How to Write a Discussion Section That Editors Don't Skip

The discussion section is where papers go wrong most often. Not because the science is bad , because the writing breaks down in specific, predictable ways that signal to reviewers the authors haven't thought carefully enough about what their findings actually mean.

Here's the structure that works and the mistakes that don't.

The Structure That Works

A strong discussion section has five parts, in this order:

1. Restate the main finding (1 paragraph)

Don't start with "In this study, we showed that..." Don't repeat the abstract. Do restate the primary finding in one or two sentences that connect directly to the question you posed in the introduction. This paragraph answers: what did we learn?

Example of what doesn't work: "In this randomized controlled trial, we demonstrated that treatment A significantly reduced the primary composite endpoint of cardiovascular death, myocardial infarction, or stroke compared to placebo (HR 0.78, 95% CI 0.68-0.89, p<0.001), confirming our primary hypothesis."

That's just the abstract, repeated.

What works: "Treatment A reduced major cardiovascular events in patients with established coronary artery disease. The 22% relative risk reduction was consistent across all prespecified subgroups, suggesting a broad rather than population-specific benefit."

The difference: the second version adds interpretation (consistency across subgroups = broad applicability) that wasn't in the results section.

2. Situate your findings in the existing literature (2-3 paragraphs)

Compare your findings to prior work. For each comparison, be specific about what's consistent and what's discordant , and why.

Don't write: "Several prior studies have investigated this relationship with mixed results."

Do write: "These findings are consistent with Smith et al. (2022) who showed a similar effect in a European cohort (HR 0.81), but diverge from Johnson et al. (2023), where no significant effect was observed. The difference likely reflects Johnson's shorter follow-up (18 months vs. our 36 months), which may have been insufficient to detect a treatment effect in their lower-risk population."

The second version explains the discordance rather than noting it. That's what reviewers want to see.

3. Propose a mechanism or theoretical explanation (1-2 paragraphs)

If you have data that suggests a mechanism, present it here. If you don't, briefly describe the most plausible mechanistic hypothesis based on existing knowledge , and be clear it's a hypothesis.

Don't overreach. If your paper is a clinical trial with no mechanistic data, don't spend three paragraphs speculating about molecular pathways. One sentence naming the plausible mechanism and citing the relevant mechanistic literature is sufficient.

4. Acknowledge limitations (1 paragraph)

State the main limitations clearly and without hedging. Reviewers will find them , it's better to own them and explain their relevance.

The limitations paragraph should:

• Name each limitation specifically (not "our study has some limitations")
• Explain why the limitation exists
• State what it means for interpretation of the results
• Note what future work would address it

What to avoid: putting limitations in a final paragraph at the end of the discussion and hoping reviewers skim over them. That strategy fails consistently.

5. Implications and future directions (1 paragraph)

What should change as a result of your findings? Be specific. "Further research is needed" is universally understood to mean the authors couldn't think of a concrete implication.

Better: "These findings support a change in first-line therapy for patients with [condition X] who meet the eligibility criteria used in this trial. Ongoing trials of [treatment combination] may address whether the benefit extends to patients with [different characteristic]."

4 Mistakes That Kill Otherwise Strong Papers

Mistake 1: Restating the results rather than interpreting them

The most common discussion failure. Every sentence in the discussion should add meaning beyond what the results tables already show. If you're repeating numbers and p-values without adding interpretation, the discussion isn't doing its job.

Mistake 2: Over-claiming based on the data

Strong findings invite overclaiming. If your trial enrolled a specific population, the conclusions apply to that population , not to "all patients with [disease]." Reviewers and editors flag overclaiming immediately, and it weakens the paper's credibility.

Mistake 3: The literature dump

A discussion section that summarizes 15 prior papers without connecting them to your findings isn't a discussion , it's an additional introduction. Cite prior work specifically, to make a specific point about how your findings compare.

Mistake 4: Weak limitations acknowledgment

Glossing over limitations in one sentence ("Our study has some limitations") and moving on is a signal to reviewers that you haven't engaged seriously with your data's constraints. Reviewers will elaborate on those limitations in their reports. Better to address them thoroughly yourself.

Length and Depth by Journal Tier

Different journal tiers have different norms for discussion length and depth:

Nature/Science/Cell: Short, dense discussions (600-800 words). Every sentence does work. No padding, no literature reviews , only what directly situates the finding.

