At first impression it’s much harder to spot general formulas for the discussion part of your manuscript, paper, project report or dissertation compared with other sections. Discussions vary so much between different projects (at least at first impression). Because of this apparent lack of consistency, discussions are often an area where new writers struggle. However, good news! This page will discuss some concepts that you can put into practice to make your writing easier. I will give some tips for what is worth including, what not to include, and how to get the balance right. Near the end there is also a section on differences between PhD theses / dissertations and manuscript/project reports.
Before we get into the do’s and don’t, it’s probably worth thinking about why discussions exist…
What is the point of a discussion?
The primary purpose of the discussion is to give readers specific guidance about what was accomplished in the study, the scientific/biological/medical/engineering significance. It is your chance to shape the thinking of your readers and to protect yourself from criticism for any limitations your work might have.
Remember, you (should) know more about your subject than the majority of your readers. This includes knowing more of the relevant surrounding and preceding literature. Therefore, the discussion also serves another important role; putting your findings into context with what has gone before.
Important – the discussion is not an extension of the results section.
Discussions Do’s and Don’ts
- Keep it focused on the core message
- Clearly state if your conclusions support your hypotheses
- Explain unexpected results
- Explain reasons for observed differences compared with previous findings
- Consider other ways to interpret your data
- Explain what is new (without exaggerating)
- Try to take a holistic view as far as possible
- Don’t repeat your results!!!
- Don’t go on too long – keep it focused
- Don’t make it too short – 🧐 address the things you need to
- Don’t overreach.
- Don’t introduce new terms or deep ideas.
- Don’t focus too much on limitations
- Don’t include future directions that you could have easily done!
- Don’t ignore controversies!
- Don’t discuss each experimental finding as an independent result
I’ll expand on some of these points below.
Synopsis (paragraph 1)
This discussion is the most flexible section of a manuscript or project report. Length and structure should be driven by your story. However, below is a guideline layout and some examples of the different types of paragraph from published work. Be aware that some of the following sections might connect better if rearranged or combined in different ways.
Summarise the goal of the study and how your data addresses that goal. I know this sounds like I am saying to repeat your results but what I really mean is to combine your key findings into a simple message. Make this short. A two or three sentence paragraph should do it. Assemble the phrasing to emphasise the points that you will delve deeper into the following paragraphs.
If your study is quite simple you might not need this. However, if you are delivering 7 figures of 5+ panels each this first paragraph can help to bring it all together again, reminding the reader how the pieces of the story connect and setting up your big hitting paras.
If you hit 150 words for this you have likely gone too far. Stop and parse it back. Focus focus focus.
Try to finish this paragraph (and the rest) with forward looking wrap sentences that leads your on to the next step. Now, let’s move onto delivering the big stuff (see what I did there).
Example synopsis paragraph:
The current study assessed the consequences of ACTN4 knockdown on skin cell motility and the organisation of keratinocyte matrix adhesion structures. Knockdown of ACTN4 lead to perturbation in the directionality of keratinocyte motility, resulted in an increase in focal contact surface area, and multifaceted changes in hemidesmosome-like protein complexes. Specifically, in ACTN4-knockdown cells, α6β4 integrin was mislocalised from the base to the main body of lamellipodia, β4 integrin dynamics were reduced, and there was reduced recruitment of the BP antigens to α6β4 integrin clusters.
Discuss your biggest finding (paragraph 2)
Discuss your secondary findings (paragraphs 3, 4)
Focus the next 2-3 paragraphs on the major findings. Choose a logical order to deliver these; usually the most important or impactful finding first followed by secondary findings. Consider the same things when deciding how long to make each One key point per paragraph, fully supported.
The “biggest finding” will usually be the one that you highlighted in your title, the answer to your primary hypothesis, or your main objective.
This part of your discussion is where you should deal with whether or not your data supports or refutes your hypotheses and then discuss the reasons why. Try to take a holistic approach rather than talking about individual experiments in isolation. Consider the whole data set together, especially if a single piece of data doesn’t fit with the rest.
When you are at the planning stage, focus your energies on identifying what points these paragraphs will deliver. Write down the key messages as accessible short sentence. Use those sentence as your lead in, your topic sentence, for each paragraph. Thereafter make a conscious effort to stay on message. This should help you to focus upon the important things, the things you want your reader to take away, rather than rambling on.
You will need to set up your discussion points in your introduction, you don’t want to be surprising the reader with something completely new here (intro guide). Often it is best to write/plan the discussion and introduction together.
