The discussion is the chapter that decides how your thesis is judged. Your methods can be flawless and your analyses correct, and a weak discussion will still leave a committee unconvinced — because this is the only chapter where you demonstrate that you understand what your own results mean. It is also the chapter students most often write badly, usually by summarising the results a second time instead of interpreting them.
This guide gives you eight steps in the order you should write them, with worked examples of the actual text. It assumes you have finished your results chapter and that your numbers are final. If they are not, stop — writing a discussion around results that may still change is how people lose a fortnight.
Step 1: Open by answering your research question in two sentences
Do not open with “This chapter discusses the findings of the present study.” That sentence tells the reader nothing they did not know from the heading, and it starts the chapter on a note of throat-clearing. Open with the answer.
Write two sentences: what you found, and what it means. Nothing else goes in that paragraph.
This study found that perceived supervisor support predicted lower academic burnout among graduate students, but only for students who reported low levels of financial strain. The relationship that the broader literature treats as universal therefore appears to be conditional, and the condition is material rather than psychological.
An examiner who reads only your first paragraph and your conclusion should be able to state your finding accurately. If they cannot, the chapter has already failed at its main job.
Step 2: Take each hypothesis in turn and say plainly whether it was supported
Work through your hypotheses in the same order you presented them in the introduction and the results. For each one, write a short paragraph with four elements: restate the hypothesis in words, state whether it was supported, give the key statistic in parentheses, and stop. Interpretation comes in the next step — resist merging the two, because merging them is what produces the sprawling, unnavigable discussion.
Hypothesis 1 predicted that self-compassion would be negatively associated with symptoms of anxiety. This hypothesis was supported: self-compassion accounted for a significant proportion of variance in anxiety scores, β = −.34, p = .002.
Report the statistic once, in APA 7th edition format, and do not reproduce your results tables here. If the reader needs the full model, they have the results chapter — and the conventions for reporting it are set out in the walkthrough on how to run and report a multiple regression for your thesis.
Step 3: Interpret each finding against the specific literature you reviewed
This is the core of the chapter and the step where marks are actually won. For every finding, make one of three moves, and make it explicitly:
- Convergence. Your finding agrees with prior work. Say whose, and say what your study adds — a new population, a stronger design, a Canadian sample where the original was American, a replication that was overdue.
- Divergence. Your finding disagrees with prior work. Say whose, and offer a specific, testable reason: a different measure, a different age range, a different cultural context, a different time period.
- Extension. Nobody has tested this before. Say why the gap existed and what your result now establishes.
Vague gestures at “the literature” are the most common weakness in a graduate discussion. Name the studies. You already did the work of identifying them when you wrote the literature review for your thesis, and the discussion is where that work pays off — every source you engage with here should already have appeared in chapter two.
This pattern diverges from Mitchell and Rao (2021), who reported a uniform association across their sample. One plausible explanation is measurement: their single-item financial strain indicator would not have detected the threshold effect observed here, since a threshold cannot appear in a measure with insufficient variance at the lower end of the distribution.
Notice the structure — name, disagreement, mechanism. “Perhaps cultural differences played a role” is not a mechanism; it is a placeholder, and a committee will ask you to replace it with an argument.
Step 4: Explain non-significant results without apologising for them
A hypothesis that was not supported is a finding, not a failure, and treating it as an embarrassment is what makes it read like one. Give it the same structural treatment as a significant result: state it, then explain it.
Legitimate explanations, roughly in order of how often they are the real answer:
- Statistical power. Your sample may have been too small to detect an effect of the size that exists. Say so with reference to your own planning, not as a generic excuse — the reasoning behind that calculation is covered in the guide to how many participants a thesis needs.
- Measurement. The instrument may not have captured the construct in your population, or may have shown weaker reliability in your sample than in its validation study.
- Restriction of range. Sampling university students often produces a narrow band on variables such as age, education or socioeconomic status, and a restricted range attenuates correlations mechanically.
- The effect may not exist. This is a legitimate scientific conclusion and it is chronically underused in student theses.
Two phrases to strike from your draft permanently: “approaching significance” and “trending towards significance.” A p value of .07 is not a smaller version of significance under the framework you adopted; using that language signals that you do not understand your own test, and it is a reliable invitation to a difficult defence question.
Step 5: State the theoretical contribution precisely
Name the theory. State what it predicted. State what your data did to that prediction. Three sentences is often enough, and precision matters more than length here.
These findings qualify the conservation of resources framework as it is typically applied to graduate populations. The model predicts that psychosocial resources buffer strain regardless of material context; the present results suggest instead that material resources function as a precondition for psychosocial buffering, rather than as a parallel resource.
Be honest about scale. A master’s thesis rarely overturns a theory, and claiming that it does invites exactly the scrutiny you cannot survive. Qualifying, extending or bounding a theory is a real contribution and a defensible one.
Step 6: Write implications that someone could act on
Practical implications fail in one of two directions: too grand (“universities should reform graduate funding”) or too vague (“these findings may be useful for practitioners”). Aim between them by naming who does what differently.
For graduate programme administrators, these results suggest that supervisory training initiatives are unlikely to reduce burnout among students experiencing significant financial strain. Where resources permit only one intervention, guaranteed minimum funding may be a precondition for the effectiveness of relational supports rather than an alternative to them.
Where your topic touches clinical practice, keep the boundary visible. A thesis with a non-clinical convenience sample does not license clinical recommendations, and the Canadian Psychological Association’s ethical framework expects claims proportionate to evidence. “These findings are consistent with, but do not establish” is a phrase worth keeping.
