WHY THE METHODS SECOND?

Only once I know which results I'm presenting do I move on (back?!) to the Methods section, if possible. Why not move on to the Discussion section next instead, while your results are fresh in your mind? I think that’s valid, but, personally, I have a simple reason for preferring to write the Methods next instead: it’s faster. As I’ll discuss, the Discussion section is hard to write (even if it’s fun!); the Methods section is comparatively a breeze (at least for me!). I find that quickly churning out a Methods draft next helps me build momentum that bolsters me for the really hard work that will come afterwards.

So, Let's start this section off with the "bad news:" The Methods section is the longest section (in words) of a typical scientific paper. This is by design and for good reasons I won’t belabor here (in fact, I’d argue Methods sections should be even longer then they already are, but I'll try not to get hung up on that!). Because they are so long, Methods sections can be daunting to write. They are also somewhat less formulaic than Introduction or Results sections, so they don’t “write themselves” as much.

That’s the bad news. The good news? The Methods section is, at least for me, often the most brainless to write. It is, at its heart, only a list of facts organized into a few “buckets”—unlike for other sections, there’s a lot less critical thinking needed in the Methods (we'll get to the exceptions later!). So, once your ducks are in a row, you can bang out a Methods section draft (even a long one) in just one or a few writing sessions.

And there’s one other, major advantage Methods sections have: their content is, ultimately, just what you did. You know what you did, right? Cool—that’s all you have to say here! Unless you need to justify an unusual methodology, you usually don’t need to cite any literature in the Methods, so there’s no need to integrate outside information (one of the hardest cognitive tasks writing papers requires). We just need to deliver facts; facts we know better than anyone else because they’re our facts!

Ok, but how do we do that? Well, I mentioned above that a Methods section has “buckets” our facts go into. There are three, and, in the natural sciences, they are almost always ordered thusly:

1.    The study system and/or taxa

2.    The data collection process

3.    The analysis process

Now, I'd further sub-divide that second bucket—data collection—into three ‘sub-buckets’: the “how,” the “inputs,” and the “outputs.” But more on that later.

First, there’s more good news here: That first bucket, which lays out the "essentials:" of your study system, is almost always brief—usually, around a paragraph per element, and sometimes even just a paragraph total. And that last section, on your analyses, is often short too (though this does depend on the complexity of your analyses. I'm a data scientist, so this section is often bulkier for me, but that’s peculiar to my field). Simple analyses don't require long descriptions, so this is a good reason to aim for simple analyses!

By contrast, the middle bucket is almost always the longest by far. However, it tends to be very rote. Plus, as I said above, we can break it down into subsections, which makes it even more manageable.

With this framing unpacked, let’s dive into each “bucket” in turn.

METHODS BUCKET #1: DESCRIBING THE STUDY SYSTEM/TAXA

Imagine I'm about to describe a study I performed in Minnesota on black raspberry. Imagine you’ve never been to (or even heard of!) Minnesota nor black raspberry (after all, the modern audience for science is international. If you can’t name the different states or species of Germany or Thailand or New Zealand, you shouldn't assume all your readers can do the same for your region).

If you're totally unfamiliar with my region or study species, my results are likely to mean less to you, right? Is black raspberry a tall species? Is it an annual? If an individual produces 10 fruits/season, is that a lot? Is 70 degrees Fahrenheit a hot day where the study is taking place? If the answers to these questions are important for understanding your study's design or results, then we the readers need answers to them. Without them, our Results won’t be nearly as transferable or impactful.

So, the goal of the first section of the Methods is to give the reader enough context about the study system and/or study taxa that they can at least relate them to the systems/taxa they do know well.

The only taxing mental work needed to write this subsection (and the Methods generally) is deciding exactly how much the reader actually needs to be told. It is, sadly, easy to write a Methods that is way too detailed—and it’s just as easy to write a Methods that is not nearly detailed enough (or, by relaying tons of mostly irrelevant details, do both!).

