A wise man – I’m quite sure it was Brian Wansink – once pointed out that it is impossible to both read and write a lot. So, maybe reading a post about how to write just steals time from the more urgent task of writing more. Then again, maybe it wouldn’t have hurt Wansink if he had spent more time reading rather than cranking out more papers, so there’s that.

This blog post is a collection of some pieces of writing advice for a certain use case:

1. You want to write about a rather technical topic but
2. You want readers, including non-technical ones, to be able to follow along.

The precise format of what you want to write does not matter as much as the general underlying mindset: “This is complicated but I really do want people to get it.” So, you may be trying to write a tutorial, or maybe a primer to some technical topic, or actually just a substantive paper but the methods you use are sufficiently non-standard to require more explanation.^[1]To the point that your substantive paper ends up being yet another tutorial. Been there, done that. Some may argue that all scientific publications are on rather technical topics and thus qualify.

All of the following is stuff that I have found to work well for me. Your mileage may vary – writing is a lot of trial and error anyway. None of what I say is original as I’ve read a couple of books and articles on the topic,^[2]Most prominently Steven Pinker’s “Sense of Style”, which I can highly recommend. plus I have absorbed a lot of pieces of advices that others have generously passed down to me. Most of this input has by now coalesced into brain mush of uncertain origin. At this point in human history, nobody is going to say anything genuinely new about writing anyway.

In principle, writing about technical topics in an accessible manner requires two steps:

1. Figure out what precisely you want to say
2. Figure out how to say it in an accessible manner.

And then, of course, sorry forgot that one, you still need to

3. Write the whole thing.

Here’s some advice for drawing the rest of the owl.

• Dealing with uncertainty in empirical findings
• The reader is smart but has no idea what you are talking about
• The NAEFRO Rule: No abbreviations, except for rare occasions
• Consistent terminology
• Crafting a nice logical flow
• Making things more modular
• Writing as a craft rather than an art
• Shitty first drafts followed by endless revisions
• Feedback is life, feedback is love
• Breaking conventions and finding your own voice

Figuring out what precisely you want to say

The bad news is that if you do not precisely know what you want to say, it is rather unlikely that the reader will get it. The good news is that sometimes this problem solves itself as writing is a great way to figure out what you wanted to say in the first place. Still, at some point before submitting your piece, you have hopefully figured out what precisely you want to say.

What you want to say does not have to be a single great overarching point or some narrative that dominates the whole piece. I am confident that readers of the scientific literature can bear reading something that cannot be boiled down to an elevator pitch. For example, we collected some data during my doctoral studies for which I kept procrastinating the write-up because it all was a blooming, buzzing confusion. But at some point I had settled on a set of statistical analyses and decided that I wanted to: (1) Explain to readers why longitudinal data are of particular value for the research question at hand, (2) demonstrate what a section on causal identification assumptions (usually not a thing in psychology) could look like, and (3) communicate all results with a particular eye on the high uncertainty resulting from the design we chose (which wasn’t great, in hindsight). That gave me a purpose and fixed my motivational issue. We made it through, eventually, and I’m satisfied with the outcome.

Here’s a little magical trick that I sometimes do with students. I tell them “I took a look at your draft but cannot quite figure out what point you want to make here. What do you want to say?” — and then usually their reply will be very reasonable and well-structured. So I take some notes, hand them over and tell the student to write it down exactly how they just explained it to me.

Dealing with uncertainty in empirical findings

One particular thing I see people struggle with is the uncertainty attached to empirical findings. On the one hand, you want to tell a compelling story. On the other hand, you don’t know what to make of those findings yourself either. One way to move forward here is to say precisely that. Now of course you can’t  just want to leave your readers with a shruggie emoji (or can you ¯\_(ツ)_/¯?), so first you have to find out where precisely the uncertainties lie. Is your statistical precision so low that you can neither claim an association nor reject it confidently? Explain that to your readers. Could there be plausible confounding invalidating a desired causal claim? Explain which confounding paths you have in mind. Is there any particular reason why, despite uncertainties in the data, you consider one explanation more plausible than the other? Explain how you end up at this belief. The trick is to be as precise as possible about the imprecisions. Confidently assert why one should not be confident.

