A guide to writing scientific text
Scientific writing is a lifelong pursuit. Even though our varied reading habits predispose us to various styles of writing, it usually takes years to hone the skills needed for excerpts to make sense. As we are drawn to the wonders of natural and social sciences through our senses, writing is a means to invoke the same feelings in others. Can we bring them the same euphoric sense of discovery experienced by us while researching? Besides, useful pieces of writing have the tendency to outlive us, thus keeping our ideas alive.
Strangely, many of us are never taught how to write. It is also considered the least glamorous and most challenging task to accomplish. Our text is often reviewed and graded against rubrics for clarity, novelty and impact on society. How then must we communicate our efforts to an audience that is looking for clarity, originality and impact? In the following sections, I will describe good practices of writing. The principles might help write with clarity, the novelty of your work falls upon you, and its impact is only a matter of thought and conjecture.
Forms of writing
There are only four primary forms of writing styles: expository, descriptive, persuasive and narrative (Grammarly, 2020). Scientific writing is mostly expository in nature as it aims to expose (Latin exponere: to set forth) evidence and educate an audience.
We ought to describe the details of our work. But describing experimental information, methods and limitations of data is very rarely about a character, event or place and should not be confused with the nature of descriptive writing such as poetry, fiction or journal entries.
In a persuasive essay, argumentation is typically presented in a way to convince someone of your own opinion. Academic writing need not fall for description or persuasion. Our job is to present evidence and draw a logical conclusion from a series of arguments supported by the evidence presented. There must not be any room for emotionally charged statements, opinion or influence. Such statements harm your academic credibility, unless you can back them up with proof.
How then do we connect facts and figures into a coherent story that is devoid of judgment and persuasion?
A research article is about helping the reader visualise your factual message using a well-crafted narrative. Narrative writing connects events into a gripping account and can enhance our own pleasure in writing scientifically. When someone questions your research work with “What’s the narrative?”, they usually mean how do you connect all your evidence into one detailed message. This depends on who the audience is.
Audience is key
Irrespective of the writing style, the audience is always at the centre of your writing. It is for them that we are writing so we can illuminate a group of people with rigorously worked out evidence, deductions and prose. Start by asking: Who is the audience and what goals they may have for reading your work? What do you want them to retain? Then have a clear narrative to communicate the message (Turbek, 2016).
The hourglass approach
[This section is a general outlook] The hourglass method (similar ideas have been developed by Turbek, 2016) is a straightforward approach to looking at scientific writing. There are two significant points to consider. Why are we doing something and what is its impact on a community: be it society, other scientists or decision-makers? Many scientists believe not to focus on the relevance of their research results. I do not agree with it because if our work is not relevant to society, why must the taxpayers foot our bill? The hourglass connects the ‘why’ with the ‘impact’. We start by describing the big picture and systematically narrow down to the specific gap in knowledge. This leads us to write about a particular research question(s).
Like in an hourglass (see Fig. 1), scientific content flows in one direction: top to bottom. There is no room for suspense, thrill or the ultimate reveal. The data, methods and experiments form the lens in the middle of the glass through which you address the knowledge gap. Questions like what do you do, and how do you do it is discussed here.
In the lower inverted cone of the hourglass, objectively expand on methods and results. Build one step at a time from the first set of exploratory data analysis, modelling and experimentation to the final conclusions. Finally, discuss the broad impact and opportunities for future research. Always end on a strong note.
A Loose Structure of research article / thesis / proposal
(every publication outlet has a different purpose and publication guidelines)
A title is a choice between two options. Either it is the primary method you developed or the main finding of your work. For example,
- Estimating spatiotemporal transportation demand for public transit networks using gravity models, or
- Temporal demand profiles of mobility reveal the spatial structure of a city
Both titles describe the same research. In the former, the method itself is essential. The latter possibly uses a method already existing in the literature and reveals critical insights about a particular phenomenon. If you want your article to have more impact, it is best to select a result-oriented title that helps others find your work easily. Many scientists use wordy and catchy titles like “No shit, Sherlock: AI does not solve all societal problems.” (I really hope that is not a real title). While they are fun, many journal reviewers are your primary audience. They will either let your smart or cheeky titles pass or feel offended. So choose wisely.
Write this last. Some say to write this first. So much confusion. When the outline of your article is ready, the first abstract draft could be the pointers strung together (see Introduction for details). Naturally, this will develop as we go along. After writing the outline in the abstract, you can safely move to the introduction.
