Volume 45, Issue 6 , Pages 485-487, September 2007
Writing a paper
Article Outline
You don’t write because you want to say something, you write because you’ve got something to say.
F Scott Fitzgerald
Surgeons tend to be perfectionists and have a good eye for detail. They often wish to share insights into their work, gained from research or clinical practice.
Often they are not trained in the art of communicating these insights on paper, despite having the ability to present their work at meetings in front of their peers. For instance they are often unaware of “technicalities” such as Condon's Type III error “ No data were obtained to support the conclusions expressed” – i.e. covering dubious conclusions with confusing language.1
It is perhaps appropriate, therefore, for me to outline the rules for the presentation of papers in journals. These are different from those for books, in which the personality (and idiosyncrasies) of the author are not only permissible but also important. Papers in journals are the way of communicating the maximum of information in the fewest, clearest, words. It is not an ego trip—save that for a book.
Most papers accepted by the journal are well written. Obviously writers whose first language is not English may require more help than others. However, language should always be “a precision tool, conforming to simple rules and conveying meaning logically”.2
One must remember at all times that the purpose of the paper is communication, and for this reason a “house style” is used, because the aim of such writing is not only to make yourself understood, but to make sure that you are not misunderstood.3
Having established the philosophy let me give you a brief, and obviously selective, introduction to writing a paper.
I am not going to deal with electronic submission, because each journal has its own rules. Suffice to say that you should follow them religiously.
So, assuming that you are writing a full-length paper, which is reporting a clinical trial, how do you start? Sir Austin Bradford Hill was the father of the controlled clinical trial in this country, and he laid out the order of a paper.4 He asked:
Why did you start?
What did you do?
What did you find?
What does it mean anyway?
In other words, Introduction, Patients and methods, Results, and Discussion. At this point it is necessary to get your data on to paper. There are many different ways of writing a first draft; use the one that suits you best. I write on A4 paper by hand, and then transcribe on to the computer (doing my first edit at the same time) but you may prefer to type directly on to the computer. Don’t stop to edit at this stage — just write. You can edit later. It is important to keep the flow going.
Answer Bradford Hill's questions directly, in the active voice. For example, “I/we wanted to know whether A was better than B for the treatment of X”. This is easier to read than “The purpose of this trial was to determine the efficacy of substance A and compare it in a random control clinical trial with the effect of substance B in the treatment of X”.
Do not be afraid of simple, everyday words. Readers will thank you, not to mention editors and referees. Keep it simple, and keep it as short as possible; imagine that you are going to have to pay £1 for every word you use.
Introduction
The Introduction should be just that: an introduction to the topic and why you thought that the topic was worth investigating. It usually fills about half an A4 page.
Patients and methods
The Patients and methods section should describe the comparability of the groups that you studied (sex [please not “gender”, which is a grammatical term], age, severity-as objectively as possible, current treatment, and so on) and give a brief resume of your statistical methods. Remember to mention the program that you used to analyse the data because the statistical adviser may wish to check the calculations, and programs do differ in some ways. Normally this section should fill about 2–2½ pages.
Results
In the Results section consider using tables as much as possible. If you can tabulate results, or put them into figures, do so. However, remember that one of the most common faults is to tabulate your results and then describe them in detail in the text. Don’t. Use one or the other. The ideal results section would read, “The results are shown in Tables 1–4”.
As Figures are often used as well as tables to illustrate results I will mention them here. They are, of course, partly a matter of the journal's chosen style, but there are rules. First of all, three-dimensional graphs are never acceptable in scientific publishing, because they are business graphics and designed to give only a vague impression.5, 6 They are fine as slides, because you the speaker will be there to present them and talk about them, but never in a written publication.
Each figure and table should stand on its own without reference to the text. This means that all abbreviations should be spelt out each time that they are used in the specific legend, or as a footnote. And please explain units of measurement. Remember that the United States still uses imperial measures, whereas most of the rest of the world uses Systeme Internationale (SI) units, so conversion factors are also helpful.
A word about statistical analysis. A p-value does not equal truth; there is no substitute for common sense and, if something does not make sense, it probably isn’t true. Never say that something is “NS”, meaning not significant (always give an exact p value). As Murray7 put it:
“If one obtains a ’non-significant’ result then one has failed to demonstrate a difference, which does not mean that one has established equivalence. A non-significant finding rather is analogous to the Scottish legal verdict of ’not proven’, which is generally if mischievously regarded as meaning ’not guilty, but don’t do it again.’ In other words, a non-significant finding often simply reflects a lack of evidence rather than a lack of difference.”
There is particular danger in using statistical analysis with small numbers. A useful rule is that if you have small groups and a significant difference, transfer one “event” from the experimental to the control group. Repeat your test, and if the difference is no longer significant, do not use statistical tests.8 People will have more respect for someone who says “We did not study enough patients, but the result looks promising” than for someone who tries to blind them with statistics.
