How to write a research paper
Condensing months or years of research into a few pages can be a mighty exercise even for experienced writers. Authors need to find the sweet spot between convincingly addressing their scientific questions and presenting their results in such detail the key message is lost. They must describe their methods succinctly and clearly so their experiments can be reproduced, and discuss the broader implications of their research without overselling their work.
The feeling of being exposed that comes with publishing can also get in the way of writing, says Daniela Anahí Parker Yáñez, a fourth-year materials science Ph.D. student at Linköping University in Sweden. “It is damn frightening to state something incorrect.” Especially if you are a newcomer, “one always feels like there is something one doesn’t know, or that your words are not the best.” For nonnative English speakers like Parker, there is also an additional language barrier that requires being “proactive and willing to become better,” she adds. “But at the same time, keep in mind that publishing results is very helpful for others.”
Science Careers asked early-career scientists in a range of disciplines to share their approaches for writing a research manuscript and strategies for overcoming common stumbling blocks. The answers have been edited for clarity and brevity.
How do you know when it’s time to start working on a paper? What is your overall writing process?
Ideally, by the time I start writing a paper I have a strong foundation for why I decided to research this topic, robust results from different experiments that support my idea, and a good overview of how my research advances scientific knowledge. To be more certain that I have a consistent story to tell, I also like to put my findings to the test by trying to invalidate them experimentally or see if there is anything important missing. Then, writing the paper and getting it ready for submission may take me 3 to 6 months. I like separating the writing into three phases. The results and the methods go first, as this is where I write what was done and how, and what the outcomes were. In a second phase, I tackle the introduction and refine the results section with input from my supervisor and collaborators on how we want to develop the story, which references should be included, and what the takeaway message is. In the last phase I write the abstract (and prepare a graphical abstract if needed) and work on providing a coherent ending in the conclusion.
For me, the writing stage typically reflects which phase of the research project I am in. While doing the data acquisition, I create a blank manuscript template that includes all relevant sections. (Some publications provide in-house templates.) I give the project a placeholder title and fill in an unordered list of authors that I keep updated. I also create a PowerPoint file with rough figure panels representing the data, grouping them according to topic on separate slides. As soon as I have what I believe is a final figure, I write a proper protocol so my results can be reproduced and I put an abbreviated version into the methods section. Once I feel my data converge into clear findings, I start drafting the results section by developing a rough storyline from my assembled figures. I then usually shut down my research activities to concentrate on writing the introduction and discussion. I tend to break the introduction into three smaller sections: foundational background knowledge on the topic; specific aspects pertinent to the models and results presented; and main findings and brief conclusion. For the discussion, I start by summing up findings one figure at a time and integrating the newly presented data with the existing literature. A summary figure can also be very helpful. The final paragraph I leave for speculation and clear areas that require additional research.
—William Diehl, former postdoc in virology
I first collate all the figures and data—both positive and negative—that I think are likely to be relevant to the story. I’m more a visual person, so the next step is to generate a coherent storyboard of my figures to get a big-picture view of the project. This process helps me formulate an outline of the manuscript that I can use as a guide during the writing. Once I start writing the draft—and consequently also spend time looking more closely at the data—I constantly go back and forth to the literature to make sure I’m not missing anything about my topic and that I’m citing the right studies. Sometimes during this process, the story narrative can change a little, and that’s OK! I just go with the flow and see where the data takes me.
—Jessica S. Ho, postdoctoral researcher in microbiology at Icahn School of Medicine at Mount Sinai
I feel ready to turn my research into a paper when I have a set of results forming a clear storyline addressing a scientific and societal need. I start the writing process with an outline that serves more as a wish list for how I want the flow of the paper to go, bearing in mind the overall goal, specific aims, and main takeaways of the manuscript. I usually put all the results I have into the outline in one long, running results section with a summary of what they add to the storyline. This helps me recognize how results fit together, which aspects enhance the story, which parts are redundant or otherwise unnecessary, and whether there are any gaps or outstanding questions. Then the methods section is typically the first I write, as I find it relatively straightforward. During my research, I try to keep my code for data analysis organized and documented in such a way that writing the methods and results is effectively translating my scripts from R into a storyline. I often write the results section at the same time as the methods to try and mirror the flow. In the results, I also include some text describing the figures I want to generate or preliminary, hand-drawn figures. Before writing the introduction and discussion, I take a day or two to really dive into the literature and refresh my vision of where my work fits in the current state of the science. For me, the discussion is the most difficult section because there are so many different directions it can go when placing the results into the broader context. I try to focus on the big scientific and societal takeaways of the manuscript or the specific interests of the journal readership. The final aesthetically pleasing and well-formatted figures and tables are usually one of the last things I develop. I always save the abstract for the end. Assuming all analysis is complete before I start writing, preparing a paper usually takes me 2 to 3 months from blank page to submission.