“But, Julia,” I hear you say, “won’t they reject my article if I write it like that?” 

Well I guess that could happen – but if you think about it, there is a good chance it will be rejected either way.^[3]I’m fun at parties It will depend on the journal, the reviewers, the editors, and, given the erratic nature of the peer-review process, maybe also the current planetary positions. And, of course, on whether or not your study is informative in some way, despite these uncertainties. There may still be a take-home message worth disseminating, including “if you want to do this you will need a much larger sample size than you may think” or “here is a cleverer design for future studies” or “don’t try this at home, it just won’t work.” 

In my (limited, biased) experience, the peer-review process usually starts with a manuscript that hides away a lot of stuff which is slowly unearthed as reviewers raise their concerns. Hence, the uncertainty is added incrementally and usually not in an elegant manner, because the authors have already settled on a narrative from which they’re not willing to deviate — begrudgingly written limitation sections that are at odds with the rest of the article, anyone? Personally, as an editor and a reviewer, I’m delighted by manuscripts that transparently and coherently communicate uncertainty from the get-go. They usually mean less work; they signal that the authors actually thought stuff through before sending it out for review.  

Figuring out how to say it in an accessible manner

The reader is smart but has no idea what you are talking about

A good mindset to write everything down in an accessible manner is to assume a reader who is smart – maybe just as smart as you – but who does not know anything at all about the topics that you are talking about. This may sound like taking it a bit too far. Aren’t you writing for a specialist audience? Well, maybe you are, but what we are trying to do here is counteract the curse of knowledge, which is stronger than you expect, even when you take into account the curse of knowledge. As Steven Pinker puts it in “The Sense of Style” (Chapter 3): “The main cause of incomprehensible prose is the difficulty of imagining what it’s like for someone else not to know something that you know.” So, try to assume as little prior knowledge as possible. 

But try to do it in a non-condescending manner because (1) you don’t want people who know less than you to feel stupid just because they happen to lack some background knowledge and (2) you don’t want people who know more than you to feel like you’re wasting their time. For example, you can add little definitions and explanations “en passant”, as a little refresher for those who need it. In Rohrer and Murayama (2023) we wrote: “Between- and within-persons associations can be statistically independent (i.e., they can take on different values or even opposite signs), and it is the latter within-persons associations…” – the text in parentheses is not a big detour and won’t distract those “in the known”, but just in case somebody has forgotten what statistically independent means – here’s a tiny piece of scaffolding for them so that they can climb along as well.

You also shouldn’t expect readers to connect any dots on their own. Maybe they genuinely can’t; maybe they could in principle but are just not paying that much attention when reading your paper.^[4]This, by the way, should never be a reason to get upset at readers. Of course you’ve put in a lot of time and want people to pay close attention to your carefully crafted prose; but for readers there are approximately ten gazillion papers they could have squeezed into their precious ten minutes of reading time instead. So, just carefully lay out all the points but then also spell out how they come together. Those who could have done so on their own still enjoy the intellectual satisfaction of being able to check their solution against yours.

Again, this should be done in a non-condescending manner, so avoid anything along the lines of “as can be clearly seen”, “obviously”, “of course”, and, of course, never ever: “This is left as an exercise for the reader” (Figure 1).

The NAEFRO Rule: No abbreviations, except for rare occasions

On a related note, don’t assume that readers have the memory capacity to store new abbreviations. Abbreviations that are more common than the spelled out words are of course not only fine but actually good; EEG is a lot more readable than “electroencephalography”, which I had to look up for this blog post. One of my pet peeves are substantive or only mildly methodological articles referring to statistical models by cryptic abbreviations (CLPM, RI-CLPM, ARTS, STARTS, LGCM, LCSM, HLM/MLM/REM/MEM). It’s perfectly fine to do that whenever it is easier on the reader,^[5]Easier on the writer doesn’t count. You can easily use the abbreviation while writing but then search and replace at the very end. If this results in bungled grammar, the abbreviation wouldn’t have worked very smoothly anyway. which is most likely the case in simulation studies comparing some of them, thus requiring you to refer to every single one a dozens of times. But people often use these abbreviations when they only mention each one twice or thrice, spread out over different sections of the article. In that case, it’s much preferable to just spell them out. Any reader who is not versed in the arcane arts of psychological statistical modeling will thank you. 

