To be sure: advice for writing discussions

When we write, we hopefully have a point we want to make. Brian has called on us to view ourselves as story tellers when writing manuscripts, embracing

the art of story-telling that knows where it is going and does it crisply so that it sucks us in and carries us along with just the right amount of time spent on details of character and setting. Where the characters (questions), the plot (story arc), the setting, the theme (the one sentence take home message) all work together to make a cohesive whole that is greater than the sum of the parts

In doing so, Brian says:

Every word, every sentence, every paragraph, every section of the paper should be working together, like a well-synchronized team of rowers all pulling towards one common goal. The introduction should introduce the questions in a way that gives them emotional pull and leaves us desperate to know the answer. The methods and results should be a page-turning path towards the answer. And the discussion should be your chance to remind the reader of the story arc you have taken them on and draw sweeping conclusions from it. Any freeloading sentence or paragraph that pulls in a different direction should be mercilessly jettisoned (or at least pushed to supplemental material).

In this post, I am going to disagree with Brian’s last point (gasp! blogging drama!), but, in doing so, I am motivated by the same goal. When trying to make a convincing argument, it can help to address the most obvious concern or counterargument. As you are leading the reader towards your exciting, sweeping conclusion, you don’t want some part of their brain thinking “Well, I guess they are unaware of this thing that sure seems like a problem for their argument.” If it’s something that a reasonably well-informed reader might be wondering about or distracted by, you should consider directly addressing it in the discussion. (This is also important in terms of not over-selling your results.)

When I did training with the OpEd Project a few years ago, we discussed having a “To be sure” paragraph, often as the second-to-last one in the piece, that articulates the potential opposition to the point we were trying to make and countering it. This advice is not unique to the OpEd Project. For example, this resource on writing effective op-eds from Duke University includes:

Acknowledge the other side. People writing op-ed articles sometimes make the mistake of piling on one reason after another why they’re right and their opponents are wrong… They’d probably appear more credible, and almost certainly more humble and appealing, if they took a moment to acknowledge the ways in which their opponents are right. When you see experienced op-ed authors saying “to be sure,” that’s what they’re doing.

For the OpEd Project, they recommended a few possible strategies that could be used in the To Be Sure paragraph. You can acknowledge a potential concern and then dismiss it (e.g., explaining why it’s wrong or doesn’t apply). You can validate the concern but then trump it (e.g., by arguing that the thing you are arguing for is more important). Or, if you think someone might go for a personal attack (say, you can’t be unbiased about X because you are Y), you can include that as a personal caveat, taking the wind out of their sails.

This is a strategy I had already employed (though not always fully aware of what I was doing), and it’s one I’ve paid much more attention to since spending some time writing op-eds. So, I was amused when someone who I view as a really good writer recently said his template for a discussion is:

  • First overview paragraph (one of Brian’s Five Pivotal Paragraphs)
  • Three paragraphs, each with one specific point
  • A caveats paragraph, possibly framed as future directions – what are three things that could be done next to follow up on this work?
  • Cosmic take homes (another one of Brian’s Five Pivotal Paragraphs)

In other words, he recommends a “To be sure” paragraph, in exactly the same place that the OpEd Project recommended. He noted that one advantage of the future directions-oriented version of the “to be sure” paragraph is that it raises issues that are worth considering without handing the person a hammer to whack you with.

Something I’ve noticed over the years is that many early career writers try to have the “to be sure” stuff as the main focus of the discussion. Stephen Heard has apparently noticed this, too. In the section entitled “Consider possible weaknesses” in the chapter on Discussions in his guide to scientific writing, he writes:

Early-career writers sometimes greatly overdo this Discussion component, perhaps because we emphasize it when we teach undergraduates to write lab reports. If you leave your readers with the impression that your results don’t matter much, they will feel they’ve wasted their time reading the manuscript.

