Writing Science


The writer's object is - or should be - to hold the reader's attention (Barbara Tuchman)


Who am I to pretend to give advice on writing to young researchers? After 10 years working with scientific journals as Associate Editor (Bulletin de Minéralogie, Geochimica et Cosmochimica Acta, Chemical Geology), I became Editor of Earth and Planetary Science Letters (1993-2000) dealing with ~140 manuscripts per year and then Editor-in-Chief of the Journal of Geophysical Research – Solid Earth from 2001-2004 (650 manuscripts per year). I have put this page on the web because I have the impression that some things simply do not change. The kinds of errors that I made as a PhD student and the types of problems that I have had with reviewers and editors with my own articles keep repeating themselves after 30 years – I find myself making the same comments to junior and often to more seasoned authors that drove me up the wall at 25 years of age. As an Editor, and especially as a French Editor, it is a terrible feeling to have to reject a manuscript by a junior scientist because the ideas are not expressed clearly, when other criteria indicate that the author is capable of making important contributions. Some easily irritable reviewers will readily reject a poorly written manuscript by a first-time author, even when the manuscript may contain great ideas that the author is simply not capable of expressing. Don’t forget that a reviewer irritated by inexcusable faults or by a poor organization will probably lose some of her/his initial goodwill towards the science being presented.

 If you are an experienced writer, you can forget this first paragraph. I assume that you now have all your results at hand and a full set of references.

  1. Pick up ALL the papers in turn that you think are relevant to your story, even those written by the guy who persists in not quoting your own work.
  2. Lift (copy) one, or maximum two, sentences that you think encapsulate the content of each paper and shamelessly paste them into an empty text document.
  3. Once you are done, move your thefts around and organize them into a logical sequence of ideas. If somewhere there is a break of logic it probably means that you need to gather the papers into groups each of them relevant to a particular argument. You will have to organize the groups in a star-like fashion, in fact to make a list of higher-level arguments.
  4. If this fails, it means that either you are missing some critical papers or that you are breaking the sacred rule of “one paper-one point”.
  5. If the citations do not follow a chronological order, it probably means that you left out some founding papers, which is a sign of scholarly deficiency (mistaken by older people as arrogance). A good thing is that you will probably have to forget review papers and focus on original findings.
  6. If you are still left with papers that do not belong anywhere, ask yourself if they really are part of your story and try to leave them out.
  7. In both fairness and for pleasant reading, rewrite all the lifted sentences into a creamy coherent text.
  8. The seamless result of the previous steps should be the context of your work and allow you to discuss why you actually bothered to collect all these new data.
  9. Add a couple of simple and strong liminal sentences that will capture the essence of the problem.
  10. Wow, you now have a nice introduction and you finally understood why you started working on this project!

Leaving the introduction for later is ill-advised policy, as it condones the pernicious habit we all have to sidetrack endlessly.

Now, you are in business. Let us talk about important technical points. Remember that you are writing to be understood by the readers (and first by the reviewers), not to impress them. You goal is to communicate, not to scare or bore anybody away. Although I had initially listed a number of typical problems for manuscripts written by young – and less young – French authors, most of these points actually appear in manuscripts written by scientists of all walks of life regardless of their country of origin:

