20 ways to improve your paper: a journal club introspective

​Over the past 12 months, members and visitors to the lab have been meeting weekly to discuss a topical paper from the literature. In our journal club, a different person picks a paper each week for everyone to read, and explains the paper before leading a discussion about the questions, methods, and results. Given that the journal club is taken as a class by undergraduate and masters students (who present most often), we have also focused on details of the writing—how has the presentation of the ideas, methods, results, and discussion contributed (either positively or negatively) to our perception of the quality of the paper? 

Written by James and Damien

As it turns out, how authors write their papers can have a really big impact on the general consensus about the paper: missing information, opaque methods, or over-stating ideas typically result in more negative perceptions about the work. As a result, we feel that we would be less likely to cite that work. So, to help others increase the scientific appeal of their research, we have identified 20 common ways that can help improve the presentation of scientific papers (at least for biologists).

1. Avoid jargon, political puns, or misleading statements in the title
The title should be a clear statement of the results and scope. Most importantly, it should be an honest representation of the results. Avoid making a title that suggests support for a hypothesis that isn’t directly tested (e.g. don’t say B causes C if only presenting data on A causing B and A causing C). Often this mismatch is obvious and will really put the reader off side. We feel that a poor title can really diminish a good paper. Similarly, a sexy title will not improve a poor paper, and worse can convey over-confidence in the authors’ perception of their own work. Finally, taxonomic information can be helpful in the title by narrowing the scope of the title a bit (if it is very broad),  and if the paper is on a sexy species – put it in the title!
 
2. Identify a clear gap
The first sentence(s) of the abstract and introduction should provide the scope of the manuscript. Avoid going outside this scope later, but also make sure that these first sentences are centered on topic! Also, make sure that they are factually correct (amazingly that isn’t always the case). Finally, avoid being repetitive. One common mistake when writing key sentences is to have two halves of the sentence that actually say the same thing. For example: Little is known about the biology of this very interesting species, meaning that we have little information about the species.
 
3. State the question early enough
It should be clear from the very first paragraph what the question in the paper will be about. Sometimes, these are simply stated. Good writers will lead the reader to the question without necessarily making it explicit. Ultimately, the aim should be that the ‘Here we …’ sentence towards the end of the introduction should not come as a surprise. Good writing will make the reader anticipate the questions and results.
 
4. Answer the question!
This one is surprisingly common, even in high impact journals. We have read many papers where, upon reading the title, abstract, or opening paragraph, we realise that the authors never actually answer the question they set out to test. This is often a balance between the hypothesis and questions – it’s fine to outline broader hypotheses which inspire the research question, but avoid confounding the hypothesis with the question. State the question in a way that it can be answered, and answer it.
 
5. Present the results as clearly and honestly as possible
Missing background or (worse) broad generalisations related to the findings of a paper are off-putting. For example, many studies claim to study trade-offs between costs and benefits, but only actually explore one side of the equation (either a cost or a benefit). Similarly, a study might claim to be important for understanding selection on a trait, but without ever quantifying the underlying variation in the trait or whether that variation changes over time. From the perspective of the reader, these kinds of statements detract from the perceived value of the study. Note that this isn’t about not putting studies in a broader context, but more about avoiding overstating the scope of the results.

6. Tell a clear story
This always sounds harder than it actually is. A good paper will keep the reader engaged from start to finish.  Be succinct. Avoid the “shotgun approach” of throwing too many ideas at the reader at once. Make sure that topic sentences (the opening sentence of each paragraph) are clear, linked to one another, and relevant(!). Avoid long sentences and complicated terminology. Short sentences with simple wording win the day. Keeping the reader engaged is important. Avoid giving the reader a chance to question what has actually been done (e.g. ‘I didn’t understand this – oh wait, I actually didn’t understand any of it’).
                                 
7. Avoid randomly introducing concepts in the paper
This is one that is typically harder for researchers at the start of their career. Having done so much reading, it is natural to want to share every idea you’ve had. After all, you only have one paper to do it in! Having lots of ideas in a paper is great, but if an idea is not completely tied into the question, then it should not be in there. Stick to the topic – avoid tangents and anecdotes. This is particularly common in the introduction and discussion, which should not include concepts that are not central to the manuscript (see the point about mirroring).
 
8. Treat your variables well
Take care to be consistent and explicit about variables throughout the paper. Describe the connection between variables and their biological contexts. Failing to do so can easily cause the reader to lose connection with the paper. Also make sure to provide the units of measurement when first introducing the variable, and then how the values were calculated. A few words can help a lot. For example, instead of ‘we calculated emergence time’, say ‘We calculated emergence time, which we defined as the number of days from the hatching until the individual left the nest. This represents the development speed of the individual.’
 
9. Keep constant terminology throughout
This is one I (Damien) learnt the hard way. After having my first paper rejected from a high-impact journal, I realised that I’d been inconsistent in the use of patches, sites, and feeders throughout (they were all used interchangeably to refer to single feeders, feeders within experimental treatments, and to experimental treatments). To avoid such situations, make a glossary at the start. The glossary doesn’t have to be in the final publication, but just having one while writing will help keep the writing clear and consistent. Also, don’t think that repeating the same word is a bad thing – it might feel like it sounds less clever but actually it really helps with clarity. Avoid using a thesaurus! (Having ‘smart sounding’ words might make you feel smart, but it will cause the reader to have to stop – and thus lose momentum on the paper).
 
