WRITING A RESULTS SECTION FOR A SCIENTIFIC RESEARCH PAPER
WRITING A RESULTS SECTION FOR A SCIENTIFIC RESEARCH PAPER
Yes, when I formally start writing a scientific paper, I really try to start with the Results. Why? Well, first, I think the Results is the most important section. After all, our data are the new things we're adding to the scientific record; the Results section is where we show those off. So, I like starting with the Results—I want the most time to polish it so my data shine.
I also like starting with the Results because I find it the easiest section to write. To start with the Results is, for me, the "path of least resistance." Look...the hardest part of writing, for virtually everyone, is going from an empty page to something. The Results is where I can make that leap with the least pain.
I chalk that up to three main factors. First, it’s often (though not always) the shortest section (in terms of words). Second, it’s generally the most straightforward in content. If the Methods is “what I did,” then the Results is “what I found.” It takes finesse to describe how things were done, and it's hard to separate essential details from extraneous ones. But what was found? That’s usually more obvious once the analyses are done and you're ready to write.
Third, I always come into the analysis phase of projects with an outlined plan, but that plan almost always changes. Maybe I discover a problem with my model, and I need to pivot. Maybe I find an unexpected pattern and decide to probe it deeper. My analyses almost always include branching paths, intended and not. I often have no clue which path will ultimately be the final one until I’ve explored all those available.
This is all to say that I find starting with the Methods section (which is otherwise where I’d start) to be inefficient. I prefer to decide which results I’m going to present and how first! If I try writing the Methods first instead, I all too often have to start it anew when the Results section is done.
The final reason I like to start with the Results is that I find making figures, tables, and captions (the beating hearts of the Results section!) to be both critical to nail and draining to create—they are things I enjoy and value so much that I want to ensure there’s plenty of gas in my tank to give them my all.
A meme depicting a character from "The Matrix." The text says "What if I told you the most important part of a scientific paper isn't something you write."
Now that you know why I prefer to write the Results first, let’s move onto how I write do it. Here, I can sum up my keenest insight with the meme above. See, we think of papers as things we write. They're made of words, and we write those words!
...But most of those words aren’t “new.” Many of our ideas and words are fundamentally derivative--such is the nature of science.
What is guaranteed to be new are your data. Largely, scientific papers do not present data in words; data are most often presented (and consumed) in tables and figures. So, in the Results, your words take a backseat; your focus should instead be your figure and tables, the real stars of your show.
There is a lot I could say about figures (and to a lesser extent tables) but covering them in depth is outside the scope of this guide. For way more on designing great figures, see my other guide on the topic.
Let's instead stick to the core ideas here--A good scientific figure (or table) has three parts:
1. The figure or table itself,
2. Its supporting caption, and
3. Text, in the paper's body, that interprets or explains the data/patterns in it.
These three components are necessary for every figure and table, but only the first two belong wholly in your Results section. Remember: A Results section is a neutral presentation of your data—what you think the data mean (i.e., your subjective interpretation of their importance or meaning) should not appear in your Results. Discussing is what the Discussion is for! 😄
I'll admit it: I used to find the separation between data (in the Results) and interpretation (in the Discussion) frustrating. However, I’ve realized it’s intentional and valuable: we want readers to be able to make up their own minds before we influence them with our ideas.
That said, there is Results-section text that should accompany every figure and table you include that answers these types of questions:
What is the pattern?
How strong is it?
What is the evidence for it?
How can I understand/read/find it?
These questions can all be answered objectively. Objective explanation—where you describe the strength, direction, or features of your data—does belongs in the Results.
So, we must eventually consider a Results section's text, but, for now, let's focus on components #1 and #2 above, since they are the most pivotal.
As noted above, teaching you everything about crafting great figures or tables is outside this guide's scope. That said, here are some good things to keep in mind whenever you're creating one:
Keep the design simple. Respect your reader’s eyeballs. Don’t overwhelm them with too many colors, line styles, shapes, etc. Using grayscale is perfectly fine (even preferable!). Gridlines are almost always unnecessary. While too much white space is wasteful, some is essential for ease of interpretation. Make data points and labels big enough to be easily read. Science is for everyone; make your data as clear and accessible as possible.
Clearly label your figure axes and table rows/columns. Include units.
Ensure captions do everything your figure or table can’t or don’t do. For example, you can label the y-axis of your graph “leaf toughness,” then use the caption to explain what exactly that means, if that's easier. If your graph or table uses any codes, abbreviations, formatting, symbols, etc. to convey meaning, make sure your caption clarifies them. Use the caption to provide the "when, where, and how" context a figure or table usually doesn't include.
