William O. Hancock
Pennsylvania State University
Member, Biophysical Society Publications Committee
“I have never thought of myself as a good writer. But I’m one of the world’s great rewriters.”
James A. Michener
Part 1 of this series covered the task of transforming data in your lab notebook and thoughts in your head into a first full draft of your manuscript. The next task is to convert this rough draft into a polished manuscript that you can publish and be proud of. This process requires streamlining your message, honing your logic, and achieving clarity and conciseness in your prose. You will likely work through a number of drafts, and revising will probably take significantly longer than writing your first draft, but this effort is essential to create a publication-quality manuscript. Here I detail the key steps of this process.
Revisit your story
Ask yourself: Have I achieved my goal of presenting a compelling story for a specific audience? Don’t worry that the topic may have drifted far from where you started when you first sat down to write. Your story should be presented as a logical progression of experiments that build upon one another to convince the reader of your main point. Hence, consider the logic and try to think from the point of view of the reader. You may decide at this point to significantly re-sequence your figures and the subsections that make up the Results section. Don’t be afraid of “major surgery” as moving big pieces is easy, and a smooth and logical flow is essential. You may also realize that one (or more) figures contributes little to the essential narrative and can therefore be deleted or demoted to Supplemental Information. If you find yourself holding on too tightly to your hard-won text or plots, keep in mind the following quotes:
“In writing, you must kill all your darlings.”
“The more you leave out, the more you highlight what you leave in.”
Before setting out to revise your first draft, consult the Guide for Authors for the journal you are targeting, and follow word count, formatting, and figure guidelines. Doing this in advance will save you a lot of later work during the final journal submission steps.
Hone your writing
Now it’s time to pick apart your text and to tighten up your writing to maximize the clarity and impact of your message. There are many good writing resources available, but here I’ll highlight some key points:
- Each paragraph should make a single point that is ideally presented in the first sentence (the topic sentence). The last sentence of a paragraph should link it to the topic of the next paragraph. Some people write outlines with the first sentence of each paragraph written out and write a draft from there. That is a good practice, and when revising you can do this retroactively to track the overall organization of the manuscript.
- When writing, strive to be clear as well as terse. Don’t use extra words (instead of “at this point in time” use “now”; instead of “a large majority of” use “most”). Don’t use pompous language (replace “utilize” with “use”; avoid the phrase “needless to say”). Never use the word “believe” in scientific writing. Watch out for the word “prove”; instead use “suggest,” “indicate,” or “are consistent with.” It is also best to use the active voice when writing.
- Avoid lab jargon. Consider renaming your constructs or methods from the terms that you routinely use in the lab to more specific terms that readers can understand and remember, and that are consistent with previous use in the literature.
- Minimize acronyms because, although they save space, they are one more thing the reader must keep in their mind. So, err on the side of clarity and inclusiveness (broad readership), and when possible write them out.
Think about your audience
As you hone your writing, maintain a focus on educating and informing your reader — try to make it easy for them. In the Introduction, think of the essential background material they need to know in order to understand your study. In the Results, clearly explain what the data do and do not say and emphasize the most important data. In the Discussion, clearly explain the implications (as well as the limits) of the work and how it relates to what has been done before.
One way to help your reader understand and remember your message is through repetition. There is a useful old saying: “Tell ‘em what you’re gonna’ tell ‘em … tell ‘em’ … tell em’ what you told ‘em.” In the structure of a scientific manuscript this means that in the last paragraph of the Introduction you need to preview the results, in the Results you need to clearly present the findings, and in the Discussion you need to reiterate and expand on the findings.
A second strategy is to build up from the highly believable (established or simple) to the less believable (new) (Senturia, 2003). At the level of the entire manuscript, this means the Introduction sets up what is known (believable) and the Discussion allows for your speculation and making links to other work (less believable). This idea also applies to the Results — you should generally start with the simplest results and build up to the most novel and surprising. You are establishing the readers’ (and reviewers’) trust and providing them with a firm foundation on which to interpret your most exciting findings.
A final point is: Don’t overestimate how much information a reader can absorb and remember. There is always a temptation to present all of your data and make as many points as possible. However, more data can paradoxically reduce the impact of a paper by diluting the message. If your results revolve around a single central point of the paper, you have a good chance of having the reader come away with that point and remember it hours, days, or weeks later. If you are trying to make three loosely related points, your odds go way down. Hence, consider cutting and demoting some data to Supplemental Information — or in extreme cases — even splitting a paper that is bursting at its seams into two.
Make your figures beautiful
Revisit your figures to ensure that they are informative and uncluttered, and that they connect tightly to the text in the Results section. Every panel of every figure should be referenced in the text (if you don’t reference a panel, cut it). Think of the key point you want to get across in each panel, and use that to guide precisely how you want to plot your data. Can you remove non-essential data? Change symbols or add labels or lines to emphasize the key point? A few points to remember:
- Make your symbols sufficiently large to see, and make them consistent throughout the manuscript. Are the axes clearly labeled with sufficiently large fonts (keep in mind that figures may be reduced in size by the journal)? Consider the range — ideally start with zero at each origin and choose a maximum value on each axis that highlights the important variation of the data and also shows any plateau effect.
- Are you plotting the data in the optimal way? Bar plots are notorious; not only do they distill a distribution down to a single mean but, because of equal spacing on the x-axis, they can obscure important time and concentration dependencies. For measurements that depend on a quantitative variable, consider an x-y scatter plot. Or, instead of presenting a simple mean or a “bar and whiskers” plot, consider using a “Bean Plot” for moderate N values to show every individual measurement, or a “Violin Plot” for high N values to show their distribution (Weissgerber et al., 2015; Spitzer et al., 2014).
- All images should have scale bars that are labeled with units on the figure or in the figure legend. Ask yourself whether you should crop to emphasize the key element in the figure. Avoid nonlinear contrast enhancement in images, gels, and blots.
- Consider what data to put into Supplemental Information. Are there raw data that can be presented that are informative? Are there key control experiments that are important but don’t fit particularly well in the main results? The phrase “data not shown” should be avoided if possible (some journals even prohibit it), and the data instead should be included as Supplemental Data. However, avoid the temptation of putting extra data into Supplemental just because you did the experiments and you want to put it somewhere.
Honing specific sections
Does your first paragraph set up the paper? It should not be overly general background information; instead it should focus the questions being addressed. Is referencing correct throughout the Introduction? Apart from the most general statements, any time you state that something is “known” or you are stating a “fact,” you need to reference it (using original research articles rather than reviews where possible). Avoid excessive self-referencing. Avoid long strings of references; a general rule of thumb is that no more than three references are needed for a given point. Finally, the last paragraph of the Introduction should briefly summarize the key results (“Tell ‘em what you’re gonna’ to tell ‘em”), and should serve as a transition to the Results section, and it should tie to the first paragraph of the Discussion.
Materials and Methods:
The theoretical goal is that the methods you write out should provide sufficient information for others to repeat your experiments, but this is difficult to do in practice. Minimize text by referencing previous work and by describing any alterations in the protocol(s) you used. Consider putting detailed methods and derivations into a Supplemental Methods section.
- Generally, every symbol in every figure should have an error bar that is defined in the figure legend and in the text. Standard Deviation describes the scatter in the sample, Standard Error of the Mean is used to determine statistical significance.
- Beware of R-squared, which is a statistical measure of how close the data are to the fitted regression line. It does not denote statistical significance and is inappropriate for nonlinear curve fits. Consider an F-test.
- Significant Digits (General Rule of Thumb): Experimental precision limits the significant figures. To allow for later calculations, present uncertainty in a measurement (SD or SEM) with two significant digits and present the mean with one significant digit beyond the largest digit in the uncertainty. So, 3.4471 +/- 0.238 should be 3.45 +/- 0.24.
The first paragraph of the Discussion should briefly summarize the Results (“Tell em’ what you told ‘em”), and it should set up the entire Discussion that follows. You should strive to extract as much insight from your data as possible by: (1) making links between different results that you present, (2) connecting your results to published work, and (3) modeling, simulating, or carrying out further analysis of your data, where possible. You have license to speculate, but it has its limits. Be sure to note the limitations of your study and your methods. Be sure to properly cite your colleagues and competitors, and to site all relevant studies that came before. In the concluding paragraph avoid a generic call for more research, and instead place your work into a larger perspective and relate it to the original questions stated in the Introduction.
Before submitting your polished manuscript to a journal, give it to lab mates and colleagues and solicit their feedback. Don’t be defensive in responding to their constructive criticism. If there are key points that they do not understand, expect reviewers to have the same problems, and work to clarify your message. Finally, before submitting your manuscript, make sure that pages are numbered. And good luck with your submission!
References and Resources
S.D. Senturia. How to Avoid the Reviewer’s Axe: One Editor’s View. J. Micromechanical Systems, 12(3):229–232 (2003).
- A paper full of sage advice on organizing a paper and persuading your reader.
G.M. Whitesides. Whitesides’ Group: Writing a Paper. Adv. Materials. 15(16): 1375–1377 (2003).
- An excellent guide that advocates generating paper outlines early and building them into full manuscripts.
W.A. Wells. Me Write Pretty One Day: How to Write a Good Scientific Paper. J. Cell Biol. 165:157–158 (2004).
- Gives good overview of structuring a paper and developing a narrative.
- Spitzer, J. Wildenhain, J. Rappsilber, and M. Tyers. BoxPlotR: A Web Tool for Generation of
Box Plots. Nature Methods, 11(2):121–122 (2014).
- Advocates for using bean and violin plots to show distributions, rather than bar charts with means or box and whiskers plots.
T.L. Weissgerber, N.M. Milic, S.J. Winham, V.D. Garovic. Beyond Bar and Line Graphs: Time for a New Data Presentation Paradigm. PLoS Biol, 13(4): e1002128.doi:10.1371/journal.pbio.1002128 (2015).
- Demonstrates how much information about distributions and outliers is lost when using bar graphs, and suggests alternative approaches.
Navigating peer review and the publication process will be the subject of Part 3, published Friday, March 10.
Interested in learning more about writing a good biophysics article? Want a chance to ask questions? Dr. Hancock will present a webinar on this subject today, March 10, at 1:00 PM ET. Register at http://www.biophysics.org/Education/Webinars