Read the reviewer reports on papers sent back for major revision and the comments on the Discussion repeat a handful of patterns: “the discussion largely restates the results”, “findings are not situated in the literature”, “implications are too general”, “limitations are listed but not analysed”. Those four comments map onto four jobs the Discussion must do and often does not.
If Results say what was found, the Discussion answers “so what?”. It is where you demonstrate your contribution, and also where it is easiest to go beyond your data.
Six steps of a Discussion
- Answer the research question in the first paragraph, in one or two sentences, without repeating detailed numbers.
- Compare with the literature: which findings agree, which differ, and why they might differ.
- Explain unexpected or non-significant findings, offering plausible interpretations.
- Theoretical contribution: how the findings change, extend or bound existing understanding.
- Practical implications for specific audiences.
- Limitations and future research, linked to each other.
Steps four and five may swap depending on the field. What does not change is that step one comes first: readers need the answer before they hear you discuss it.
Weak and strong sentences
| Job | Weak | Stronger |
|---|---|---|
| Answering the question | The results showed that five factors influence… | Instructor feedback, not the quality of materials, was most closely tied to satisfaction. |
| Comparing | This result is consistent with many previous studies. | This matches studies of fully online courses but differs from blended-learning studies, where materials mattered more — possibly because… |
| Implication | Managers should pay more attention to quality. | Programme leaders might prioritise cutting feedback group sizes below 30 before investing in more video production. |
| Limitation | The study has a limited sample size. | The sample covered three public universities only; private institutions, with different class sizes, may show a different role for feedback. |
Comparing with the literature: explain the differences
Writing “this finding agrees with Smith (2019) and contradicts Lee (2020)” and stopping there is half the job. The other half is explaining why. Common explanations include differences in context (country, level, type of institution), measurement (the same construct name but different instruments), timing (before or after a major change) and sample (full-time or part-time students). Choose explanations that evidence supports, and flag them as conjecture.
Non-significant results are not failures
When a hypothesis is not supported, do not skate past it. Consider three possibilities: the effect is genuinely absent or tiny; the study lacked the power to detect it; or the measure did not capture the construct. Use the confidence interval to talk about plausible effect sizes: “the 95% confidence interval of −0.05 to 0.15 suggests that any effect is small”. That sentence is far more informative than “there was no effect”.
Limitations: specific, directional, addressed
A good limitations passage is not an apology but an analysis. Each limitation should have three parts:
- What the limitation is, concretely: not “small sample” but “only 14 male teachers among 180 participants”.
- Which way it could bias results: “so gender comparisons were underpowered, and the non-significant difference may reflect low power”.
- What you did about it, or what future work should do: “a sensitivity analysis excluding this group gave similar results; future studies should stratify sampling by gender”.
Avoid the empty line “future research should use a larger sample”, which is true of every study. Pick the two or three most important limitations and analyse them properly instead of listing ten one-liners.
Implications with an address
A practical implication should state who should do what differently, based on which finding. “Schools should pay more attention” has no address. “Subject departments could reserve fifteen minutes of each meeting for early-career teachers to present a difficult classroom situation, since colleague support showed the strongest association with intention to stay in this sample” has a clear one. Keep implications within what the design supports: for correlational work, say “may” rather than “must”.
Errors that try reviewers’ patience
- Introducing results that did not appear in the Results section.
- Using causal language for a correlational design.
- Citing a whole new body of literature absent from the Introduction, so the two sections read like different papers.
- Ending with a general statement about the topic’s importance instead of the main message.
A quick outline
Paragraph 1: the main answer. Paragraphs 2–4: one finding each, compared and explained. Paragraph 5: theoretical contribution. Paragraph 6: practical implications by audience. Paragraph 7: two or three well-analysed limitations with next steps. Final paragraph: one take-home message, written so that someone who reads only this paragraph knows what the paper contributes.
Next step: highlight every sentence in your current Discussion that merely repeats numbers from Results. Delete them, replace each with an answer to “so what?”, and check whether each limitation states the direction of possible bias.
Câu hỏi thường gặp
What should be included in the Discussion section?
Typically an answer to the research question, comparison with prior literature, explanation of unexpected findings, theoretical contribution, practical implications, limitations and directions for future research.
Should I repeat my results in the Discussion?
Only summarise the main findings briefly to answer the research question. Avoid repeating detailed numbers already reported in the Results section.
How do I write the limitations of a study?
State each limitation concretely, explain how it could bias the results and describe what you did or what future work should do. Focus on the two or three most important ones.
How do I discuss non-significant results?
Consider whether the effect is small, the study underpowered or the measure weak, and use confidence intervals to describe plausible effect sizes. Do not simply conclude there is no effect.
How do I write practical implications in a research paper?
Say who should do what differently, based on which finding, and keep claims within what your design supports. Avoid generic statements such as “more attention is needed”.
