Papers are the currency of future employment in science. They are the second thing that most people look at when evaluating resumes (after where you were trained and what your skills are). High numbers of publications signal to potential employers that you are highly productive and a “finisher” of projects. Publications in well-respected journals signifies that your work is creative and meaningful to other scientists. Completed patent applications demonstrate that your work is useful beyond academic curiosity. Thus, it is imperative that research gets documented and published. How, then do you get a paper? There are a few guiding principles that will help you get higher quality papers more quickly. (see also this infographic, these 11 points to getting your paper taken seriously, and a helpful online short course on writing)
- Write daily. At the end of each day (at least), document what you did, what you learned, what questions remain, and how you will answer those the next day. This paper trail not only allows you to know exactly how you got your results (you WILL forget), but it also gives you figures and text to add to a paper. A research paper really is just a progress report, which includes set of questions, analyses, and resulting answers, and so this daily writing will quickly lead to papers.
- Think carefully about what goes where: These essays provide some useful insights into what goes where: writing science as a story, and the parts of a paper
- The title: It should contain key words that will catch the eye of those you want to read it. It is often best to have it be a short statement of the key discovery in your paper. This is often written last.
- The abstract: a few sentences that describes the whole paper. 1 sentence of background, 1 sentence stating a dilemma or open question. 1 sentence describing how you answered that. A couple sentences stating your results. 1 sentence describing the broader impact of your results. Abstracts should be short… no longer than 200 words. Some journals require them to be as short as 150 or 125 words. Keep it short and clear. (this link has some additional insights)
- The results: this section is all questions, analysis, and interpretation. Specifically, each sub heading will be a discovery. Then the text will start with a transition, mentioning the previous finding you presented, and then leading into the obvious next question. There might need to be a few sentences of additional background you will add. Then the rest of each subheading then shows the analyses that answers the question, allowing you to make the statement you used for your header. The results section is a narrative for the data and analysis in your study, complete with context, analysis and interpretation. Avoid just listing facts, since that requires a reader to remember those details until they are told what they mean and why they care in the discussion, which doesn’t really work for most readers. Thus, break the results sections into subsections where each has a header that states the take-home message of the following results. Then within 1-2 paragraphs you will provide context, the hypothesis, the rationale for the analyses/experiments, the results, and then a couple lines about what it means. This usually breaks down to 1-2 lines for the background/context for the analyses/experiments and the question/dilemma/hypothesis. Then you’ll have several lines describing the approach and data. Then you end with a couple sentences interpreting those results.
- The discussion/conclusion: In this section you take your results and place them in the greater context of the field. Given what is known, what are the implications of your results? What are the remaining open questions? How might your findings be useful? The discussion section is a space where you’ll describe the results more as a whole, and place them in the broader context of the scientific literature. This is where you handle more of the extrapolation or mild speculation.
- The methods: Go into enough detail that someone can reproduce what you did. If the journal requires methods before the results, be brief and have a supplementary methods section with greater details. As a general rule we publish our code on our lab GitHub account, to provide transparency and facilitate reproducibility.
- Figures and figure captions: Figures are the core feature of your paper. They should clearly show the data. Usually each plot should have only one message, and in selecting how to present your data, seek out visualizations that reduce distractions from the finding you are presenting. For the figure captions, assume the reader may not have actually read the main text of your paper. Thus, try to concisely explain what the plot shows. Be sure to explain all abbreviations and present the statistics, error bars, and sample size (n).
- Write your results section first. This could just be pages from your daily research diary. Some people prefer to start from figures or plots they made. Others prefer to start with text. Either way, ALWAYS start from your results. The results section will determine which methods you need to document. The results will dictate the discussion section… i.e. how your results fit in the context of the greater scientific community. Then the results, methods, and discussion will dictate what background information is needed for the introduction.
- Organize your paper by questions you can answer and resulting discoveries. For every result in the results section, make sure there that you spend a couple sentences stating the open question or motivating your analysis. Then describe your result, and provide a short summary of how the result answers the question. If you document your research results daily, you this will be easy to do since you’ll have the results and your thinking about the results already written.
