SETTING THE STAGE

Like so many scientists I've known, I have long aspired to be a writer—of fantasy novels, of poetry, of reference non-fiction…In fact, I fell in love with creative writing long before I fell in love with science. Who doesn’t enjoy a great story, well told? Based on what I said about creative writing on the Preface page (that little of what creative writers do works well for science writing), you might assume scientific writing and creative writing are nothing alike. And you’d be right…almost!

While it's generally true that creative and science writing are not remotely similar, a Discussion section is as close to creative writing as science writing gets, and I believe the best Discussion sections absolutely benefit from some “creative writing thinking.” It’s in Discussions that you, the writer, not only get to exercise more creative choice than in any other section, but you also get to finally tell us what you think. These are probably the two most powerful reasons I enjoy writing Discussions.

The other reason is that Discussions are so readily chopped into bite-sized chunks. Good Discussions often feel more like several “mini” sections expertly weaved together instead of a monolith. So, even though the Discussion is often the longest (or second-longest, after the Methods) section, it needn't feel that long to write because you can write it one “mini-section” at a time. If you’re someone who struggles to write more than a few paragraphs at a time, the Discussion is perfect for that.

Before I explain what I mean by all this, though, I do need to first point out some inconvenient truths about Discussions to keep in mind:

• Discussions require the most critical and scientific thinking to write well of any section. Simply put, even at their easiest, Discussions are unavoidably cognitively taxing to write.

• As noted above, Discussions tend to be physically long too. While they're easier to write than Introductions word-for-word (in my opinion), Discussions are usually roughly double to triple an Introduction's length, so I think that makes them harder to write overall for most writers. 

• You need to be deeply familiar with your sources (and have enough!) to write a good Discussion draft. So, don’t plan to start your Discussion draft until you’ve (nearly) completed your literature review!

• Discussions are the next-most impactful section after the Results, so the stakes are high for Discussions. Even though your Results are even more important, if your Discussion doesn’t deliver on its promise, your paper’s impact will definitely suffer.

So, that’s all the "bad news." But here’s one more piece of good news: you only need to master three core concepts to write a Discussion that delivers. Those three things are:

1. What “stories” a Discussion can include,

2. How to order your Discussion “stories,” and

3. How to structure each “story.”

What follows is an explanation of each of these concepts in turn, starting with the kinds of stories a Discussion can tell.

CORE CONCEPT #1: TAXONOMY OF DISCUSSION SECTION STORIES

At its core, a good Discussion is more or less just a collection of “stories,” generally 1-3 paragraphs long, arranged in a logical order. I call the pieces of a Discussion "stories" because it feels the truest word to me, but you can think of them as arguments, insights, or reflections instead—they are self-contained chunks that, when combined together, pay off the overarching narrative we've been building throughout the paper (and, by extension, our project).

While the range of stories a Discussion might tell is huge, thus I could not exhaustively list them, I've found that the majority you'll encounter fall into just a handful of categories:

#1: ANALYZING A KEY RESULT

These stories unpack a pattern or result (or a set) and give it the 360-degree treatment. What does this result mean, in theoretical terms? In applied terms? What does this result not mean? Does it even make sense, based on what we thought we knew? Should we trust it? Does it suggest something is true we didn’t previously know? If the result is valid, how should it change how we think, act, or do future research?

While the mix of stories you tell in a Discussion is largely up to you, it’s uncommon to not tell (at least) one story centered around each of your study's key results. What's the point of your study if it wasn't to produce such results?! Also, it’s into these stories you can dump all the interpretation you wanted to put in your Results but weren’t allowed to!

#2: COMPARING/CONTRASTING YOUR DATA WITH OTHERS'

Presumably, yours wasn't the first such study. If you have sources that did similar research and present similar data, how similar are your results? If all results so far are consistent, what's that tell us? If the results are inconsistent, in what ways? What could explain the discrepancy? 

