Scientific writing is hard. It is easy to forget this, we tend to think doing the science is the hard part. However, the coding/analyses/fieldwork etc. is only half of the scientific process. Writing up results into a narrative, one that provides a reader with some new knowledge and tells them how it fits into what was known before, is as much part of the science as the experiments themselves. Despite this, writing scientific English is often the part of the process we have had little formal education in. For this reason, I have written these notes to help guide our thinking about scientific writing.
What follows are some principles, some advice, and some (useful) pedantry. I will fall short of the principles outlined below every time I write. Indeed, only by falling short of these principles have I, slowly and with the forebearance of supervisors, co-authors, and reviewers, learned some effective approaches to scientific writing.
When I am trying to improve a piece of writing, the preferences and opinions expressed below are what I am trying to implement. I think they are principles that will generally help improve a piece of scientific English, but personal style will dictate when they should be ignored. Topics covered are:
Principles
(1) One idea per paper
Papers are artificial: the research process has taken you far and wide, but in the end the insights gained often need to be communicated in a single paper. Constructing this single paper is a case of picking back over what you have done for the results that can build a coherent and focussed narrative. This may not be the same story you thought you would be telling when you begain the work in the first place. Be willing to leave that imagined story behind and to tell the one you actually have in front of you.
The reader needs the paper to communicate one main idea very clearly. Anything that is not directly a corollary of the main idea is likely to be lost. So, be brave, resist being vain, and be willing to leave results out of the paper. Much better than putting these in the current paper, is to use these as the seeds of your next publication.
(2) Know what you are trying to say
The worst writing occurs when you do not know what you are trying to say. If you do not know what you are trying to say, stop and think. Do not just continue writing. All you will be communicating to the reader is that you do not know what you are trying to say.
Introductions attract the worst of writing without seeming purpose. They are hard to write. They are pulling you out of the details of doing the science and into having to motivate the science and place it in the context of the wider literature. Remember that, ultimately, all you are trying to communicate is why this field/area is interesting/important, what the problem is, and what opening you have found to contribute some new knowledge.
This problem of knowing what you are trying to say is fractal. It can strike at the level of an introduction, a paragraph, or a sentence. Learn to recognise when not knowing what you are trying to say is causing the problem, and to stop and think how this section/paragraph/sentence fits into the narrative you are constructing.
(3) Make the reader do as little as possible
Your aim in scientific writing should be to do everything for the reader. If you leave it to them, you leave it to chance that they will have understood what you mean, and to their patience that they don’t grow frustrated with the cognitive load of deciphering what you have written.
i. Word limits do not equal word targets. You should be writing as clearly and succintly as possible to use as little of your reader’s time, energy, and patience, to communicate everything you want to say and no more.
ii. Be explicit not implicit in your writing.
iii. Do not leave it to the reader to have to carry information around from previous sentences, or worse, previous paragraphs, to understand what is being written about. Repeat the subject/object word or concept rather than overly relying on use of ‘this’.
iv. Keep sentences short. Reread paragraphs and break up sentences with multiple clauses.
v. Use the simplest language possible. Prefer short words to long words, except where the short word lacks beauty.
(4) Do not take shortcuts, the reader might not know what you mean…
…and the reviewer certainly will not know what you mean. To be confused is the role of a reviewer, so that others do not have to be.
It is tempting when writing to rely on the fact that the reader ‘will know what you mean’. We rely on this in our conversations all the time. Indeed, more often than not, our interlocutor will know what we mean. That understanding is, however, rooted in the common and immediately shared cultural context. Do not make this assumption in scientific writing.
The reader of your scientific writing could be over the other side of the world reading in 50 years time. You still want this future person to understand you, even though they are having to reconstruct the field’s concerns and assumptions from scratch.
Scientific writing, like a contract, should leave little room for interpretation, or for the reader to have to bring their own cultural knowledge of the field to bear to fill in the gaps.
Clearly, it is fine for the reader to need to know what an electron is, a planet, or gravity. If we had to go back over fundamental physical concepts we would say nothing, and these are specific terms a reader can search for the meaning of. The problem is when a whole sentence is loosely written.
For example, “We will make observations in the accessible region of the atmosphere to constrain the sulfur abundance.” What is the ‘accessible region of the atmoshpere’? Is this where we can get to? Where the instrument can get to? Where our instrument can see? With context, the reader might know what this means, but what we should have written is something more like, “We will make observations of the mbar region of the planet’s atmosphere, where transmission spectroscopy has sensitivity to constrain its sulfur abundance.”