An introduction is the most critical and challenging part of your work. As the name suggests, introduce the reader to the big picture of the problem you are addressing, and slowly narrow down to the knowledge gap, a paragraph at a time. Substantiate your narrative (exposition) with references. End it with identifying the research questions and give a brief summary of what you do to address the knowledge gap. Here, write 5-7 paragraphs (just a suggestion and not a golden number, see Fig. 2), each talking about three main things.
Every paragraph (see Fig. 2)
- starts with a topical sentence introducing the main message being conveyed,
- the body of the paragraph elaborates on the main message, and
- ends with either a concluding remark or a connection with the next section.
An excellent way to start writing this is to lay an outline with 5-7 significant points. This way, you can be sure that your message is following a coherent narrative (a story of connected events following a proper sequence) and is meticulous in nature. As you elaborate on the sub-points of each bullet, your paragraph starts to take shape. This also helps you in identifying the main message for every section. The main points can even be thought of as an elevator pitch. All that remains is to remove the bullets and use your literature survey to substantiate the argumentation.
When and where should I write the literature survey? A survey should be a rough draft separate from the main article. What you want is an outline of every paper you have read that is relevant for your research. Tables, paragraphs, rough sketches, all suffice in this matter. As the years will pass, you will start to make mental maps of old literature. The process of outlining the main points of your article following the hourglass approach utilises the literature survey but avoids writing in the format,
- Ref a did this
- Ref b shows …
- Ref c illustrated …
- Ref d was not up to the mark
- And we show this…
This approach helps to develop a story that can be sharpened later and much, much more comfortable for any other scientist to understand your research.
Explain the sources, the composition and the properties of the data. Often, you have to explain how you constructed or designed a study for collecting your own dataset. Keep in mind that you want your research to be reproducible, understandable and shared (it is not in anyone’s benefit to tuck away these details). One could argue anonymity, privacy and ethics sometimes do not allow data to be open and freely accessible. That is okay. Then you must clarify why that is the case and what could be an alternative route to carry out research that builds upon your work.
In this section, you describe how you tackle the problem. Over the years, the practice that I follow and preach is to document methods when you write code for exploratory data analysis, modelling, designing algorithms, a conceptual model, or experimentation. Find sync between performing your modelling, simulation and analysis work and note down all the points that will help you elaborate on the methods later. Even if you can find 30 minutes a week to write about data and techniques, you will save yourself days of writing at the end of the project. Feel free to use subsections to organise the details of your methods. The goal of this section is to provide another scientist in your field the means to replicate your study and possibly build upon it. The methods should be sound, even if the novel, describing the details that are necessary for replication. For example, unless implicitly mentioned in your algorithms (surveys, experiments, etc.), always specify the seeds, settings or libraries that are important to run your study. If the methods have been designed in other studies, cite the source and describe them in enough detail that is necessary for the readers to understand your findings.
This section warrants accurately presenting your critical findings. First, idealise a rough sketch of data, tables and graphs that you consider essential for informing your audience. This forms your guiding hand that may change through the project but keep your research focussed on the goals you started with. Once you implement the methods and visualise your key findings, write descriptive sentences that explain each result in detail. If you are answering multiple questions, divide this section into subheadings (Lorenz-Spreen, 2019) to provide the reader with a structured narrative. Each figure, table or dataset could present trends and outliers. Make sure your summaries cover these instances. The readers need to find a guide into your figures. If there are limitations and possibilities for errors in your data sources or modelling, refer the reader to a Supplementary Information. Such a section should have details of a robustness or sensitivity analysis you have carried out.
This section is a complement to the (#introduction). Subjectively interpret the results in the context of the research question(s). The purpose of this section is to provide the reader with a recap of the main finding(s), the contribution to the domain of knowledge and future prospects. Start by mentioning the main conclusion(s). Address the knowledge gap stated in the introduction using subjective interpretations of the principal outcome (s). Do not hesitate to provide alternative explanations of your results. It is in everybody’s interest to broaden the possibilities of future research. It is important to situate your findings in the existing literature. Your colleagues and other scientists in similar fields are likely going to read your work. They would like to know how your results are related to their previous findings and how everything may come together for the future growth of the field. There is no reason to state that someone else’s work was not on point while your own was up to the mark. Under general consensus, all work is peer-reviewed, and so you have likely provided additional insight into the field. However, it is good to reflect on results that do not agree with previous findings of other people and identify possible explanations for it.