For example, if studies to date focused on vines have all produced positive results but all those focused on trees have produced negative results, that tells us vines and trees may operate differently in an important way, right? The Discussion is where synthesis stories like this belongs! Remember: Your paper should ideally serve your reader and the field at large—you can use a “story” in your Discussion to distill all the papers you’ve read down into a single “trendline” so your readers can see the "big picture" as clearly as you now can but without having to read everything you did to get there.

Again, it’s uncommon for a Discussion to lack a story focused on comparisons with other studies. In fact, it’s very common to discuss a result and compare it to others in the same story.  

If you find yourself thinking you won’t be able to find data comparable enough to yours, I’d be skeptical. You may just be thinking about your data too narrowly—for example, maybe no one’s done a study exactly like yours on your specific species before, but maybe they've done a similar one on a related species. Is there something cross-cutting there?

#3: DEBATE AN IDEA—OR PROPOSE A NEW ONE

Imagine your study produced the exact outcome you were expecting, which means it provides great support for your hypothesis. However, despite this, perhaps your hypothesis is not the only one out there. Maybe another researcher has previously proposed a different hypothesis, and you can think of a third. 

In that instance, you could spend a story in your Discussion laying out the logic of and argument for/against each hypothesis. How much support does each have? Is there some evidence that could falsify any of them? Which one seems like it’d hold in more situations? Is there a test that could determine which explanation is likeliest? 

With Discussion stories like these, you get to be the "lead scholar," steering the conversation in your field forward! These are my all-time favorite stories to read in Discussions. What can I say—I’m a nerd!

#4: TELL US ABOUT SOMETHING SURPRISING

By this I mean something you, as a reasonable scientist, weren’t expecting to observe, but you did. This can be something you happened to notice in the field (like a bunch of strange bugs all over your plots), something you encountered by accident (how samples reacted when the ingredients were mixed in the wrong order), a result that seems totally backwards to you (like growth rates going down with more sunlight), etc. 

How or why did it happen, do you think? What do you think it means? Has anyone else encountered it before? What should be done about it—do our methods need to change, e.g.? In science, it’s often unexpected results that we learn the most from! These "unplanned" stories can often be the most fulfilling and eye-opening to tell.

#5: TELL US ABOUT SOMETHING NEW

Imagine, while collecting data on seed germination, you took pictures of the seeds to track their growth. However, later, you realized you could analyze their color too, using those same pictures, and their colors varied a surprising amount. You search the literature, and you can’t find anyone who's published seed color data before…

...Maybe you should be the first! Maybe, if you do, someone else will find your seed color data and see a pattern you didn't, or develop a question that makes them curious, and a new line of inquiry is born! This may not be something you intended to study, and it may not be a “key result,” but if you’ve got “new” or “rare” data, they may inspire questions we didn’t know we even needed to be asking.

#6: CRITIQUE SOME METHODOLOGY

To get data, we have to do stuff. How did that go for you? Was there a challenge you encountered but overcame? Did something introduce bias into your study that could've been prevented? Does your method seem better than others in some way? Science is not just about getting new data—it’s also about developing better ways to get data. You can absolutely use a story to advocate for new or improved methodology in your Discussion.

#7: HIGHLIGHT A (SO-CALLED) "NEGATIVE RESULT"

Sometimes, what we expect isn’t what we observe. While many scientists have been wrongly socialized to view getting results that don’t match their predictions (so-called “negative results,” such a terrible term!!!) as a failure, it isn’t! If you, as a reasonable scientist, expected one thing and found another, that’s noteworthy! The world does not, it seems, work in that particular, perfectly reasonable way. How could that not be worth discussing?!

By demonizing negative results as much as we have, I fear science is engaging in dangerous self-sabotage. Here’s what I mean: Imagine you do a test expecting X, and you observe Y. 

One possibility is that your hypothesis was flawed. I, another researcher, need to know this, or I could waste my precious time, money, and energy (even my entire career!) researching your same incorrect hypothesis again! 

Another possibility is that your methods were flawed. If you don’t warn me about your faulty methods, I could repeat your same mistake! 

Sometimes, we need to be told we're wrong by the Universe pointedly enough that we get off the wrong paths and onto the right ones. While it may not feel “sexy” or “fun” to call attention to results that didn't match your expectations, these can be some of the most interesting stories of all, at least for those of us willing to listen!

