22
How to Write a Good Scientific Paper
Introduction
Simply putting words down on paper is hard for most people. Besides being difficult, it’s very dull. Most of us would much rather think of new experiments, or carry out experiments, or analyse results of experiments. We all know, in our heart of hearts, that there isn’t much point in doing experiments if the work is never published. It is human nature to avoid difficult tasks, and writing papers ranks high on the list of difficult things. Yet writing a scientific paper is really not all that hard. We are not talking about Great Literature. Writing a major work of fiction might take 10 years. Scientific papers shouldn’t take more than a month or two. Why? Largely because most of the work has already been done. We will shortly discuss this in some detail.
But there are two aspects to writing a good scientific paper. The first is writing a scientific paper, which we will take up momentarily. The second is writing a good scientific paper, which is to say, writing well. Unfortunately, this second aspect is sadly neglected. Much scientific writing is awkward and graceless. It frequently obscures the subject and even fails to describe experiments clearly. Even worse, we have become so habituated to bad writing that we often associate good science with bad writing. We are uncomfortable with good science expressed with good writing. In many scientific journals, the most important papers are so hard to understand that the editor feels obligated to place an ‘editorial’ in the front of the issue in order to explain the paper so that other scientists can follow what is being done. And yet, scientific writing is a form of communication. Bad writing, by definition, is poor communication. Therefore, once we have discussed how to write a scientific paper, we will discuss how to write it well.
Structure of a Scientific Paper
To focus the discussion, I have proposed a basic rule of writing a scientific paper, specifically
The Basic Rule: Write the Paper Before You Do the Study!
Doesn’t this seem a bit… backwards? How can you write the paper before you know what the studies have shown? But, in fact, writing the paper first is relatively easy. In fact, most scientists have already done most of the work required. Notice that the Basic Rule doesn’t say to write the paper without doing the study. That would be very wrong, and regarded with disfavour. It is certainly true that you cannot finish the paper until the experimental studies have been done. But it is also true that you have to do most of the work of writing the paper before you can start the experimental studies.
A scientific paper is simply a piece of writing which describes a specific instance of the scientific process. Consider the nature of the scientific process. There are four defined stages: (i) identify a problem, (ii) formulate a hypothesis, (iii) conduct experiments and (iv) interpret the results.
To carry out a basic science or clinical study, one first identifies an issue or a problem. That may require some work, perhaps a great deal of it. There must be a thorough literature search. After all this preliminary work, the investigator knows enough to formulate a hypothesis. But the preliminary work is not over. The scientist must then design experiments (or must plan clinical studies) to test that hypothesis. Going into the laboratory and amassing data without any hypothesis is generally regarded as ‘phenomenology’, and produces the sort of paper which will be rejected. In short, before work is done in the laboratory (or on the wards, or in the medical record department), the investigator has done a great deal.
In fact, most of the material which will appear later in the paper is now available. The Introduction defines the problem and states the hypothesis. The Materials and Methods section outlines the plan for the experiments (or the clinical studies), provides details of critical assays, discusses data collection and analysis and documents all relevant approvals. If these two sections cannot be written before beginning the study, the planning process is faulty. Both may of course require modification later, based on what is learned during the study. The third section of a paper, the Results section, must obviously be written later. But much of the fourth section, Discussion, can be done at the time the work is being planned. In a grant proposal, there is usually a section called ‘Significance of Anticipated Results’. That’s very similar to the final Discussion section. Much of it deals with prior work or work by others, and that will not change. In fact, putting down on paper what others have found, and how the proposed experiments will be interpreted in view of the findings of others, may be very helpful in refining the experiments themselves. So it’s a very good idea to write at least a first draft of the Discussion early in the process. The Conclusion section usually has to wait on the experimental results. But the Acknowledgements can (and probably should) be written fairly early.
Sections of a Scientific Paper
Whether the scientist has written a draft of the paper early, or whether the paper is being written a year after the experiments are done and the investigator can barely remember what he or she was thinking when the experiments were planned, the paper itself has to be written. The several sections are quite stylised. The author doesn’t have to worry much about how to organise a paper, since there is a very standard format.
Title
The title comes first, but should be written last. Actually, it’s best to use a draft title, and then modify it into the final title after everything else is done. The title should describe the paper accurately and precisely. It can take a number of forms:
All of these are seen commonly. The important thing about a title isn’t its form, but its content. The title should communicate the specific topic of the paper. A paper titled ‘Drug Metabolism’ doesn’t communicate very much. A title such as ‘Isoniazid Metabolism in Patients with Heart Failure’ tells the prospective reader what the paper is about. Finally, the title shouldn’t be excessively dramatic. A title such as ‘Apoptosis: The Body’s Secret Weapon’ belongs to the tabloid newspapers.