  1. As we already alluded to: one paper, one point! One idea, one article. Don’t write an article containing a number of different overlapping or parallel ideas. If you have two ideas, write two articles and do not write your magnum opus that no one will understand. Papers of the type Part 1, 2, 3, etc. are very unpopular and perceived as arrogant because they are based on the assumption that all of the parts will be accepted in the format decided by the author (in defense, an author purportedly said, ‘why not, Beethoven wrote 9 symphonies, didn’t he?’).
  1. Put yourself in the position of the potential reader of your paper and ask yourself two questions: is it worth my time to read this paper, and what will I get out of it? Most readers of your papers (who themselves represent a tiny fraction of the scientific community) will perhaps read the abstract and conclusions, look at a couple of figures (without reading the captions) and check whether they have been cited. All of this takes about 45 seconds. If you want to be read by more than the five colleagues around the world who work on exactly the same topic as you, make it short and make it readable to a broad readership. Chances that readers will reach the end of your paper are better when the text is short.
  1. English is often a scapegoat: 4 times out of 5, an incomprehensible sentence in English will not translate in your own language because the very basic formulation of the idea is wrong, not the words. In other words, don’t blame the language. If you feel that you can’t find the proper word to express your idea, try: (1) leave this word out and see whether the sentence reads well anyway, and (2) use an alternate word with almost the same meaning and use a thesaurus, the writer’s best friend, to zoom on the right term.
  1. Clean up after yourself! Spell-checkers and grammar-checkers are part of all word processors (even LaTeX). There is nothing that ticks off a reviewer more than dealing with a large number of errors when a single check can catch. Not using a spell-checker and a grammar-checker is essentially the same as dropping candy wrappers on the street and expecting someone to pick them up for you.
  1. Title: it may look prestigious to be general in the title, but turgid headlines tend to make people shrug. Avoid ‘A model of the world: example of my backyard’. Don’t write an article entitled ‘A global model for the seismicity of the Earth: example of the Santa Monica event’, but rather ‘The moment tensor of the Santa Monica earthquake’. Avoid two-part titles that are separated by a colon, they are pompous.
  1. Names: Napoléon Bonaparte and not Bonaparte Napoléon. Even in French, one says ‘pré-nom’ for first name, but nobody listens: North American telephone books are full of Dr Philippe, Pr Jean-Paul, and Mrs Georgette, etc., who confused the ‘first name’ (prénom) and the ‘last name’ (nom de famille). Putting the last name in capital letters is perturbing, tacky, and a waste of time.
  1. The Abstract provides a summary of the paper. It needs to be concise, because if it is not, the reader will become discouraged and will move to the next article in the journal. The abstract needs to be informative: writing ‘a theory has been presented and the implications discussed’ is simply a waste of time for everyone involved. Get straight to the facts!
  1. The Introduction introduces the paper: ask the correct scientific question, discuss the current state of the problem in the literature, and explain the reason why you have the key to the problem that others do not.
  1. The Results section describes the data, quickly reviews expected regularities and highlights oddities. Use a reasonable number of plots to show how the new results fit with existing knowledge and what makes them distinctive. The section provides a context for comparing the new results with literature data. Save trees, though! As Al Hofmann correctly points out, writing that ‘87Sr/86Sr ratios in samples 5 to 9 vary from 0.702985 to 0.703117, whereas, quite to the contrary, they do vary from 0.703017 to 0.703315 in samples 1 to 4’, which most normal people can get from the table, is a total waste of space and time and immensely boring. Telling that Ba increases from sample 1 to 10 while Rb decreases in sample 8 to 15 is boring.
  1. The Results and Discussion sections are rigorously distinct. The reader must have all of the available data by the beginning of the discussion. Don’t pull out new results or data out of your sleeve in the middle of the discussion to support an idea. A layered cake of results and discussions is one of the surest ways of having your paper rejected.
  1. Presentation style (Part I): go for the linear progression, and not for the crime story line. Avoid building suspense, because if you succeed in hiding your arguments for too long, the white knight will arrive a few seconds too late to rescue the prospective mother of his children from the villain. Your paper will pass away before the end, and your story will fall into undeserved oblivion. The first sentence in a paragraph always describes the motivation of the paragraph – don’t take the reader down a hidden path. It’s better to be understood in 30 seconds than to lose the reader after 15 minutes of struggle.
  2. Presentation style (Part II): use sub-headers with moderation, they each betray a break of logic. The ideal article could have none. My high-school teachers enjoyed the sight of a full table of contents, but a segmented structure rarely serves a strong scientific discussion. Rather that breaking your work into separate pieces, try to streamline the sequence of ideas. Show causal or temporal relationships by rephrasing consecutive sentences to this effect. Empty words like ‘indeed’ or ‘in fact’ do not connect anything, certainly not adjacent ideas, and look like meaningless objects casually dumped in the crack between two unrelated statements.
  3. The paragraph! Avoid woolly style, with no memorable beginning, no foreseeable end, and endless sidetracking. Follow the thread of your idea in the most possible linear way. Although a new step in the development of an idea calls for a new paragraph, you should work hard to keep the transition smooth.
  4. Samba, two steps forward, one step backward, is great for communicating feelings, but is no help for scientific ideas. If you develop your idea in the first part of a sentence, do not ruin it in the second part. Such a trick signals that you are afraid of being forthright and will certainly leave the reader confused about what you really meant to say and how confident you are about your story. If a confrontation between ideas is really needed, leave the reader on the idea that you think you should support. Compare the impression left by ‘I like beer, but I really need to watch my waistline’ versus ‘I know I shouldn't, but I could really use a cold beer now’ and make up your mind about which message you wish to send around!