#8: PROPOSE THE NEXT STEP

Where do you want to see the field go next, now that we know more than we did before, thanks to you? What is the next “Small Unknown” we need to sort out? What is the next set of hypotheses to test? Maybe most importantly, how could our behaviors change as a result of your findings? 

You can use a story in your Discussion to not only suggest the next study that should be done but also to issue a call to action—how could your readers apply what they’ve learned from your study?

It seems to me that all the best Discussions end on a “next-step” story. Science is continuous—at the end of one story, it just feels right to look forward to what the next one could or should be.

There are undoubtedly many other types of Discussion stories out there, but this list should hopefully give you plenty of ideas to get started and feel inspired.

CORE CONCEPT #2: ORDERING YOUR DISCUSSION SECTION STORIES

Once you’ve decided what your Discussion stories are—maybe you've even drafted them!—we need to order them. 

I think there're are many valid ways to do this; the choice is personal and should match your "creative vision." I’ve had mentors that have encouraged me to order my stories very differently than I’d have ordered them myself, and I’m sure I’ve done the same thing to my mentees over the years!

However, to my knowledge, the most commonly used ordering strategies are:

1. By “importance.”
   
   While the stories you choose to tell in your Discussion are up to you, some stories are probably more “mandatory” than others. Critiquing a methodology might be important (to you), but it’s unlikely to be more important than comparing your results to those of others, for example. So, placing a “comparison” story ahead of a “critiquing” story is logical.
   
   It’s also strategic to order by importance, though. How? A helpful device for writing every part of your paper is to imagine your reader is going to get bored any second and stop reading.
   
   Yikes! You want to make sure they don't stop before reading your most important stories, right? In that case, we need to make sure those stories come first.
   
   What makes a story "important," though? That’s mostly a subject-matter-expertise question for you and your team to wrestle with. But, if I had to throw out an idea, I’d say the more you have to say about a story, the more "important" that story probably is. Interest, prior research, conceptual challenge—these things are all volumetrically correlated with importance. So, do a bulleted list of everything you want to say about each story you have and see which ones are longest.
   
   

2. By a previous order you’ve signposted in your paper.
   
   Readers like consistency. If you presented your Methods and Results in the order "X, then Y, then Z," discussing your findings in that same order will likely make sense to your reader. This approach ensures your Discussion section doesn't feel like it's “jumping around,” which can absolutely happen when ordering stories by importance.
   
   You can also more readily employ subheadings (e.g., “Effects of X on seed germination”) to help the reader follow along if you order things this way.
   
   However, of course, ordering this way may mean your most interesting stories come later than less interesting ones do (that said, if you wrote your Results first and Methods second, you may be able to write your Discussion both in "prior order" and in order of importance because you planned it that way!).
   
   Also, relying on a previous order can mean there’s no obvious place to put certain kinds of stories, like those critiquing methodology. In those instances, you often need to create a new subheading just for these stories, generally towards the end, and that can sometimes be awkward to pull off.
   
   

3. According to your questions/hypotheses/predictions. 

For example, if you laid out three predictions in your Introduction —one about seeds, one about flowers, and one about fruits—you could order your Discussion stories in that same way. All the relevant stories for seeds come first, the stories for flowers next, etc.

Again, this makes using subheadings easier. However, remember your Introduction was a while ago to the reader by this point, so, if you go this route, you may want to use a “roadmap sentence” or two at the beginning of the Discussion to refresh their memory before continuing.

4. By “category.”
   
   Imagine that three stories you want to tell involve biology, two involve methodology, and two involve debates in the literature. Lumping and ordering by these categories can at least make it easier for your reader to find the parts that are most interesting to them. Plus, it means transitioning from one story to the next or drawing connections between stories within each category will be easier.  

This list is not meant to be comprehensive, but, again, I hope it gets your brain churning about how to add a logic to the sequence of stories you tell so they can tie together in your readers' minds. No matter which ordering mechanism you choose, your goal will be the same: to guide the reader through your highly curated interpretations in a way that feels logical, readable, and satisfying.