There may also be edge cases where it makes sense to use an abbreviation locally (e.g., in some more technical paragraphs in the method section) and later spell it out again (e.g., the first time you mention it in the discussion section). This probably flies in the face of multiple style guides, but I’d rather take into account your audience reading habits than worship the idols of APA (more on breaking conventions later).

Also, try your best to avoid abbreviations in tables and figures. Tables and figures should be able to mostly stand on their own and abbreviations don’t help with that. Now you may of course spell out the abbreviations in a table of figure note, but I wouldn’t operate on the assumption that people enjoy reading these when they only want to get the gist (which they normally do). Usually better, more creative solutions are possible (e.g., header rows that extend over two lines; manually tweaking figure labels so that everything fits when fully spelled out). As a side note, I think it’s almost always preferable to label figures within panels (e.g., add some label right next to the line that it is referring to) rather than relying on the legends automatically added by whatever you use to plot your data. This usually adds a bit of “manual” tinkering, but I think it’s well worth the effort given how much easier it makes to read figures. And figures are often the one thing that gets shared on social media and re-used in future talks and lectures on the topic, so you’re investing the time into something that may outlast the rest of your writing.

Consistent terminology 

This one is really a low hanging fruit. Readers have to understand what you are referring to, and this gets a lot easier if you always refer to it in the same way. So, for example, if you frequently refer to a particular statistical model, you will have to give it a name (such as “Bayesian censored location-scale model”), and then just stick with it. You may also repeatedly refer to a problem (“the age-period-cohort identification problem”) or to a controversy in the literature (“the debate on…”) – just settle on a name, introduce it, and then stick with it throughout. Of course, sometimes there are multiple common and reasonable names. In that case, it makes sense to point to them the first time you use your preferred term (“multilevel models, also known as hierarchical models or mixed models,…”) and then consistently use your preferred term throughout. That way, (1) readers who have learned a different terminology can make the connection, (2) readers who are hearing about this for the first time get a bit of preparation for terminology throughout the literature, and, most importantly (3) in your article it is still crystal clear that you are always referring to the same thing. There’s a bit of flexibility here in that you may introduce a longer name and later shorten it a bit if it does not result in any ambiguity (“the age-period-cohort identification problem” becomes “the age-period-cohort problem”). Sometimes I am not sure which terms I want to use throughout and switch back and forth; in that case I try to add comments or highlight all occurrences to remind myself that I still need to replace them later for consistency.

Crafting a nice logical flow 

Readers also have to understand why you are saying what you are saying. It’s quite frustrating if somebody just dumps information on you without telling you how that relates to anything else. Every section, every paragraph, every sentence and every sentence part fulfills some sort of function in the context of what you are writing – if you feel like the logical flow does not quite work yet, it’s helpful to think about these functions and the relations more explicitly.

When I edit somebody else’s text, I often spend a lot of time reverse-engineering why they are saying the things that they say in that specific order, and then I try to make the connections more visible for readers. The logical flow may look like this: “Here is a problem. Here is some elaboration on the precise nature of the problem. Here is a specific example. Here is why somebody may think this is not a big problem. Here are three reasons why it is actually a big problem that we should care about.” And then you need to pick the right words to make the relations clear: “Here is a problem…more specifically…for example…however…but: first, second, third,…” Steven Pinker refers to the logical connections as coherence relations (similarity, contrast, elaboration, exemplification, generalization…) and even provides handy tables with typical connectives (similarity: and, similarly, likewise, too; contrast: but, in contrast, on the other hand, alternatively…). If you don’t have the time reading his whole book, I think just repeatedly asking yourself “what has that to do with anything?” and then adjusting the text to reflect the answer can go a long way.