Step 7: Write limitations that earn credit instead of losing it
Most students write limitations as confession. Examiners want analysis. The difference is a single added element: the direction of the bias. Do not just name a weakness — say what it likely did to your result.
Cover four categories, and give each one a direction:
- Design. Cross-sectional data cannot establish causation, and reverse causation is plausible here — burnout may erode the perception of support as easily as low support produces burnout.
- Sample. Who is missing, and would including them likely strengthen or weaken the association?
- Measurement. Self-report shared-method variance inflates correlations; a subscale with weak internal consistency attenuates them. If a reliability estimate came in low, address it here rather than hoping nobody checks, and see the discussion of what counts as an acceptable Cronbach’s alpha for the defensible framing.
- Analysis. Assumption violations, multiple comparisons, listwise deletion of a non-trivial share of cases.
Because both predictor and outcome were assessed by self-report in a single session, shared-method variance is likely to have inflated the observed associations. The reported coefficients should therefore be treated as upper-bound estimates.
That sentence demonstrates methodological maturity. “A limitation of this study is that it used self-report measures” demonstrates that you can recall a phrase from a textbook.
Step 8: Make future research specific enough to be a proposal
“Future research should explore this relationship further” is filler and reads as such. A good future-directions paragraph names a design, a population and a measurable outcome — it should read like something a reader could actually apply to fund.
A longitudinal design following students across a funding transition — for example, from a funded first two years into an unfunded third year — would allow the temporal ordering of financial strain and burnout to be established. Sampling across both thesis-based and course-based master’s programmes would additionally test whether the effect depends on the extended timelines characteristic of thesis-based study.
Two or three of these is enough. A list of nine reads like padding.
What must never appear in a discussion chapter
- New results. If an analysis is worth discussing, it belongs in the results chapter. Introducing a number here tells the examiner it was run after you saw the outcome.
- A wall of new citations. A small number of new sources is acceptable when a finding was genuinely unanticipated. Twenty new references means your literature review missed the field.
- Causal language from correlational data. “Predicted” is acceptable in a regression sense; “caused,” “led to” and “resulted in” are not, and examiners catch every instance.
- A rerun of the results chapter. If a paragraph could be moved into chapter four without editing, it belongs there.
How long should it be, and how should it be structured?
For a Canadian master’s thesis in psychology, expect roughly 3,000 to 5,000 words, typically fifteen to twenty per cent of the total. Confirm against your own school of graduate studies guidance and your supervisor’s expectations, because both vary by department and by whether you are writing a thesis or a course-based major research paper.
Use subheadings that name findings rather than generic labels. “Financial strain as a boundary condition” tells a reader where they are; “Discussion of Hypothesis 2” does not. Whether limitations and future directions appear as separate chapter sections or as parts of a combined discussion is a departmental convention — check a recently accepted thesis from your own programme in the institutional repository, which is also the fastest way to resolve formatting questions your methodology chapter raised earlier.
The defence connection
Your discussion chapter writes your defence. Every unhedged claim is a question you will be asked to defend, every unnamed limitation is one the committee will name for you, and every mechanism you assert without evidence is a place where an examiner will ask “how do you know?” Read your finished chapter once with a single question in mind: where would I attack this? Then address those points in the text, pre-emptively. The typical shape of that questioning is set out in the guide to what questions are asked at a Canadian thesis defence.
Write it faster without writing it worse
The discussion is the chapter where students stall longest, because it demands judgement rather than procedure. Tesify helps you get a structured draft down — hypothesis by hypothesis, with your findings, your literature and your limitations in the right places and consistent APA formatting throughout — so your supervisor’s time goes to the argument rather than to fixing the scaffolding. The thinking stays yours; the blank page stops being the obstacle.
Draft your discussion chapter with Tesify
Frequently asked questions
What is the difference between the results chapter and the discussion chapter?
Results report what you found, with statistics and no interpretation. The discussion explains what those findings mean, how they relate to prior research, and what follows from them. If a paragraph contains a new number, it belongs in results; if it contains a “because,” it belongs in the discussion.
Can I combine results and discussion into one chapter?
In qualitative theses this is common and often preferable, since themes and interpretation are difficult to separate. In quantitative psychology theses the two are usually kept apart. Ask your supervisor, and check accepted theses from your own department before deciding.
None of my hypotheses were supported. Can I still pass?
Yes. You are assessed on the quality of your reasoning and execution, not on whether nature cooperated. A thesis of well-explained null results with a rigorous method passes; a thesis of significant findings the candidate cannot interpret does not.
How many limitations should I list?
Three to five, each analysed properly with a stated direction of bias. A list of ten limitations mentioned in one line each reads as a candidate hedging against every possible criticism, and it dilutes the ones that genuinely matter.
Should the discussion have its own conclusion section?
Most Canadian psychology theses end with a short concluding section of two or three paragraphs that returns to the research question and states the contribution without introducing anything new. Some departments require a separate conclusion chapter instead — confirm the local convention.
Can I cite sources in the discussion that were not in my literature review?
A few, where an unexpected finding requires literature you had no reason to review beforehand. If you find yourself adding many, the honest fix is usually to strengthen the literature review rather than to bolt the missing field onto the end of the thesis.
How do I discuss an unexpected finding I never hypothesised?
Label it exploratory and say so explicitly. Present it after your hypothesised findings, offer a tentative explanation, and state that it requires confirmation in an independent sample. Presenting an unplanned result as though it had been predicted is a serious integrity problem, not a presentational choice.
Is the discussion different for a major research paper than for a thesis?
The structure is identical; the scale is smaller. An MRP discussion typically runs shorter and engages a narrower body of literature, but the same eight steps apply, and the same weaknesses cost the same marks.