How to Write the Discussion Chapter of a Psychology Dissertation (2026): Step by Step, With APA Examples

The discussion chapter of a psychology dissertation has a job description that other fields do not impose: it must state, hypothesis by hypothesis, what the data did to each prediction; interpret effect sizes rather than p-values; place every finding against the replication record of the field; and confine its claims to what the design can support. The APA Journal Article Reporting Standards (JARS-Quant, Appelbaum and colleagues, 2018) set those expectations for published articles, and psychology committees read dissertations against them. This guide walks through the chapter in nine steps, with the sentence patterns that pass and the ones that get sent back. For the generic architecture of the chapter in any field, see our guide to how to write a thesis discussion chapter; this one is about what psychology adds.

Step 1: Answer each hypothesis in the first paragraphs

Open with a one-paragraph restatement of the aim, then take the hypotheses in the order they were numbered in the proposal and say for each whether it was supported, partially supported or not supported, quoting the effect once. A committee should be able to read the first page and know the outcome of the study. The pattern is: Hypothesis 1, that participants in the mindfulness condition would report lower anxiety at week eight than waitlist participants, was supported: the adjusted difference was 2.1 points on the GAD-7, d = 0.41, 95 percent CI [0.12, 0.70]. The hypotheses themselves should already exist in the proposal in exactly this numbered form; if they do not, our guide to research questions, hypotheses and variables for a psychology dissertation shows how to write them so that the discussion can answer them one by one.

Step 2: Interpret the effect size, not the p-value

A significant p-value tells the reader that an effect is unlikely to be zero; it says nothing about whether the effect matters. Psychology has spent a decade moving from significance to estimation, the shift Cumming (2014) called the new statistics, and committees now expect the discussion to interpret the size and precision of each effect. Three habits follow. Compare the effect with benchmarks from your own literature rather than Cohen’s generic labels: a d of 0.40 is unremarkable for a mood induction and large for a brief app-based intervention. Read the confidence interval as a range of plausible effects, and say when it includes values too small to matter. And treat a large but non-significant effect in a small sample as inconclusive, not as absent. The mechanics are in our guide to effect sizes and confidence intervals in a dissertation; the discussion is where the numbers get their meaning.

Step 3: Compare with prior findings and explain the differences

For each hypothesis, name the two or three studies your literature review set up as the comparison and say whether your result agrees. Where it disagrees, offer a specific explanation rooted in a difference you can point to: a different population (clinical versus student), a different measure (clinician-rated versus self-report), a different design (between-subjects versus within), or a different dose. Avoid the sentence that explains every discrepancy by culture or sample size without evidence. Psychology’s replication record is relevant here: the Open Science Collaboration (2015) attempted to replicate 100 published findings and obtained significant results in 36 percent of replications, against 97 percent of the originals. If your comparison study has failed to replicate, or is a single small study, say so; a dissertation that agrees with a fragile finding is not stronger for it.

Step 4: Handle null and unexpected results honestly

A hypothesis that was not supported needs the same care as one that was. Work through the candidate explanations in order and rule each in or out with evidence from your own data: was the study powered to detect the expected effect (report the achieved power or the smallest effect the sample could detect); did the manipulation work (the manipulation check); was the measure reliable in this sample (alpha or omega); did the design leave room for the effect (ceiling or floor effects, restricted range). If the confidence interval is narrow and centered near zero, you can say the study found no evidence of an effect of practical size, and an equivalence test can formalize that claim. What the chapter must not do is recast the null as a directional prediction that was confirmed, or promote an unplanned significant result to a headline finding. Exploratory analyses are reported as exploratory, in their own paragraph, with the caveat that they require replication.

Step 5: State the theoretical implications within the limits of the design

Return to the theory named in the introduction and say what the results do to it: consistent with, inconsistent with, or extending. The constraint is causal language. A cross-sectional survey that found social comparison mediates the link between social media use and loneliness supports a mediation model; it does not demonstrate a mechanism, because the mediator and the outcome were measured at the same time. An experiment with random assignment can use causal verbs for the manipulated variable and for nothing else. Committees flag the mismatch between design and verb more than any other sentence-level error in this chapter.

Step 6: Write practical or clinical implications proportionate to the evidence

A student sample and an eight-week app do not license a clinical recommendation. Write implications as conditional and specific: the effect was of a size that, if replicated in a clinical sample, would justify a trial of the app as an adjunct in university counseling services. For applied sub-fields, name who would act on the finding (school psychologists, human resources practitioners, clinicians) and what they would do differently. One paragraph is usually enough; three paragraphs of implications from a single study read as overreach.