Don’t be afraid to break up a large or complicated figure or table into smaller, simpler ones.
Only present as much data your audience can handle. Figures and tables should serve your narrative (what you want the reader to take away). If something distracts, overwhelms, or "misses the point," don't include it.
Those are the major things I wanted to get off my chest about figures and tables. Now, we need to have a heart-to-heart. In fact, I think you better sit down. Are you comfortable? Good. Because we really need to talk about your figure and table captions. Of all the parts of a scientific paper, figure and table captions are likely among the most undervalued, in my assessment. For many, they are an afterthought, but that that's just not acceptable--they're much too important!
To explain why, first recognize that a good scientific figure should “stand alone.” That is, you should be able to rip it (along with its caption) out of the paper and present it by itself and it should still make good sense to a reader. In other words, a reader shouldn't need the rest of the paper to extract solid meaning from your figure and table.
This means that even thoughtfully crafted figures and tables need thoughtfully crafted captions—our captions must fill in all the “gaps” between the figure or table itself and the rest of the paper it comes from.
As such, I find writing good captions to be hard. They have to do a lot in a very tight space. What exactly a caption needs to say is so context-dependent that I can’t give a definitive list. However, the general rule to use is: “What do I need to tell the reader so they can read this table/figure roughly as well as I can, even though I made it and they didn’t?”
So, if there's something you can demystify about the figure or table, that should go in the caption. Here's a list of some things that might qualify:
Explanations of any colors, symbols, shading styles, point types, etc. AKA the figures "symbology" or "coding," to the extent a legend doesn't already do this.
The differences between subpanels (if any).
Definitions for all abbreviations, “nicknames,” shorthand, and jargon. Yes, even if these are also found in the text!
The gist of how the data were collected (you're not reprinting your Methods, just summarizing the most relevant “what,” “where,” and “how” elements as briefly as you can).
Any specialized equipment used to yield the data.
Summary of any statistics used to analyze the data, if their trappings appear in the figure or table in any form. For example, if there’s a best-fit line, where'd it come from? Was it statistically significant? The rule is: If you ran statistics, they should be in your figures/tables so readers can evaluate them. And if they're in your figures/tables, the captions should contextualize them.
The gist of why the data were collected and/or presented. What argument does this figure/table advance? This isn’t always necessary, but it can be good to remind your reader what they should be looking for, specifically.
If that sounds like a lot to include in each caption—it is! Figure captions have a hard job, and they need the space to do it well. I often see 1-3 sentence captions, but I rarely see captions that length that are actually minimally effective! I strongly believe it'd be better for science to move towards papers with fewer, better figures, with better captions. This would be long-overdue reform, in my opinion, and reform we can all choose to start participating in right now.
But I digress. Here's a list of things that, conversely, should generally not appear in your captions:
A sentence saying what type of figure it is, unless it’s unusual. If your graph is something common like a scatterplot, don't start your caption with “A scatterplot of…”. Personally, as a reader, I'd take that as a jab at my intelligence. 😉
An interpretation of the pattern. Remember, what you believe has no place in the Results section—don’t put opinions in the caption. Save those for the Discussion; stick to the facts in the Results.
Irrelevant or entirely "new" information. A caption is not the place to present something entirely new. If you feel something needs to be said, find a place to say it in the text too.
Something evident from the figure/table already. You needn’t rehash a legend, for example (if you feel you should, it’s a sign your legend isn’t clear enough).
Writing a good caption is a balancing act between thoroughness and brevity, one that takes practice to master. But, without good captions, your figures and tables are unlikely to have the impact they otherwise would.
Generally, table captions follow the same rules as figure captions. However, remember that, because there’s less to look at in a table, seeing patterns in them is harder. Thus, helping a reader “see” patterns in a table by giving them a sense of what to look for in the caption can be especially helpful.
A great question without an obvious answer! For me, the part of writing a Results section that takes the most thought is deciding how to divvy up my content between these three media. While there often must be some overlap, the best Results sections seem to use them complementarily.
That’s probably because each has its own strengths and drawbacks:
Good figures take a long time to design and can take up a lot of space relative to the data volume presented. However, they're often the most visually compelling and easiest-to-digest way to present data
Good tables can be very data-dense. This is good, but it also means they can be hard to interpret and very dull to “read.”
Text is flexible to adjust and fast to generate, and it can convey certain complex ideas more easily than tables or figures can. However, text doesn't “pop” like a figure, it isn't as dense as a table, and it must be "hunted through" for details of interest in ways that figures and tables don't necessitate.
