The discussion section explains what your results mean, how they answer your research question and how they compare with earlier studies. Some findings will come out unexpected, and here you explain why. You also say how far your conclusions can go and what follows for theory and practice. The job of the discussion is to interpret the results, and repeating them is not its job.
Use this guide when you are writing the discussion section of a journal paper or the discussion chapter of your thesis. You will get a paragraph order that works well, and worked examples that show reporting and interpreting side by side. Limitations and implications get their own part, because examiners and reviewers read those two most closely. If you need the overall shape of a paper, our research paper guide has it. The thesis structure guide shows where the discussion comes in a thesis.
Last reviewed September 2026.
What should the discussion section do?
Examiners and reviewers open the discussion mainly to see if the scholar has understood the meaning of his own findings. The results section may be fully accurate and still not answer that. Answering it is the work of the discussion.
The discussion has five jobs to do. First it answers the research question in plain words. Then it places that answer inside the existing literature. Whatever you did not expect, it handles openly. It also marks the limits of what you can conclude, and finally it says why the work matters, for theory, for practice, or for the next researcher.
| Results section says | Discussion section says |
|---|---|
| Engagement was higher among employees with 2–3 office days a week (M = 3.9) than among fully remote employees (M = 3.4), p < .01. | Some office time, in moderate amount, seems to support engagement. Our data go with earlier studies here, which argue face-to-face contact keeps team identity alive. Fully remote staff may be missing the informal chats, and engagement depends on them. |
| The moderation effect of supervisor support was not significant (β = 0.06, p = .21). | Against H3, supervisor support did not change this relationship. Maybe the scale captured general support and not support specific to remote work. Another reason may be limited variation, because most respondents gave their supervisors high ratings. |
| Accuracy of the proposed model was 94.2% against 91.8% for the best baseline. | The 2.4-point gain is modest. But it is there in all five folds, not one or two. Most of it came from fewer false positives on minority attack classes, and in practice that matters more than overall accuracy. |
Compare the two columns. Most of the numbers drop out on the right side. In their place come words like “appears”, “may” and “one possible reason”. And every result now connects to an explanation and to other people’s work.
What order should the discussion follow?
No one fixed structure is there for the discussion. For most empirical papers and thesis chapters the order below is fine, and reviewers don’t get lost in it.
State the main finding
Answer the research question in one or two sentences, with no p values in them.
Compare with earlier work
Show where your result matches published studies and where it does not. If some published study got the opposite result, say why you think the two differ.
Explain the surprises
Take up the unexpected or non-significant results. Give plausible explanations and clearly call them possible reasons only.
Limitations
Say what limits your conclusions. For each limitation, mention which way it may push the results.
Implications
Say what the findings mean for theory, for practice or policy, and next research.
Start with the answer, not the background
In the first paragraph itself, give your main finding in one or two sentences, without statistics. Many scholars begin by discussing about the aim and method again. The reader has just finished reading those parts. So open with what you found.
One finding per block
Suppose you have three or four main findings. Give every finding a separate block of one to three paragraphs. In one block the finding comes first, earlier work next, and your explanation last. Order of findings is your choice, by importance or as per objectives. Examiners like the objective way more, as they are ticking off each objective one by one.
Limitations and implications at the end
A few journals ask for a separate “Limitations” heading. Some others want “Theoretical implications” and “Practical implications” as sub-headings. Read author guidelines of the journal first. Also see how two or three recent articles in your target journal have done it.
How do you compare your findings with earlier studies?
Here the examiner sees whether you really know your field. Most cases fall under three patterns. (The author names in the examples are only placeholders.)
Agreement. “This is consistent with Sharma and Rao (2021), who found a similar effect among bank employees.” A line like this is useful, but do not end the paragraph there. If the study you agree with was done in a different setting, your result extends what is known, so explain how it extends. And if your finding only repeats what five earlier researches already found, then your contribution is in some other place. Tell the reader where it is.
Disagreement. “In contrast, Lee et al. (2020) reported no relationship.” After this, try to account for the difference. Maybe the population was different, or the measure, or the period, or the analysis. Many times this turns out to be the most interesting paragraph of the whole discussion, so don’t hide it.
Nothing to compare with. Sometimes you genuinely find no earlier study on this relationship. Then say so with care (“We are not aware of earlier studies that…”) and compare with the closest related work. Examiners start doubting the moment they see “no study has ever…”. Search the literature well before writing that line.
Cite only those studies which you have actually read. People copy citations from another paper’s discussion without opening the paper, and one wrong citation keeps travelling like that. Viva examiners also do ask what a cited study found.
How do you discuss unexpected or non-significant results?
All real studies have a few unexpected results. Write about them openly and the reader believes the rest of your discussion also.
- Report them plainly. Write “H3 was not supported.” and leave it visible. Hiding it in a side clause of a long sentence does not work, examiners notice.
- Offer one or two plausible explanations, and mark them as possibilities. The reason may lie in the measurement, the sample, the context or the timing. Or the theory itself may not fit this setting.
- No new theory just to save the hypothesis. If one non-significant path is taking you a full page to explain, stop and step back.
- Say what would settle it. Write a line like “A study using a remote-work-specific support scale would show whether…”. Now the next researcher has something to pick up from your disappointing result.
Many scholars are reading non-significant as “no effect”. It only means your data could not show an effect of that size with confidence. When the sample was modest, mention that the study may have been underpowered for small effects.
- Repeating the results. If a paragraph can move to the results section without changing a word, it does not belong in the discussion.
- New results in the discussion. Any number you talk about here should already be there in the results.
- Citing only studies that agree. Examiners know the literature well. If you leave out the study with the opposite finding, it looks careless, or it looks selective.
- Overclaiming. A cross-sectional survey shows association only, so mostly “X leads to Y” has to become “X is associated with Y”.
- Explaining away every non-significant result. Some hypotheses are just not supported. Accept that, give one or two plausible reasons, and go to the next point.
- A list of limitations with no consequence. Writing “The sample size was small” and stopping there is not enough. Tell the reader how much this limits trust in the findings, or how far they can be generalised.
- Generic implications. After reading “Managers should take note of these findings”, which manager will do what? Name the person and the change.
How do you write limitations and implications?
Limitations
Examiners know very well that every study has limitations. They are checking whether you understand your own. So for each limitation, write what it is and how it may affect the results. Then add what you did about it, or what a future study can do.
Weak: The study was limited to Chennai.
Better: Data came from IT employees in Chennai, and most firms in the sample required at least two office days a week. Where fully remote work is still common, engagement may differ. So the findings on fully remote employees come from a smaller subgroup (n = 64), and they should be read with caution.
Usually three to five genuine limitations are enough. Scholars sometimes fill a full page with limitations, small ones included, and then the thesis looks weak. Some “limitations” are really design choices that you can defend. Leave those out of this list and defend them in the methodology chapter.
Implications
Theoretical implications tell what your findings add to the theory you used, and where they qualify or go against it. Suppose the framework predicted X and in your data X came only under condition Z. That is your contribution to theory, write it in plain words.
Under practical implications write who has to change what. Name the actor, for example HR managers in mid-sized IT firms, district education officers, or hospital infection control committees. Stay inside what your data can carry. A survey of 300 employees in one city cannot advise national labour policy.
Future research suggestions should come out of your limitations and surprises. A stock list copied from other papers is not useful. Two or three from your own study are better than ten vague lines.
How is a thesis discussion chapter different from a paper’s discussion?
The logic remains the same in both. What changes is the scale and where the discussion is placed.
- Placement. In some theses there is one discussion chapter after all the results chapters. In others, mostly in science and engineering, each results chapter ends with its own discussion. After that comes a short overall discussion or a “summary and conclusions” chapter. Go by your university’s format and the custom in your department. You can see recent theses from your department on Shodhganga.
- Length. A thesis discussion may go up to 20–40 pages. That space is for depth on each objective. Repeating the literature review there wastes pages.
- Integration. Findings from different chapters or studies have to come together in the discussion. For example, chapter 4 may be a survey and chapter 5 may be interviews. The discussion is the place to show how the two fit together, or why they do not.
- Conclusion. In the conclusion chapter, you answer each research question in short and state the contribution. Starting a fresh discussion there is not correct.
Mostly the examiners prepare their viva questions from the discussion and the limitations. Whatever you claim in this chapter, be ready to defend it in the viva-voce. The viva questions guide lists the questions that come most often from this chapter.
Language that helps
Write what you found in past tense (“engagement was higher”). For what it means, and for what is already known, use present tense (“this suggests”, “earlier work shows”). Hedge only as much as the evidence needs. For a single survey “suggests” is enough, and keep “shows” for strong, replicated evidence. Scholars writing in a second language sometimes hedge every sentence, and sometimes they don’t hedge at all. One level of caution per claim is enough. Three levels in one sentence is too much.
Want a specialist to help with this in your own work? Research Paper Review Before Submission: a free first consultation, then a written scope and fixed quote.
Sources
FAQ
Questions scholars ask
Can I combine results and discussion?
Yes, many journals accept a combined “Results and Discussion” section. You see it a lot in chemistry, engineering and materials science. Check the author guidelines of your journal. In a combined section also, in each paragraph give the number first and then what it means.
Should I cite new literature in the discussion that was not in my literature review?
Yes. An odd finding many times sends you to studies you never thought of citing. A few such citations is fine, not twenty. If one of these sources becomes central to your argument, consider adding it to the literature review also.
How long should the discussion be?
For a journal article it is often about the same length as the results, commonly 1,200 to 1,800 words. In a thesis, write till every objective gets its proper depth. Forty pages that say little will not make any examiner happy.
What is the difference between the discussion and the conclusion?
In the discussion you interpret and argue. The conclusion only states briefly what was found for each question and what the study contributes. It has no new argument and no new citations. For papers, literature reviews and the final thesis chapter, see our conclusion guide.
My results contradict my supervisor’s earlier paper. What do I do?
Your guide’s paper is one more earlier study here, so report the difference and the possible reasons the same way. Speak to your guide about it early. Most supervisors respect a well-argued difference much more than a result that was adjusted to fit.