(5) Avoid redundant words
The use of redundant words often indicates an uncertainty, hesitancy, or pleading in the writer. Maybe for this reason, and maybe just because more words is more words to have to trudge through, redundant words undermine the point you are making.
For example, saying something is “certainly plausible”, only leaves the reader feeling less certain than if you had just written “plausible”.
“Very” is another word that tries too much and achieves too little in scientific writing.
Ask: what do you want that adjective or adverb to express? Can it be expressed using a magnitude or quantity? If it can, use the magnitude or quantity. If it can’t, resist using the adjective or adverb at all.
Writing
How to start
It is generally much easier to write the article once all the figures are in place, with captions. Once these are in, but prefereably as you are adding the figures, add sections and subsections. Of course, in writing the article, you will discover that you want to say something for which there is no supporting figure/calculation. This is fine, and in fact why writing is such a valuable part of the scientific process. Go back and make the figure/perform the calculation, then carry on.
I prefer to write an article from start to finish, title, abstract, introduction… as a reader would read it. I understand the desire to shave off easy writing first, the methods and results, for example. However, the article’s narrative tends to be the tightest if there has not been too much going backwards and forwards between sections.
Writing straight through from the introduction tends to mean that a strong narrative focus gets carried through to the subsequent sections – i.e., it helps you write sections that know their role in the story, a role defined in the introduction. This is much harder when writing sections without the story having been setup yet by the introduction and title.
Structure
Structure exists at all levels of an article. At the top level, you need to make a decision about whether you are following a standard Introduction-Methods-Results-Discussion format, or deviating from this. And do deviate from this structure. It is often completely inappropriate to follow ‘IMRaD’, and forcing your science into this format will produce something unreadable.
One level down from sections exist subsections; for example, populating your methods with details of the field campaign in one subsection, the sample analyses in another, and numerical methods in a third.
It is easy to forget the structure that exists below the level of sections/subsections, however, this lower-level structure is as important, if not more important. This lower-level structure exists at the level of paragraphs and sentences.
Paragraphs need to communicate one coherent point. They are not an arbitrary break in the flow of text. Each paragraph will need to follow its own arc, from an introduction sentence making a point the reader is familiar with, through to a conclusion where a single new concept/set of linked ideas has been communicated.
Within a paragraph each sentence needs to pick up the meaning from the last sentence and carry it forward, advancing the narrative and providing a nice setup for the following sentence. Think of relay runners passing the baton from person to person; the runner is the sentence, and the baton is the idea or narrative.
Scientific writing often become difficult to read not because the subject matter itself is too complicated, but because the paragraphs and sentences no longer obey the principles outlined above. Common deviations from the principles above, that lead to a difficulty in reading, include:
- a sentence that makes too large of a conceptual jump in one go;
- a sentence that does not have an obvious conceptual link to the preceding sentence;
- a paragraph that attempts to make too many new points, or, more likely, fails to make a single point.
Structure at all levels in the manuscript helps with…
Keeping the narrative alive
The reader will benefit hugely if they always know where the article is taking them. This requires the article at a section, subsection, paragraph, and sentence level to make clear how it fits into the narrative. I have a mental picture of this narrative focus as being like a sawtooth graph: when the line is high, we are zoomed out onto the main narrative, but within sections and paragraphs we need to progressively zoom in on details.
I’ve sketched how I visualise this below at the level of the whole article, but we could draw the same spiky graph zooming and in and out at the paragraph level as well. A paragraph begins connecting to the information that came before, then makes a new specifc point of its own.
Even if you keep the narrative alive by directly writing something like “We just showed XYZ, which raises the question PQS. We now explore PQS here”, this is much better than assuming the reader will understand why ‘PQS’ is now being talked about.
Once you get into the mindset of not dropping the narrative thread of the article, you will find many inventive ways of ensuring the reader always understands where they are going and why.
Approaches to good writing
(1) Read good writers. Whether of fiction or non-fiction. Some writers with a sharp and sparing use of language, which can be helpful for science, are Hemingway, Ferrante, Harari. To really see every word count, read good poetry.
(2) Read what you have written aloud. It forces you to see and hear words that your mind might now just be scanning over. Especially if you have written and re-written the sentence a few times already.
Minutae
Small things that matter. That matter to me, perhaps more than they should… though, know that presentation of the results is going to matter a lot to the reader, however much we might like the science to speak for itself.
Words
- ‘Firstly’, ‘Secondly’… Can all be written more simply as ‘First’, ‘Second’…, without losing any meaning.
