Your results chapter is finished. The next heading is empty, and it has been empty for three weeks. Everything before it had a template — the proposal had sections, the literature review had a protocol, the results had tables. The discussion has none of that, and it is the chapter examiners read most closely.
The cost of leaving it is not abstract. In South Africa an unfinished dissertation means another registration cycle, fees due again, and for anyone on a bursary, a funding window that does not extend to match. The chapter that stands between you and submission is usually 4 000 to 6 000 words, and it has a shape even though it does not have a template.
Start your discussion chapter in Tesify — free to begin
Why this chapter, specifically, stalls
Because it is the first chapter that asks you to make claims rather than report procedures. Up to now you have described what the literature says, what you did and what came out. The discussion asks what it means — and that is a judgement you have to make and defend, not a fact you can cite.
Two failure modes follow, and they are opposite. The first is a discussion that repeats the results chapter in sentences instead of tables: “Table 4.3 showed a mean of 3,42.” That is not a discussion, and an examiner will say so. The second is a discussion that floats free of the data entirely, making broad claims about the sector that no finding in your study supports. The structure below is designed to prevent both.
The five parts of a discussion chapter
- A short opening that restates the purpose and the headline finding — two paragraphs, no more. The reader should know your central claim before the first subheading.
- The findings, discussed one at a time, in the order of your research questions. Not in the order the software produced them, and not in order of how interesting you find them. Objective 1, then Objective 2, so the examiner can check coverage.
- Implications — theoretical, practical and, where the study warrants it, for policy. Keep these tied to specific findings rather than to the topic in general.
- Limitations, written as boundaries on the claim rather than as apologies.
- Recommendations and future research, each one traceable to a finding or a limitation you have just stated.
Whether conclusions form a sixth section or a separate Chapter 6 is a faculty convention. Check your guide, because a merged chapter submitted where a separate one is expected reads as an omission.
The four-move paragraph pattern
This is the whole engine of the chapter. Apply it to each finding in turn and the chapter writes itself at a predictable rate.
- Restate the finding in one sentence, in plain language, without statistics. “Respondents in branch offices reported markedly lower access to development opportunities than those at head office.”
- Interpret it. Why would that be so? Offer the most plausible mechanism, and say that it is an interpretation rather than a demonstrated cause.
- Place it against the literature. Name at least one study it agrees with and, where one exists, one it does not. This is the move students skip, and it is the one that turns a report into a contribution.
- Say what follows from it. One sentence on why this matters for theory, practice or the population studied.
Four moves, roughly 200 to 350 words per finding. Six findings gives you the body of the chapter. If a finding cannot fill four moves, it is a detail belonging in the results chapter, not a finding.
The literature you place findings against should already be in your review — if you are reaching for sources you have never cited, either the review has a gap or the finding is outside your framework. Our guide to writing a literature review for a South African dissertation covers the synthesis matrix that makes this retrieval quick rather than a fresh search in week eleven.
What to do when your results contradict the literature
Students treat this as a disaster. It is usually the most interesting paragraph in the chapter, provided you work through the possibilities honestly and in order:
- Is it a measurement difference? Different instrument, different scale, different wording. A construct measured differently produces different numbers legitimately, which is why the validity argument set out in our guide to proving your questionnaire is valid earns its place here.
- Is it a population difference? This is the most common and the most defensible answer for South African work. A finding established on North American or Western European samples has no obligation to replicate in a South African organisational, educational or health setting, and saying so precisely — naming what differs — is a genuine contribution rather than an excuse.
- Is it a design difference? Cross-sectional against longitudinal, self-report against observed, one sector against another.
- Is it a power or sampling artefact? A non-significant result in a small sample is not evidence of no effect. Say that plainly rather than claiming you disproved something.
- Only then: is the earlier finding wrong? Possible, occasionally correct, and the claim requires the most evidence. Do not lead with it.
Write the contradiction into a full paragraph of its own. A finding that diverges and is discussed thoroughly reads as competence; the same finding mentioned once and moved past reads as something you hoped nobody would notice.
If part of your analysis had to change because an assumption failed, this chapter is where that decision is interpreted rather than re-justified — the decision itself belongs in the methodology, as our guide to what to do when your data fails the normality test sets out.
Limitations that strengthen the chapter
A limitations section written as a list of apologies invites the examiner to agree with you. Written as boundaries on the claim, it does the opposite — it shows you know exactly what your study can and cannot support.
The pattern for each limitation is three parts: name it, say what it constrains, say what would resolve it.
The sample was drawn from one metropolitan municipality, so the findings describe an urban public-sector workforce and cannot be generalised to rural district municipalities, where staffing structures differ. A multi-site design across at least one rural district would establish whether the pattern holds.