So, what’s the secret to knowing what’s essential and what’s filler? If you find out, let me know! 😊 I wish I had a perfect rule to offer. I do, however, have some suggestions for how to think about the difference and find the line between the useful and the extraneous:

• Pretend you're writing a short “biography” for your system/taxa, like one for a movie star in a magazine interview. It’s short, terse, and focused on just the most interesting highlights, particularly elements the reader may not already know but find curious.

• Ground the decision about what is relevant firmly in your research question. If we’re investigating spring seed germination, don’t discuss autumn leaf drop patterns—those'd be extraneous, even if your sources regularly mention them when discussing the taxon.

• Know your audience. Botanists may not need “fruit” defined, but citizen scientists or statistical modelers might.

• Get someone else to tell you what they need. The best advice I can give is to tell you to find someone willing to review your drafts. Why? We are often simply not the best people to judge what is essential vs. non-essential to say about our own work. We’re too close to it, and we can’t “unknow” what we know (this is called the “curse of expertise”). We need an outside perspective to help us assess that.

• Err towards exclusion. The second-best piece of advice I can give is “when in doubt, don’t include it.” I bet you weren’t expecting me to say that! Yes, I’m actually saying you should do less work! Wait and see what reviewers tells you is missing rather than trying to guess or including everything “just in case.”

This subsection will probably be tricky to write at first, especially if you’re drafting your Methods before your Discussion, because you may be harder to know what's relevant to your story when your story isn't solidified yet. As such, plan to come back to this subsection once your Discussion is done and thus you have a better sense of what's critical context for your audience to know.

METHODS BUCKET #2: UNPACKING THE DATA COLLECTION PROCESS

This subsection is usually the longest part of the Methods—but it doesn’t have to be the most painful to write! I'll break down it down into discrete chunks to make it more manageable for you. 

First, though, let’s back up for a second to consider: What is a Methods section for? Why must we have one? I think there are a few valid answers to this question:

1. Readers must know our methods to be able to scrutinize them. Did I make a mistake that'd invalidate my results/conclusions? If I don’t present my methods in enough detail, no one'll be able to tell, and I'll thus have polluted the scientific record with “junk.” No one, including me, should want this! If I’ve made a mistake, it’d be a bummer, but I should want that mistake to get caught and corrected. That means the reader must get enough detail about what I did to be able to say whether it’s what I should have done.

2. Readers must know our methods to be able to trust them. If they're going to cite our research as evidence for a claim, or if they're going to use our results to decide how to invest millions of dollars, or if they're going plan their next multi-year study around what we’ve found, our readers must know our work was honest, logical, and well-envisioned. If they can’t be certain our work is legit because our Methods reads like a riddle or mystery novel, our paper will become “scientific junk”—no one'll ever act on what we’ve found! Good science poorly communicated is too hard to distinguish from bad science—most won't bother to try. 

3. Readers must know our methods to be able to replicate them. Resolving science’s “Large Unknowns" will require us to repeat studies. If our Methods is too vague or undetailed, no one'll be able to repeat what we did (in spirit—it’s not essential, and can even be detrimental, to repeat research to the letter!). Methods described irreplicably means our results can't be verified, and they're more likely to die a slow death as “scientific junk.”

Hopefully, you’re sensing a pattern: Our Methods is where we prove we’re legit; our science is not junk—it’s trustworthy and actionable and well-intentioned and repeatable. Because that's so important, we use the bulk of the Methods to accomplish it.

So, that's what this second bucket is for, in a nutshell. For me, the easiest way to write it is to work backwards. Remember: The next section of your paper is the Results, where you’ll present your data. Your Methods (and this part, specifically) needs to set up those data. If I (the reader) do not understand what data you're gathering, how you got them, or why you got them, I'm not ready for you to present them. Thus, your Methods needs to answer these "what," "how," and "why" questions for the reader. This is why I prefer to write my Results first; doing so narrows down exactly which questions the Methods must answer.

There're probably lots of ways to organize the data collection subsection, but the one I’ll explain works for me: “Setup,” then “inputs,” then “outputs.”

1. Setup: How'd you achieve the proper "conditions" for your data collection? (What were the essential components of your test?).

