Illustration of empty paper sheet with pencil and pen on workdesk

Credit: vladwel/Getty

I’ve managed the review of more than 1,000 climate-science papers at Nature over the past 11 years. My decision of whether to send a paper out to review is guided by the novelty and importance of the research, not by how the work is presented.

But there are many aspects of presentation that routinely lead to confusion or misinterpretation and generate substantial delays in the review process. In the interests of minimizing hassle during review — and making it as easy as possible for editors and reviewers to focus on the work’s scientific content — I offer the following suggestions (modified from a Twitter thread). Some are more relevant to climate science than to other fields, but most are generic.

Formatting and layout

Use wide line spacing and big fonts. Make the text easy to read. Small fonts and single spacing are fine when the final piece is published in multiple columns in print. In the two-column layout used in Nature’s print version, for example, one line in a single column will normally include about ten words. But the same font and spacing for the one-column layout of most initial submissions will yield 20 or more words per line, which is difficult to parse. Double spacing with 15-point font is also overkill. A reasonable compromise is 1.5-line spacing and a font size that will produce about 12–15 words per line of text in a single-column layout.

Use continuous line numbers. Reviewers like to refer to specific line numbers, and frequently comment on their absence. If the paper makes it to a second round of review, continuous line numbering will also make it easy for you to point out the location of your revisions.

Provide inline figures and legends. At least for the purposes of initial review at Nature, we encourage authors to provide figures and legends as part of a single Word, PDF or LaTex file, with the figures placed as close as possible to the relevant text.


Avoid subjective wording. Reviewers will often object to words or phrases such as ‘unprecedented’, ‘paradigm shift’, ‘amazing’, ‘dramatic’, ‘interesting’ and ‘remarkable’. It’s best to present your results with few adjectives, and let the readers make up their minds about the magnitude of the scientific advancement.

Avoid acronyms and abbreviations. Although it is true that these shorteners leave you more room for other material, they can also create a whiplash effect as the reader goes back and forth to match the term to its original definition. A good rule of thumb might be to use a shortener when a term is used five times or more in your paper. Do use common shorteners; in climate science, for example, these might include ‘SST’ for ‘sea-surface temperature’ or ‘CCN’ for ‘cloud condensation nuclei’. Avoid, if at all possible, inventing shorteners that are unique to your paper.

Avoid directionless words, such as ‘influence’. Instead, state the direction of the effect you’re describing. So, instead of writing, ‘Precipitation influences net primary production’, write, ‘Precipitation increases net primary production’. Better yet, be specific: ‘A doubling of annual precipitation in arid ecosystems increases net primary production by 20%’.

Avoid using ‘significant’ in almost all cases. If you mean ‘big’ or ‘major’, then don’t use ‘significant’, because it might be confused with the results of a statistical test. And if you are reporting the results of a statistical test, it’s better to report numerical results specifically.

Use a declarative title. Reviewers (and editors) will often recommend a declarative title that states, rather than suggests, and that provides a sense of the main conclusions of your paper. Instead of a hypothetical title, such as, ‘Trends in groundwater storage’, try: ‘A doubling of groundwater loss since 2010’. Be cautious — a declarative title might flirt with overstatement or obscure nuance. By avoiding subjective words, and carefully focusing on the paper’s actual findings rather than potential implications, you should find it possible to craft a readable, informative and interesting title that doesn’t exaggerate your findings.

Follow a template for your abstract. The abstract, or first paragraph, of Nature papers is fairly standardized, and it might be helpful for all authors to follow our template.

Data and figures

Define uncertainties. In many initial submissions, error bars and uncertainties are not fully defined. Are they 95% confidence intervals, ranges, 2σ? Is the boxplot showing the interquartile range, or something else? Tell us in the figure legend.

Consider statistics. When possible, include a statistical or numerical data analysis. Although visual interpretation of your data might be justified, this is usually not the case. For example, if you are proposing that an acceleration of ocean circulation warms European climate two years later, reviewers will probably expect a quantitative analysis as opposed to a visual representation. Nature research journals maintain a collection on best practices in statistics.

Show and provide underlying data. If possible, display all data, rather than simply providing summaries. Although it is fine to use bar charts to count simple data (such as the number of fish in a pond), bar charts can obscure statistical nuances. Instead, consider using scatter plots for small data sets or boxplots or violin plots for larger data sets (free tools are available online). Before accepting a paper, Nature and many other journals require adherence to the Enabling FAIR (findable, accessible, interoperable and reusable) Data initiative, for Earth-science research, and submission of many other types of raw data (including electrophoretic separations, seismographs, genomes and those derived from animal models). Reviewers are coming to expect open data, so it can be well worth the effort to include your data at the time of initial submission (the Nature Research journal Scientific Data maintains a list of potential data repositories).

Follow a colour code. Try to use colours in a way that won’t create problems for the many readers with visual conditions. Good discussions and design suggestions are available, as are online tools for checking the accessibility of your figures. In particular, avoid rainbow colour schemes — rainbow colours are popular in many fields, but they are problematic for a number of reasons. For more discussion, and for alternative colour palettes, see cmocean, NASA GISS and ColorBrewer. Helpful changes in major software packages are afoot — R, a common computing and graphics platform, has moved its default colour scheme from rainbow to viridis. Use divergent colours when appropriate: for graphs in which data diverge from a central value, it’s often helpful to use white for the central value and then ramp up to two colors. In Figure 4 of a recent Nature paper1, the mean change in temperature is set to white, increased cooling is shown with darkening shades of blue, and warming is shown with darkening shades of brown.

Simplify figures. Remove anything that doesn’t contribute to information transfer. Extraneous effects, such as 3D, shadowing and unnecessary colours, should be eliminated. If you are displaying a temperature map using colours, for instance, don’t display the same information using an overlaid contour map.

Provide short titles for figure legends. Sometimes these can be descriptive, such as, ‘Study area’. But, if possible, make the title declarative; for example, hypothetically: ‘Ocean heat content in the western Pacific Ocean has increased by 15% since the late twentieth century’. Alternatively, provide one or two sentences in the legend that highlight the figure’s main conclusions.

Some Twitter commenters have suggested that a few of the above suggestions, particularly those relating to line spacing and font size, could be solved or made easier for authors by providing a LaTex template. This is true, but many issues — such as subjective wording, rainbow colours or acronyms and initializations — can be solved only by the author, not by the technology.

At Nature, we have made the barrier to submission as low as possible, particularly because we publish fewer than 10% of initial submissions. If a submission is ultimately accepted, then our in-house team of expert subeditors will handle the final formatting.

Except in extreme circumstances, adherence to the suggestions I’ve provided would neither increase nor decrease the chances of a paper being sent to review. But my experience with reviewers and readers suggests that following these guidelines will enhance your work’s clarity and transparency — and thereby smooth the path for reviewers and readers.