Step 7: Write the limitations psychology committees expect to see

The limitations section is not a list of apologies; it is an assessment of which conclusions are safe. The limitations that recur in psychology dissertations, and the way to write each one, are these.

• Sample. Undergraduate and online-panel samples are Western, educated, industrialized, rich and democratic, the WEIRD problem Henrich, Heine and Norenzayan (2010) named; state which conclusions depend on the population and which are likely to travel.
• Measurement. Self-report for both predictor and outcome invites common-method variance; a single-item or adapted scale invites unreliability. Say what you did about it and what remains.
• Design. Cross-sectional data cannot order cause and effect; a waitlist control cannot separate the intervention from attention and expectancy.
• Power. If the achieved sample fell short of the a priori calculation, report the effect the study could detect and read the null results in that light.
• Procedure. Demand characteristics, attrition that differed by condition, and attention-check exclusions in online samples all belong here if they occurred.

Each limitation should end with its consequence for a specific conclusion, not a generic call for caution. The structure of the section, and the difference between a limitation, a delimitation and an assumption, is set out in our guide to writing a research limitations section.

Step 8: Derive future research from the limitations

Every limitation implies a study. A cross-sectional mediation result implies a longitudinal or experimental test of the mediator; a WEIRD sample implies a named alternative population; an underpowered secondary hypothesis implies a preregistered replication with the sample size you now know is needed. Write three or four specific studies, each with a design, rather than a paragraph saying more research is needed.

Step 9: Close, and decide whether the conclusion is separate

End the chapter with a paragraph that answers the research question in plain language and states the contribution. Whether that paragraph is the end of the discussion or the start of a separate conclusion chapter depends on your program; the functional difference between the two, and how to avoid saying the same thing twice, is covered in our comparison of the discussion and conclusion chapters. The rest of the chapter structure, and the APA 7 reporting rules the discussion must respect, are in our guide to how to write a psychology dissertation.

The JARS elements, mapped onto the chapter

A worked opening paragraph

The present study tested whether an eight-week mindfulness app reduced anxiety in first-year undergraduates and whether baseline severity moderated the effect. Hypothesis 1 was supported: participants assigned to the app reported lower GAD-7 scores at week eight than waitlist participants after adjustment for baseline, a difference of 2.1 points, d = 0.41, 95 percent CI [0.12, 0.70], an effect comparable to brief self-help interventions in student samples and smaller than that reported in trials with clinically anxious adults. Hypothesis 2 was not supported: the condition by baseline-severity interaction was small and its confidence interval included zero, although the study was powered to detect only a medium interaction, so a smaller moderating effect cannot be ruled out. The paragraphs that follow interpret each result against the literature, consider the limitations of a single-site waitlist design, and set out the studies that would establish whether the effect holds in clinical populations.

Frequently asked questions

How long should the discussion chapter of a psychology dissertation be?

Commonly 15 to 25 percent of the total word count, which is about 1,500 to 2,500 words in a 10,000-word undergraduate dissertation and 3,000 to 5,000 in a master’s thesis. Check your program handbook, and give the discussion at least as many words as the results.

Should I repeat statistics in the discussion?

Quote each key effect once, in the sentence that states whether the hypothesis was supported, using the effect size and confidence interval rather than the full test statistic. Do not re-present tables or repeat every result from the results chapter.

How do I discuss a non-significant result?

Report the effect size and confidence interval, state the smallest effect the study could detect, check the manipulation and the reliability of the measure, and then say whether the result is evidence of no meaningful effect or simply inconclusive. Never reinterpret it as support for a prediction you did not make.

Can I discuss exploratory analyses?

Yes, in a clearly labeled paragraph that states they were not planned, that they are hypothesis-generating rather than confirmatory, and that they require replication. Keep them out of the opening paragraphs and out of the abstract.

What is the difference between limitations and future research?

Limitations say which conclusions are constrained and why; future research says what study would remove each constraint. Writing them as paired sentences keeps both sections specific.

Do qualitative psychology dissertations follow the same discussion structure?

The logic is the same, organized around themes rather than hypotheses: what each theme adds to the literature, where it agrees or disagrees with prior work, the reflexive limits of the sample and the researcher’s position, and implications. APA also publishes reporting standards for qualitative research, which your committee may expect you to follow.