Three or four limitations handled that way is enough. Ten is a chapter arguing against its own study. And do not list a limitation you could have fixed — “the sample was small” without a reason reads as a resourcing failure, while “recruitment was constrained to a single organisation because access was granted by one gatekeeper” is a research condition.
For qualitative studies
The four moves still apply, with the middle two adjusted. A theme is restated in one sentence, illustrated with a short quotation you have not already used in the results chapter, placed against the literature, and then interpreted. The most common error is a discussion that simply re-presents the themes: if your Chapter 5 headings are identical to your Chapter 4 headings, you have written the results twice.
Coding and theme construction belong in the earlier chapter, worked through in our guide to thematic analysis for a South African dissertation. Chapter 5 is where the themes stop being findings and start being an argument.
Write it in this order, over five sessions
- Session one: list your findings in research-question order, one line each. This is the chapter’s skeleton and takes an hour.
- Session two and three: write the four moves for each finding. Do not edit. Two or three findings per session is a realistic evening’s work.
- Session four: implications, limitations and recommendations, in that order, each traced to a finding.
- Session five: write the two-paragraph opening last, once you know what the chapter actually argues. Openings written first are always wrong.
Five sessions is a fortnight of evenings for someone in full-time work. That is the realistic distance between an empty heading and a chapter — far shorter than the three weeks you have already spent not starting it.
What the examiner is checking
At most South African universities there is no public defence; your examiners assess the document in writing, which our explainer on external examination in South Africa sets out in full. The practical consequence for this chapter is that you cannot explain yourself afterwards. Whatever the discussion says is the whole of your argument.
So check four things before you submit it. That every research question is discussed. That every claim in the chapter traces to a finding in the results. That correlational findings are described as associations rather than effects. And that nothing appears in the discussion that never appeared in the results.
Getting the chapter written
The discussion is hardest when the material it needs is scattered — the literature in one folder, the results in another file, the interpretation in your head from three months ago. Tesify holds the chapter structure and the sources you cited beside the findings you are interpreting, so a session that starts at nine at night starts on the writing rather than on the search. There is a free tier, so you can put the skeleton in tonight and decide about anything paid later.
What it will not do is decide what your findings mean. That judgement is the chapter, and it is yours.
Write your discussion chapter in Tesify
Frequently asked questions
How long should a discussion chapter be?
Commonly 4 000 to 6 000 words for a master’s dissertation, though your faculty guide governs. A more useful measure is 200 to 350 words per finding, plus implications, limitations and recommendations.
What is the difference between the results and the discussion chapter?
Results report what the data showed, without interpretation. The discussion says what it means, places it against the literature and states what follows. If a sentence contains a statistic and no interpretation, it belongs in results.
Can I repeat numbers from the results chapter?
Sparingly, and only to anchor a point. Restating a finding in plain language is better than repeating the statistic — the reader has the table, and they need your reading of it.
What do I do if my results contradict previous studies?
Work through the explanations in order: measurement differences, population differences, design differences, power and sampling, and only last the possibility that the earlier finding is wrong. For South African work, a population difference is frequently both the honest answer and a genuine contribution.
How many limitations should I list?
Three or four, each named, each with what it constrains and what would resolve it. A long list reads as an argument against your own study.
Should conclusions be a separate chapter?
That is a faculty convention rather than a rule. Check your guide, and if it is silent, ask your supervisor in writing — submitting a merged chapter where a separate one is expected reads as an omission.
Can I introduce new literature in the discussion?
A little, where a finding takes you somewhere the review did not anticipate. If you are citing many new sources, the literature review has a gap that should be filled rather than patched here.
Is it acceptable to say my hypothesis was not supported?
Yes, and it is expected. A non-supported hypothesis honestly discussed is a normal research result. What is penalised is reframing the study afterwards to look as though you predicted what you found.
Is using an AI tool for my discussion chapter against my university’s policy?
Your institution’s policy is the binding document, and South African universities differ. The safe line is the one that applies to any writing assistance: the interpretation and the judgements must be yours, and you declare assistance where your policy requires it. Check your faculty’s rules before you start, not after.
What does Tesify cost in rand?
There is a free tier you can start on without paying anything, which is enough to build the chapter skeleton and see whether it suits how you work. Check the current plans and the rand amount on the site before committing to anything paid — and never pay for a tool you have not used on your own draft.
Is my unpublished data safe in an online writing tool?
Ask that of any tool holding unpublished work. Read the privacy terms, check whether your content is used to train models, and keep raw participant data in your own analysis files rather than in a drafting tool — your ethics clearance conditions on storage apply to every copy of the data, including the convenient one.
How do I start when the page is completely blank?
Write the list of findings in research-question order, one line each, and stop there for the evening. The blank page problem is a structure problem, and a six-line skeleton solves it in under an hour.