- Linking words. ‘Moreover’, ‘Furthermore’. These are linking words. They are helping you to get from one sentence to the next and I am sympathetic to wanting to use them. However, there is almost always a better way of getting from one place in your writing to another than using these words, which in and of themselves rarely carry any additional meaning. If you have found you have written one of these words, first just try deleting it. Often the sentence is fine without it.
- Punctuation for ‘i.e.’ and ‘e.g.’ should be ‘i.e.,’ and ‘e.g.,’, with a comma after each and with full stops inbetween the letters. The trailing comma follows from the meaning of ‘i.e.’ and ‘e.g.’ as ‘that is’ and ‘for example’ respectively, which if you used in place of the Latin abbreviation would require a comma after them.
- Use of ‘Impossible’ is probably best avoided. Most of what we want to say is impossible is often just impractical or unlikely.
- ‘We argue that…’. I think arguing is generally sub-optimal, I prefer suggesting. But more importantly, there is little need to announce our intention to argue, we can just go ahead and say what we think.
- Comparative words: ‘More’ ‘Less’… When using comparative words like these, be explicit about the thing(s) the subject is being compared too.
Punctuation
- A range is given using an en-dash, which in LaTeX is written by putting two dashes next to each other:
-- - Parenthetical statements separated by ‘dashes’ uses em dashes:
---. Try not overuse parenthetical statements, nor em dashes as a way of demarcating them. Parenthetical statements add a cognitive load to the reader, as they have to recall where the sentence broke off to pick it up again, all whilst assimilating the new information contained within the parentheses. - An in-sentene list should probably have a comma between the penultimate and final item, so as to make clear that the final two items are not linked. For example, “We varied A, B and C”, allows that B and C may have been varied simultaneously, whereas, “We varied A, B, and C”, makes it less ambiguous each parameter was varied separately. Of course, use of this penultimate comma could be avoided by being more explicit, “We independently varied three parameters A, B and C”. Generally, I would advise both being more explicit and using the penultimate comma so there is no chance of misunderstanding.
- Units should be separated from numbers (in LaTeX) by use of
\,. Goodness knows how or what Word does.
Equations
- Units and chemical elements should be in regular font, not the italicised LaTeX maths font. Use
\mathrm{}to protect text from being italicised. - Equations typeset on a line of their own, as would be the case in LaTeX with the use of
\begin{equation}, are likely to need punctuation after them, often a comma. - If a comma is used, or any punctuation that indicates that the sentence continues, then the text after the equation should not be indented as if it were the start of a new paragraph. I.e., avoid leaving a blank line after the equation in LaTeX.
Figures
- The contents and text of figures should be large enough to read.
- Figures should always be referred to from the main text. They cannot just exist in isolation from the main text, as then the reader does not know when to look at them or how to link them to the narrative.
- Figures should appear in the paper in the order they are referred to from the main text.
- It is often quite efficient and elegant to refer to a figure parenthetically. You don’t always have to write sentences like “Figure 1 shows that…”, which explicitly discuss the figure’s contents. A parenthetical reference to the figure can sometimes do everything that is needed, allowing the main text to focus on the results.
- Where possible, avoid using a legend in preference for labelling the content of the plot. This also helps disambiguate points/lines that could otherwise be hard to tell apart for someone with colour blindness.
Citations
- When making a point that generates a series of citations, cite in order oldest to youngest.
- Prefer parenthetical citations for brevity, where substitution of a explicit citation does not change the meaning.
- Explicitly talking about the community/previous work/views/knowledge has to carry a citation. Especially in introductions, it is tempting to write sentences like “Habitability has been defined as the ability of a planet to host liquid water at its surface.” You know this is true, the reader knows this is true, even so, because writing the sentence in this way explicitly refers to the work of others (“…has been defined as…”), the sentence has to end with a citation to the work doing this.
Feedback, understanding it and responding to it
After drafting a manuscript the next thing that will happen is that we receive feedback on it. This will first be from co-authors, maybe next a wider circle of collaborators, before finally submitting it for peer review and receiving reviewer feedback. After all the effort that has gone into the science and writing at the point feedback is received, it is well worth ensuring we make the most of the feedback to garner the article maximum impact and recognition.
However, this process of receiving feedback on something that we have worked so hard on can be challenging. Some feedback might seem insensitive to the huge investment of time and emotion that has gone into the paper by this point. This can make it difficult to positively engage with the feedback received.