Introduction Section
The Introduction section should be short, and usually consists of only two or three paragraphs. Do not bring in a long literature review. The most common error in the Introduction is to make it entirely too long, and to carry out the Discussion section in the Introduction. How much prior work should be put into the Introduction? The purpose of the section is to define the experimental hypothesis or the question being asked. This doesn’t mean an extensive discussion. The Introduction should make it clear why the work was done, and what it was designed to prove. References to previous work are entirely appropriate, if they make it reasonable for the reader to understand the present work.
Here is an example:
‘Most authors have assumed that hypoxia depletes energy stores (ref, ref). But Blake, Chu et al., have reported a recent study which showed that energy stores in hypoxic cells may be maintained at nearly normal levels (ref). The present series of experiments have been designed to resolve this question’.
The conflict in the literature is briefly pointed out, to put the study in perspective, and then the hypothesis is stated. The above example doesn’t represent the entire introduction, but it does summarise the essential elements.
Two paragraphs will suffice for an introduction. The first discusses why the problem or question was selected for study. The second paragraph presents the experimental hypothesis and outlines the reasoning behind it. Adding a third or even a fourth paragraph can be acceptable, especially for a complicated study.
The introduction should always contain an explicit statement of the question. This can be phrased, for example, as ‘The hypothesis of this study is that cellular adenosine nucleotides are depleted in hypoxic cells’. Or, could be put in the form of a question, as’the question this study should answer is…’ A new author especially should avoid a long introduction. I have sent many papers back to their authors for revision over the years. The most common criticisms of the Introduction section are excessive length and failure to state the hypothesis or question. Many reviewers have criticised papers for too long an introduction. Few, if any, have even criticised an author for too short an introduction.
Materials and Methods Section
The Materials and Methods section is very specific, very descriptive and—unfortunately—very dull. It should be of great interest to other investigators in the field, of mild interest to more casual readers, and will be read completely by very few. It should be designed to provide factual information only. There are several elements. First, this section must provide an outline of the experimental design (or clinical study design). This part of the section is most commonly unclear. At least a third of papers sent back for revisions have failed to explain the experiment clearly enough.
Later paragraphs should discuss how data were collected, how assays were done and what sort of statistical analysis was used. Many writers use subsections, like ‘Experimental Protocol’, ‘Data Collection’, ‘Biochemical assays’, and so on, which makes it easy for the reader to focus on those portions of the section which are of interest. There should always be a description of the statistical methods. For complicated analyses, this might be a whole subsection. For most short papers, the first paragraph of the section should contain details of the study design; the last paragraph should contain a description of the statistical analysis; and in between will be descriptions of critical, unusual or complicated assays.
One final item is important. Somewhere in the section should be a paragraph, or perhaps one or two sentences describing compliance with the ethical requirements of research. This should provide documentation of approval from the institutional review board for clinical studies, or from the local animal care and use committee for animal studies. If this detail is not present, most journals will return the paper for the author to include it before publication.
Results Section
The Results section consists of nothing but data. There should be no interpretation, nor references to other work. It’s best to use tables and graphs, if it makes the data clearer or easier to understand. The conventional standard for scientific papers is to report each piece of data only one time. That’s the goal, but a totally pure implementation of this standard may make a paper difficult to follow. It often makes the results clearer to use a graph of data which has been presented in the text or in a table. That being said, most writers use too many graphs, and not enough tables.
Photomicrographs or other photographic results (as, of ELISA results) should be used very sparingly. Frequently, they don’t show findings as well as the data in the tables. For photomicrographs in particular, the magnification, brightness and size should be adjusted so that the experimental and control pictures are similar. Access to a photo-editing program, such as Adobe Photoshop (Adobe Systems inc, 345 Park Ave, San Jose, CA, USA), Gimp (http://www.gimp.org/) or Paint. net (http://www.getpaint.net/), is essential for preparing photographs or photomicrographs for publication. They are invaluable to adjust size, brightness and even magnification to make pictures comparable. In particular, a good photo-editing program will usually allow colour photographs to be converted to black and white, while actually enhancing the points to be made. However, such programs should not be overused. It is painfully obvious that illustrations of spots on paper—such as Western blots—can be manipulated to show almost anything.
Graphical presentation varies by individual. Some measure of variability is essential, usually standard error or standard deviation. Assuming the data permits, all graphs should contain either standard error or standard deviation bars. Editors tend to have a preference for one or the other. As an editor, I have a slight preference for using standard error, because it makes it much easier for the reader to judge whether or not the results are significant. Others feel that standard error bars make the data look less variable—that is, ‘better’—and prefer standard deviations. But they are mathematically equivalent, and it makes no scientific difference at all. However, it is vital to make it clear on the graph or the legend which is being used. The significance (p value) should also be indicated, at least for p less than 0.05 (or 0.01), to make it very clear which differences are significant by statistical tests.