In my experience, early career writers not only include too much of this, but put it too early – including having this as the first paragraph of the discussion. So, I think the op-ed advice to include a “To be sure” paragraph as one paragraph that is the penultimate one in the manuscript is good advice for scientific manuscripts, too.

Do you include “To be sure” paragraphs in your writing? What framing do you find works best for it? Do you notice and appreciate them in other people’s writing? How do you balance between not overselling your work and not leaving your reader feeling they’ve wasted time reading your manuscript? I’d love to hear more about what people think works (and doesn’t!) for acknowledging potential limitations and weaknesses of the work!

22 thoughts on “To be sure: advice for writing discussions

  1. Good post. I really struggle with this, both in papers and blog posts. If anything, over the years I’ve become less sure when I should anticipate and address possible criticisms in my papers: Looking forward to learning from others’ comments on this.

    In blog posts, I’ve increasingly been writing defensively, anticipating whataboutism and other objections and misunderstandings from a certain segment of ecology twitter. But I’m not sure why I bother trying to anticipate and address all possible criticisms. For many posts it’s probably not necessary. Either because I’m anticipating criticisms nobody would actually make, or because the very few people who would make those criticisms aren’t likely to see the post. And even if I do correctly anticipate the criticisms a few people on Twitter will have, defensive writing is probably pointless. Readers who want to engage in whataboutism and other groundless objections are going to do so anyway, no matter what I say. Especially if they don’t even bother reading the post first, which as best I can tell they often don’t. Finally, writing defensively is tiresome and annoying for me; it makes the writing process less fun for me. So what’s the point in burdening my posts with a lot of defensive writing? Unless maybe it makes me seem careful and thoughtful to readers who wouldn’t themselves engage in whataboutism?

  2. Nice post, as always. Our lab calls it the “caveats paragraph”, and it arrives at about the 2/3rds mark in the Discussion. Too much earlier and you take the steam out of your conclusions. Too much later and the reader with that caveat in mind is getting antsy. In the caveat(s) paragraph you either attempt to defuse the caveat (yay!) or frankly admit this is a weakness that can be addressed in future work (less…yay!).

    One more thing about the Discussion re Brian’s 3 “specific points” paragraphs. Often, such paragraphs are in the form of an argument: (“This result, in the context of other work, suggests this interpretation.”). An ideal summary sentence to that paragraph is a prediction deriving from the argument (e.g., “If true, then we predict that adding X to the system, and vigorously stirring, will generate whirled peas.”)

  3. I agree with the themes. A caveat paragraph in the discussion is usually a good idea.

    But more often then not, I see them done poorly. They should not
    1) Run on for multiple paragraphs taking up most of the discussion. One paragraph (or sometimes two short ones is always enough.
    2) Never be the first or last paragraph of the discussion – the consensus here seems to be about 2/3 of the way through which I agree with. But never never as the last paragraph! Please!
    3) Use it just to list everything you could have done but didn’t. The main goal is to head off reviewers concerns, and if you can phrase that as a future direction with a tack on of why it was not essential great.

    It overall seems like there is a lot of agreement about this. The only question is Jeremy’s about how much you should anticipate. My own approach is to present my work as a presentation a few times and then respond to the commonly raised concerns.

  4. I agree most enthusiastically!

    Another option is to address potential concerns as they become relevant in the discussion. Advantages to this approach are: a) you don’t waste words restating the argument; b) you provide the full weight of your argument is positioned immediately adjacent to the counterweight of potential opposition, making your argument stronger, c) you don’t compile a handy and succinct list of caveats at the end of the piece that 1) lacks the full weight of your argument; 2) is easy for people to find so they can 3) use it to bash your argument.

    • I agree the option of interweaving the caveats with where they first arise (intro or methods) is an underrated option that has the benefit of diffusing the short comings of the research and addressing the issues immediately when they are raised.