2. Inputs: What variables were manipulated and/or observed that ought to dictate outcomes, and how did you measure them? (What were your test's independent variables?).

3. Outputs: What variables were measured in response to variance in your input variables and how? (What were your test's dependent variables?).

Since we’re already working backwards, let’s start with our “outputs.” These are the data you're hoping will match your predictions. On your graphs, they're probably on the y-axes (i.e., they're your dependent variables).

In the "outputs" part of this subsection, you outline:

1. What your test’s outputs were, and

2. How you obtained their data (what specifically did you measure, in what units, and how?).

I’ll explain how you do those things in a moment. First, though, let's step backward to our “inputs.” These are the data you varied to (hopefully) observe a predictable response. In other words, these were probably your experimental or independent variables and are probably on the x-axes of your graphs.

In the “inputs” part of this subsection, you outline:

1)    What your test’s inputs were,

2)    How you obtained their data were measured (what specifically did you measure, in what units, and how?), and

3)    How you chose the "levels" of these variables, how you achieved those levels, and why you picked them (how much variation were you aiming for? Why? How'd you manage it?).

Let's take one more step back to the “setup.” Presumably, your inputs and outputs didn’t arrive in the mail! You had to do something to create the conditions under which you could gather those data. Here, you describe the test that yielded your data.

For example, maybe you had to plant seeds (setup) before you could water them differing amounts (an input) to measure how fast they later grew (an output). In the "setup," you'll tell us about each preliminary step needed to even get to the input stage, in enough detail that someone could replicate your work in spirit. 

Sometimes, your "setup" and your "inputs" feel so intertwined it doesn't make sense to separate them, but I'd bet this is rare for experiments. How can you water seeds before you plant seeds? How can you plant seeds before you have growing spaces and growth media? In the "setup," you tell us about everything that needed to happen before you could even manipulate anything.

[Clarification]: While I’m suggesting you work backwards by writing this subsection in “outputs, inputs, setup” order, because I think this is the easier order, they should appear in the opposite order in your manuscript!]

DIALING IN THE LEVEL OF DETAIL

Now, I bet you’re wondering: How should I describe how I collected my inputs/outputs? As briefly and simply as possible. Bear in mind the three things a Methods section needs to do (summarized above) and, at the same time, the #1 rule of scientific writing: “Make it as long as it needs to be and no longer.” My Masters advisor taught me that!

Every detail needn’t be included—as I said earlier, studies don’t need to be replicated exactly. In fact, that would usually be pointless! Instead, we need to know only the substantive details…whichever those might be! For example, what reagent you used is probably essential, especially if no other reagent would work the same way; what brand of flask you mixed it in is probably not—some other brand would probably have worked just as well! 

When I say we need to include the substantive details in our Methods, the former kind is what I mean. If someone else didn't do more or less exactly what you did or use exactly what you used in this regard, would they be replicating your experiment? Or doing a different one? If the former, include it. 

This discussion has all been pretty abstract so far—what does a paragraph in this subsection look like? Let’s look at an example from a paper by one of my mentees (Thexton & Bajcz 2021):

We measured leaf toughness similar to Lowman (1984) and Feeny (1970). We constructed an apparatus out of wood blocks and metal dowels (Fig. 3). We pressed the center leaflet from a randomly selected leaf from each cane between the apparatus’ two halves, taking care not to place the midvein over the central hole. We then placed the metal plunger in the top of the central hole down to the leaflet surface. We placed an empty beaker on the plunger and slowly filled it with sand until the plunger dowel punctured the leaflet. Three beakers (50mL, 100L, 250mL) were used to ensure a wide range of leaves could be measured because larger beakers alone could puncture weaker leaflets. The combined beaker, sand, and tape roll (used to increase beaker stability) mass required was recorded as leaf toughness (in g of total mass). The mass of the plunger was assumed to be constant and was not included. We used digital calipers (Precision Measuring Digital Caliper 150mm) to measure the basal diameter of each cane segment (in mm) immediately following harvest.

This was one of Brady’s “output” paragraphs, but much of what I’ll say about it holds true for “input” and even "setup" paragraphs as well. I would argue this paragraph is pretty much the perfect length—each sentence serves a purpose, and there’s nothing obvious missing or extraneous included (to me, at least).

Let’s dissect what each sentence is doing:

1. In their first sentence, Brady notes prior studies that describe similar methods to theirs, where the reader could go for more details than we provide. This builds trustworthiness—Brady shows they're not inventing methods, and they’re also not hiding where their methods derive from. This also enhances replicability, since we’re providing places the reader can go to learn more about how to emulate our methods more exactly.

2. In their second sentence, Brady explains the essential materials and equipment needed (they also provided a picture of them, a nice plus!). This enhances replicability; we’re saying “this is the core stuff you would need to properly repeat what we did--otherwise, you'd be measuring something other than what we measured.”

3. In their next few sentences, Brady explains their measurement approach, noting three key details in the process: 

   1. We chose leaves randomly (to reduce bias), 

   2. We always measured the center leaflet (to increase consistency), and 

   3. We avoided leaflet midveins (to reduce variance). 

This all relates to how we expected our methods to be scrutinized—we’re showing we thought through our work, foresaw potential missteps, and worked to avoid them.

4. The underlined sentence indicates how our final, analyzable data were derived. Now, when the reader sees the term “leaf toughness” later, they’ll hopefully picture the mass of an amount of sand large enough to puncture a leaflet! Note we include units too.

5. The next sentence anticipates a question a critical reader might have: what about the mass of the plunger? Brady tells us to ignore it.

6. The last sentence (kind of an extra to avoid a single-sentence paragraph), does a lot. It outlines the entire approach used to quantify another output; it notes a key detail a reader could scrutinize (that we measured stems right after harvest so they didn’t have time to dry out and shrink), and it provides model information for the measurement tool used, so the reader could buy the same on, in case each brand operates differently.

This paragraph works not because it says everything; it works because it says the right things: it builds trust, supports replication, anticipates critique, and tells a clear story about what we measured, how, and why (though, mostly, this is answered in an earlier paragraph). 

I’d argue Brady didn’t relate any trivial details, ones that wouldn’t really matter if someone were to replicate our study. As just three examples, Brady didn’t specify the time of day they did things, or the species of wood used for the blocks, or how exactly they chose leaves at random.

Understanding which details are trivial and which are vital takes practice, unfortunately! But I hope this demonstrates there is some common sense to it. Does it really matter what size of beaker someone uses? If you can’t explain why it would matter, then it probably doesn't!

Brady’s entire output section was simply a sequence of paragraphs like this one, with the following sentence at the beginning to set them up: “To characterize physical defense allocation, we quantified prickle length, prickle density, leaf toughness, and stem basal diameter.” (That's the "why" part I alluded to earlier.)

That’s just our list of output variables in the order in which we then describe them. I call sentences like these “roadmap sentences” because they help the reader anticipate exactly where we’re headed. While it's possible to overuse “roadmap sentences,” using them judiciously can be really powerful. 

Let’s be clear: This paragraph was great, but it took a lot of work to get there. Brady and I went back and forth at least eight times to get this paragraph to this point. And I’m sure reviewers improved it further, since we probably couldn’t perfectly see what was essential vs. non-essential ourselves.

If you think your science writing could be perfect without feedback and editing, think again. The earlier you start writing, and the sooner and more often you pursue feedback, the better your writing will be and the faster you'll learn to write well.

METHODS BUCKET #3: DESCRIBING YOUR ANALYSES

Imagine you’ve generated your results, and they seem to match your predictions. That’s what your eyes are telling you, anyway. But how can we be sure that there's something “special” happening in our data? How should the reader assess whether you're just seeing what you wanted to see in your data?

These questions are why scientists use statistics. Ahh! I said a scary word! So many of us have been conditioned to faint at the “S” word. In fact, >80% of science student report having an aversion to, if not an outright hatred for, statistics. I’d imagine that continues into their professional lives, sadly. For many, doing statistics is scary.