Report normally distributed data as mean (SD) and skewed data as median (range) or (interquartile [25–75]). A good rough guide as to whether data are normally distributed or not is to see whether the SD is more than half the mean. If it is, they are skewed.9 Avoid the SEM; prefer the 95% confidence interval (CI) of the mean instead. Do not use the ±sign.6 Use a parametric test for normally distributed data, and a non-parametric test for skewed data.
Always report whole numbers together with percentages, never percentages alone. The reason for this is that for numbers of less than 20, percentages are misleading.8 For example, in an animal experiment 33.3% of the rats died, 33.3% survived, and the third rat ran away. It is difficult to estimate the length of the results section, as it depends on how much you have used Tables and Figures.
The Discussion is difficult to write, because there is such a temptation to mention and discuss everything that you have found. Restrain yourself. Pick out only your major points (particularly if they are controversial) to explain and elaborate.
Again, avoid the passive voice. “The skin was incised” is, for example, shorter and clearer than “An incision was made through the skin”. Use simple words: ”do“ is better than ”perform“ (you are not a seal) or ”carry out“ (you are not removing a corpse); and the use of nouns as adjectives should be avoided as it makes a paper difficult to read
It is always a bonus for the reader if you avoid clichés – prefer “clinically” to “in the clinical situation”, and “daily” to “on a daily basis” – and also vogue words such as “determine” and “enhance”. Table 1, Table 2 list some examples of tortured English and superfluous words.10
Table 1. Examples of tortured English (all of which have been taken from original manuscripts submitted for publication to various journals)
| Tortured English | Meaning |
|---|---|
| Excessive interposed adipose tissue | Fat |
| Mobility of the leg was present | The leg moved |
| In the author's opinion | I think |
| On account of the fact that | Because |
| One out of every two | Half |
| Surgical procedure | Operation |
| Exogenous nutrient elements | Food |
Table 2. Superfluous words (in bold)
| Positive action |
| Distinct possibility |
| Active consideration |
| Absolute end |
| In actual fact |
| Personal opinion |
| Real danger |
| Agonising reappraisal |
| Diametrically opposed |
There is nothing wrong with the words “killed” or “died” if that is what you mean; the euphemism “sacrifice” puts one in mind of what may have happened on the altar of a temple.
Again, use as few words as possible. You will find this difficult. Orwell11 wrote “It is easier – even quicker once you have the habit – to say ‘In my opinion it is a not unjustifiable assumption that.’ than ‘I think”’.
Conclusion
There are people who are natural writers, and there are those who are not. Even natural writers, however, find writing difficult and have to revise their manuscripts again and again.
Above all resist the temptation to “pad out” a submission with repetition and flowery prose. Remember that one of the most important scientific papers of the 20th century, Watson and Crick's description of the structure of DNA was published as a one page letter to Nature.12 If it can be said in 600 words and 2 tables do not be tempted to use 2000 words and 6 tables.
If you want to get a rough idea how readable your prose is, apply the Gunning Fog Index. The formula: 0.4
×(mean number of words in sentence+percentage of words of more than two syllables).13 The score ranges from 5 (very easy), 6 (easy), 7 (fairly easy), 9 (standard English), 12 (fairly difficult), 14–16 (difficult), to 16+ (very difficult). We should aim for a score of 7–9. The score of this paragraph is 7.
In summary, you are writing to communicate as clearly and succinctly as possible what you did, why you did it, and the conclusions that you have drawn from the evidence. Your peers will remember good practice, carefully investigated and set out, with judiciously drawn conclusions, for longer than they will wordy prose illogically and incorrectly presented with woolly and incomprehensible analysis.
References
- . Type III error. Arch Surg. 1986;121:877–878
- Layton PR. Grammar from a one-sided view. The Times 1989; June 26:10 (col 3).
- . Plain English: a user's guide. Harmondsworth: Penguin; 1987;
- . The reasons for writing. BMJ. 1966;ii:870–871
- . The glitter of the t table. Lancet. 1993;342:27–29
- . The visual display of quantitative information. 2nd ed.. Cheshire CT: Graphics Press; 2001;
- . Statistical aspects of research methodology. Br J Surg. 1991;78:777–781
- . How to report statistics in medicine. Philadelphia: American College of Physicians; 1997;p. 48
- . Statistical guidelines for contributors of medical journals. BMJ. 1983;286:1489–1493
- . Writing for publication. Br J Oral Maxillofac Surg. 1998;36:161–164
- . Inside the whale and other essays. London: Penguin; 1957;pp. 143–158
- . Effect of peer review and editing on the readability of articles published in Annals of Internal Medicine. JAMA. 1994;272:119–121
- . Nature. 1953;171:737–738
PII: S0266-4356(07)00357-9
doi:10.1016/j.bjoms.2007.07.198
© 2007 The British Association of Oral and Maxillofacial Surgeons. Published by Elsevier Inc. All rights reserved.
Volume 45, Issue 6 , Pages 485-487, September 2007
