Folio

How to write a discussion section

Updated 2026-11-17 · 4 min read

On this page
  1. The shape of a good discussion
  2. 1. Open with the answer
  3. 2. Interpret and compare
  4. 3. Deal honestly with unexpected results
  5. 4. Theoretical implications
  6. 5. Practical implications
  7. 6. Limitations
  8. 7. Future research
  9. 8. Conclusion
  10. A short example of one discussion paragraph
  11. Language that helps
  12. Common mistakes
  13. Writing the discussion before results are final
  14. Final check

The discussion is where your study becomes a contribution. The results section says what you found; the discussion says what it means, how it fits with what others have found, why it matters and how far it can be trusted. Many readers skip straight from the abstract to the discussion, so it deserves as much care as the introduction.

The shape of a good discussion

Think of the discussion as the mirror image of the introduction. The introduction moved from the broad problem to your specific question; the discussion moves from your specific answer back out to the field and the world. A dependable sequence is:

  1. Restate the main findings in relation to the research question.
  2. Interpret them and compare them with previous research.
  3. Explain unexpected or non-significant results.
  4. Spell out theoretical implications.
  5. Spell out practical or policy implications.
  6. Acknowledge limitations.
  7. Suggest future research.
  8. Conclude (sometimes a separate section).

1. Open with the answer

Begin with a short paragraph that answers the question you asked in the introduction, in plain words and without statistics: "This study found that higher remote-work intensity was associated with greater exhaustion, and that this association operated partly through blurred work–home boundaries rather than through reduced autonomy." Do not open by restating the background or the methods.

2. Interpret and compare

For each major finding, give it its own paragraph or subsection:

Return to the sources you discussed in your literature review. A discussion that cites none of the studies from the review misses the point; one that introduces many new studies may suggest the review was incomplete.

3. Deal honestly with unexpected results

Non-significant or surprising findings are not failures; they are information. Report them plainly, consider alternative explanations (low power, measurement issues, a genuinely different mechanism), and avoid spinning them. Reviewers trust authors who treat inconvenient results seriously.

4. Theoretical implications

Explain what your findings add to theory. Do they support one model over another, identify a boundary condition, reveal a new mechanism or show that a well-known relationship does not hold in a new context? Be specific: name the theory and the proposition affected. "This contributes to the literature" is not an implication; "This suggests that boundary theory's integration–segmentation distinction explains more variance in remote-work strain than job autonomy does" is.

5. Practical implications

Translate findings into what someone could do differently: managers, practitioners, educators, clinicians or policymakers. Keep recommendations proportional to the evidence. Correlational data support "organizations could consider…", not "organizations must…".

6. Limitations

Every study has limitations, and stating them shows judgment. Focus on the ones that matter for interpreting the results:

For each, briefly say how it might affect the findings and, where possible, why it does not undermine the main conclusion. Avoid a long apologetic list; four or five well-explained limitations are better than twelve generic ones.

7. Future research

Link future research directly to your limitations and findings: a longitudinal design to test causality, a different population to test generalizability, a qualitative study to explore a mechanism you inferred. Specific suggestions are more useful than "more research is needed".

8. Conclusion

End with a short paragraph that states the main takeaway and why it matters. Do not introduce new information or citations here. If your journal or program requires a separate conclusion section, keep it to a few paragraphs.

A short example of one discussion paragraph

The mediating role of boundary integration is consistent with boundary theory, which predicts that weak separation between work and home roles increases the cognitive cost of switching between them (Author, Year). It also helps explain why earlier surveys found mixed effects of remote work: studies that did not measure boundary practices may have averaged across employees who segment successfully and those who do not (Author, Year). However, because both variables were self-reported, part of this association may reflect shared method variance, a possibility the time-lagged design reduces but does not eliminate.

The paragraph states a finding, connects it to theory and prior studies, offers an explanation for the inconsistency in the literature and qualifies the claim, all in five sentences.

Language that helps

Match your claims to your design. Experimental evidence can support causal language; observational evidence usually cannot.

Common mistakes

Writing the discussion before results are final

If you are drafting before your analysis is complete, you can still plan the discussion: for each hypothesis, outline what it would mean if it were supported and what it would mean if it were not, and which prior studies each outcome would agree with. Keep the actual claims as clearly marked placeholders until the results exist. This keeps the draft honest and makes revision quick once the numbers are in.

Final check

Read your introduction's last paragraph and your discussion's first paragraph side by side. The question in one should be answered by the other. Then check that your abstract states the same main finding and implication as the discussion. When those three pieces agree, the paper holds together. For how the discussion fits the overall structure, see how to structure a research paper.