Don’t refer to data figures in your discussion
You’ve already described your data, you don’t need to reference the figures again. If you find that you are restating your findings just using different phrasing you should stop and ask yourself why you are doing this! If the reason is because you didn’t write your results clearly enough the first time then you should go back and fix that rather than trying to make up for poor delivery in the discussion (need some tips of results writing? comprehensive guide here, or short version here).
The discussion is for putting the data into a wider context. You might want to consider bringing in your updated model/diagram at this point. Make sure the diagram helps though, and that you don’t over interpret the inferences you can make from your data.
Don’t ignore controversial findings
One thing for sure; your reviewer won’t appreciate it if you have ignored their seminal work in the field!
Whether your data fits or disagrees with others, you should be prepared to cite and discuss. Consider why differences may have occurred: What have you added/changed? Why is your study better? Or even what other studies have been published in between that, combined with yours, leads you to a different interpretation. Think about how your findings compare with studies of related proteins/drugs or in other tissues/disease?
These key finding paragraphs will inevitably have lots of primary data references. You might find that you need to them to be a little longer than the standard ~200 word paragraph you might find in the rest of your writing. Be comprehensive but still stay on message (don’t go crazy though – I’m still a little uncomfortable about the length of my example below!).
Often papers get criticised for not being “novel”. Therefore, the middle paragraph should definitely emphasise what aspects of your findings are new. Don’t overstate the novelty though; that 1980s paper that did the same sort of work with earlier technology may have already established the paradigm. Your study may well be an improvement, it might add mechanism or important validation, but don’t try and sell it as bigger step forward than it really is. Also a big shout out for validation studies. A study doesn’t have to be completely novel to have value.
Example key finding paragraph:
A focus of the current study was to assess the role of Col XVII in migration and compare the motility phenotype of keratinocytes with a knockdown in Col XVII expression against that of BPAG1e- and β4 integrin-deficient cells. The data reported here revealed that Col XVII-knockdown cells exhibit the same motility and lamellipodial defects as the latter cells, although the underlying mechanism is subtly different. In this regard, the role of Col XVII in migration is somewhat controversial. Evidence has been presented that the extracellular domain of Col XVII impairs keratinocytes ability to close scratch wounds (46). Conversely, the shed ectodomain of Col XVII has been demonstrated to promote squamous cell carcinoma cell transmigration (49, 53). At the cell level, one group has shown that GABEB cells lacking apparent Col XVII expression exhibit a migratory defect whereas another study indicated that Col XVII knockdown inhibits migration via reduced p38 MAPK signalling (11, 22). The data here are consistent with the latter report. They indicate that Col XVII loss induces motility and signalling defects in keratinocytes. Intriguingly, these new data also provide key insight into the mechanism via which Col XVII might regulate signalling to p38 MAPK. It likely does so by mediating both the localization and activation of Rac, a critical regulator of the p38 MAPK pathway (47).
Limitations and alternative interpretations (paragraph 5)
A problem I have encountered when grading student papers is that the descriptions of the limitations section are overly long and overly negative. Don’t go on and on here. Similarly don’t be too superficial. Your marker knows you had a time limit, they know you might have limited numbers of biological repeats, they certainly know that the data might be more reliable if gathered by experienced researchers (I’ve seen this included as a limitation on a student paper a lot more often than you would expect). These superficial and obvious problems don’t add any valuable insight, therefore, if you feel it’s necessary to include, do it in a way that is short and focused. “Sample size limitations preclude robust statistical inferences” (better still, indicate what sample size you would actually need – calculator here).
To avoid problems, aim for a balanced approach that emphasises the strengths of your study while still acknowledging where any limitations will mean that you can’t be as confident. To be more positive about your findings, try to phrase your sentences so that they focus on what you can interpret rather than what you can’t.
Compare the following:
- Negative version: In this study, traction force generation was not measured, therefore it can not be ruled out that the observed wound closure defects were due to a failure in force generation
- Positive version: Although it is not yet possible to rule out the contribution of force generation to the observed phenotype, the obtained data strongly support a mechanism where the measured changes in proliferation are a primary determinant of the wound closure defects
The versions above deliver the same point but with different tone; the positive version acknowledges the limitation but takes ownership of the findings and is assertive.
Your limitations parts should not be longer than your main discussion points!