My personal approach, then, is to view these three options as a hierarchy, one that recognizes and embraces their strengths:
1. I feature my cleanest, most relevant results in figures. However, I assume I'll have space for just a few, so I've gotta be choosy. Ask yourself: If you could only show, say, four results graphically, which ones would most deserve that treatment?
2. I go with tables for relatively voluminous data that aren't already in a graph and can be represented tabularly (like metadata from my statistical tests).
3. For all else, I use words. That includes describing small pots of data that don’t justify a table because they're not voluminous enough. It also includes “blow-by-blows” of what each analysis found (e.g., “This analysis found X, but this one found Y”).
“Higher-level” analysis is where the Results text can really shine. Imagine I ran ten statistical analyses, 5 on leaf data and 5 on root data. I might use the text to describe trends larger than any one analysis. For example, I might say “Out of our ten analyses, four were significant. Three of these involved leaf data, and all those were negative.” While figures are great for showcasing specific results from specific analyses, words can be better for painting a broad picture.
While it’s good for the text, figures, and tables to complement rather than repeat one another, you should generally “repeat” the key findings of your analyses in words, even if they are also in figures or tables. For example, if your study focuses on the hypothesis that intense sunlight exposure harms seed germination, and you show that negative relationship in a figure, that result is worth repeating in words.
However, smaller details contained in figures or tables often don’t need to be repeated in the text. For example, if you present p values in a table, you don’t necessarily need them in the text too—just point your reader to the table for the deets!
As I said earlier, once you have final results, Results sections (at least, their text) are relatively easy to write because they are super formulaic and brief. At its core, the text is just a list of all those facts not already sufficiently communicated by the figures and tables. Still, it’s useful to organize these facts into sensible paragraphs and to order those logically somehow (like the order you listed your predictions in your Intro section, or the order you discussed predictors in your Methods) so the reader can more easily skim to find the results that interest them.
Much of the rest of what I'll say about writing a Results section requires some examples. First, here's a sample Results paragraph from one of my publications (Bajcz & Drummond 2017).
Vegetative traits were frequently altered by flower removal (Table 1). Mid-season canopy light transmittance was higher in control plots (0.074) than in removal plots (0.058; p = 0.018; Fig. 1a; transformed and least-squared means are provided in Table 1), indicating a denser canopy in removal plots. Mid-season vegetative mass/stem was also higher in removal plots (1.65 g) than in control plots (1.41 g; p < 0.001; Fig. 1b). Removal plots tended to have higher mid-season surface area/leaf values as well (1.43 vs. 1.29 cm2 in control plots; p = 0.107) but not higher leaf fresh:dry mass ratios (2.58 vs. 2.55 in control plots; p = 0.379). Average leaf nitrogen content was higher (1.41%) in control plots than in removal plots (1.37%; p = 0.005; Fig. 1c), and average leaf chlorophyll content was also lower in removal plots (1.44 ug−1) than in control plots (1.64ug mg−1; p = 0.009; Fig. 1d).
Here're some things I think this paragraph does well:
It gives a higher-level overview of what this subset of analyses turned up (in the topic sentence, the first [and most crucial] one).
It references (in parentheses) some key metadata (such as p values and group means, the former central to how we drew conclusions and thus worth repeating, the latter not already presented elsewhere) but also points to tables and figures where more relevant data can be found.
It recaps only significant results in detail. It omits many (though not all) non-significant comparisons (readers can find the rest in the tables).
It provides group averages (with units!) for the two groups being compared so the reader can see how big the differences were. Using a whole table or a figure just to present means would've been wasteful; meanwhile, they slot nicely into text, where they are skippable for readers who want to stay focused on the "headlines" and skimmable for those interested in some but not others.
Besides some variance to keep things bearable to read, each Results sentence is largely the same—a “bullet point,” but in sentence form. In fact, even though the paper referenced above is a whopping 14 pages long and reports on 30+ analyses, the Results section is just seven paragraphs of text—it summarizes only the data and patterns it must in order to pay off the “story,” which is that flower removal affects plant behavior, and then moves on.
Next, let’s look at a sample table from a paper authored by an undergraduate mentee (Marshall et al. 2018):
Table 2. Soil physical (top half) and fertility-related (bottom half) property metadata (means, ranges, and standard errors (SEs)) collected by analyzing soil samples taken from the 10-20 cm depth interval of the restored (Area 3) and unrestored (Area 6) parcels of the Biocore Prairie in Madison, Wisconsin, USA. ANOVAs (right side) were used to evaluate the strength of differences in soil properties between the two land parcels. The t and p values are for the Area main-effect terms (i.e., these data have been omitted for the intercepts), and significant differences at α = 0.05 are noted in bold. SOM = Soil Organic Matter (% by volume). Soil texture data are reported as percentages. Elements are represented by their atomic abbreviations.
