The discussion section answers the “so what?” question. The results should be summarised and interpreted, but neither overly generalised nor extrapolated. This section is typically more or less a copy of the Introduction, but highly compressed, and focusing much more on the interpretations of the results and looking forward. Some useful sub sections are listed below.
- Validity Threats: Are there any reasons to question the results? How general are they? This section seemingly weakens the importance of the paper, by saying that the results can’t be trusted. However, it also shows maturity by the researchers, and that they know the boundaries for when their method would work, and when it would not.
- Lessons Learned: Sometimes the results provide unexpected surprises to the researchers. Or formulated differently, this is the place to put “previously we were wrong” stuff. This can be the most fun part to read.
- Related Work: There are normally two major groups of related work. First, you have other solutions to the same problem. Second, you have solutions to similar, but not identical, problems.
- Future Work: Most likely, the presented solution still have room for improvement, or could be further generalised. Providing these ideas here again show we are aware of the limitations, and could give other researchers a starting point and ideas for new directions.
- Conclusions: This is the place for giving concrete advice, in the light of the new results and their interpretation.
Some reviewers prefer to see one or more of these sub sections as top level sections instead. That does not really change the structure of the paper, so this is mostly a matter of taste and tradition.
Dr Dean R Hess has some further suggestions.