How to Write the Discussion Section — The Hardest Part of Any Research Paper

Most researchers find the Discussion harder to write than any other section of a paper. The Introduction has a known structure. The Methods is a record of what you did. The Results is a report of what happened. The Discussion is different. It is the only section that asks you to think independently, to interpret rather than describe, and to place your findings in honest conversation with everything that has been published before you.

It is also the section most commonly written badly. After editing dozens of manuscripts from South Asian researchers, the same problems appear in almost every Discussion: too much repetition of the Results, not enough engagement with the literature, vague conclusions, and a complete absence of explanation for unexpected findings. This post addresses all of them.

What a Complete Discussion Must Cover

1 Opening Restate the main finding in interpretive terms — not copied from Results

2 Interpretation What do the results mean? What mechanism explains them?

3 Literature comparison Consistent with or different from published work — and why

4 Unexpected findings Acknowledge them, propose an explanation, identify what further work would resolve them

5 Limitations Honest, specific, and proportionate — not self-destructive

6 Implications What the findings mean for the field or for practice

7 Closing The contribution, the unresolved questions, the specific next study needed

What the Discussion Is Supposed to Do

The Discussion has four things to accomplish, in roughly this order.

First, it tells the reader what your results mean, not just what they are. The Results section reported what happened. The Discussion explains why it matters and how to interpret it.

Second, it places your findings in the context of what other researchers have found. Are your results consistent with the literature? Different? If different, why?

Third, it acknowledges what your study cannot claim: its limitations, the boundaries of its applicability, the questions it raises that it cannot answer.

Fourth, it closes with the implications of your work — what the field or practice should take from it.

A Discussion that does all four is complete. Most submitted manuscripts do only the first, partially, and skip the rest.

How to Open the Discussion

The opening sentence of the Discussion is one of the most important sentences in the paper. It sets the analytical register for everything that follows. It should restate your main finding in interpretive terms, not copy it from the Results.

✗ Wrong — copying the Results

"The results showed that T4 had the highest soil organic carbon at 1.67%, which was significantly higher than all other treatments (p < 0.05)." This sentence belongs in the Results section. It has already been said. The Discussion needs to move beyond reporting.

✓ Right — restating with interpretation

"The combined compost and biochar amendment produced the largest improvement in soil organic carbon among all treatments tested, suggesting that the two materials interact synergistically in enhancing organic matter retention in paddy soil." The number is not repeated (the reader has just seen the table). The sentence states the finding in terms of what it means: a synergistic interaction, a mechanism, an interpretation.

Engaging with the Literature

Every significant finding in your Results needs to be compared against published work. Not cited — compared. There is a difference.

✗ Weak — a citation, not a comparison

"Several studies have also reported that biochar improves soil organic carbon (Ali et al., 2022; Kalu et al., 2022)." This sentence tells the reader nothing about how your results relate to those studies. It is name-dropping, not analysis.

✓ Stronger — a real comparison with explanation

"The 83% increase in soil organic carbon observed under combined compost and biochar amendment is consistent with the 70 to 90% range reported by Ali et al. (2022) in flooded paddy systems, and higher than the 55% reported by Kalu et al. (2022) under dryland wheat, a difference likely attributable to the slower decomposition rate of biochar in anaerobic conditions that preserves organic matter in waterlogged soils." The comparison is specific (83% vs 70-90% vs 55%), directional (higher than), and mechanistically explained. That is what comparing with literature looks like.

For results that differ from established findings, the Discussion is where you explain the discrepancy rather than ignore it. A result that contradicts published work is not a weakness. It is an opportunity to advance the field's understanding, if you engage with it honestly. State the difference, propose an explanation, and acknowledge what would be needed to resolve the question.

Explaining the Mechanism

One of the most important things the Discussion can do is explain why you got the results you did. Not just that treatment T4 outperformed T3, but the underlying mechanism: why, chemically or physiologically, this happened.

For a soil amendment experiment, this means drawing on soil chemistry: cation exchange capacity, microbial biomass carbon dynamics, nitrogen mineralisation rates, or pH buffering effects. For a fertilizer trial, it means drawing on plant physiology: nitrogen uptake patterns, root architecture, source-to-sink relationships during grain fill.

You do not need to have measured these mechanisms directly to discuss them. A well-reasoned mechanistic explanation, clearly framed as interpretation rather than demonstrated fact, strengthens the Discussion.

The key phrase is "may reflect"

"The higher nitrogen use efficiency under slow-release treatment may reflect reduced initial-burst volatilisation, consistent with the lower ammonia flux reported in comparable incubation studies." This is a legitimate mechanistic interpretation even when your experiment did not directly measure ammonia flux. Proposing a mechanism is different from claiming you proved one.

— ❧ —

Handling Unexpected Results

Many researchers omit unexpected results from the Discussion because they are afraid reviewers will see them as weaknesses. This instinct is wrong. Experienced reviewers notice when a paper's Discussion does not address a result that appears in the tables.