There are a lot of other nifty things you can do to enhance the logical flow of the text.

When you are talking about things that are logically parallel, make the sentences parallel as well: “In condition one, participants did this and that. In condition two, participants did that other thing.” If you talk about all the things that are done to a thing or a person, by all means, use the passive voice to ensure coherence across sentences (“Participants were first greeted…they were then told to…”). If you instead talk about all the things that a thing or a person does, use active voice instead (“We first imputed missing values…we then conducted regression analyses…”). Sometimes, you may just want to list a couple of points that follow no particular structure but still need to be said. In that case, you may just as well include them as a list (either an explicit one, or an implicit one, like this somewhat winding paragraph). Sometimes, you want to make a bigger point that just needs to be made, in that case you might as well make it transparent for readers (even if it isn’t the most elegant solution): “Apart from these concerns, another topic one has to consider…” And sometimes, you may just want to break the logical flow and pick up elsewhere. That can be made more visible by starting a new paragraph. Or why not go one step further and even include a subheading?

In the end, the logical flow is the most important thing to ensure that reader can follow along; it’s a lot more important than “superficial” features such as sentence length. A very long but logically coherent sentence can be perfectly understandable with little cognitive effort. A very short but unconnected sentence may leave readers utterly confused.

Making things more modular

Creating a perfectly logical flow for a scientific article that covers everything which needs to be said is a daunting task. Luckily, you don’t have to squeeze everything into one continuous text. Maybe that’s not even a good idea because most people will skim read your writing anyway, if they look at it at all (in which case you can consider yourself lucky). So, here are a couple of things to make your writing more modular. 

If you have advanced content that more technical readers (including reviewers) may want to see, but that at the same time may distract readers who can barely follow the “main plot”, apply some containment protocol. For small technical detours, content can be added in footnotes or in parentheses at the end of a sentence (which tells readers that it isn’t essential  for following the rest). Boxes are a level up from that and my favorite display piece because they can go either way: you can include a box to provide more detailed technical content for advanced readers, or you can include a box to provide additional handholding (e.g., a worked example, additional explanations). Glossaries can also be helpful – although now that I know that boxes are usually an option, I tend to prefer them as they are more flexible. If you want to publish something somewhere that does not allow for boxes (boo!), strategically placed subheadings can fulfill a similar function and tell less technical readers that they can skip certain sections.  

Last but not least, you can always add supplemental material. In that case, make sure that it is called out and summarized in a reader friendly manner in the text. One thing I have seen quite often are references to additional robustness checks in the supplement without any summary of what those checks actually show. I understand how these things happen – reviewers demand checks that you may not consider important, and then maybe one check isn’t actually quite as unambiguous as you wanted it to be – but they still reduce my trust in authors. Please don’t use the supplement to hide stuff. If you cannot easily summarize the results of robustness checks in a couple of sentences (à la “the estimated effects deviated from the main result by less than 0.1”), consider a figure displaying the central quantity of interest across various robustness checks. This makes everything transparent without distracting readers who don’t actually care about the robustness checks.

Write the whole thing

Writing as a craft rather than an art

When I read writing advice from novelists, I often get the impression that the whole task an ordeal – you have to wait until you are kissed by the muse, then you agonize about how to express some deep truth about either the world or your innermost self, and then in the end nobody actually pays attention and you may end up in the gutter. Luckily, if you are in the business of scientific writing, you don’t have to put yourself through all of this (maybe except for the very last part). What you are doing is a lot more mundane because you are not trying to create art. The editor-in-chief of our student newspaper used the catchphrase “Schreiben ist Handwerk” – “Writing is a craft/trade/handwork/handicraft/” (sorry it’s not that catchy in English). Over time, anybody can learn the right motions and become able to produce something of solid workmanship, which is really all it takes (no transcendental truths or self-exposure required). But as a craft, it does require practice to get there.^[6]Unless you are naturally talented; life is unfair like that.  