Note the formatting is simple: Straight borders separate row and column headers from body cells, unnecessary grid lines between cells in the same section were omitted, abbreviations used are explained in the caption, the headers are bolded for easy skimming, etc. The description explains what the bolding means, where the statistics came from, how significance was determined, where the data came from, etc.
I would say, in hindsight, we could've included the year the data were collected (the “when” context), and we could've been clearer that these data were gathered, analyzed, and presented to assess whether soil differences could explain differential restoration outcomes between these two areas (the “what’s the point?”). Still, I’d argue this table and caption are better than average.
One other thing: All the data of the same type have the same significant digits, something you probably haven’t thought about since grade school! For example, consider pH: All these data have exactly 2 significant digits. This is because Krista’s pH meter measured to the nearest tenth (e.g., 6.4). The rule is: However precise your instruments are (the significant digits of their readings), that's how many significant digits you must report your data to. By comparison, Krista’s soil texture analysis determined clay, silt, and sand contents to the tenth of a percent (e.g., 20.3%), so she must report those results with 3 significant digits instead.
Next, let’s look at an example figure from that same paper:
Figure 2. A heat map of total nitrogen (% Kjeldahl N by mass at 20cm depth) for the unrestored parcel (Area 6) of the Biocore prairie in Madison, Wisconsin, USA versus the restored parcel (Area 3). The 32 grid points sampled (see Figure 1) were interpolated to smooth the heat map, producing a gradient of colors based on total nitrogen values (highest values = reds, lowest values = greens).
What does this figure and caption do well? I’d say it clarifies where the data were collected well (but, again, not when). It also does a good job of clarifying what all labels, abbreviations, and visual channels are. The use of color, while rarely necessary, does make the figure captivating to look at too.
Meanwhile, I don’t think we needed to say this is a “heat map” in the caption. I’d argue, now, this is a familiar enough graph type that the reader doesn’t need to be told what it is (your mileage may vary). I think the argument this graph is advancing (that nitrogen differences could be playing a role in differential outcomes) is not as clear as it could be either. I think we also functionally repeat the legend in the caption. But the biggest regret I have here is, yikes, that color palette! It is not colorblind-friendly. Whoops. Plus, the legend labels could've been much bigger too for accessibility.
So, on balance, this figure’s caption is missing some key elements and yet is still probably longer than it needs to be at the same time. This figure is one of our paper’s “stars,” and I think we could've given it more careful attention. It’s a shame to realize that in hindsight!
In conclusion, as you begin to write your Results section, start by sorting your results into a hierarchy. What is really essential or “juicy” and thus can be presented as a figure? Make those figures and their captions first. Really pour a lot of energy into them--they're your stars!
Next, consider what the text can say about the data in your figures that they or their captions doesn’t or can't say. Draft those sentences—don’t worry about where to place them yet.
Then, consider whether you have any remaining pots of data that are structurally regular, essential enough, and voluminous enough to deserve a table. Tables are great for site metadata, statistical results, units and methods used for predictors, etc. Unless you’ve got a tiny data set, don’t use tables (or, goodness forbid, figures!) to only convey raw data. Share your data files via a repository, if that’s your goal! Tables and figures are “premium products,” best used only to communicate big ideas and “vibes.”
Once you have your tables and their captions, again consider what the surrounding text can add to them that isn't already apparent and can thus add value for the reader and draft those sentences.
Finally, consider what has not yet been said but that could be said objectively. Maybe you need to provide evidence that your methods worked, or summarize descriptive statistics, or highlight themes that emerge only at the across-analysis scale. Look for "holes" in your story that text can cleanly fill.
Then, cluster all your drafted sentences into logical paragraphs, with strong topic sentences as their glue (much more on those later!), place the paragraphs in a logical, predictable order, and you’re well on your way!
At this stage, I try to remind myself: the job of a Results Section isn’t to bludgeon my reader with my project's mass--it's to clearly communicate my work's “story." If my figures, tables, and surrounding text do that, I should stop adding and fussing and move on!
Here are some other great resources about writing Results sections.
The Results - Organizing Academic Research Papers - Research Guides at Sacred Heart University
How to Write the Results Section of a Research Paper - Expert Journals