How to write discussion in research paper: with an example

In most research papers, the results section reports what you found and the discussion explains what it means. The results are about your data. The discussion connects your data back to your research question, to other studies and to the world outside your study.

It is also the section most likely to go wrong. In a 1999 BMJ editorial, Michael Docherty and Richard Smith called the discussion "often the weakest part of the paper where careful explanation gives way to polemic". They wrote that authors commonly "go beyond the evidence they have gathered and draw unjustified conclusions". Farhan's first draft did exactly that.

Some journals and many conference papers combine the two into one "Results and discussion" section, and some theses put a conclusion inside the discussion chapter. Follow your journal's instructions or your university's thesis format. The thinking stays the same either way. For where the discussion sits among the other parts, see the structure of a research paper.

There is no single required order, but a structure helps you avoid rambling. Docherty and Smith suggested five parts for the discussion, and many guides still use a version of them:

1. Statement of principal findings.Your main answer to the research question, in a sentence or two.
2. Strengths and weaknesses of the study.What your design did well and where it falls short.
3. Comparison with other studies.Where your results agree or disagree with earlier work, discussing particularly any differences.
4. Meaning of the study.Possible reasons for your findings, and what they could mean for practice or policy.
5. Unanswered questions and future research.What your study could not settle, and the next study that could.

Many papers put the comparison with other studies before strengths and weaknesses, and that is fine. What matters is that each paragraph has one job and the reader can tell which job it is doing.

Docherty and Smith advised that the discussion should begin with "a restatement of the principal finding. Ideally, this should be no more than one sentence." This is the sentence the reader came for, so make it plain.

Do not repeat your results. The reader has just seen your tables. A good opening gives the answer in words, with at most one key number.

Notice what the rewrite does. It says who was studied and where, it uses the words "more common among", which describe a link and not a cause, and it does not claim anything about all of India. It is less dramatic and much easier to defend at a viva.

If your study had a hypothesis, say here whether the findings supported it. If you are not sure how a hypothesis and its result should line up, read how to write a hypothesis.

Next, place your findings next to what others have found. Go back to the papers you cited in your literature review. For each main finding, ask three questions:

1. Which earlier studies found something similar?
2. Which found something different?
3. What might explain the difference: the place, the people, the method, the year, the way something was measured?

Differences deserve more space than agreements. Docherty and Smith asked authors to discuss "particularly any differences in results". Farhan's link between phone use and poor sleep was weaker than in two of the studies he had cited. He noticed that both had surveyed school students, while his were in college and many had late classes. He wrote that as a possible reason, not as a fact.

Only compare with studies you have actually read. A finding you cite second-hand from someone else's discussion may not say what you think it says.

Every study has limitations, and examiners know it. What they want to see is that you know yours, and that you can say how much they matter. A limitation you name yourself reads as care. One the examiner has to point out reads as a gap.

Farhan listed his honestly:

• Design. A cross-sectional survey measures everything at once, so it cannot show whether phone use led to poor sleep or poor sleep led to phone use.
• Measurement. Phone use and sleep were both self-reported, and students may remember them inaccurately.
• Setting. Three hostels of one university may not represent students elsewhere.

For each limitation, say which way it might push the result, and if you can, how much. Then mention real strengths too, such as a good response rate or an established questionnaire, without boasting.

Do not end with a long apology. Docherty and Smith also warned against inflating "the importance and generalisability of the findings". Limitations are where you set the honest boundary around your claim, so the reader knows exactly how far it reaches. The limits of your design come from your methods chapter; see how to write a research methodology.

Now say what your findings could mean, for practice, policy or theory. This is where Farhan's first draft asked for a government ban. Docherty and Smith described this part of the paper as the place "where authors often run amok", and advised authors to "leave readers to make up their own minds on meaning".

Keep your implications the same size as your evidence. Farhan rewrote his as a modest suggestion: hostel wardens and student counsellors could include phone use at bedtime in sleep awareness sessions, and the finding was worth testing in a larger study.

The last line on the left appears in thousands of papers and tells the reader nothing. Name the exact question still open, and the kind of study that could answer it.

Students often mix up the discussion and the conclusion. Editage Insights puts the difference simply: the discussion is "a detailed interpretation of the data reported in the Results section", while the conclusion gives "a snapshot of your study implications" without going back into the details.

Whether the conclusion is a separate section or the last part of the discussion depends on the journal or your university. Neither should contain new data.

These are the mistakes Farhan's first draft made, and the ones reviewers mention most:

• repeating the results, number by number;
• bringing in new results that were not in the results section;
• claiming cause from a design that can only show a link;
• comparing with other studies by listing them, not by explaining the differences;
• hiding limitations, or listing them with no comment on how much they matter;
• recommendations far bigger than the study, and an empty "further research is needed".

Farhan's second discussion was four pages long, a page shorter than the first. At his viva the external examiner asked whether phones caused his students' poor sleep. Farhan said his study could not tell, explained why, and described the study that could. The examiner wrote something down and smiled.