If you know how to write a discussion chapter for a thesis, you already know the hardest part is not the writing itself. The hard part is deciding what your results actually mean. The discussion chapter is where you interpret findings, explain why they likely happened, compare them with theory and previous research, set out the implications, and admit what your study could not do. It introduces no new data.
Students consistently rate it the hardest chapter of a dissertation, and the reason is almost never ability. It is that the results chapter rewards accuracy while the discussion chapter rewards judgement. A table either confirms a test at p < .05 or it does not. A discussion has to say what that means, and how confident you are.
Below is the sequence I would follow, tested against how supervisors and examiners actually read this chapter. It takes a finished results section and turns it into an argument. Budget two to four weeks for a moderate-length thesis rather than assuming you will knock it out in a weekend.
Table of Contents
- 1What You Need
- 2How to Write a Discussion Chapter for a Thesis: Step-by-Step
- 31. Review your objectives and research questions
- 42. Summarize the main findings without repeating the results
- 53. Interpret what each result means in a discussion chapter
- 64. Compare your findings with theory and previous research
- 75. Discuss the practical and theoretical implications
- 86. Explain the study limitations
- 97. Write recommendations and future research
- 108. Edit, tighten and connect the chapter
- 11Common Mistakes
- 12Frequently Asked Questions
- 13How long should a discussion chapter be?
- 14Can the discussion and conclusion be combined?
- 15What should not be included in the discussion section?
- 16How do I write the discussion when results are not significant?
- 17Should I use first person in my discussion chapter?
- 18How do I stop rewriting the results chapter in different words?
- 19Conclusion
What You Need
Before you type a single sentence, gather these five things. Most stalled discussions are stalled because the writer is working from memory.
- Your completed results. Every table, figure, test output and theme, in one place, with the exact numbers you intend to report.
- Your objectives, research questions or hypotheses. Write these out again from scratch. The discussion must answer each one, in order.
- Your conceptual framework. The theory you set out to test, in two or three sentences you could say out loud.
- Your literature review notes. Not the whole chapter, just the studies you will compare against, with the direction of each finding.
- Your methodology details. Sample, design, measures, analysis. This is what you will draw on when you write limitations honestly.
Check your university’s formatting rules too, particularly how many words the discussion chapter may carry and whether conclusions are a separate chapter. That single formatting decision shapes your whole outline.
How to Write a Discussion Chapter for a Thesis: Step-by-Step
Most theses put the discussion chapter straight after the results and before the conclusion. A discussion chapter typically runs 2,500 to 4,000 words at Master’s level and 5,000 to 12,000 words at PhD level, which is roughly 10 to 20 percent of the whole document. Some institutions combine discussion and conclusion into one closing chapter, so check yours before you outline.
| Degree level | Typical discussion length | Share of thesis | Realistic time to write |
|---|---|---|---|
| Honours or undergraduate project | 1,000 to 2,000 words | 8 to 12 percent | One to two weeks |
| Master’s dissertation | 2,500 to 4,000 words | 10 to 15 percent | Two to four weeks |
| MPhil thesis | 4,000 to 7,000 words | 12 to 18 percent | Three to six weeks |
| PhD thesis | 5,000 to 12,000 words | 15 to 20 percent | One to three months |
The pattern across all four is the same: the discussion chapter grows with the number of research questions, not with the length of the thesis. Four research questions and four findings means four interpretation blocks, no matter how short the thesis is.
1. Review your objectives and research questions
Do this before you draft anything. List each objective, question or hypothesis in the order your methodology chapter set them out, and next to each one write the single result that answers it. You now have your outline, and it should be the same order as your chapters.
If a research question has no clear answer in your results, that is itself a finding worth discussing, and it belongs in this chapter rather than buried in an appendix. Examiners notice gaps. Students who raise them first look more careful than students who quietly skip them.
A useful discipline here: organise around findings, not around tests. A thesis that reports twelve analyses but answers three questions needs three interpretation blocks, not twelve. If your supervisor asks why a particular test is missing from the discussion, that test was not addressing a research question.
2. Summarize the main findings without repeating the results
Write one short paragraph, roughly 150 words, that names each major finding and states its direction. No new numbers beyond the headline ones: report the key statistic or the core theme, not every coefficient in the model.
