
- Does not simply restate the abstract
Common Mistakes to Avoid
- Repeating the Results section. If a paragraph is mostly numbers, it is in the wrong place.
- Citation dumping. Listing studies without synthesizing them adds length, not substance.
- Hiding limitations. Reviewers find them anyway; concealment reads as either naivety or evasion.
- Introducing new data. Anything worth discussing should have been reported in Results.
- Overclaiming causation from correlation. The single most common reviewer complaint in observational research.
- Ending without implications. A discussion that stops after “future research is needed” wastes its most persuasive moment.
- Contradicting your own abstract. Your abstract’s conclusion and your discussion must make the same claim at the same strength.
If reviewers have already returned comments on your discussion, our guide to responding to reviewer comments covers how to revise without conceding more than necessary.
Frequently Asked Questions
What is the difference between the discussion and the conclusion? The discussion interprets your findings in depth — comparing them with prior work, acknowledging limitations, and drawing out implications. The conclusion compresses that into two to four sentences answering “so what?” without introducing new reasoning. Some journals merge them into a single section.
How long should a discussion section be? Usually 800–1,500 words, or about 25–30% of the manuscript body. If your discussion is shorter than your results section, you have probably under-interpreted your findings.
Should I include limitations in the discussion? Yes, in almost all cases. State each limitation along with its likely effect on your results. Authors who identify their own limitations are consistently viewed as more credible than those who leave reviewers to find them.
Can I include new results in the discussion? No. Every finding you discuss must already appear in your Results section. Introducing new data at this stage suggests the analysis was incomplete or selectively reported.
How do I discuss results that contradict previous studies? Address the contradiction directly and propose a plausible explanation — differences in population, design, follow-up duration, or measurement. Reviewers respect a well-reasoned disagreement far more than a paper that ignores conflicting evidence.
What if my results were not statistically significant? Report and discuss them honestly. Explain whether the study was adequately powered, and distinguish between evidence of no effect and simply no evidence of an effect. Null findings are publishable and scientifically valuable when discussed rigorously.
Should the discussion have subheadings? It depends on the journal. Longer discussions in medical and clinical journals often use subheadings such as “Comparison with other studies” and “Limitations.” Check your target journal’s recent articles and author guidelines.
Conclusion
A strong discussion does four things reliably: it states what was found, explains what it means, admits what remains uncertain, and says why any of it matters. Move from narrow to broad, synthesize the literature rather than listing it, treat limitations as evidence of rigour rather than confession, and keep every claim proportionate to your design. Get that right, and reviewers spend their time engaging with your argument instead of correcting it.
Is your discussion making the strongest case for your research?
Our subject-matter editors work on argument structure, claim calibration, and clarity — the elements reviewers scrutinize most. Start with a free 500-word sample edit, no obligation.
Request your free sample edit → or explore our scientific editing services and substantive editing support.
