Why do strong results chapters so often end in weak discussion chapters? Because reporting a finding and arguing about it are different skills. The results chapter follows conventions; the discussion asks you to think in public: to say what each finding means, why it may have occurred, how it sits with what others found, what it cannot show and what should follow from it.
This article is part of our guide How to Write a Master’s Thesis by the Mastermind PhD team and relates to our Master’s Thesis support service. Sources last checked: November 2026.
Students who have spent months on data collection often rush the discussion chapter, repeat the results in prose and add a generic list of limitations, and examiners notice, because this is where the mark for critical analysis is earned. This guide gives the chapter a structure, shows how to interpret without overclaiming and explains how to write limitations and implications that strengthen the argument.
What is the discussion chapter? The discussion chapter interprets the findings reported in the results chapter in the light of the research questions, the theoretical framework and prior research. It explains what the results mean, compares them with earlier studies, states the limitations and their effect on the conclusions, and draws out the implications for theory, practice and further research. It is usually chapter 5 and, with the conclusion, carries about 30% of the thesis.
What does the discussion chapter do that the results chapter does not?
The results chapter says “onboarding structure was negatively associated with turnover intention, r = -.42”. The discussion says what that means for the question you asked, whether the size of the association matters in practice, how it compares with results from large on-site firms, why it might differ in scale-ups, what the cross-sectional design prevents you from concluding, and what a manager or a future researcher should do with it. Five verbs summarize the chapter: interpret, compare, explain, qualify, imply. A paragraph that does none of these is a results paragraph in the wrong chapter.
How should the discussion chapter be structured?
The order below works for most empirical theses and maps directly onto what examiners look for.
| Section | Purpose | Share of a 3,000-word discussion |
|---|---|---|
| Introduction and summary of key findings | Restate the aim; summarize the main findings in words, without statistics | 300 words |
| Interpretation by research question | One subsection per question: what the finding means and why it may have occurred | 1,000 words |
| Comparison with previous research | Where the findings agree, disagree or extend prior work, and why | 600 words |
| Theoretical implications | What the results say about the framework or theory used | 300 words |
| Practical implications and recommendations | Who should do what differently, stated specifically | 350 words |
| Limitations | Each limitation, its effect on the findings and how it was mitigated | 300 words |
| Directions for future research | Tied to the limitations and the unanswered questions | 150 words |
Where the handbook asks for a separate conclusion chapter, move the answers to each question, the statement of contribution and the final recommendations there, and keep this chapter to interpretation, comparison and limitations.
📱 Talk to the Mastermind PhD team — You are told the specialist’s field and degree before work starts, and a meeting is scheduled to review your requirements.
How do you open the discussion chapter?
Begin with the aim in one sentence, then summarize the key findings in plain prose: “This study asked whether structured remote onboarding is associated with turnover intention among junior developers in Dutch scale-ups and whether supervisor support moderates that association. Turnover intention was moderate. Structure was associated with lower turnover intention, and the association was weaker where supervisor support was high.” No statistics, no citations, no new information. The summary tells the reader which findings the chapter will discuss and in what order, and it is the paragraph examiners quote in their report.
How do you interpret a finding without overclaiming?
Match the strength of your language to the strength of your evidence. The design sets the ceiling: a cross-sectional correlational study can say “associated with” and “predicts” in the statistical sense; it cannot say “causes”, “leads to” or “improves”. A small sample or a marginal effect size calls for stronger hedging than a large, well-powered result.
| Strength of evidence | Appropriate verbs and hedges | Avoid |
|---|---|---|
| Large effect, replicated in the literature, adequate power | “shows”, “indicates”, “is consistent with” | “proves”, “confirms beyond doubt” |
| Significant but small effect, or single study | “suggests”, “appears to”, “is associated with” | “demonstrates”, “establishes” |
| Non-significant or mixed results | “did not find evidence that”, “may”, “it is possible that” | “shows there is no relationship” |
| Causal language in a correlational design | “is associated with”, “predicts” (statistical sense) | “causes”, “leads to”, “the effect of” |
Hedging is not weakness; it is precision. Hyland’s analysis of research writing shows that experienced authors hedge systematically to signal how much confidence a claim deserves. A thesis that overclaims reads as one whose author does not understand the design.
How do you compare your findings with previous research?
For each main finding, place it against the two or three studies that come closest, in one of three relationships: it agrees, it disagrees or it extends. Agreement needs a sentence, not a paragraph. Disagreement needs an explanation, and the explanation should come from concrete differences: the sample (juniors versus all employees), the measure (intention versus actual turnover), the context (scale-ups versus large firms), the time (post-pandemic remote work) or the analysis. Extension is the most valuable: your study tested a known relationship in a new population or added a moderator, and you say what that adds.
Use the literature you reviewed in chapter 2; the examiner expects the discussion chapter to return to it. A new source is acceptable when an unexpected result needs an explanation you could not have anticipated, but a discussion built on sources absent from the review signals that the review was incomplete.
How do you discuss non-significant and contradictory results?