End by briefly explaining the significance of your results. It is imperative to discuss the societal impact of your work and to address policy and decision-making such that new questions can be considered for future research. Be careful though: it is not our prerogative to explicitly mention what decisions need to be taken by policymakers. Our work focuses on computational (rigorous and quantitative) urban science (of cities and large sets of populations) and policy (analysis of decisions in the frame of urbanisation). Hence, refrain from prescribing policy solutions. Identifying different strategies and presenting them as they are is acceptable.
This section elaborates on rigorous details about the data, methods, statistical analysis, experimentation and robustness or sensitivity analyses. Often, the length and information of this section depend on the form of work and the audience. For example, while writing a master thesis, you must consider your thesis committee and the details they may need for evaluating your work. If they are network scientists and you write a lengthy chapter on how to build a network from nodes and links, consider if this is really the best use of your time. Perhaps focus more on the new algorithm you are developing for analysing a network in space or time and hide the network details with a short introduction to the literature.
Note: If you are writing a kickoff document or a research proposal, consider cutting out information not relevant to the content of your work. Meta research takes a lot of time to write and hardly helps in actually carrying out the research. There is no need to write about the process of research: how to conduct a literature survey, followed by desk research, data collection, modelling and finally, a reflection on your results. If we are working together, I will provide you with guidance on the process of research, and I am not judging on how well you can write such a section. Instead, focus on evaluating the data you need, the sources you will consider for this data, and the requirements of the model. All of this should be brief and help you as a guide to carry out your research. Remember, the top half of the hourglass is still relevant here. Start from the big picture and narrow your focus on to the concrete knowledge gap, subsequently identifying the main research question(s) and how you plan to tackle them.
Dos and Donts
- Always try to write in the active voice. However, use passive voice when the predicate of a sentence is more important than the subject. For example, consider a paragraph that is describing the low-income minority groups. Then, “…the low-income minority groups are overlooked by policy instruments addressing the equitable distribution of resources” is a better choice of words than “…policy instruments overlook low-income minority groups..”.
- Use simple words, short and clear sentences.
- Find your own voice. It takes years to get better at writing. There is no one way to describe your scientific work.
- Seek feedback. If it is constructive, incorporate it as you see fit. The feedback that says “this is wrong or delete this section” without explaining why is often useless. As a student, ask your supervisor why they have made such a remark, it will only help improve your writing skills.
- Pass your work through Grammarly before sending for publication or evaluation to correct grammar and spelling. Although the world is dominated by written and spoken English, many of us do not use it as our first language. It is completely okay to seek help for grammar and spelling from someone other than your evaluator.
- Instead of using words like lots, many and recently, provide accurate numbers and dates. (Reif-Lehrer, 1996).
- Never mention a symbol or formula that is not immediately described.
- Do not use sentences like, “Later you will see how this is connected…” If it is later, then it must not be mentioned here. The reason is simple: if the reader fails to understand your point because they have to go later in the text to grasp some meaning, it is your failure, not theirs.
- Do not excessively use conjunctive words like and, yet, but, because etc. to lengthen sentences unless necessary for drawing comparisons or stressing causal relationships.
- Minimise use of words like thus, therefore, moreover, furthermore and however. These tend to be overused and distract the reader from the content (Reif-Lehrer, 1996).
- Never say any form of the sentence, “We all know that society is …“, unless you have evidence or references to support it. For example, “We all know gravity exists as described by the general theory of relativity (Einstein, 1915).” Without referring to Einstein, your fact becomes conjecture.
- Grammarly. Accessed 29 April 2020. https://www.grammarly.com/blog/types-of-writing/
- Turbek, S. P., Chock, T. M., Donahue, K., Havrilla, C. A., Oliverio, A. M., Polutchko, S. K., … & Vimercati, L. (2016). Scientific Writing Made Easy: A Step‐by‐Step Guide to Undergraduate Writing in the Biological Sciences. The Bulletin of the Ecological Society of America, 97(4), 417-426.
- Lorenz-Spreen, P., Mønsted, B. M., Hövel, P., & Lehmann, S. (2019). Accelerating dynamics of collective attention. Nature communications, 10(1), 1-9.
- Reif-Lehrer, L. (1996). Following instructions is critical to success of a grant application.