  5. Agree entirely with the need for a caveats paragraph (and/or have caveats interwoven). But if it is the penultimate paragraph, then the last paragraph needs to be *good*. It can’t just be a milquetoast summary of the paper, but rather should really remind people of what the paper is adding to the literature, the gap that’s being filled by this work. I’m pretty sure I’ve not done this especially well in the past, but I aspire to it…

  6. I have two thoughts about this interesting topic. First thought: by far the biggest hindrance to effective story telling in scientific writing is clinging to the outmoded and obsolete IMRD structure (introduction – methods – results – discussion) as an effective way of writing. Even Brian falls for it in his exhortation:

    “The introduction should introduce the questions in a way that gives them emotional pull and leaves us desperate to know the answer. The methods and results should be a page-turning path towards the answer. And the discussion should be your chance to remind the reader of the story arc you have taken them on and draw sweeping conclusions from it.”

    Second thought: reviewers often want a “discussion” section to address every thought that comes into their (the reviewer’s) head while reading the manuscript. I recently fought against this with reviews that said I really needed to discuss how the study related to evolutionary models, to conservation and management, to integral projection models, to parameter estimation, and on and on. None of which were directly relevant to the paper. Most of which we had thought about and decided that we had nothing worthwhile to say about them. Effective story telling comes not from cramming in every thing possible; it involves carefully choosing what to include and what to leave out.

    (My response to the laundry list of suggestions in that set of reviews was to ask, “What ever happened to the idea that a scientific paper should pose a question, answer the question, and then shut up?”)

    All that being said, structuring scientific writing so that it leads the reader on a journey is hard work, but ultimately very, very worthwhile.

    • “First thought: by far the biggest hindrance to effective story telling in scientific writing is clinging to the outmoded and obsolete IMRD structure (introduction – methods – results – discussion) as an effective way of writing. ”

      Very interesting remark Hal. Is there an alternative structure you would like to see adopted? Or would you prefer that papers not have a standard structure?

      If you have further thoughts on this you’d like to write up as a guest post rather than a comment, send me an email!

      • Hi Jeremy — there seems to be some interest in my opinions on this, so I will take you up on your invitation. Email to you later.

    • Hal, I hope you’ll take Jeremy up on his offer because I’d love to read your longer-form argument for dumping IMRAD. Mind you, I’m expecting to disagree with it, but that’s what will make it interesting. To me, we use IMRAD whenever we can for the same reason that we use standard conventions in graphic design and writing (on a map, north is at the top, water is blue, etc.). It removes one kind of cognitive load from the reader; it’s a universal finding system. If I want to know what the methods were, I know to look in the Methods; if I want the caveats, I know to look in the Discussion; etc. It’s far easier to communicate successfully when there’s an agreed-upon road map that all readers share.

      Of course, there are a few papers where IMRAD doesn’t work well (some theory papers, for instance). I used to think there were lots where it didn’t work, but I now believe that a *writer* may be tempted to reject IMRAD, but that’s because IMRAD is making the *writing* harder. It’s still making the *reading* easier, and that’s what matters; if anything, the fact that IMRAD is making the writing harder might be a good signal that it’s especially valuable to the reader!

      Finally: could not agree more with “Effective story telling comes not from cramming in every thing possible; it involves carefully choosing what to include and what to leave out.” Well put!

      • We’re on the same page Steve. IMRAD is not great. It probably is suboptimal. And almost certainly it is true there is no one size fits all for writing. But IMRAD is the convention. And that matters a lot for the reader. And even more in this day and age where most people only read maybe 1 in 10 (maybe 1 in 100) papers from start to finish and skim the rest of the papers they “read”.

        But I’d love to hear a thoughtful defense for freeform writing of scientific papers!

      • Did any of you read this recent Chronicle piece? Part of what made me think of it in the context of this discussion is this section: “Writing that is overly complicated, requiring too much thought to process, taxes readers unnecessarily. That’s bad in itself, but it’s also disrespectful, as the ideas weren’t presented in ways with which readers could readily engage — and, potentially, criticize.”