A non-significant result has at least three possible explanations, and the discussion chapter should weigh them: the effect may not exist in this population; the study may have lacked power to detect it (report the achieved power or the width of the confidence interval); or the measure may not have captured the construct (a scale validated elsewhere, a ceiling effect). Say which explanation the evidence favors and why. Do not reinterpret a non-significant finding as support for the hypothesis, and do not drop it.
A contradiction with the literature deserves the same treatment: name the difference in design or context that could account for it, and state what study would settle the question.
How do you write the limitations section?
Write each limitation as three parts: what the limitation is, how it could have affected the findings, and what you did to reduce its effect or how the reader should weigh it.
“The sample was self-selected through professional networks, so developers who were already disengaged may be over-represented; this would inflate the mean turnover intention but is unlikely to change the direction of the association, and the analysis controlled for tenure and company size” is a limitation. “The sample size was small” is a confession. Keep to the three to five limitations that matter for your conclusions, and separate them from delimitations (deliberate choices of scope), which belong in the introduction.
How do you write implications and recommendations?
Implications name an audience and a change. For theory: which proposition of the framework the results support, qualify or contradict. For practice: who should do what, in operational terms (“scale-ups without dedicated HR staff should document a four-week remote onboarding sequence with a named mentor”), within the limits of the evidence (a correlational study supports “consider”, not “must”). For policy, where relevant: which decision the evidence informs.
Recommendations for future research should follow from your limitations and open questions and be specific enough to act on: a longitudinal design measuring actual turnover at twelve months, or a replication outside the Netherlands.
Should the discussion and conclusion be one chapter or two?
Follow the handbook. A combined chapter suits a short thesis with one or two questions and ends with a conclusion section that answers each question in a sentence and states the contribution. Two chapters suit theses with several questions, where a separate conclusion of 5 to 8% of the word count answers every question plainly in one place. In either case the conclusion contains no new results, literature or recommendations; everything in it has been argued in the discussion chapter.
Which mistakes do examiners flag most often?
- Repeating the results chapter in prose, with the same statistics.
- Causal claims from correlational designs.
- Comparisons that list agreements without explaining disagreements.
- New findings or analyses appearing for the first time.
- A literature review’s worth of new sources.
- Generic limitations (“small sample”) without their effect on the findings.
- Implications addressed to no one (“organizations should pay attention to onboarding”).
- Research questions from chapter 1 never answered in the words they were asked.
Before submission, read chapter 1 and chapter 5 together: every question asked should be answered, every hypothesis resolved, and the significance promised delivered or honestly scaled back. A reviewer can check the discussion chapter for hedging, structure and alignment with the results; the interpretation, the position on contradictions and the recommendations are the author’s, and they are what the examiners will ask about.
📱 Ask on WhatsApp now — 5 years of experience across all disciplines and 14 academic services. Two-stage payment (80/20); revisions are free until supervisor approval, within 30% of the chapter.
Frequently asked questions
How long should the discussion chapter be?
About 20% of the thesis when the conclusion is a separate chapter (3,000 words in a 15,000-word dissertation), or around 30% when they are combined. A discussion chapter much shorter than the results chapter usually means the findings were reported but not interpreted.
Can I cite new sources in the discussion?
Yes, when an unexpected finding needs an explanation you could not have anticipated. Most of the sources should come from the literature review, because the discussion’s job is to return to that literature with your results in hand. Many new sources suggest the review missed part of the field.
Can I use the first person in the discussion?
If your department allows it elsewhere in the thesis. The discussion is where an author’s voice is most legitimate, because it is an argument; “I interpret this as” is clearer than “it is interpreted that”. Apply one convention throughout the thesis.
What if my results contradict my hypotheses?
Report that the hypotheses were not supported and discuss why, weighing a true null effect, insufficient power and measurement problems. A well-argued negative result earns marks; a result reinterpreted to rescue the hypothesis, or quietly omitted, loses them.
What is the difference between the discussion and the conclusion?
The discussion interprets and argues: meaning, comparison, explanation, limitations, implications. The conclusion answers each research question in a sentence or two, states the contribution and ends. The conclusion contains nothing that was not argued in the discussion.
Sources
- Swales, J. M., & Feak, C. B. (2012). Academic Writing for Graduate Students: Essential Tasks and Skills (3rd ed.). University of Michigan Press.
- Murray, R. (2017). How to Write a Thesis (4th ed.). Open University Press.
- Evans, D., Gruba, P., & Zobel, J. (2014). How to Write a Better Thesis (3rd ed.). Springer.
- Hyland, K. (1998). Hedging in Scientific Research Articles. John Benjamins.
- Dunleavy, P. (2003). Authoring a PhD: How to Plan, Draft, Write and Finish a Doctoral Thesis or Dissertation. Palgrave Macmillan.
- American Psychological Association. (2020). Publication Manual of the American Psychological Association (7th ed.). https://apastyle.apa.org/
Sources last checked: November 2026. This article is general educational guidance and does not replace your university regulations or your supervisor’s instructions. How we prepare our content.