—Marissa Kosnik, group leader in environmental toxicology at the Swiss Federal Institute of Aquatic Science and Technology
I start to think about the publication when I am still carrying out the research because, depending on the audience we plan to reach, the datasets that we use to develop our model and the experimental evaluation that we present can vastly differ. The target journal is chosen long before the writing process begins, and putting the paper together usually takes me 3 to 6 weeks. But sometimes, we realize during the writing that we need to do a few further experiments to make a strong point, which can prolong the process.
—Niklas Gebauer, Ph.D. student in machine learning at the Berlin Institute of Technology
In my theoretical field, I begin by deciding how to split the content into sections. I usually think about accompanying tables and figures to illustrate my theoretical results and prepare them before even starting the draft. I very often go back to the literature and evaluate how my work fits into the broader research context to develop the scientific story. Then I work on the more technical core of the paper and only later write the introduction, broader perspectives, and conclusions. For me, the hardest section is the introduction, where the main theoretical questions need to be made clear while avoiding jargon and technicalities as much as possible. I usually list the relevant references for the scientific context and the key results of the work, and then build around this. The abstract is the very last thing I do.
—Valentina Ros, researcher in statistical physics at Paris-Saclay University in France
Any further specifics about what should go into each part of the manuscript, or how it should be presented?
The introduction should describe the motivation for the work, the contribution to the field, and the key findings with enough detail to convey your message, but not so much to confuse or bore the reader. The discussion/conclusion section should discuss the implications of the work, its shortcomings, and future studies. Design your figures and tables so they are visually pleasing and can be understood without reading the main text, for example by putting labels on all axes and columns and describing the content in the captions. Guide the reader through the manuscript by using transition words to organize your arguments and beginning new paragraphs when starting new thoughts. Use simple, concise phrasing and avoid extremely long sentences. Bear in mind that, depending on the journal’s audience, you may need to adapt the language and explain some terms or concepts that would be common knowledge in other fields. For the title, I usually do a brainstorming session with all my co-authors, which is very effective and can be a lot of fun. Declarations are usually quick to write, but they should not be done last minute as it is important to consider conflicts of interest and to not forget any acknowledgments and funding.
The abstract is a sort of elevator pitch where you should be as clear and concise as possible so a prospective reader can judge whether they want to read your paper more fully. I approach it as two to three sentences covering each of the main sections. I have recently become more aware of how important keywords also are. I used to just throw together some words and phrases that I found relevant, but now I try to ensure that the keywords I use are concise, do not overlap with words in the title, and are common terms that a reader may search for when scouring the literature. Then, for the rest of the manuscript, it is crucial to balance what should be in the main text versus supplemental material so the storyline is clear. Within the manuscript, I describe the results that directly address the main research question and the core aspects of the methods, and I move any extra protocol information and supporting results to the supplemental material. Likewise, one must determine what information enhances the paper versus being superfluous. The introduction must provide just enough background that readers have a clear and convincing justification for the objectives of the manuscript, together with sources that readers can turn to if they want more information. Regarding the style of the paper, some authors prefer to use passive voice to avoid the use of “we” in the paper, but I prefer the active voice as it makes the manuscript feel more conversational. I try to give myself a lot of room for creativity, especially in figure generation, but here again, there is a balance between being creative and making something needlessly complex.
In the results section, in addition to describing the data I give a brief explanation of the rationale, hypothesis, and set up for each experiment to help the reader follow the logic of my work. The target journal does sometimes influence the writing and preparation of the manuscript, mostly in terms of the figures or dataset requirements, word counts, and reference formats.
