How to Write a Discussion Chapter

The chapter that earns the highest marks, interpreting findings against the literature, explaining why, and stating implications without repeating results.

The discussion is the intellectual centre of a dissertation and where the best marks are won or lost. Its job is not to repeat the findings but to interpret them: what do they mean, how do they relate to existing knowledge, why did they occur, and what follows? Weak discussions restate results. Strong ones make sense of them.

Start by Answering the Research Questions

Open by directly addressing each research question with your key findings. This anchors the chapter and reassures the marker that the study did what it set out to do. Everything after elaborates and interprets.

Interpret Each Finding Against the Literature

For every major finding, do three things: state it, relate it to the literature from your review (does it agree, contradict, or extend prior work?), and explain why the result occurred. This is the analytic move that separates a discussion from a results chapter.

Consistent with Smith (2020), engagement predicted retention. However, unlike Smith's campus-based sample, the effect here was moderated by remote-work isolation, suggesting the mechanism is social contact rather than engagement per se.

Explain the Unexpected

Surprising or contradictory findings are opportunities, not embarrassments. Offer reasoned explanations, methodological differences, contextual factors, sample characteristics. Engaging seriously with what does not fit demonstrates exactly the critical thinking being assessed.

State the Implications

Draw out what your findings mean for theory and for practice. Theoretical implications advance understanding. Practical implications guide action. Be specific and proportionate, a single study rarely overturns a field, and overclaiming undermines credibility.

Acknowledge Limitations Honestly

Every study has limitations. Hiding them fools no experienced marker. State them precisely, explain their likely effect on the findings, and, where relevant, how future research could address them. Honest limitations build trust in the claims you do make.

What the Discussion Must Not Do

Checklist

The Move From Results to Discussion

Results say what happened. Discussion says what it means. The boundary sounds obvious and is crossed constantly, in both directions, and both crossings cost marks.

Belongs in ResultsBelongs in Discussion
The statistic, the theme, the numberWhy the result came out that way
Description of the patternComparison with prior studies
Tables and figuresWhat follows for theory or practice
Direction and magnitudeWhether the finding is surprising, and why
No citations neededCitations throughout

A quick test for any sentence: does it need a citation? If yes, it is discussion. If it can be verified only by looking at your own data, it is results.

A Reliable Paragraph Pattern

Most strong discussion paragraphs follow the same five moves. Once you see the pattern, the chapter becomes far less daunting.

  1. State the finding in one sentence, without repeating the full statistic.
  2. Relate it to the literature, saying whether it aligns, extends or contradicts.
  3. Explain why, offering a mechanism or an interpretation.
  4. Consider an alternative explanation, and say why yours is preferable.
  5. State what follows for theory, practice or the next study.

The third and fourth moves are where marks concentrate, and they are also the two most often skipped. A paragraph that reports a finding and cites a study that agrees with it has described a coincidence, not explained a result.

Handling Findings That Contradict the Literature

Students often treat a contradictory result as a failure to be minimised. It is usually the most interesting thing in the dissertation, and burying it wastes the best material available.

Work through the possible explanations in order. Is it a sample difference, such as a different population, culture, age group or setting? Is it a measurement difference, where you and the earlier study operationalised the construct differently? Is it a design difference, such as cross-sectional against longitudinal? Is it a time difference, where the phenomenon has genuinely changed since the earlier work? Or is the earlier finding simply less robust than its citation count suggests?

Then say which explanation you find most persuasive and why. That paragraph demonstrates exactly the critical judgement examiners are looking for, and it is impossible to write without having read the earlier studies properly.

Limitations That Help Rather Than Hurt

A limitations section works when it shows judgement, and fails when it either apologises or ritually lists constraints without consequence. The difference is whether you say what the limitation actually does to your claims.

WeakStrong
The sample size was small.With 48 participants the study was powered to detect medium but not small effects, so the null result for the secondary outcome should be read as inconclusive rather than negative.
Self-report data may be biased.Because engagement was self-reported, social desirability may have inflated scores, which would attenuate rather than create the observed association.
The study was conducted at one university.Findings reflect one institution with a distinctive intake, so they may transfer to comparable institutions but should not be generalised across the sector.

Notice that each strong version specifies the direction of the problem. Saying whether a bias would strengthen or weaken your observed effect is a small addition that reads as real methodological understanding.

Calibrating Your Claims

Overclaiming is the fastest way to lose credibility in a discussion. Match the verb to the design.

DesignSupportsDoes not support
Cross-sectional surveyAssociation, co-occurrenceCausation, direction of effect
Longitudinal observationalTemporal ordering, predictionCausation without controls for confounding
Randomised experimentCausal claims within the sampleGeneralisation beyond the sampled population
Qualitative interview studyMeaning, process, participant perspectivePrevalence, statistical generalisation
Case studyAnalytic generalisation to theoryStatistical generalisation to a population

Common Mistakes and Fixes

MistakeFix
Repeating the results chapterState each finding once, briefly, then interpret
New data appearing in the discussionAll data belong in results. Move it back
Literature cited but never engagedSay whether your finding agrees, extends or contradicts, and why
Causal language from correlational dataUse "associated with", "predicted", "was accompanied by"
Limitations listed with no consequenceSay what each does to the strength of your claims
Implications too grand for the evidenceScale them to your sample and design
Research questions never directly answeredAnswer each one explicitly, ideally under its own heading

Where Ethical Support Fits

The discussion is a good chapter to talk through with a supervisor, because interpretation benefits from someone testing whether your explanation is the only one available. Proofreading for clarity is fine, and hedging language is often where non-native English writers legitimately want a second read.

What must stay yours is the interpretation. Deciding what your findings mean is the intellectual centre of the dissertation, and it is the part a viva will test hardest.

Frequently Asked Questions

Should results and discussion be combined?

It varies. Many qualitative dissertations combine them so each theme is presented with its interpretation. Most quantitative ones separate them. Check your department's structure.

How long should the discussion be?

Commonly 20 to 25 per cent of the total word count, and it is usually the longest analytical chapter. Check any specified split.

Can I cite new literature in the discussion?

Yes, and often you must, particularly to explain unexpected findings. Sources that only become relevant once you know your results legitimately appear here.

What if my findings are not significant?

Interpret them. Consider power, measurement and whether the effect may genuinely be absent. A well-discussed null result is a perfectly respectable dissertation.

Where do recommendations go?

Practice recommendations usually sit at the end of the discussion or in the conclusion. Recommendations for future research follow the limitations, since each should address a specific one.

Should I discuss findings one at a time or thematically?

Organise by research question or by theme rather than by statistical test. A discussion structured around your analysis output follows the software's logic, not your argument's, and it makes the chapter far harder to read than it needs to be.

How do I avoid repeating the results chapter?

Refer to findings in words rather than restating the statistics, then move immediately to why. If a paragraph contains numbers and no citation, it has probably drifted back into results.

Your Next Step Today

Take each research question and write the one-sentence answer your data support. If any question has no clear answer, that gap is what your discussion needs to address first, and it will also be the first thing a viva panel asks about.

Trusted Sources

Chapter structures differ by department. Where this guide and your handbook differ, follow your handbook.

Related services

Related guides

All academic writing guides or browse our academic writing services.