Statistical analysis itself can be complex. Simple experiments can be managed with means, standard deviations or standard errors, and t-tests. Once the experiment goes into three or four groups, analysis of variance is used. The statistical test used to evaluate significance should be specified, certainly in the Materials and Methods section, and in the tabular or text results. As a general rule, if the senior investigator is not absolutely sure of the statistical validity of the tests being used, then consultation with a statistician should be obtained. Many journals, in fact, will send a manuscript back to the author with a specific request that statistical consultation is required.
Finally, managing the paper with multiple experiments can be difficult. If the paper includes more than one type of experiment and analysis, it is best to divide the Results sections into subsections.
Discussion Section
The Discussion section should contain an analysis of the results together with a review of the relevant literature. The most important element is the interpretation of the results. The section should also show how previous studies relate to the present study. A useful guideline is to write the first paragraph of the section to interpret the findings of the study in terms of the problem, and should state whether the hypothesis has been proven or rejected. Some authors start the first paragraph with a general discussion of the problem, and spend three or four paragraphs going over the literature before finally delivering the interpretation of their results. Either way works. The last paragraph of the Discussion section should outline the conclusions that have been reached from the study, and possibly indicate where the authors intend to go from there. Some authors put this paragraph in a separate section, as Conclusions. Most discussion sections are too long, and a request to shorten the Discussion section is one of the most common reasons that papers are returned to the authors for revision.
Giving Credit
Finally, it is good manners to include an Acknowledgements section at the end. This should recognise contributions of others who were not co-authors, thank companies for contributing drugs or other materials and acknowledge financial support from grants or other sources. This is especially important, and is a requirement of many grant providing agencies.
References and Citations
Reporting references is highly variable, and each journal will have its own subtle variations. The most definitive reference for this is the AMA Manual of Style, currently in its 10th edition.1 Many scientists use custom computer programs to manage their references, usually ones that maintain the references in a database, and can adjust the actual reference format to suit the needs of a particular publication.
More important than how to prepare references is when to use them, and how. In general, any statement which cites the previous work of another should have a reference. Statements such as ‘most investigators believe’ or ‘previous studies have shown’ are very unimpressive, unless they are backed up with two or more references. Some authors will place references in the middle of a sentence, where the citation occurs, but my own preference is to place all references at the end of a sentence. If it isn’t clear just what the reference covers, then it should be obvious that the sentence is both too long and too complicated. For a subject that has expanded to a full paragraph, it is best to place the citation at the end of the paragraph. It is also acceptable to put the citation at the end of the topic sentence of the paragraph.
Writing the Abstract
Abstract preparation can be a minor art form in itself. It is best done last in the process. Like the paper itself, abstract writing is a highly formalised process and is basically quite easy. First, it is important to recognise limits on length. Frequently, abstracts should be no longer than 250 words. In fact, Medline truncates long abstracts after the 250th word, even if that is in the middle of a sentence. Second, many journals require structured abstracts. These contain separate headings as (i) Background, (ii) Methods (iii) Results, and (iv) Conclusions. Structured abstracts are a bit easier to follow, and they have the advantage (from the editor’s standpoint) that the author isn’t able to leave out a critical element. They are just good practice. I would recommend writing all abstracts in the structured format. No journal has ever turned down a paper because the abstract is too organised.
The easiest way to write an abstract is to go through the paper, write a sentence for each paragraph and then edit the excess. Each heading in the abstract contains only one to three sentences, so there is no room for elaboration. The Background should contain a statement of the hypothesis. Methods should detail the study design and any critical assay techniques. Results should include only the essential points. Conclusions should state what the findings mean, in one sentence, and whether the hypothesis was supported or rejected.
Some journals require a précis. This is a two- or three-sentence summary, usually intended for the table of contents. One sentence should define the problem and one or two sentences should summarise the findings. Assume the précis will be printed immediately below the title and do not repeat any of the material in the title. The précis is supposed to allow the reader to decide whether he or she wants to read the whole paper.
How to Write Good Scientific Prose
Scientific writing is a distinctive form of English which has unusually strict requirements for precision. The distinction between ‘nitrate’ and ‘nitrite’ would be totally unimportant to most English speakers, but may be very important to the reader of a scientific paper. And style is considered secondary to precision, sometimes, very secondary. In fact, an English professor would probably class the term, ‘scientific literature’ as an oxymoron. It is uncomfortable but true that most scientific writing is barely readable. If, as a scientist, you want people to actually read what you write, you must find a way to be both precise and readable.
