Most research articles in journals have a standard structure with sections entitled “Introduction,” “Methods,” “Results,” and “Discussion.” Each has a clear remit except for the Discussion, which, if you’re a less experienced writer, may seem a hopelessly vague description. The occasional alternative of “Conclusion” or “General Discussion” isn’t much better.

Uncertain what’s needed, some authors offer a summary of their results, or even of the whole work, though both are really covered by the abstract.

What, then, should you include?

Basic elements
The following can all help the reader get more from your Discussion—and more from your article:

• a reminder of the research problem
• an objective review of the results
• the immediate implications of the results
• the relevance to the field of the results
• the limitations of the results
• a succinct conclusion

But not new results or methodological detail, which would be out of place here.

Research problem
Opening with a reminder of the research problem sets the context for the Discussion and accommodates the different ways readers approach a research article, as explained shortly. But don’t repeat the wording from the Introduction. You’ve made progress since then, and it should be reflected in the way the problem is described.

Here’s an example from a civil-engineering article. First, in the Introduction, a simple statement:

Chloride-induced rebar corrosion is one of the major forms of environmental attack [on] reinforced concrete.

And then, in the Discussion (actually the Conclusion in this example), an elaboration:

Chloride ingress into concrete is a complex process, which in many environments is further complicated by the temperature cycles and wet-dry cycles experienced by the reinforced concrete structures. While there are numerous existing experimental or modeling studies …

This opening makes sense whether the reader works linearly through the report, jumps straight to the Discussion to decide if it’s worth reading the earlier sections, or moves from one section to another according to their interests at the time.

Review and implications
Examine the Results section as if you were a neutral observer. Focus on the main findings, noting strengths and weaknesses, and any immediate implications. You don’t need to be exhaustive: some results will already have been routinely processed, for example, in control measurements; others may have been placed in an appendix or supplement, as required by the journal.

In considering the implications, strike a balance between claiming too much, which looks like boosterism (and alienates readers), and overprotecting what you do claim with too many instances of “may,” “might,” “can,” “could,” and so on, leaving readers with little sense of substance.

Depending on your research discipline, you may have applied statistical tests to decide the reliability of your results. If so, give the same prominence to the tests that didn’t reach statistical significance as those that did, not least to counter the historical bias in the literature.

You may also have listed specific research objectives in the Introduction. If so, identify those that were realized and explain any that weren’t. That knowledge could encourage others to build on your work.

Relevance to the literature
Along with any comparisons you’ve already made between your results and those of other researchers, it’s useful to stand back and relate your findings to the field as a whole. Here’s an example from a cell biology article published in 2012:

Regional clusters of mutations in cancer have occasionally been observed in experimental models, although not at the mutation density observed here (Wang et al., 2007). … Furthermore, they are closely associated with regions of rearrangement and occur on the same chromosome and chromosomal strand over long genomic distances, suggesting that they occur simultaneously or … over a short time span (Chen et al., 2011).

By making connections of this kind, you recognize the priority of other studies, albeit sometimes with caveats, but also show how your own work advances those studies in a coherent way.

Limitations
Having pointed to the strengths of your work, you should also note its limitations and possible failings. Make it clear when the reader can safely apply your findings—and when they can’t. Being open about their applicability increases the authority of your work and its potential impact.

When deciding what limitations to mention, concentrate less on the obvious ones, the finite sample size, say, which the reader already knows from the Methods, and more on the subtle ones, for example factors you couldn’t control for some alternative explanation.

In practice, it may not be enough to state the limitations for the reader to appreciate their consequences. Here’s an example from a computer-science article, edited to reduce its identifiability:

We specified default values for the tools based on pretesting. It is possible that different values for these tools could affect the results.

It would help to know, for example, how representative the default values are and what happens when other values are used instead. In this way, your findings might be applied in areas you hadn’t considered.

Concluding statement
The closing paragraph of the Discussion—or a separate short Conclusion—should complement the opening of the Introduction. Keep it short, simple, and relevant. You could mention the new understanding that’s been achieved or the discovery of longer-term implications or new research directions.

There’s no need, however, to share your own research plans. They’re important to you but not necessarily to other researchers, who might misinterpret them as signalling your claim to the research area.

Finally, if you can, end with a brief take-home message, something to remember your findings by. But not “More research is needed”—which may have the opposite effect.

Photo by Unseen Studio on Unsplash

David H. Foster is Professor of Vision Systems at the University of Manchester and formerly Director of Research in the School of Electrical & Electronic Engineering. He has served as journal editor or editor-in-chief for over thirty years. He is the author of A Concise Guide to Communication in Science and Engineering.