High-IF field journals (NEJM, Lancet, JCO): 800-1,200 words. Structured with clear subsections sometimes. Clinical implications discussed explicitly.

Mid-tier specialty journals: 1,000-1,500 words. More tolerance for literature review context, but same structural expectations.

PLOS ONE, Scientific Reports: No specific norm, but editors flag discussions that run over 2,000 words as padded.

When in doubt, cut. A tighter discussion reads as more confident and more competent than a long one.

Signaling Scientific Honesty

The best discussions share a quality that's hard to fake: they're honest about what the data shows and doesn't show. Reviewers who see a discussion that acknowledges genuine limitations, avoids overclaiming, and situates negative or null results honestly tend to be more favorable even when the findings themselves are modest.

Papers that try to overclaim, minimize limitations, or ignore conflicting prior work tend to receive harsher peer review , and sometimes desk rejection on resubmission after revision, because the reviewers lose confidence in the authors' ability to evaluate their own work.

Common Mistakes to Avoid

Most authors lose time in this topic for one reason: they optimize the wrong variable first. They spend hours polishing language while leaving structural issues unresolved. Editors and reviewers evaluate structure before style.

In practice, the recurring mistakes are predictable:

1. Using generic claims instead of specifics. Replace vague statements with concrete numbers, study details, and explicit scope boundaries.
2. Ignoring fit and audience. A strong manuscript sent to the wrong journal or framed for the wrong reader still fails quickly.
3. Treating revision as proofreading. Revision is where argument quality, methodological clarity, and limitation handling should improve meaningfully.
4. Skipping process checks. Formatting, references, checklist compliance, and data statements look administrative, but they're part of editorial quality control.

A useful rule is to run one final pre-submission pass that checks only these operational risks: scope fit, claim strength, methods clarity, and policy compliance. That pass catches most avoidable rejection reasons before they become reviewer comments.

If you're deciding between two valid options, pick the one that improves clarity for an external reader who has no context besides your paper. Clearer framing beats denser writing almost every time.

What "Future Directions" Actually Means

Most discussion sections end with a future directions paragraph that is too vague to be useful. "Further research is needed to investigate the mechanisms underlying these findings" tells the reader nothing that wasn't already obvious.

Future directions that work:

• Name a specific unanswered question that your data raises
• Identify the specific experiment or study design that would answer it
• Optionally, note whether that work is ongoing or planned

Example of what doesn't work: "Future studies should investigate the role of this pathway in other disease contexts."

Example of what works: "Whether calcium-dependent FOXO3a activation operates in primary tumor tissue, rather than cell lines, remains an open question. A prospective analysis of calcium channel expression in resected tumor specimens from doxorubicin-treated patients would test whether the in vitro findings translate to the clinical setting."

The second version tells the reader exactly what the next study should look like. That specificity is what separates a discussion that contributes to the field from one that pads word count.

Revision: How the Discussion Gets Better

The discussion is often the section that improves most substantially through peer review. Reviewers who read your paper carefully will identify:

• Claims that go beyond what the data supports
• Discordances with prior literature you didn't address
• Mechanisms you implied but didn't fully justify
• Limitations you minimized that are more significant than acknowledged

This feedback is genuinely useful. The authors who benefit most from peer review are those who approach reviewer comments on the discussion as an opportunity to sharpen their interpretation, not as objections to overcome.

The practical implication: write your first-draft discussion knowing it will probably change substantially. Don't invest enormous effort in perfecting language in the discussion before you've received peer review. The structure and argument quality matter more than the prose in the first draft.

The Bottom Line

A strong discussion section restates the finding in context, compares it specifically to prior work, proposes a mechanism, acknowledges limitations honestly, and states concrete implications. It doesn't restate the results, overclaim from the data, or dump the literature review. Keep it tight , most strong discussions run 800-1,200 words, and cutting is almost always an improvement.

See also

• How to respond to reviewer comments
• How to write a research paper abstract
• Pre-submission checklist

• PLOS ONE journal profile

Sources

• EQUATOR Network reporting guidelines (equator-network.org)
• ICMJE recommendations for manuscript preparation (icmje.org)
• Pre-Submission Checklist