Personally, I prefer to order my stories by importance. What do I think has the most “wow factor?” That goes first. Do I get very nerdy about a minor detail? That story goes last, so only the most diehard readers need to stick around for it.

However, if you feel like you’re a relative novice, I recommend not ordering by important; it may be hard for you to judge importance accurately (I’m not sure I do it all that well either!). Instead, I recommend using an order you’ve already established in your previous sections. It’s not always the most effective, and it can sometimes feel a little “stiff,” but it is the most straight-forward, and it can have quite a few benefits for the reader.

HOW MANY STORIES? HOW LONG OF STORIES?

Another two questions I often get, at this point in the mentorship process, is “How many stories do I need, and how long should they each be?”

In general, each “story” should be ~1-3 paragraphs long. Usually, it’s closer to one, but certain types of stories (like debating an idea) may need more room to roam. I’ve seen single stories many pages long, but it’s uncommon! This doesn't apply to me (thankfully) but I’d guess long stories are more common in rapidly developing fields or when researching really high-stakes or high-profile subjects, where you may have a lot to say and to defend about each aspect of your work.

So, my advice is: Read other papers in your discipline, delineate their stories, gauge how long they tend to be, and mimic! But don’t “stretch” stories to their breaking points or pad them with fluff; say what really needs to be said and move on.

Also, even if your stories are short, try to ensure the transitions between them feel intentional—your Discussion should feel cohesive because it has a uniting, overarching narrative (e.g., “we’re getting a little closer to a unified model of seed dormancy with every study”). Try to sum up the "main takeaway" of your entire paper into a sentence or two, then seed that idea at least a little into every story you tell. 

As for how many stories you need…again, I’ll invoke the golden rule of science writing: “make it as long as it needs to be and no longer.” How many stories do you have? Don’t stretch something out to the length of a “story” if it isn’t one, and don’t invent stories just to pad out your word count!

That said, there is definitely such a thing as having too few stories too. After all, you’re supposed to be an expert on your subject, so there’s an expectation you'll prove it by telling us a reasonable number of good stories!

At a bare minimum, I’d think you would have at least one “key result” story and/or one “compare-and-contrast” story. You'll probably have at least one “idea for the future” story too. Beyond that, I think it’s not unreasonable to expect you'll have 2-4 more stories. 

In general, then, I think having around 5-8 stories total is a solid goal to aim for. You might end up only having two really meaty stories or 15 bite-sized ones. There’s room for experimentation here! But 5-8 is a good initial target, and probably where most papers settle by the end.

As another measure, the Discussion is very rarely shorter than the Results or Introduction sections. The wildcard is the Methods, which can be much shorter than, much longer than, or the same length as the Discussion. It really depends on the study! In any event, the Discussion should probably not be longer than the other three sections combined. Hopefully, those heuristics gives you another way to gauge how many stories to should plan for.

Oh, and one other thing: No, you don’t need to discuss every one of your results in your Discussion! Some results are just not that interesting, surprising, or unexpected. Let those results speak for themselves; only go deeper into the results you can add real value to in your Discussion. 

CORE CONCEPT #3: THE BUILDING BLOCKS OF A DISCUSSION STORY

You ready for the best news yet? The paragraphs of a “Discussion story” are almost always just “sandwich paragraphs" (as are the ones in the Introduction, as we'll discuss).

These paragraphs are so important I'll cover them again in the Introduction section of the guide, but, for now, here’s what you need to know: In a “sandwich paragraph,” your first sentence, called the topic sentence, states a fact, opinion, or observation (“the point” of the whole paragraph). Then, the bulk of the paragraph (the “meat” of the “sandwich”) is evidence from your own study and others to support, explain, expand on, or contest that “main point.” If it doesn't directly relate to your “main point,” it either A) doesn’t belong in this paragraph (or maybe anywhere!), or B) the topic sentence needs revision to make it relate.

Then, your last sentence does one (or more) of three things, as befits the scenario. It...

