Writing a Scientific Paper
Writing a Scientific Paper
Copyright 2006-7, Michael K. Gilson
It is worth writing well. The better your paper is written, the more people will read it. Don't let poor writing drive away scientists who should be reading and appreciating your scientific work. This updated (10/17/2007) version of "Writing a Scientific Paper" begins by discussing the sectioning of a paper, followed by guidance regarding style and process.
Parts of a Scientific Paper
The Abstract is a very short summary of the paper which must stand on its own because many journals and databases provide the abstract without the rest of the paper. The reader then uses the abstract to help decide whether or not it is worth the effort of looking at the main part of the paper.The Abstract should state the field of the work, the topic of the study, the methodologic approach, the main results, and the main implications. Avoid unspecific summaries, such as "important insights were gained"; or if this seems unavoidable, try at least to squeeze in a sample insight. Similarly, instead of merely saying a quantity was found to be "large", give the actual value or a range of values. It is usually easiest to write the Abstract after the rest of the paper is done.
The Introduction section should immediately let the reader know the field in which the paper lies, and the specific topic or question you work addresses. It should then briefly summarize prior work addressing this topic, making sure to give appropriate citations to other researchers. When the prior literature is extensive, citing recent review papers can be a good way to keep your introduction from growing too long. Key papers that focus on issues addressed by your paper should be noted specifically, however. Then explain briefly the approach your study takes to the problem at hand, and close with a terse description of the structure of your paper.
The Methods section is a technical description of the methods used in your study. Ideally, it is complete enough to allow another skilled scientist to replicate your work. If the methods are complex, including some material in Supplementary Material can be a good way to be complete without making the print version of the paper overly long. Try to start the Methods section with a paragraph explaining the overall strategy of the methods. Then provide a series of subsections that describe each method in detail. Each subsection should provide a very brief, intuitive explanation of the method, followed by the details. This structure allows the reader quickly to get a sense for what you did. He or she can read the details later if needed. To quote Ben Schneiderman, "Overview first, then details on demand." Occasionally, a few results may be put into the Methods section; for example, the results of a study establishing the parameters of a method. When in doubt, however, keep all results in the Results section.
The Results section should crisply present your new data, and normally contains most or all of the paper's tables and figures. It can include a very small amount of interpretation and reasoning. For example, Results can say, "Good agreement was obtained, as shown in Table x, which presents..."; or: "In order to evaluate the sensitivity of the results to variable y, the calculations were repeated with two additional values of this parameter, as shown in Figure z." However, more extended analysis belongs in the Discussion section. Avoid the common mistake of allowing methodological details to infiltrate the Results section. This happens most commonly when the Results section becomes a narrative which describes the need for studies beyond those originally envisioned and detailed in Methods, and proceeds to describe additional methods and the results of their application. This makes for a sloppy and confusing paper, so if you realize this has happened, move the additional methods back to the Methods section. The Results section can still explain -- briefly -- the reason for the additional studies.
The Discussion section explains what you have concluded from your results and why, and explains the implications for the field. It should not present further data, but it can include a figure or two to help explain a new model based on the data, for example, and it can reference additional papers as it puts your work into perspective.
The Introduction and Discussion are the two most flexible parts of the paper, and the Discussion is more flexible than the Introduction. Once you are comfortable with the basics of paper-writing, you can start working on writing lyrical and thought-provoking Introduction and Discussion sections. Keep in mind, though, that your comments should be connected with the data you presented, and should be well-reasoned.
A Conclusions section can be used to briefly summarize the main points of a complicated paper. However, this section is optional and should be omitted from a straightforward paper whose main points are already apparent.
A place to thank the individuals and organizations that played a supporting role in the research and paper-writing. Check whether your grantor requires a specific format or disclaimer; the NIH does.
Style and Process
Make your points.
Decide what are the key conclusions of your work and state them clearly and succinctly in the Abstract and in the Discussion and/or Conclusions sections. Far too many papers say that "important insights were gained", but do not explicitly state the insights. This leaves the reader suspicious that no insights were, in fact, gained.
Be mindful of the context of your work.
A research project takes place within a specific intellectual and social setting, and a sophisticated paper is written with this context in mind. The consequences include exploration, brief or lyrical, of potential implications of your results; acknowledgement, muted or forceful, of related controversies; and well-chosen terminology.
Beware of writing long, boring descriptions of your results. Focus on the results that support your conclusions. The other details can be summarized in tables, figures, and/or supplementary materials. Paragraphs should almost always be less than 1 double-spaced page in length.
Scientific papers are intrinsically hard to understand; consistency can make yours easier to read. For example:
- Use consistent terminology even at the risk of seeming repetitive. Clarity is more important than elegance. (Thanks to Barry Honig for this paraphrased advice.)
- The order of the rows of data tables should match the order of presentation in Results. Then your reader can follow the text by going down the rows one by one.
- The column headers of your tables should be consistent across tables. For example, if you use G for free energy in one table, use it in the other tables as well.
Use correct grammar.
Incorrect grammar is jarring to the reader; more importantly, it obscures your meaning. Do not expect your reader to finish reading a paper whose every sentence must be deciphered. If English is not your native language, don't hesitate to get help! Many universities employ professional grant writers who can also help with your papers.
Be sparing with italics and bold font
Use italics or bold font only when powerful emphasis is needed. Overusing these font changes will make your text cluttered and sloppy and will also deprive them of their ability to focus the reader's attention. Avoid the temptation to use italic or bold fonts to rescue unclear writing. It doesn't work. Italics and bold font read as a raising of the voice and are similarly ineffective at generating clarity.
Make your figures clear and simple.
A figure is a way to help your reader to understand your data more easily, so strive for clarity and simplicity. Give the most significant or authoritative curve in a graph the most authorititive appearance; typically, this will be a heavy, dark-colored, solid line. The other lines can be made dotted, dashed, lighter in color, etc. The most significant or authoritative data might be from the most reliable experiments or the highest-quality calculations, or from a novel method with which other methods are being compared.
Unless you are skilled at embedding tables and figures in just the right places in the text, put them at the end of the paper, subsequent to the citations: tables first, and then figures. This will allow your reader to find them easily. For in-house drafts, it is helpful to place the table captions and figure legends directly below their respective tables and figures. However, you may need to group the legends in a separate section in the version that is submitted for publication.
Color figures are appealing but may be costly to print and they do not add genuine value to a figure if it can be rendered equally clearly with black and white. For example, three curves in a graph can be readily distinguished by making one solid, one dotted, and the third dashed. When you do use color, avoid gaudiness. The default colors provided by Excel are neither visually appealing nor particularly clear, so do not rely on them for presentation-quality graphics.
Cite prior work
For the sake of documentation, not to mention courtesy to your colleagues, almost every statement of scientific fact -- or supposed fact -- in a paper should be accompanied by a citation. Exceptions are typically made for time-honored textbook material, such as Coulomb's law, or possibly other information that may be regarded as generally known; but when in doubt, cite.
Provide supplementary materials.
A paper is supposed to allow the reader to reproduce your work. This ideal is not always achievable, but providing supplementary files of data, structural information, methods, graphs, etc., can go a long way. It is also an act of good scientific citizenship.
Put the date below the title and author list each draft of a work, so that a reader with two versions of it can be sure of reviewing the latest one. Include page numbers so colleagues and referees can easily reference sections of section when discussing the paper with you. Make sure Figure and Table number matches the text correctly, even in a rough draft, so that your colleagues do not waste time struggling to determine which figure or table you are describing in the text.