Shitty first drafts followed by endless revisions

In a rather famous chapter from her book “Bird by Bird”, writer Anne Lamott reassures us that all good writers write shitty first drafts, “This is how they end up with good second drafts and terrific third drafts.” This seems like a good mindset to get started, but I may be biased because my first drafts are shitty as well. No, you can’t take a look at them.^[7]In a writing workshop during my PhD programme, Ulman Lindenberger was actually so brave and shared a first draft of a grant proposal with us. It was rather messy and quite incomprehensible. I presume one needs to rise to the rank of a Max Planck director to be willing to flaunt one’s first drafts like that. In any case, that was a very good way to adjust our expectations for first drafts.

Starting from a shitty first draft already implies that some degree of revision will be needed. Usually, it’s a lot of revisions, to the extent that actually the rewriting part is a lot more important than getting anything on the page in the first place. William Germano has written a whole book just about the business of revising, but it feels a bit like cheating, because it’s essentially just a book about writing (minus the part where you write the shitty first draft).

Two hard questions one has to ask oneself during the revision process:

1. Am I getting anywhere through incremental changes, or do I need to overhaul something to make it work?
2. Am I there yet, or is it worth investing more effort?

Considering (2), just from the manuscripts I see getting submitted, people most definitely have different cut-offs when to stop revising. You should definitely get past the zone in which your writing actively hinders understanding. This gets you into a neutral zone. If being understood is really a high priority, you can push further and try to get to the zone where your writing helps readers rise to a level of understanding that would otherwise be out of reach.

Considering (1), once one has settled on a certain way to say something, it can be very hard to decide to throw everything over. Maybe you’re stuck with a structure you just can’t get to work. Maybe you’ve fallen in love with a certain phrase but then it doesn’t really fit with the rest. Or you’re really settled on making a particular argument, but it’s actually not necessary and even distracts from the main message. Nobody likes killing their darlings and major overhauls aren’t exactly fun. But often, they’re absolutely worth it. This is the point where feedback from other people becomes really important — they will let you know if what you are trying to do just doesn’t work. They don’t care about your darlings one bit.

Feedback is life, feedback is love

And love hurts, but sometimes it’s a good hurt.

If you can convince others to read your stuff and tell you what they think about it, that’s tremendously helpful. The general idea here is that the reader is never wrong, even if they are literally wrong. They may have barely skimmed what you wrote, missed central pieces of information that you clearly provided in the manuscript, gotten all worked up because what you wrote relates to one of their hobby horses, and then missed the central take-home message although you explicitly spelled it out. But the average reader later on probably won’t pay more attention, won’t have better working memory, and won’t have fewer hobby horses. If anything, the average reader later on will exert less mental effort and extend less goodwill than anybody who volunteers to provide feedback.

So, if they got something wrong because they only skimmed what you wrote, maybe you can make your text more skimmable by ensuring that the important stuff is communicated through subheadings and attractive display pieces (boxes, figures, tables). If they couldn’t follow along because they missed something you said in an earlier part of the manuscript, maybe you can add some pointers or a tiny repetition to refresh their memory. If they got worked up because they took something wrongly, maybe you can adjust your writing to pre-empt misunderstanding and bad feelings. In general, don’t expect readers to fix things (unless they are co-authors); their job is just to provide hints for you to work with.

Mindset-wise, I think it is helpful to give up the notion that what you are writing has some intrinsic value that is under attack when somebody criticizes your writing. This notion leads to all sort of dysfunctional behavior, such as starting arguments with people who were so kind to volunteer their time to help you with your draft (don’t do that). What you are writing has value because of what the intended audience can get out of it. If somebody is not getting much out of it, that provides valuable information which may either lead to changes in the text, or to the decision that certain people are actually not part of your intended audience. Which is fine as well, no text can be optimized for every potential reader.^[8]For example, when I published my primer on graphical causal models, I got some flak because it didn’t say anything new and didn’t talk about what one actually needs to draw causal inferences: instrumental variables. I briefly took that personally until I realized that I indeed didn’t say anything new  – that was never the plan – and that I wasn’t writing for people who strongly believed in instrumental variables. I was writing for people who had never heard of instrumental variables.