The balance comment extends to how much space you devote to limitations. If your limitations extend to much more than your major findings then you will leave the reader with the impression that the major take away from the study is that your experiments weren’t very good. Not much chance of convincing an editor to publish if you even you aren’t convinced the findings are strong enough to be believable.
What I like to do here is describe the different ways you could interpret your data based on the data I have. Rather than present these alternate interpretations as a problem, I detail how one could differentiate between those interpretations in the future. Present arguments for why you are willing to accept your limitations can also help reduce their negative impact: “To resolve this dilemma would require 100 extra human samples, which is unethical to request at this time.”
Be careful! If a limitation or alternate interpretation could have been addressed easily then a reviewer can, will, and should, get you to do the experiment!
Example paragraph discussing limitations:
The participants in this study were drawn from University students and staff members and therefore there is a selection bias toward well-educated participants. Moreover, the participant demographics reflected the local University population and were skewed toward lighter skin types. This has two important implications for the generalizability of the findings. First; one would predict that this cohort would have higher baseline cancer-risk awareness levels and as such are likely to be vigilant when applying sun protection [43,46]. This may suggest that the disproportionately poor coverage of the eyelid region could be even more striking in the general population. It should be noted, that in the questionnaire just under half of the respondents did not cite a specific reason for missing the eyelids so, despite a perceived general awareness, their application was still relatively poor. Second; this cohort may respond more to the information-based intervention strategy used here than a broader cross-section of the public would, therefore we may be over estimating the magnitude of the improvement. However, information-based intervention is a standard and proven efficacious approach in wide variety of contexts in diverse groups of patients/participants and as such we do not believe this to be a major limitation here [47,48,49].
Wider biological implications (paragraph 6).
As you progress through your discussion you should get wider and wider with what your data means (i.e. the mirror image to your introduction). You should end by putting your findings into real-world context.
This is the place where you can talk about the next big question(s). A specific agenda for future research based on the questions generated by the current findings is much more valuable to a reader than vague suggestions. I often try to integrate these phrases into my limitations section, e.g. stressing the value of making an animal models to counter a limitation of absence of immune cells in a culture model.
As with limitations, don’t go too far here. A long list of what needs to be done next could leave a reader thinking “what have they actually done!” Reviewers also appreciate how specific research recommendations will advance the field. For biological research this can include the wider implications for clinical needs. Can it open an avenue toward therapeutic development?
Example paragraph combining biological implications, next step and conclusions.
The critical interactions through which the N-terminus of laminin α3a might mediate mesenchymal cell signalling are as yet unknown. Possibilities include a direct interaction with a cell surface receptor expressed by mesenchymal cells or alternatively the domain might interact with other extracellular matrix molecules and the resultant macromolecular complex may influence cell behaviour. Granulation tissue is a major clinical problem, not only in LOC syndrome but in a range of much more common conditions such as rheumatoid arthritis, chronic venous leg ulcers, transplant surgery and others. A fuller understanding of the role of laminin α3a in this sometimes undesirable biological process might lead to the identification of novel targets for therapeutic intervention.
Conclusions (paragraph 7)
Round it out. Tie it up with a neat little bow.
In positive language, describe your core finding and what it means to the big picture. Refer to your equivalent section of the intro to make sure you are covering the correct points.
Likely this short paragraph will be quite similar to the last line of your abstract (or vice versa) and will pick up on one or more of the bigger reasons for your research you covered in the first paragraph of your intro. In some journals, the conclusion is viewed as being so important that gets its own separate section. So, nail it. It’s the last chance to influence your reviewer’s and reader’s opinions.
PhD thesis discussions
Whereas a manuscript or project report thesis should only be the length required to deliver the points that you need, your thesis examiner will expect you to be comprehensive in your doctoral work.. That’s not to say it should be long for the sake of being long! Although the freedom to express yourself at length is fundamentally a good thing, each individual paragraph and each individual sentence should still be succinct and focused.
Despite the option for increased length, your discussion still should only discuss things that are relevant. However, there is no real reason for you not to address anything that is relevant and also to go deeper into concepts and raised. This is especially true what discussing what else could be done. Indeed, as all good science generates questions I would expect to see a more developed next steps section (go bold; cost, time, sample availability are not an issue here, just the quality of the questions!).
Can you go wider and still stay relevant? Well yes, you probably can. A common problem in thesis writing is that you have got so far into your niche that you struggle to get back out again. In your discussion you can go wider; talk about other model systems, talk about different tissues, talk about the Mongol invasion of the Punjab in the 1300s*. Well, maybe not the last one, but your discussion can be a chance to have fun and really explore all aspects of your work and not just the stuff that was directly related to the core hypotheses.