10. For modelling papers, make the model part of the paper
Often a modelling paper will be presented almost like an empirical study, with the model thrown into the appendix. Make the model part of the paper! Start by laying out the assumptions of the system in the introduction (they are, after all, part of the background). Make the variable names meaningful – but be careful to avoid overlap with names elsewhere in the paper (e.g. avoid ‘resources’ as a variable if discussing resources in a broader context). Finally, include a table that lists every variable and what they represent (in clear unambiguous terms).

11. Cite the correct papers
Often what is being cited can say a lot about how much care the authors have put into thinking about their work. Two common things to avoid are over-use of books, and citing reviews as evidence for empirical results. Technically, both can be correct, but they can indicate a degree of laziness on the part of the authors. Books are important for citing key over-arching concepts, but usually not for citing more specific background. Thus, they are acceptable in the opening paragraph, but rarely elsewhere. Similarly, some reviews explicitly evaluate the support for or against key hypotheses, but generally it is much more agreeable to the reader when statements about empirical evidence cite the actual papers giving that evidence.
 
12. Make sure that the evidence is supported empirically
Fundamental assumptions that are central to a paper should be stated in the introduction and supported by evidence. One issue here is that results from modelling studies are presented as if they were empirical evidence. These are not the same. If the only support for a hypothesis comes from the results of models, be honest about this, and don’t ignore contradictory empirical results. Similarly, avoid constructing hypotheses without any empirical evidence to support them (unless explicitly laying out a new hypothesis).
 
13. Avoid building a straw man
A straw man is an argument that is constructed in a way such that it can too easily be refuted. For example, a well-known biologist once presented the argument that group selection cannot be true. Their argument was that number of people smoking in many states of the US has decreased. A group-selectionist view of this phenomenon is that states which had lower smoking rates invaded states with higher smoking rates, and thus outcompeted them. This is clearly not true, therefore group selection is clearly untrue. We are not arguing for group selection, but the evidence here was a straw man, it ignored many other known facets of human behaviour and biology that clearly make the argument invalid (such as the fact that these changes occurred within the same generation).
 
14. Take care with anecdotal evidence
Anecdotal observations or studies are important. Many of our most important concepts in biology are built on what was originally an anecdotal observation. However, that does not mean that every anecdote is important! Sure, write a short communication about it, but don’t blow it out of proportion. It is also best to avoid anecdotes when writing a more substantial study, as they can only serve to distract from more important results.
 
15. Mirror your writing
Mirroring ideas throughout the paper is one of the main ways to improve clarity of the overall study. If concepts are introduced in the introduction in a certain order, maintain that order throughout. For example, if a paper addresses several questions, present these in the same order from the introduction through to the methods and results (and ideally in the discussion as well). This can be facilitated with subheadings (see below).

16. Use subheadings (appropriately)
Subheadings are a really useful way of structuring a paper, but only if they are done in a consistent way. In general, every subheading that states a question or problem in the methods should appear in exactly the same way in the results. In doing so, also think about how the heading would work both as a way of capturing what was done (the methods) and what was found (the results). The use of subheadings can also flow through to the discussion.
 
17. Be selective with statistical tests
It is often tempting to test all variables against each other. However, statistics (i.e. hypothesis testing) should only be used to test and support hypotheses laid out in advance. Not everything has to be tested. If there are potentially interesting relationships, then perhaps present these in a supplement, but avoid clogging the paper with loads of stats that aren’t obviously addressing what was initially outlined (i.e. contributing to the statements in the first or second paragraph of the introduction). Further, it is also best to prioritize a priori hypothesis testing, and to be clear about what counts as an a priori hypothesis test and what does not. Avoid post-hoc testing (unless clearly appropriate). It is also important to note that not all procedures constitute hypothesis testing (nor does everything need to be hypothesis testing). Avoid huge tables (e.g. full pages) of statistical results. As a rule of thumb try to only conduct one test per question. More P-values does not mean a better paper.
 
18. Figures are key
Make use of figures. A Figure 1 showing the timeline, design, or overview can be very useful in passing information to the reader in an efficient way. Similarly, a figure summarizing the results can really help readers remember the findings of the paper. In terms of key results, avoid building the support for a question using multiple figures. Each figure (or each figure panel) should present the relevant information for validating (or refuting) one prediction or question. Finally, figures should stand alone (see the final point).
 
19. Tell the reader what is in the supplement
Almost no one will just read the supplementary materials, yet often some of the most useful information is contained in there. Tell the reader what it contains by referring to it often and in detail. Use statements such as ‘Supplementary Materials section 1 contains a detailed overview of each step in our experimental procedures.’ Highlight why it is important for the reader to look there. Take time to make the supplement clear, attractive and sharp. But, never bury crucial information in the supplement. The supplement can provide meaningful context that didn’t fit in the main paper, but if it’s a key part of the bigger picture, it belongs in the main paper. Finally, avoid referencing it in the discussion.
 
20. Satisfy both the skim reader and the detailed reader
This is perhaps the most challenging goal. Everyone reads papers differently. To an extent, the author can guide the reader through the use of structure, headings, and—most importantly—topic sentences. A good topic sentence will provide the skim reader the broad strokes of the whole story. However, the paper should also avoid just being a repository for big ideas, which will really put off the detailed reader (and likely also the reviewers). It is important to flesh out each idea for the detailed reader. Thus, finding a good balance between opening and explaining ideas, while remaining succinct, is the goal. Finally, use the topic sentence to convey as much of the relevant broad information as possible, it is surprising how often the first sentence of the abstract is identical to the first sentence in the introduction (which will completely put off the skim reader).