A template for (nearly) painless writing in 10 steps.
Writing a scientific paper can be a daunting task, particularly because a student's reading hasn't really prepared them for translating their experience conducting research into the production of a paper. Students sweetly imagine that the familiar narrative of a scientific paper reflects the order in which work was conducted. How do you squeeze all of those data into this form? Here's a step-by-step template that can get you through writing a scientific paper. It's not the only way to do it, but it's a useful guide to give you some structure.
1. Write short, modular descriptions of methods, results and their interpretations for each figure as you generate each bit of data.
Before you gather any data, write down the design of your experiment and the details of your procedures. This will go into your Methods section. As you generate data for a figure and visualize these results, write a short legend to your figure, and a short paragraph that describes what is in the figure. Write down your interpretations of the figure, mentioning other work from the scientific literature you think is informative.
2. Make rough drafts of your figures.
Some of your preliminary figures won't be as informative as you hoped. Others will form logical groups that make an important point. Group your individual figures and re-imagine them as panels within a larger figure. I find it useful to give each of these larger figures a one-sentence title that summarizes its main point. I spend a bunch of time sketching them in different configurations to help me visualize the story they will tell before I actually put them together. Order the figures in a way that makes it easy to explain what you found.
Don't worry about the exact order in which you did an experiment. It's often useful to have a methodological figure in the beginning. For students in my lab, that might be a bit of neuroanatomy, or representations of acoustic stimuli. The point is to let the reader visualize things that are difficult to describe. Expect to make lots of revisions. As you revise, ask yourself: Are there data that you need to add? Are there data that should be taken out, perhaps provided in a table? Do the figures follow logically from one another? The figures will define the entire narrative of your paper.
3. Write a detailed, hierarchical outline. Include a "word budget."
Look up the instructions for authors for your target journal. Count the number of words in a line, and the number of lines on a page. How long is each section of a paper in this journal? How long do you want yours to be? Write an outline that estimates the length (in words) of each section. Use a few words to describe each major point and subpoint you want to make. An outline is not a series of sentences that highlight some important points you want to make. Those are notes. An outline is a formal roadmap that will help you remember where you want to go and how to get there.
When I'm writing, I usually have a separate document for each the major sections of an article -- one for the Introduction, another for the Methods, etc. I paste the corresponding section of my outline at the top of the clean page. I then use the details of the outline as I write each paragraph. Sometimes I stray from the original outline. The writing is never good at this point, but it's a start. There may be some sentences I like, but they probably won't survive revision. That's OK. After all the sections are done, I will put them together into a bigger document.
4. Write your Methods.
You should have a lot of methodological details already written down. Arrange those details so they align with the order in which the data are ultimately presented. Use subheadings to help organize this section. I generally combine all of my statistical analyses together in one section at the end of the Methods. See how easy that was?
If you didn't write your methods in advance this will be more work. Tedious too. You'll get through it, but make a note to do better next time.
4. Write your Results.
Just like you did in your Methods section, take the short pieces you wrote as you were generating your data and group them so that they follow the order that data are presented in your figures. You will have a lot of details that are not present in your figures. That's normal. Generally, useful but less informative findings get reported in the text but do not show up in the figures. If you have results that are redundant with results presented graphically, but you still think they are worth reporting, they might belong in a table. After you've organized your first pass, focus on each figure. Your Results section should describe everything about the figure a reader might want to know, including your summary statistics, P-values, etc. Most readers will skim your results, but a reader should be able to use your data in a meta-analysis, or catch statistical details that could indicate a mistake (or worse, fraud). By presenting these details in this section, you're ensuring transparency.