To be fair, I’ve just included depth for completeness! Your manuscript discussion will be deep too so this isn’t really a difference. Make sure whatever you cover, in whatever writing format, is deep enough to be useful; superficial coverage of a comment or assertion won’t add value to your work.
A note of figure references; above I said not to refer to figures in your discussion and that generally holds true for a thesis too, despite the longer format. In your thesis you might choose to have chapter discussions, essentially the same as a manuscript style, and then perhaps a more complete discussion which discusses the combined findings. As a concession to the longer format, I would be relatively happy to see backward references to specific chapters or collections of subsections. If you do decide to include these references, make sure to focus the sentence on the message rather than where the info is (ref the subsection in brackets). You don’t want your reader to feel like they should flip back through ~100 pages to find something that you should have explained fully the first time!
As you progress through your doctoral studies every person you ever meet will tell you that they wished they started writing earlier. Mostly, people think that it is only the results, methods and literature review that can be covered as you go. However, that’s not totally true. You can chip away at paragraphs within your discussion as you go. Each time you read a key paper think about how it connects with what you are investigating. Your study might add or might challenge these findings. As you progress to later in your studies you will have a better idea of which way the work is going and so your notes should start containing relevant details about which ways your approaches differ and why this can explain what is going on.
*My PhD was on laminin alpha3; specifically on a rare genetic skin disease (LOCS) and a new family of laminin splice isoforms (LaNts). However, the Mongol invasion of the Punjab did end up in my discussion. It was relevant but somewhat random in the grand scheme of things! We had a fun time discussing it during my viva. It continues to amuse me that that bit made it in and I wish we published it somewhere.* EDIT; I got asked about this part of my thesis, so have posted it here as a blog post.*
Or should that be “conclusion”?
A discussion in a manuscript or project should be the length in needs to be!
I’ve seen 3 paragraphs that worked really well and 11 paragraphs that really got after it and didn’t feel long, and those were both from papers from the same author. So don’t force it; use the style that works for your story. If something needs discussed, discuss it. If things stand alone, are complete and clear in their own right, then let them lie.
My biggest tip for discussions, as with other sections, is to get a draft down without worrying too much about content and then edit edit edit. Make sure everything works for you. Don’t be afraid to play with order or to remove things that aren’t adding value.
Like the rest of science writing it gets easier with time so Don’t Panic.
My biggest tip for discussions, as with other sections, is to set out a framework of topic sentences and then get a draft down without worrying too much about writing the perfect sentence. Once you have something in place, it is easier to edit, edit, edit until the message is clear and compelling. Some things to check:
Have you addressed the hypothesis
If you set out to test a hypothesis, then your discussion should come to a conclusion about the outcome from your tests. Either the data supported or refuted your hypothesis. This should be written in a clear and understandable way.
Avoid discussing single experiments on their own, in order
Your discussion should be looking at the cumulative knowledge throughout. When you are editing, check that any time you mention a specific experiment you are doing so in a way that highlights its contribution to the combined story. Remove all references to primary data figures.
Make sure it’s not too narrow
I find that new writers often spend too much time focused only upon their findings, failing to connect those new data with the wider world. The only place you can be self-centred is in the synopsis and conclusion. During the editing stage, make sure you have added enough connections between your work and the literature that everything is connected.
Check that you aren’t being overly speculative
A little speculation is good, but you can go too far. When you are editing, look at the balance of what you have written and try to avoid over-interpreting what you have found. Try to avoid including things just to show off your knowledge; every point should add value.
Watch that you haven’t identified too many unknowns
If more than half of your discussion is saying “we don’t know this” in a variety of different ways, then the overall feel to the reader will be that the study is incomplete. It isn’t surprising to have unknowns, good research often raises more questions than answers, but do you really need to discuss them all? My advice in editing is to reduce the unknowns into a single paragraph and set yourself a rule that whenever you say “we don’t know” you also include some statement of how you would address the gap in the knowledge.
Check that you haven’t added new information
Make sure that your introduction and discussion are connected. You shouldn’t be bringing in completely new material into the discussion. It might be appropriate to fix this issue by editing the introduction rather than the discussion.
Links and downloads
Download this page as a pdf – [the chapter from my “life in the lab” book.
This page is part of a series, click the buttons below for more. There is also more good stuff coming, so follow the blog to keep up to date.
Found this useful? You’ll love this book: