Examiners read the discussion to find out whether you understood your own study. It is the one chapter where you are allowed to think out loud, and it is the chapter most often sent back, because it is written as a second results chapter with the numbers spelled out in words.
What the discussion is for
Results answer what happened. The discussion answers five different questions: what your main finding means, whether anybody else found the same thing, why it might be so, what could have distorted it, and what should happen next. Those five questions are the chapter. If a paragraph answers none of them, it belongs in the results, the introduction or the review of literature.
Open with the principal finding, once
The first paragraph states the single most important thing you found, in words, without the table. Give direction and magnitude rather than a p value: mean fasting glucose fell by this much more in one arm than the other, or the proportion with the complication was this against that. Then one sentence on what that means for a patient or a clinician. Do not begin with a restatement of the aims, and do not begin with a definition of the disease. Your examiner has read the first chapter. Opening a discussion with throat-clearing wastes the strongest position in it.
Put your finding next to the literature, not after it
The commonest weak structure is a run of paragraphs, each summarising one previous study, with your own result nowhere in sight until the end. Invert it. Lead every paragraph with your own finding, then bring in the studies that agree and the studies that do not. The unit of the discussion is your result, and published work is the evidence you argue with, not the subject of the chapter.
Disagreement is more useful than agreement, and it is where the chapter earns its marks. When your result differs from a published one, give a reason you can defend in the viva: a different population, a different definition of the outcome, a different assay or cut-off, a different duration of follow-up, a hospital-based sample against a community one, a different era of treatment. The sentence that a difference may be due to methodological differences is not a reason. It is a way of avoiding one, and an examiner reads it as exactly that.
Offer a mechanism, and mark it as a proposal
After the comparison, say why your result is biologically or clinically plausible. One paragraph is enough, and the language must stay conditional: may, could, one explanation is. An observational design cannot establish causation, so a discussion that starts by reporting an association and ends by asserting an effect has quietly overstated the study. That slide is easy for an examiner to spot and hard to defend once it is on the page.
Limitations, written without dismantling your own work
Two opposite failures are common. The first is the one-line dismissal, usually that the only limitation was a small sample size, which tells the examiner you have not thought about bias at all. The second is the confession, three pages of everything that could conceivably be wrong, which leaves the reader unsure why the conclusion should be believed.
The form that works is three parts, one sentence each: name the limitation, say in which direction it could have pushed your result, and say what you did to contain it or why it does not overturn the conclusion. Written that way, a limitation demonstrates control of the method rather than weakness in it.
- Single centre, so the findings describe patients reaching a tertiary referral hospital and may overstate severity; the setting is therefore reported in full and the conclusion is stated for that setting.
- The sample size was calculated for the primary outcome, so the secondary comparisons are exploratory and are reported as such rather than as findings.
- The outcome was assessed by the treating clinician, which risks observer bias; a fixed written definition and a single proforma were used to limit it.
- Follow-up ended at six weeks, which cannot capture late recurrence, and no recurrence rate is claimed for that reason.
Never write that a limitation does not affect the results. If it truly does not, it is not a limitation. Say how much it could affect them, and in which direction.
End with what follows
Two short paragraphs close the chapter. The first says what the study adds that was not already known, in one sentence, without inflating it. The second says what the next study should do differently, specifically enough that another resident could take it up as a topic: a multicentre sample, a longer follow-up window, a randomised allocation, an objective endpoint in place of a clinical impression. The phrase further studies are needed, unattached to any detail of what those studies should change, adds nothing and is noticed.
A test for repetition you can run in twenty minutes
Print the results and the discussion and lay them side by side. Mark every number that appears in both. A number earns its place in the discussion only if the sentence around it interprets or compares it. If the sentence could be moved back into the results chapter without losing anything, move it. This one pass cuts a first draft noticeably, and what remains reads as argument rather than recital.
Length and proportion
Most departments do not prescribe a length for the discussion, but some do, so confirm with yours before you plan it. As a working proportion, a discussion usually runs a little longer than the results and shorter than the review of literature. If it has become the longest chapter in the thesis, it is probably reviewing the literature a second time.
Before you hand the draft to your guide
- The first paragraph states one finding, with direction and magnitude, and no aims.
- Every comparison paragraph starts with your result, not with somebody else's study.
- Each disagreement with published work has a named, concrete reason.
- Causal verbs appear only if the design supports them.
- Limitations are specific, directional and contained, and none is described as not affecting the results.
- The closing recommendation names what should change in the next study.
- Every reference cited in the discussion is one you have actually read.