1. Provides a “recap” of what you’ve just discussed (i.e., it reiterates the “main point” in different words than the topic sentence used).

2. Transitions from the current idea towards a new one (especially relevant when we change between story types).

3. Provides a “so what,” a take-home message your audience should walk away with from this paragraph (especially valuable when the contents of a story are cerebral and the reader might need help tying them together).

Like any sandwich, a topic sentence, this last “summation sentence,” and the “evidential meat” is often not quite sufficient to make something truly "digestible." You might need some “mayo” to smooth things out…transitions, flourishes, your own reasoning, reframing devices…whatever is minimally needed to make things smoother. These things often add themselves naturally during revision, so I don't worry much about them at first. Often, you can start by just drafting the three core elements of each paragraph and worrying about the rest in your second draft.

Because almost every discussion paragraph will be a “sandwich paragraph,” and these are predominantly "evidential meat," you absolutely need sources to write a Discussion draft meaningfully. Your sources are where your evidence and insights of your stories largely come from. So, there’s little sense in even trying to build a Discussion story without enough source material to make each story “filling” (am I riding this metaphor too hard??).

But, at the same time, that every Discussion story is just a series of sandwich paragraphs is why Discussions can be written piecemeal—you can treat every story, or even every "sandwich," as its own little “meal” you can “cook.” The largest chunk of a Discussion you should ever need to write in one sitting is three paragraphs (i.e., one story of three sandwiches). Writing three paragraphs in a sitting doesn’t sound so bad, does it?

WHAT WE MEAN BY "EVIDENCE"

Let me get the megaphone out for a hot sec—there’s one last thing I really need to harp on about Discussions, and that is what we scientists really mean by “evidence.”

In science, evidence is (virtually) never what we say; it's what we demonstrate. I can claim anything (so long as the reviewers let me!). That doesn’t make it true, though! What are objectively true are the data I present (assuming I've produced and presented them honestly!).

That is to say, in science, data are the only evidence that count. Obviously, past interpretations of data, as might be found in Discussion sections, are additionally relevant to some stories we might want to tell or arguments we want to advance, but even then, the data themselves are everything—without them, there would be nothing to interpret!

If the only evidence that counts is data, then the evidence you present in your Discussions should come overwhelmingly from the Results sections of your sources—not their Introductions or Discussions! I shudder to think how long it took me to appreciate this fact; if you learn nothing else from this guide, learn this!

Relatedly, we must differentiate between what I'd call “soft evidence” and “hard evidence.” Soft evidence is a statement that summarizes past work superficially, such as “In a study by X wherein they did Y, they found that seed germination rates were significantly higher.” 

Ok, this references a specific study and its specific results, but not very specifically! How much higher? Without hard data (“numbers”), we are left to take your word for it that “significantly higher” means “a lot higher—no, really, you’d think so too.” Plus, here, the reader might wonder: Does “significantly” higher refer to statistical significance or biological significance? Those are not the same thing!

Compare the example above to this one: “In a study by X wherein they did Y, they found that seed germination rates were significantly higher (mean difference: 11.5%, p = 0.048).” As a reader, this small addition opens up much more room for consideration! I can now decide for myself: Hmm, is 11.5% a lot, biologically? How do I personally feel about a p value so close to 0.05? 

Hard evidence, then, is a statement that doesn't merely summarize past work—it brings the "receipts," i.e., the hard data generated by that work. In my experience, developing science writers present far too much soft evidence; early thesis drafts I've read often contain no hard evidence at all. This means these works were asking us, the readers, to "take their word for it" in a way that we really shouldn't have to. In science, we’re not supposed to take each other’s word! Show me the money data; let me decide just how good of evidence they really are for the point you’re making!

SECOND OPINIONS (DISCUSSION SECTIONS)

Here are some other great resources about writing Discussion sections. 

• How to Write Discussions and Conclusions - PLOS

• How to Write a Strong Discussion in Scientific Manuscripts (biosciencewriters.com)

• 6 Steps to Write an Excellent Discussion in Your Manuscript (elsevier.com)