If your data produced something you did not predict, say so. State the result, propose an explanation, and note what further work would be needed to confirm or reject that explanation. This approach demonstrates that you understand your own data well enough to know when it surprises you — which is evidence of scientific rigour, not weakness.

✓ How to handle a surprising result

"Contrary to the hypothesis, nitrogen harvest index did not differ significantly between the compost and biochar single-amendment treatments (p = 0.23). This unexpected result may reflect compensatory nitrogen redistribution from straw to grain under the more gradual nitrogen supply associated with biochar, a mechanism that has been proposed but not conclusively demonstrated in previous work (Gamage et al., 2021). Further studies with direct measurement of post-harvest nitrogen partitioning would be needed to test this." Three moves: acknowledge the surprise, propose a mechanism with a citation, identify the next study. Reviewers respect this approach far more than silence.

Writing Limitations That Are Honest Without Being Self-Destructive

Every study has limitations. The question is how to state them without implying the entire paper is invalid. A limitation that affects one part of your conclusions is not a reason to abandon the rest.

✗ Vague and unhelpful

"This study has limitations and further research is needed." This says nothing. Every study has limitations. What are they? Further research into what? Reviewers will send this straight back.

✓ Specific and functional

"The pot experiment conditions, while allowing controlled nitrogen management, may not fully replicate the spatial variability of nitrogen uptake in field-grown rice under intermittent irrigation. The agronomic efficiency values reported here should therefore be treated as indicative rather than directly scalable to production-scale recommendations without field validation." States exactly what the limitation is (pot vs field), why it matters (spatial variability), and what it means for interpretation (indicative, not prescriptive). The work is not dismissed — its scope is defined.

Implications

After interpreting the findings and acknowledging the limitations, the Discussion should state what the research contributes. This is different from the Conclusion, which states your direct answers. The implications section explains why those answers matter.

For applied agricultural research, implications are often practical: "If the synergistic SOC improvement observed under combined compost and biochar amendment can be reproduced under field conditions, it would suggest a feasible local input strategy for improving soil health in Bangladesh's intensively cropped paddy systems, using materials that are currently either discarded or burned." That is an implication: a practical consequence of a finding, conditional on its scope.

The implications paragraph should not overstate. "This study will transform fertilizer policy across South Asia" is not an implication. It is an overclaim. Implications are conditional, specific, and proportionate to what the experiment actually demonstrated.

For more fundamental research, implications may be theoretical: "These results suggest that the relationship between soil carbon-to-nitrogen ratio and microbial biomass carbon in waterlogged systems is more sensitive to carbon source quality than to total carbon input, which has implications for how amendment recommendations are formulated in saturated soil management guidelines." Both forms — practical and theoretical — are valid. What matters is that they are proportionate.

The Closing Paragraph

The Discussion's final paragraph is often where papers become weakest. Researchers frequently resort to: "In conclusion, this study demonstrated that X. Further research is needed to Y." That closing restates what the reader already knows and identifies no next step beyond a vague gesture.

A better structure: acknowledge the contribution, identify the outstanding questions the findings raise, and indicate what kind of research would be most valuable to pursue.

✗ The generic closing to avoid

"In conclusion, this study demonstrated that combined compost and biochar application significantly improved soil organic carbon. Further research is needed to confirm these findings." No contribution statement. No unresolved question. No specific next study. This closing adds nothing to the paper.

✓ A closing paragraph that does work

"These findings contribute to the growing evidence base for biochar-compost co-application as a soil health management strategy in tropical paddy systems. The interaction effects observed at the treatment level raise questions about optimal application ratios that this study was not designed to answer. A dose-response study examining multiple compost-to-biochar ratios across several soil types and seasons would be the most productive next step toward translating these pot-study findings into field-applicable recommendations." Three sentences: the contribution, the unresolved question, the specific next study. Clean and forward-facing.

One Structural Note

The Discussion is not a second Results section. If you find yourself writing "As shown in Table 3, T4 had the highest SOC" in your Discussion, you are repeating, not interpreting.

The Discussion should refer to your results in summary terms ("the combined amendment treatment outperformed all single-amendment options") and focus on what those results mean, why they occurred, how they compare to the literature, and what they imply. The numbers live in the tables. The reader has read them. The Discussion's job is to make those numbers mean something.

If you are working on a manuscript and want structural feedback on the Discussion before submission, the manuscript editing service on this site covers this section specifically, with tracked changes and written notes on where interpretation is needed and where claims are running beyond the data.

Sajjadur Rahman

Manuscript Editor · MSc Researcher · University of Dhaka

NST Fellow and active researcher certified in scientific communication through Stanford Writing in the Sciences and the Wiley Researcher Academy. Edits research papers across the natural and applied sciences for grammar, structure, tone, and journal formatting. Based in Dhaka, Bangladesh.

Manuscript Editing IMRaD Guide Grammar Guide Contact