However, allow me to plant a different fear in your mind: If this fear of statistics among scientists continues, science is doomed. 

Okay, okay, “doomed” is hyperbole, but not by a lot. Here’s why I think this is a serious problem:

• We humans see "patterns" even where there are none. We're hard-wired evolutionarily to spot “something interesting” even amidst randomness—like when five dice all roll five. That feels special, but it’s not actually that improbable.

• We humans miss patterns that are real but subtle or complex. Nature is messy. Without statistics, we might miss the signals hiding amidst that noise.

In our Results section, we show all the patterns we’ve found in our data and convince try to convince the reader (most likely using statistics) that they’re real patterns. Whatever statistics we present in the Results have to be set up in the Methods; that’s what this last Methods subsection is for.

To do this well:

1. Start by listing the tools you used to store and analyze your data (Excel and R, for example).

2. Then, list the analyses you did (and why, if they need defending). So, three linear regressions, for example.

3. Next, explain which diagnostics you did to confirm your statistics were successful and appropriate (residuals plots, leverage plots, etc.) and what you subsequently concluded.

4. Lastly, specify the “rule(s) of thumb” you used to decide if a result was meaningful (usually, what p values you deemed statistically significant).

That’s really it! For a relatively simple study using simple statistics (like ANOVAs), one or two paragraphs may be plenty to do all of this. If your study is more complicated, you can often follow this same template but for each analysis you performed. If there are any commonalities, summarize those all first, then summarize any relevant distinctions. 

Here is a sample paragraph from this part of the Methods from another mentee of mine (Angell et al. 2024):

We built and ran a binomial mixed-effects regression model, including random intercepts for each participant to account for potential pseudoreplication. The response variable, percent AIS removed, was regressed against six categorical fixed effects (three main effects and three two-way interactions). The three main effects were participant type (boater, inspector, or decontaminator), AIS type (plants, zebra mussel, spiny water flea, or residual water), and region (metro or outstate). Boater, plants, and metro, respectively, were coded as zeroes to be the default levels for comparisons. The three interaction terms were participant typexAIS type, participant typexregion, and regionxAIS type. We initially included the three-way interaction terms between these three variables in the model, but all associated terms were not significant (all p values > 0.83), so all were removed to simplify the analysis. Only final model results are reported. A post hoc Tukey’s test was conducted to evaluate the differences among levels of a factor if a significant interaction or main effect was detected. Statistical significance was assessed using an alpha of 0.05. The location the AIS was placed on the boat was identified as a partial confounding variable in the comparisons between removal and AIS type. These results should be considered in light of that potential interaction. 

About the only thing missing from this paragraph that probably ought to have been included was a statement about how we diagnosed the model's fit to ensure it met the assumptions. I suspect that part got cut at some point and I didn't notice. Whoops!

However, I share that example to prove the point that, of all the subsections in the Methods, this one is by far the most formulaic. Every statistical test you’re using has been written about, in a subsection just like yours, a bajillion times! As such, there’s absolutely no benefit to “reinventing the wheel.” If you want my rebellious and dangerous advice? Find a paper that uses the same test you’re using and that describes it well and steal the thrust of their language, swapping in your specifics for theirs. I've often pointed other mentees to the above paragraph with that very advice.

To be clear: I'm not advocating for, nor describing, plagiarism; I'm advocating for finding what works and using it! Don't emulate substance, obviously, but style? Absolutely. 

Really, much the same could be said about the Methods as a whole—learning to write one is a lot about picking up on the patterns (mostly unspoken and assumed!) we scientists tend to use to convey the “whats,” “wheres,” "whys," and “hows” of our research and then learning to mimic them. The more times you do this, the more patterns you will pick up on, and the easier writing Methods sections will become!

SECOND OPINIONS (METHODS SECTIONS)

Here are some other great resources about writing Methods sections. 

• How to Write a Strong Methods Section Fast and Well | Research Paper Tip (youtube.com)

• Writing The Methods Section For Your Research Paper (Proven Framework) 

• How to Write Your Methods - PLOS