The following section will, of necessity, concentrate on English. First, it is the international standard for scientific publications. Second, it is the native (or at least a major secondary) language of most of those who read this chapter. Third, this author has spent most of his life learning to write English, but has no such background in other languages.
The biggest problem with poor scientific writing is not simply that it is poor communication but it also reflects adversely on the writer. Anyone who writes poorly must be suspected of thinking poorly. One of the masters of the English language was George Orwell. About 60 years ago, he published an essay on the use of language.2 His point in this essay was that how we use language not only indicates how we think, and how well we think, but also shapes our thinking. Those who write clearly and precisely also tend to think clearly and precisely. Unhappily, the converse is also true.
To digress a moment, what characterises a ‘master of the English language?’ Let us consider one of Orwell’s most famous lines (from Animal Farm). Orwell was obsessed not only with language, but with its use in the political realm. Animal Farm marked the course of the animals’ revolution by changes in the Seven Commandments, which were painted upon the wall of the barn.3 Consider the last of the changes:
All animals are created equal but some animals are more equal than others.
It is composed of simple words and even repeats some of those words. Yet this one sentence perfectly captures the inherent contradiction of a supposedly egalitarian society. While Animal Farm satirises Russian communism, the meaning of this sentence is completely separable from that target. It can be—and has been—applied to any society that professes egalitarian ideals. At the same time, the sentence fits into the plot of the novel, and advances it. Now, that’s mastery of the language.
Scientific writing, first and foremost, is a distinctive style of writing. Most people have different writing styles for different occasions—one for a term paper, one for a personal letter, another for email and so on. Some ‘writing styles’ aren’t even writing, strictly speaking, such as text messaging on cell phones. Like any other writing styles, scientific writing has its own conventions. Some are more useful than others, but it’s important to know the rules. The main reason is so that when you break them, you are aware of doing so, and you do it for a reason.
The first person singular is rarely used in scientific writing, while the first person plural is often used. This is the scientific version of the ‘royal we’ used by European monarchs, back when there were European monarchs. One supposes the Queen of England still uses it. It makes a certain amount of sense, as most scientific papers have several authors. The second person, plural or singular, is almost never used. It can be used in instructional writing. You must have noticed the use of the second person in this chapter.
The third person is most commonly used in scientific writing, which is appropriate for expository prose. There’s nothing wrong with using the third person. Third person in the active voice is strong and readable. But third person in the passive voice is not. Its use is the most commonly committed sin in scientific writing.
What do we mean by ‘strong’? Editors and critics call prose ‘strong’ or ‘weak’. Besides using the active voice, strong prose is characterised by short words, short sentences, direct expressions, few qualifiers and few subordinate clauses. But out of necessity, scientific writing has to use long words with many qualifiers, including subordinate clauses. That’s simply the nature of scientific writing, and a major reason for calling scientific writing a separate and identifiable style. Nobody expects a scientific paper to read like a Hemingway short story.
That said, scientific writing still should be readable. It is one thing to use long, Latin-derived words, because they are the only precise words available. It is another to use them in preference to simpler words. How many times do authors use ‘demonstrate’ or ‘indicate’ or ‘reveal’ rather than ‘show’? Sentences in scientific papers commonly run over several lines. I recall one that ran to 20 lines in a manuscript. That’s nearly 500 words. No sentence should be that long. Subordinate clauses abound. Authors frequently use so many as to become lost in them, desperately placing a period at some arbitrary point so as to escape from their own complexity. Often, this is from a well-meant but failed effort to increase precision.
Consider the following excerpt from a Conclusion section.
‘The data which were produced during the course of these experiments are entirely consistent with the initial hypothesis that serum levels of free digoxin-equivalent activity can be found in an inverse relationship with the photometrically measured apparent concentration of serum albumin’.
Compare with
‘The data support the hypothesis that low-plasma digoxin correlates with high serum albumin’.
There are 41 words in the first, 14 words in the second. There are 9 subordinate clauses in the first, none in the second. The meaning is unchanged. The phrases “digoxin-equivalent activity” and “photometrically-measured apparent concentration” refer to details of the assays used, and would have been explained in the Materials and Methods section or in the Discussion. Their use here isn’t necessarily wrong, but interferes with the clarity of the stated conclusion.
A clear and direct writing style communicates meaning more completely and prevents ambiguity, besides enhancing readability. What are the rules of good style? Any brief treatment such as this must be both incomplete and dogmatic. That said, here are some rules to simplify and clarify.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