        That piece appeared after this post was already in the queue. I debated whether to delay this post until I had time to reflect more on their piece and how it relates, but didn’t in the end. But this section clearly relates, “In a lot of academic writing, the goal is to persuade the reader to believe something. You’re expecting at least some of those readers to change their minds. So be as explicit, substantive, and powerful as possible. Also, be frank about the opposing argument — all the reasons your readers might not change their minds. That’s the honest thing to do — and it also can involve humility, if your reasons are more limited or weak than you wish they were.”

      • Trying to imagine a defense of freeform scientific papers…first thing that comes to mind is something like Andrew Gelman’s argument that complicated, eye-catching graphs can be good because they draw the viewer in. The viewer *wants* to spend 30 seconds (or whatever) figuring out what the heck the graph is showing. And then understands the data better for having been obliged to think, rather than having been spoon fed by a clear transparent graph. I can imagine an argument along these lines for entire papers rather than single figures. Not saying I buy this argument (or that I don’t buy it), but it’s at least a plausible argument. Anyway, really hoping we get to hear your argument, Hal! 🙂

        Against this, the other day I happened to see on Twitter the argument that it’s *unethical* to write scientific papers in anything other than the conventional format (and using anything more than the fewest possible words), because you’re inflicting a cognitive burden on readers. Forcing them to spend their scarce time and attention against their will. Frankly, I find that argument pretty silly. No one’s ever obliged to read any particular paper*, and anyone who starts is free to stop any time. And not just “free” in the sense of “no one’s putting a gun to your head” but “free” in the sense of “there will be absolutely no negative consequences to you whatsoever if you don’t read paper X”. No, not even if paper X is by prominent authors and published in Nature and about the exact topic you work on. So if you don’t like how paper X is written, you can just, you know, not read it. Just like how, if you see a stranger on the street wearing an unconventional outfit that you don’t like, you can just, you know, not look at her. Her sartorial choices aren’t *unethical* just because they’re unconventional, or because you don’t like them!

        *Ok, unless you were assigned to read it for a class or something. But if you think the paper you’ve been assigned to read for class is a waste of your time, your beef is with the instructor, not whoever wrote the paper.

      • The comment was directed in particular to the idea that a paper should tell a compelling story. Don’t move the goalposts to whether IMRD is a convention. It is, but “we always do it that way” is not a sure-fire argument to support continuing to do something that way. And in particular, it is no counterargument to the claim that IMRD, convention or no, gets in the way of telling compelling stories. And, no, putting all the methods in one place does not make life easier for readers. Not even a little. I must write more on this apparently.

      • Seems to me that telling a story serves the immediate purpose of the paper and writer but it doesn’t serve the larger enterprise of science especially well. In the larger enterprise every section (intro, methods, data especially) has potential uses beyond your particular story and should be accessible to others outside the context of your story.

        It might be hard work to write well, but it’s not hard work to write well because of the standard model. The standard model makes a hard job easier by providing a fixed broad structure that hooks together the various pieces so those relationships don’t have to be reconstructed or reinvented every time, either for you or the reader.

      • A lot of times the motivation for an experiment or analysis emerges from previous results and its methods can only be understood after knowing those results. Yet, we are “forced” to group all the methods in the same place as if we had concocted all the experiments and analyses at the same time. I think that’s one way the conventional structure could be modified, to become one that allows a series of linked mini-papers.

      • Richard, I suspect you’re on to something! I’m no expert, just someone who’s read lots of papers recently, but several have been laid out to reflect the way different experiments refined the authors’ focus. Once I got over the “shock” of the unfamiliar strategy, I found it very helpful in understanding the context of the study. (Just my two cents.)

  7. Pingback: TCTS weekly: от магистратур до онлайн-преподавания | THINK COGNITIVE, THINK SCIENCE

  8. Pingback: Pick and Mix 33 – resilience, entomophagy, entomology, the windscreen phenomenon and writing habits | Don't Forget the Roundabouts

Leave a Comment

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.