Here is the test for whether you have drifted into summary syndrome. Cover the paragraph and read it aloud to someone who has not seen your results. If they could have reconstructed your results chapter from what you said, you have restated rather than interpreted. If they cannot, but they know roughly what you found and in which direction, the paragraph is doing its job.
A second test works better than most: try deleting the paragraph entirely. If nothing in the chapter stops making sense, the summary was decorative.
3. Interpret what each result means in a discussion chapter
This is the core of the chapter and the part students rush. One workable pattern is four sentences per finding, sometimes called claim, evidence, meaning, boundary. State the finding as a claim in your own words. Give the evidence briefly, pointing the reader to the table or figure. Explain the meaning in substantive terms, in the language of your field rather than statistical language. Close with the boundary, the condition under which the finding holds.
The difference between a claim and evidence shows up immediately in the sentence structure. “Flexibility was significantly related to retention (β = .34, p < .01)” is evidence. “Greater schedule flexibility helped organisations hold on to staff, with a moderate positive relationship” is a claim written about the world.
For quantitative work, report the effect size alongside significance, and describe it in plain terms: small, moderate or large for your field. A p-value tells your reader the finding is unlikely to be chance; it says nothing about whether it matters.
For qualitative work, interpretation looks different. You move from a code to a theme, then ask what that theme means in context and how it relates to your framework. Quote sparingly, and only where the wording carries meaning a paraphrase would lose.
For mixed methods, you have the hardest job because you must say whether the strands agreed. Convergence is a finding worth stating plainly. Divergence is a finding too, and it needs an explanation rather than a footnote.
4. Compare your findings with theory and previous research
For each major finding, name two or three specific studies and say whether your result agrees with them, disagrees, or sits between them. Then account for the difference. Sample population, measurement instrument, setting, timing and analytical method all produce divergence without any of them being an error.
Do not narrate the literature review here. If a paragraph would make sense in a chapter with no findings attached to it, it belongs in your literature review instead. The test is simple: can you point to a table, figure or theme in your own study that this paragraph is about?
When your finding contradicts an established result, write that down early rather than leaving it for a footnote. Say what differs about the context, what it implies for the theory, and what it would take to settle the disagreement.
5. Discuss the practical and theoretical implications
Keep these separate. Practical implications cover what a manager, clinician, policymaker or teacher could do differently on the basis of your results. Theoretical implications cover what your study adds to, narrows or challenges in the field’s conceptual framework.
Every implication traces back to a specific finding. If you cannot name the result it follows from, it is an opinion and it does not belong. For example, a finding that workload predicts retention among early-career nurses supports a practical implication about scheduling, and a theoretical implication about how procedural fairness shapes attachment to an organisation.
Use cautious language here. “The findings suggest” and “may indicate” are not weakness; they tell an examiner you understand the difference between what you measured and what you infer. Reserve “proves” and “demonstrates” for claims your design genuinely supports.
6. Explain the study limitations
Pick three to six real limitations and group them by category: sample and selection, measurement, design, analysis, generalisability. Then state the likely effect of each on your findings, not just its existence.
Group them by category and give each one a direction. “The sample was small” tells an examiner nothing. “Because the sample was drawn from a single organisation, the findings may not transfer to sectors with different incentive structures” is a limitation with a consequence attached. Examiners flag vague limitations, not honest ones.
Then name what you would do differently. A limitation paired with a fix reads as methodological maturity rather than an excuse. The three hardest categories to write well are selection bias in the sampling frame, validity of the measurement instrument, and how far the design lets you claim generalisability.
7. Write recommendations and future research
Recommendations come from findings plus limitations. If your study measured a single point in time, a recommendation about tracking change over a year is a recommendation. If your sample came from one region, a recommendation about replicating elsewhere is a research direction rather than a recommendation, because it belongs to future work.
Future research should follow from the same place. Each gap you identified in your limitations or your comparison with literature should produce one direction, stated in a sentence or two.
Keep a short sentence of “so what” at the end. Students often write an entirely accurate chapter and leave the reader with no sense of why it mattered. That sentence closes the loop back to the research problem you stated in Chapter One.
8. Edit, tighten and connect the chapter
Read the chapter in one sitting and check four things: does every paragraph earn its place, does each claim trace to a result, do transitions carry the argument rather than just the sequence, and is the tense and voice consistent.
On voice, use first person where your discipline expects it and third person where it does not. Business, education and many applied health fields accept “we surveyed 312 nurses”; pure and life sciences often prefer passive or third person. Match the published papers in your field.
Do not introduce new references just because they are relevant. The reference list for this chapter should mostly be sources you already cited in your literature review, plus the few specific comparators your discussion genuinely turns on.
Finally, read only the first sentence of each paragraph in sequence. If those sentences alone form a coherent argument, your structure works. If they read as a list of unrelated observations, reorder the blocks.
Common Mistakes
These six failure patterns account for most of the discussion chapters I have seen marked down. Each has a specific fix.
| Failure pattern | What it looks like | The fix |
|---|---|---|
| Summary syndrome | The chapter restates the results in different words | Delete any sentence that only reports. Keep sentences that explain |
| Literature dumping | Long summaries of studies with no link to your findings | Ask for every citation to name a table or theme in your study. If it cannot, cut it |
| Speculation without evidence | Explanations for a result that your data cannot support | Label it as a possible explanation and say what evidence would test it |
| Ignoring contradictory evidence | Selective comparison with studies that happen to agree with you | Address the strongest disagreement directly and explain the difference |
| Shopping-list writing | A flat list of findings with no cumulative argument | Link findings. Show how the second result qualifies or extends the first |
| Hedged into silence | Every sentence loaded with may, might, possibly, arguably | Hedge the inference, not the finding. The result itself is a fact; state it plainly |
One more habit worth breaking: writing the discussion as a list of implications in the singular. One practical implication and one theoretical implication per finding reads thin. Trace the chain instead, where one result supports a practice change which then raises a question for theory.
Frequently Asked Questions
How long should a discussion chapter be?
A Master’s discussion chapter typically runs 2,500 to 4,000 words, and a PhD chapter commonly runs 5,000 to 12,000 words. That is roughly 10 to 20 percent of the whole thesis. Length should follow the number of research questions rather than a target: one interpretation block per question, plus a limitations and implications section. Check your institution’s regulations before you start, because some set a hard cap.
Can the discussion and conclusion be combined?
Yes, and many institutions require it. A combined chapter works when your conclusion stays short: two or three sentences that state what the study established, for whom, and what follows. Combining becomes a problem when the conclusion simply repeats the discussion’s final paragraph, because examiners then cannot tell whether you synthesised or restated. Many supervisors prefer two separate sections, so ask before you outline.
What should not be included in the discussion section?
Do not report new data, reintroduce full statistical output, repeat the results chapter in summary form, or narrate your literature review without linking it to your findings. Do not present recommendations that no result supports, and do not apologise for your study’s limitations at length. Keep raw methodology detail in the methods chapter. The discussion should contain interpretation, comparison, implications, limitations and synthesis.
How do I write the discussion when results are not significant?
Report the null result clearly, then do the same three moves you would do with a positive finding. Describe its direction and size rather than its p-value alone. Compare it with what prior studies expected. Then consider explanations: insufficient statistical power from a small sample, a ceiling or floor effect in your measure, or a genuine absence of effect. A well-interpreted null result is a stronger contribution than a buried one.
Should I use first person in my discussion chapter?
It depends on your field. Business, education, nursing and applied health research commonly use first person plural, because the researcher is part of the design. Pure and life sciences usually prefer passive or third person. Match what the published literature in your discipline does, and stay consistent throughout the chapter. Switching voice mid-chapter reads as a lack of control, so decide early.
How do I stop rewriting the results chapter in different words?
Apply the deletion test. For each paragraph, ask whether it would still make sense if your results table were removed. If it would, you are summarising, and it should be cut or rewritten. A genuine interpretation paragraph contains something your results table cannot contain: an explanation, a comparison, a consequence or a judgement about confidence.
Conclusion
Start with your research questions in one column and your findings in another, then work down the list asking what each result means, where it agrees or disagrees with the literature, and what follows from it. Write the interpretation before the summary, keep the summary under 200 words, and let every limitation carry its likely direction. That sequence is what turns a set of tables into a discussion chapter in 2026.