5. Write your Discussion.
By now you should have a clear sense of your narrative, and you'll want to revise that outline you made early on. Look at the interpretations you made of the figures as you generated your data (step 1). Reorganize them and group them in an order that reflects your figures. Since you organized your figures at the outset, you already have a logical order to your discussion. What are the most interesting findings? What literature is most informed by your data? Now organize your outline as follows: Your first paragraph is a couple sentences that gives a bird's eye view of what you did and why. The next paragraphs will march through each of your data figures, describe what you found and how you interpret it. Relate those interpretations to the literature. As you do that, you may insert a paragraph elaborating on a particular point and connecting it to the literature, Your last couple of paragraphs will help the reader step back and see the bigger picture, bringing together evidence from different figures. The very last paragraph is often a summary of your key findings, followed by a statement that connects your work to the most general questions you think your data address. If it is a study of female choice, that might be the evolution of elaborate signals or speciation or cognitive ecology. If it is a molecular study, the bigger picture might have to do with epigenetics, genotype-to-phenotype mapping, or the molecular substrates of complex behavior. Be sure you are speaking to the readers of this specific journal. Now that you have all these details outlined, use your word budget to get writing. Understand that your first pass will be very rough, and your transitions will be abrupt. That's normal. Try to write about 10-20% more than your desired length. You may be surprised that you accidentally wrote something longer.
6. Write your Introduction.
It feels strange to write your Introduction last, but it can be easy to write a lot about topics that your data end up not really addressing. (This is also the part that most terrifies graduate students.) One common template for a scientific paper is sometimes called the "hourglass model." The idea is that the scope of the paper generally has an hourglass shape -- broadest at the beginning and narrowing toward the end of your Introduction, staying narrow and focused through your Methods and Results sections, then opening up and becoming most broad again at the end of your Discussion. You don't have to take this too literally, but it's a pretty good summary. Following that model, outline your Introduction so that you begin at the broadest possible vantage point. What is the biggest question that motivates your paper? I generally like to end my first paragraph with a sentence that states that we'll investigate this problem by looking at what we're looking at, and by studying the species we're studying. This let's the reader know what you'll be doing right away, but doesn't get bogged down in details until you've had a chance to convince the reader that the paper is important.
The next two paragraphs or so will summarize literature you think the reader needs to know about in order to understand the big picture, and to appreciate your results. (This is one reason it helps to do your Introduction last.) Next you can get into any specialized information the reader needs to appreciate your paper. This may be the species you are studying, and why it is useful for the questions you are posing. You don't need to be exhaustive; it's not a sin to leave important work uncited. Your last paragraph will be a summary of the specific questions you ask, and the experiments you are about to describe in detail. Because your Methods, Results and Discussion all follow the general order and logic of your figures, you can do that in this last introductory paragraph too. Now you've reverse-engineered an excellent paper: You've led the reader from the big picture to your specific experiments, along the way providing them with the information they'll need to appreciate your findings and their significance. You've laid out a map for the rest of the paper, summarizing the main questions and the order in which they're addressed. The end result will probably be pretty different than your original vision of the study. If this is your first paper, you might feel like you're doing something dishonest. You're not. The point of a paper isn't to faithfully describe the path by which you came to a specific insight -- it's to efficiently convey that insight, the evidence, and its conceptual significance. As with the Discussion, begin with a clear outline, expect to write a little more than you need, and realize that your first drafts will be very rough.
7. Revise.
Your paper will require many rounds of revision. If your adviser tracks changes, you may feel like there is nothing left of your original writing. Don't be discouraged. Feedback is often focused on improving the manuscript rather than congratulating you for the excellent work you did, so if you don't get praised, that doesn't mean you didn't do a great job.
It's good to get feedback at the outline stage, and again after the figures have been made, so that major changes in your narrative can be done before you've done a lot of writing on your Introduction and Discussion. In the first round of revisions, I try to focus on what I would call "structural" problems. Does the order of the paragraphs make sense? Does each paragraph have a coherent point? At this stage, what I find most frustrating is when a student clearly has not used an outline. This means that the section will meander back and forth, with the sentence-to-sentence transitions making sense, but with a logic that doesn't really move forward. Revisions to this are brutal (and sometimes cranky).
