SciSpace 指南：如何在研究论文中撰写讨论部分（10 步附示例）

The discussion is where you explain what your results mean. It is often the hardest section to write, because it asks you to interpret, compare and admit limits rather than just report.

In short: a good discussion section opens by answering your research question in one or two sentences, then interprets the main findings, compares them with previous studies (including ones that disagree), explains unexpected results, states the implications, acknowledges limitations and their consequences, and ends with specific directions for future research. It should not repeat the results or claim more than your design supports. The manuscript writer agent can help you structure a draft from your own results and references.

What the discussion section is for

The results section says what you found. The discussion says why it matters. Readers and editors look for three things in it:

• Did the study answer its question?
• How do the findings fit with, extend or challenge what was already known?
• How far can the conclusions be trusted and applied?

How to write a discussion section in 10 steps

1. Start your discussion notes early. Keep a running file of papers that support or contradict your results while you analyse the data, so the discussion is not built from scratch at the end.
2. Answer the research question first. Open with one or two sentences that state the main finding in plain words, without statistics.
3. Interpret, do not repeat. For each key finding, explain what it means. Refer back to a result in a few words rather than restating the numbers.
4. Compare with previous studies. Say where your findings agree with earlier work and, just as important, where they differ.
5. Explain differences and surprises. If your results contradict other studies or your hypothesis, suggest plausible reasons: population, setting, measures, design or timing.
6. Stay on your objectives. Discuss the findings that relate to your research questions. Interesting side results can be mentioned briefly or left out.
7. State the implications. Say what the findings mean for theory, practice or policy, and for whom.
8. Acknowledge limitations and their consequences. Do not just list limitations. Say how each one might affect the interpretation.
9. Suggest specific future research. Replace "more research is needed" with a concrete next study: which population, design or question.
10. End with a clear take-home message. Close with the one idea you want the reader to remember, either as a short final paragraph or a separate conclusion, depending on the journal.

A simple structure to follow

How long should it be? It depends on the journal and field. A common rule of thumb is a length similar to, or a little longer than, the introduction. Check the author guidelines and a few recent papers in your target journal.

Example sentences

• Opening: "This study found that students who received weekly feedback improved their writing scores more than those who received feedback only at the end of term."
• Agreement: "This is consistent with previous work showing that frequent feedback improves revision quality (Author, Year)."
• Disagreement: "In contrast to Author (Year), we found no effect among first-year students, possibly because our sample had less prior writing experience."
• Limitation with consequence: "Because data were self-reported, the effect may be overestimated if students over-reported their revision time."
• Specific future research: "A randomised study in schools with different feedback cultures would show whether the effect holds outside university settings."

These are illustrative sentences; adapt them to your own findings and citations.

Common mistakes reviewers flag

• Repeating the results with the same numbers instead of interpreting them.
• Introducing new results that were not reported in the results section.
• Overclaiming. Writing "causes" when your design shows an association, or generalising beyond your sample. Journal editor Prof. Roger Watson covers how to write a discussion that does not overclaim in his desk-rejection webinar.
• Ignoring studies that disagree with your findings.
• Generic limitations and future research that could apply to any study.

Use AI to find comparison studies, then write it yourself

AI is most useful in the discussion for finding studies that support or contradict each finding and for checking that a cited paper really says what you claim. The interpretation should be yours. This SciSpace webinar walks through a 10-step discussion workflow:

Key moments: 4:24 learning from example discussions · 18:07 how long a discussion should be · 20:54 discussing results without repeating them · 23:00 finding studies that support or contradict your findings · 35:27 ways to order your points · 37:02 a discussion checklist before submitting.

Discussion section checklist

• The first paragraph answers the research question.
• Every key finding is interpreted, not repeated.
• Each finding is compared with previous studies, including contradictory ones.
• No new results appear.
• Claims match the study design (association vs causation, sample vs population).
• Each limitation says how it affects the conclusions.
• Future research suggestions are specific.
• Every citation has been checked against the source.

Watch: from research question to peer review

The discussion depends on every section before it. This SciSpace webinar follows a paper from the research question through drafting to peer review.

Related guides

Frequently asked questions

注意转载授权 · 阅读原文