Where can you get feedback in the first place? Helpful sources are (hopefully) co-authors, peers, and people who owe you (because you previously gave them feedback on their manuscript). If you need more input and can’t find anybody in your closer circle, you can consider asking on social media if anybody is interested in taking a look. And you can absolutely try to cold email experts whose input may be valuable. I’ve both been contacted that way and contacted others, usually with positive outcomes. Make sure that your email clearly communicates that they’d be doing you a huge favor because their unique expertise would render their feedback extremely helpful. Nobody likes to feel like you think you are entitled to their free labor; everybody likes to feel like they have something special and valuable to contribute. And, of course, make sure that the manuscript is in good shape. Then the worst case scenario is that they don’t reply, which is fine as well. Academics ignore emails all of the time.

One strategic note: Anybody who provided substantial feedback should of course be acknowledged in the manuscript and some editors generally won’t consider people mentioned in the acknowledgments as reviewers. Make of that what you will.^[9]I suspect I was at least once invited to take a look at a manuscript to reduce the chances I’d end up reviewing it. I still ended up reviewing it. So much for keeping your enemies closer.

As a side note, it should be acknowledged that not being a native speaker of the language you’re writing in puts you at a disadvantage. The only solace I can provide is that to create a nice logical flow, general reasoning skills probably matter more than your vocabulary. A lot of native speakers aren’t great at it either. And let’s all appreciate the fact that we don’t live in the alternative timeline in which German has become the leading language of science. The thought alone is furchteinflößend, even for me as a native speaker.

Breaking conventions and finding your own voice

Academic writing brims with conventions, from how to structure your article to which words are admissible. Sometimes those conventions can get in your way. For example, something may technically belong to the method section but really only make sense in the context of the introduction or the results section.^[10]As Tim Morris pointed out in the comments, another good reason to break conventions is to lower readers’ defences and engage them with an element of surprise. I fully agree with that! 

Here are two things to consider before you break a convention.

First, conventions shape readers’ expectations of what comes next. Thus, if you deviate from conventions, some readers may end up lost. The way to prevent such confusion is to ensure that you still have a crisp logical flow that guides readers through everything. If that’s not enough, you can also add some signposting; that is, you can tell readers what you are going to explain in which order. Signposting can be annoying if you lay out the obvious (“After motivating the research question, the research methods will be described in detail, followed by the results”), but it can be helpful if you prepare readers for the non-obvious.^[11]And chapter 2 of “The Sense of Style” has some advice for how to make it sound less clumsy.

Second, sticking to conventions is a great way to communicate that you do know the conventions, which is important in many academic contexts. Whether you like it or not, writing and even formatting send subtle signals about who you are. Now I do believe that many people take conventions too seriously and thus produce cumbersome prose, but even I would raise an eyebrow if somebody broke with conventions too flagrantly (e.g., by using a completely different article structure, four-letter words, or vertical lines in tables). Thus, if you break with conventions, you do want to make sure that the rest of your writing signals that you know your shit. 

Which leads me to my very last point. There seems to exist a belief that professional scientific writing needs to have a certain tone – impersonal, dense, abstract. I often see it in students who try to imitate that drab style, but this is confusing a descriptive norm with an injunctive norm. A lot of academics do write that way; nobody thinks that academics ought to write that way.^[12]Maybe except for people who care more about appearing academic than about actually contributing intellectually. Don’t try to make those people happy. They know no joy. In fact, editors and readers usually very much prefer if you don’t write that way.^[13]Helen Sword has written a whole book about that topic titled “Stylish Academic Writing.” 

So, by all means, feel free to try different voices until you find something that feels sufficiently authentic and helps you get the job done. Of course, you do want to appear sufficiently professional for the occasion at hand. But what really matters for that is that your writing has a clear logical structure and that you deliver your points in a precise and clear manner. Anything beyond that is a matter of taste.