One useful exercise can be to outline a section you've already written. That can help you figure out the logical flow you have, or should have, and can guide a major revision.
8. Revise
Once you have the pieces in order, the next problem is usually that the manuscript is too narrow -- that it refers only to results from our lab, or only to findings from the same species, for example. The opposite problem happens when a student feels that they have to put all of their background reading into the paper. Before writing a story, reporters do many interviews that do not end up in a finished article, or do a lot of reading that allows them to ask the right questions and understand which points need to be emphasized. This is what much of your background reading will be -- it will give you the expertise to understand why your work is important. Unfortunately, that often ends up in just a few sentences of your paper. In any case, don't be surprised if you have log into the library to do more reading at this point. And don't be discouraged if all that reading ends up producing just a few well-crafted sentences.
9. Revise
The next revision will focus more on filling in gaps, smoothing transitions between paragraphs, and simplifying the structure of your sentences. At this stage, you might review some good advice on writing. This classic piece by Gopen and Swan is very useful, as is this brief summary of Gopen's advice on revision. Reading a paragraph aloud can help you hear if a passage is too convoluted.
It can be intimidating to change writing you have already spent a lot of time on. Try not to worry about that one sentence you thought was really good. It's only really good if it does its job in the context of the surrounding sentences. When I get nervous about a major revision, I save the document as a new draft before I start the revision -- that way I can always go back to an earlier version that still has those great sentences. The irony is that I never do, but having the old version to fall back on gives me the confidence to make major changes. The result is that I may have 20 or more "drafts" saved. Who cares? Similarly, when I'm struggling with a particular paragraph, I often copy and paste it into a new document. I just start writing and revising several versions below it. That lets me focus on getting that paragraph in shape. When it's close enough, I paste it back into my draft.
As you revise your figures, focus on minimizing visual clutter. Imagine that you are a printer and ink is extremely expensive. Where can you get rid of it? Are there lines you can remove? Similarly, less important information can get less contrast with the background, using shaded lines instead of solid lines, for example. Here's my favorite piece of advice on scientific graphics: line shit up. You'll probably have different panels laid out in a figure. Aligning edges and minimizing ink makes it easy to find and follow information. These days you're probably using color too. Come up with a color code that serves as a cue to let a reader quickly know what is happening in a figure. For example, different shades of a color can be used to reflect different doses of a drug, or different numbers of alleles in a genotype. Once you have a code, be as consistent as possible from figure to figure. Try to limit the number of colors you use to just a couple.
Even after all of this, it takes multiple rounds to get a paper in good shape. The good news is that it gets easier as you get more papers under your belt. The only way to get there is to start writing.
10. Manage your energy.
Different tasks take different kid of attention, and are taxing in different ways. Many professional writers do their writing at the very beginning of the day, before checking email or anything else. This is a particularly good time for new writing. Editing takes a different kind of effort, but still requires focus. In contrast, I can listen to music or a podcast while I make a figure. Alternating between these tasks can keep you moving forward on your paper. You may do a revision of your Methods as you're writing your Discussion, or tweak your figures while you're getting feedback from someone else. The key is to keep it moving forward.
On a related note, don't put off writing until you have large blocks of time, because you rarely will. You will spend a lot of time gearing up to write, reminding yourself of exactly what you did and why, and it will take a lot of energy. If you have a good outline, you can write in small amounts, as little as a paragraph a day, and make significant progress on a paper. Randomized trials find that people who write just 30 minutes a day are ultimately able to write much, much more. Frequent writing in small amounts makes it less stressful, and keeps the material familiar so that you can become productive within a short time slot. One popular guide to productive writing is this book by Paul Silvia, How to Write a Lot.
Don't be afraid of your writing. Get it done as you go, and your dissertation will largely be written by the time you defend. You'll also be a lot more marketable for your next position if at least some of your work is already published. In my (brutal) experience, trainees almost never publish more than one paper after they've left the lab. That's a real tragedy for the student, for the lab, and for the people who funded the work.