Understanding feedback
My most fundamental recommendation in dealing with feedback is to
digest feedback over a few days.
This gives you the time to move beyond any initial reaction to the feedback and to see it from the reader’s perspective. We want to develop a sympathy for the reader’s point of view.
Remember, if someone is providing feedback they have likely earnestly tried to understand the article. They will have given it more time than another reader would, so any confusion they exhibit will likely only be a lower bound on what someone just scanning the article would experience.
Below are some things one might think in response to feedback and how I suggest these thoughts be most productively responded to.
- “I am annoyed”. Come back to the feedback later. Come and talk to me.
- “I am upset”. Come and talk to me.
- “The reader’s comments are overly blunt”. Feedback should not be rude. It is worth us remembering though that giving feedback has required a lot of the reader’s time, so the comments we receive might be quite direct simply because couching them in softer language would have required more time on the part of the reader only to convey the same information.
- “The reader hasn’t understood anything”. It is tempting to blame the reader for this, or indeed, if this is a review, the editor for sending it to such a seemingly unqualified individual. However, it is very unlikely the reader is actually incapable of understanding the science, so the misunderstanding they are exhibiting is extremely valuable feedback. Here is how to deal with it:
- Try and see past the specifics of the reader’s comments when the comment itself is rooted in a misunderstanding or confusion.
- Instead, try and identify what has tripped the reader up. The reader themselves in their comment might not have identified the real point of misunderstanding, so it is our job now to root this out.
- Edit any and all portions of text that contribute to this misunderstanding. Remember, that the seed for the misunderstanding could have been planted in the introduction or abstract, with our ill-judged use of language or terminology.
- “I don’t want to do what the reviewer is suggesting”. There are perhaps several variants of this, with different implications. Overall though, I would advise setting the bar high for us not doing something in response to well intentioned feedback.
- “Change the science”: These comments have the chance to improve the rigour and quality of the work we have done. The reader is acting more like a collaborator with these types of comments and we should be grateful for them as they are not receiving credit as a co-author for these scientific contributions.
We do need to evaluate whether the scientific suggestion is within the scope of the article. If it is not then we can say this, perhaps even in the article itself.
If the suggestion can be implemented, then we have been able to improve the article. This is the feedback process working at its best. - “Change the communication”: Whilst potentially frustrating given the effort we will have gone to to communicate our results clearly (whether written or visual), comments like this are almost always worth doing something in response to. We will have been steeped in the science for months, potentially years, by the time we are receiving feedback, so it will be easy to forget what might not be obvious to someone else.
- “Change the science”: These comments have the chance to improve the rigour and quality of the work we have done. The reader is acting more like a collaborator with these types of comments and we should be grateful for them as they are not receiving credit as a co-author for these scientific contributions.
Responding to feedback
Once we have implented any changes because of the feedback received, we will need to provide a response. This can be quite informal if it was feedback from a colleague or collaborator. However, a review will need a more formal response.
Some advice on writing a formal response to feedback (review):
- Write the feedback when you are feeling dispassionate.
- No matter the tone of the comment, we need to remain polite in our response. Most of the time we should be explicit in our gratitude for the feedback, which is helping us improve our manuscript.
- Make it as easy as possible for the reader of the response (the reviewer(s) and editor(s)) to understand what you have done in response to the feedback and why. This means:
- Logically structuring our response, providing feedback to one reviewer at a time
- Including the full text of the reviewer’s comment so that our response is fully contextualised
- Saying what we have changed because of the comment, including line numbers, figure numbers, and where appropriate quoting the changes in the response text itself.
- Use formatting of text (bold, italic, color) and indentation to make it clear what is a reviewer comment and what is our response.
- We will want to disagree with suggestions or not implement them. This is fine, but more than ever, in such responses we want to be clear in our reasoning and be polite.
- Provide a short preamble at the beginning of the response document thanking the editor and reviewer(s) for their time and summarise how the article has changed (improved!) due to the feedback, and how we have formatted the responses. If there are common points raised by reviewers, this is where these can be covered at a high level, before the detailed responses.
The advice here is born of experience, experience gained thanks to the time of others, either as readers of my papers, or as people receiving my comments. So thank you to them for their patience. Special thanks to Steve Jones, who opened my eyes to how badly written my first paper was, and to Francis Alberède, for teaching me that there should be a limit to how much you put into a single paper. Thanks also to Sean Jordan, Paul Rimmer, and Craig Walton, who provided feedback on the advice contained here and contributed some wisdom of their own.