  1. Don’t think that if you explain things clearly, that you will be considered mentally retarded. Obviously you do not want to take the reader for an idiot, but take the time to clearly and simply state your ideas, without embellishments, without insinuations, and without pontificating. The worst thing that could happen to you is that you will be understood.
  1. A very common related problem is whether potential readers should be assumed to be on top of the literature, even in their own field. I was taken aback many times by reviewers who considered that their ignorance of papers published 5, 10, or 25 years ago was essentially none of their responsibility. Although such an attitude is morally and intellectually flawed, I recommend that you act defensively and do not hesitate repeating --with due referencing-- what you think is common knowledge. Of course, doing this for trendy concepts will be perceived as a lack of taste.
  1. Complexity is rarely a virtue: demonstrating that a problem is complex (i.e., more than one degree of freedom) is the best way to chase away a reader and in any event is not sufficient reason to write a paper. Find a thread (a simple idea) that allows you to unravel the problem so that the author, the reviewer, and the reader remain interested. Occam’s razor (principle of parsimony), though, makes good cocktail conversation but has no scientific merit.
  1. Don’t use smart innuendos if you are not 100% sure that the reader will get them. If you are on shaky grounds, be pedestrian. The archetypical example of this kind of behavior is summarizing a particular theory by simply adding the term ‘classic’. The notion of the ‘classical dynamo theory’ is translated by the average reader as ‘I will not waste my time explaining the trivial details of the dynamo that this stupid reader should know; by the way, I don’t remember what the original references are’. This is very, very unprofessional and editorially risky.
  1. Don’t use empty words (e.g. done, performed, carried out, etc.) and instead use words that have meaning: ‘The Sr isotope compositions of these samples have been measured’ should be preferred to ‘The measurement of Sr isotope compositions has been performed on these samples’.
  1. Don’t use insults, even unintentional ones: when a reviewer comes across ‘it is obvious that’ or ‘it is clear that’ at the beginning of a complicated statement, he/she will be made to feel like an idiot. Expect this reviewer to make you pay for your condescending attitude and lack of communication skills.
  1. Avoid using personal pronouns ‘I’ or ‘we’ when there is no actual action carried out by the author. ‘We mixed the residual solutions in a 10 ml beaker’ is OK, but ‘We can see that the points form a linear array’ is at best is a little absurd, and at worst an insult (see point 18). The solution is simple; just cross-out ‘We can see that …’! Use you spell checker to hound the unnecessary ‘we’.
  1. Avoid unnecessary problems related to debatable meaning: ‘quantitatively’ for ‘completely’, ‘to evidence’ for ‘to attest to’, ‘complimentary’ for ‘complementary’, ‘paper’ for ‘article’. Also avoid ‘and/or’ and split infinitives (e.g., to boldly go).
  1. The Conclusion section is an essential (and required) part of a paper: it represents what remains after the dust of the discussion has settled. No new ideas can be introduced in the Conclusions.
  1. The Acknowledgements should be dignified and include the reviewers, all the reviewers, who have spent time on your work, even if you did not really appreciate their comments.
  1. References. Dropping a significant article from your reference list to minimize its importance with respect to your own manuscript is a sure strategy for triggering a hostile reaction from the reviewers. Good Editors pick the reviewers among the most competent scientists in the field, but most of them unfortunately come with a strong ego. Be a scholar, if only to minimize trouble. Otherwise, the first paper where an idea is introduced should always be cited and a recent review article can also be useful. The absence of old references can indicate a lack of perspective, but having too many citations of old papers is pedantic.
  2. Fonts. Do not use more than two fonts for the text and avoid non-proportional fonts (Geneva, Courrier) as they are difficult to read in the form of long paragraphs. Use serif fonts (with embellishments, as Time) for the text, and sanserif fonts (Helvetica, Arial) for headings and titles. Be stingy on bold, underlined, and italic style: the human brain does not handle well too many levels of coding. Right justification is pretty but makes the paragraph bulky and unpleasant to read (the eye looses its marks). Hyphenate your text. Never forget to produce a double-spaced manuscript with wide margins, which will let the reviewer scribble annotations when he or she takes the manuscript on the plane. Pre-formatting a manuscript in the journal print style anticipates acceptance and boasts perfection: this will, most of the time, irritate both the Editor and the reviewers.
  3. Figures. Limit their number, but a text with no figures is usually unattractive. Remember that figures are often the only component that people check when they ‘read’ a paper. You will easily find web sites to help you prepare terrific illustrations. Limit the number of symbols and colors (5 of each is a good maximum). Use sanserif fonts (Helvetica, Arial) for labels and your figures will have more muscle. Likewise, limit the number of font sizes. It is good policy before submitting a manuscript to print the figures reduced to the size expected from the particular journal and check their readability.


  1. Manuscript revisions. Good words break no bones. When a reviewer indicates that he (she) has not understood something that appears obvious to you, seriously consider the possibility that he is sincere. Don’t try to reprimand the reviewer for an assumed ineptness, but instead work on better expressing your ideas (and test them out on your non-specialist colleagues). When an English-speaking reviewer takes time to correct your errors in English, don’t just ignore the suggestions, as you won’t get a second chance to benefit from this help.


Now, please don’t check my own papers in too much detail and do not ask how often my own manuscripts get rejected for breaking these ground rules. I am making progress but it is not perfection yet!



Francis Albarède


(Thanks to James Scoates for the original translation and to Al Hofmann for useful comments. French version: http://perso.ens-lyon.fr/francis.albarede/conseils.php)