In the discussion, I try to explain how the data supports my assertions and how strongly. I feel it is the most difficult section since there is always a risk of under-explaining your data. Giving some thought to how to give a twist to the title or make it unique is important so the paper can be more easily found and remembered in the tide of research manuscripts. For the acknowledgments, I usually keep a spreadsheet with all the individuals and organizations that have supported the research. Finally, I consult with my supervisors or institute's legal office to fill in the declarations about funding, potential conflicts of interest, and compliance with ethical standards and reporting guidelines. The choice of journal may affect the length of the manuscript and order of the sections. Formatting and citation style can easily be changed by working in reproducible workflows using LaTeX and RMarkdown.
—Roshan Paladugu, postdoctoral fellow at the Max Planck Institute for Evolutionary Anthropology
As the main author, at what point is it most crucial to get your supervisor and co-authors involved in the manuscript preparation?
My supervisor is involved throughout the writing process. We first make an outline of the manuscript together, as it allows us to have a better sense of what data and experiments are missing for the story and what we have left to do. Once we have this, we contact the co-authors and discuss our plan. Then we ask them to send us the finalized raw data, figures, and methods for the segments they have contributed to, and information about conflicts of interest and funding. We continue to seek their input until everyone is satisfied with the final manuscript. My supervisor and I usually select the target journal based on the novelty and direction of the story being told.
My Ph.D. mentor and I found a writing dance that worked for us. After chunks of time writing on my own, we would have 1 to 2 continuous hours of one-to-one meetings to make edits. They would explain the reasons for specific changes, which greatly helped me improve my writing skills and find my own voice. Once, in a frantic race toward submission with a collaboration, a conference call was held with three co-first authors (including me) and three co-senior authors present and each modifying different sections simultaneously on Google Docs. This was highly effective—the manuscript was ready to submit in about a month.
If I’m working on a multidisciplinary project with co-authors who were involved in the project conception, then I will look for their input earlier in the process to ensure that the storyline is clear and the interpretation is sensible to multiple disciplines. If I’m working on a paper where I’m more confident with the storyline, then I will wait until I have a structured draft with complete methods and results and at least bullet points of what the introduction and discussion will say before sending the paper around for feedback. The manuscript is ready for submission when all co-authors feel that the storyline is coherent, well-supported, and complete.
I usually meet my co-authors after data interpretation with an outline for a manuscript. All the collaborators are asked for their input on the manuscript structure and about suitable journals. Once the manuscript is complete and gets reviewed by all co-authors at least three times—with one month´s deadline—I consider it ready to be pushed off to the journal.
What writing challenges have you experienced, and what ways did you find around them? Are there some potential mistakes that you would like to warn early-career researchers against? Any further advice?
In the beginning, I can easily get tangled up in too much thinking and too little writing. I try to dive in with an idea and remind myself that at this stage it is more important to just get into the flow of writing, knowing that I can always rewrite and rearrange later. Whenever I get stuck, I will either read a bit of the literature or prepare figures for an hour. Beyond easing me into the academic style, these activities can provide new starting points for writing.
It is certainly daunting to look at a blank page knowing that a paper needs to come out of it! I find that checklists give me extra motivation and some leeway in choosing tasks to advance the paper. I also know that early in the morning with my first cup of coffee is the time when I am most productive and capable of overcoming writer’s block, and I have music that I only break out when I’m having trouble focusing that will help me get into a writing zone. But ultimately, while a manuscript could always be improved in some aspect, it is important to remember that there will never be a perfect manuscript. Falling down a rabbit hole with continuous reworking can prevent a good paper with important results from being out in the literature.
I find that taking a break or getting up to walk usually helps when I hit writer’s block. Sometimes, it can be several days before I start writing again. I use this in-between time to polish figures, collate my data sets, or complete additional control experiments.
A common mistake that novice STEM writers tend to make is to present figures in the chronological order in which the data was collected. Sometimes this can work, but there’s often a better way of presenting the data.
Do not limit the circle of people who provide comments on your manuscript to direct collaborators. Feedback from a colleague in a different discipline can help you identify issues you might otherwise overlook. Having an English native-speaker pal who can proofread your drafts and give comments can also be very helpful. Read other articles not only for their content but also to understand how they are structured. Also, make sure to check out the journal’s publishing guidelines and know the ethical standards in your field. But most importantly, be sure of what you are writing! I had to make a correction on my last published paper, where the units on a graph were incorrect. The error did not affect the calculations or conclusions, and we all know that everyone can make mistakes, but sometimes it is still embarrassing. Still, don’t let the fear take over. One should never stop trying and practice to get even better!