- Provide ample supplementary material, and ensure that tables are clearly annotated.
- Remove all jargon from your paper. Think critically of who would need additional explanation to understand what you mean by each term you use. You’ll want to avoid using terms that others outside of our immediate community would not understand. Some jargon is allowable, but only if it is clearly defined prior to its use.
- Anticipate what the reviewers will expect
a. Clarity in arguments. The questions must be addressed by the results, and the conclusions must be supported by the data. If your data doesn’t support your claims, step back and figure out (i) what must be true to make the claim and (ii) what claims would your data support?
b. Keeping your work on topic and free from tangents. Many tangents are essential validations and tests for robustness of your claims. However many of these items can be detailed in your supplementary material.
c. Data should be accessible, in processed and raw formats. There are several databases for different types of data. The NCBI GEO allows you deposit transcriptomic data. The NCBI SRA allows you to deposit large sequencing files. GitHub can host code. Synapse allows you to upload diverse project files. Wherever you host the data, it is important to have a stable accession number or DOI, and clear documentation describing the data.
d. Well organized and professional-looking documents. Tables and figures should be numbered in the order as they appear in the text. Grammar and spelling should be near perfect upon submission. Publication citations should be correct (use a citation manager such as Mendeley or PaperPile (we have a lab account), and import all of the citations using a standard approach such as through the PubMed records: https://www.mendeley.com/import/). Number the lines in your submission to make it easy for them to point out where revisions are needed.
- Your papers are often scrutinized by employers. The clarity and professionalism that is (or isn’t) demonstrated by them will impact hiring decisions. Make sure that their quality comes through.
- Cover letters are your first chance to get an editor excited enough to send your paper out for review. The cover letter is not just a formality that accompanies a paper when you submit it. It is a chance to let the editor get excited about your work, and you can often get away with stronger claims since the letters are not peer reviewed. You want to keep it succinct. It is often like an abstract in which it provides a background of the problem (i.e., the state of the art, and the big open question). Then in the second paragraph, you usually provide the name of the paper, and what you have solved in the study.
Other items to improve clarity
Paragraph structure. A well written paragraph will have a key idea, arguments that support it and a conclusion. The key idea is usually found in the first three sentences.
Other items. For a nice satirical view (that has a lot of truth in it) see this essay: http://www.sciencemag.org/careers/2012/03/how-write-scientist
Dealing with revisions
1. Carefully read the reviews. Any misunderstandings are legitimate and more often reflect ambiguities in how the paper was written and/or structured. Other readers will have similar misunderstandings if you don’t address them.
2. Determine which comments are essential to the main points of the study and which are tangential. Try to address both, but put all tangential analyses in the supplement.
3. When rebutting, cordially reply, acknowledge their comments and provide clear replies, and then highlight the strengths of your work.
4. In the resubmission, provide a version with all of the edits visible (maybe with blue text).
5. Don’t blow off the reviewer comments. If they feel it is important, it IS important. If it’s flat out wrong or if there is clear antagonism, then these issues can be brought up to the editor, not the reviewer.
A note about my (Dr. Lewis’) edits
As the researcher leading your study, you are aware of all of the intricate details. If you have been thorough, you will best understand many of the limitations and assumptions of your work. Thus, you are responsible to ensure that the reporting of your work is honest and thorough. As the PI, I am under the same obligation to ensure that anything we publish is correct. I have different experiences and skill sets, and so I might come to different conclusions than you. I also may not have seen the details of all of your analyses. Therefore, as we write a paper together, it is critical that we have open discussion about what the results mean. I might have experience that casts the results in a different light, or highlights a unique spin that you might not have thought about. While I try not to do so, I might overstate a result, since I might not be aware of some limitations you are aware of. Thus, ensure that your work is reported honestly, and don’t hesitate to talk with me if you feel I am overstating or misinterpreting something. Similarly, if I disagree with you on a point, make sure we discuss it, since I likely have reasons for my opinions or interpretations. The ultimate goal is to maximize the impact of your research while making sure it’s accurately presented.