Scientific Figure Rules: What Journals Enforce and What Only Helps
Scientific figure rules come in three kinds with very different penalties: file specifications, image-integrity rules, and readability habits. This guide sorts them, with the published numbers from PLOS ONE and Nature.
Readiness scan
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan before you submit. Catch the issues editors reject on first read.
How to use this page well
These pages work best when they behave like tools, not essays. Use the quick structure first, then apply it to the exact journal and manuscript situation.
Question | What to do |
|---|---|
Use this page for | Getting the structure, tone, and decision logic right before you send anything out. |
Most important move | Make the reviewer-facing or editor-facing ask obvious early rather than burying it in prose. |
Common mistake | Turning a practical page into a long explanation instead of a working template or checklist. |
Next step | Use the page as a tool, then adjust it to the exact manuscript and journal situation. |
Quick answer: Scientific figure rules come in three kinds, and they are not equally dangerous. File specifications (format, resolution, width, font size) are published by each journal and mostly cost you a production query if you miss them. Image-integrity rules (no selective contrast, no cloning, disclosed splices, unprocessed originals on request) can stop peer review or retract a paper. Readability rules are not enforced by any checker, yet they decide whether an editor understands the result in the minute they give it. Fix them in the reverse order of how most authors do.
Evidence basis: We read the PLOS ONE figure guidelines, Nature's final-submission instructions, the Nature Portfolio image-integrity policy, the IEEE Author Center graphics page, and Rougier, Droettboom and Bourne's "Ten Simple Rules for Better Figures" on September 30, 2026. Numbers below come from those pages. The ordering, the failure patterns, and the checklist are Manusights editorial judgment from pre-submission manuscript review.
The three kinds of figure rule
Kind of rule | Who checks it | What a miss usually costs |
|---|---|---|
File specification: format, dpi, width, font, colour mode | Editorial office or production team | A request to resupply files, sometimes a delay in production |
Image integrity: processing, splicing, unprocessed originals | Editors, sometimes with screening software | A halt in review, a correction, or a retraction |
Readability: message, labels, scale, colour choice | Nobody formally; the editor and reviewers by eye | A misread result, a skeptical reviewer, or a desk decision made on a figure nobody could parse |
The second row is the one to fear and the third is the one that most often changes the outcome quietly. The first row is where authors spend most of their evening.
File specifications: two journals, two sets of numbers
There is no universal figure standard. Two large publishers show how far the numbers drift:
Specification | PLOS ONE | Nature (final artwork) |
|---|---|---|
Accepted formats | TIFF or EPS only | Layered PSD or TIFF for photos; vector files for line art, with JPEG only as a fallback |
Resolution | 300 to 600 dpi | 300 to 600 dpi for photographic images |
Width | 789 to 2,250 pixels at 300 dpi (6.68 to 19.05 cm) | 89 mm single column, 183 mm double, 120 to 136 mm for a column and a half |
Text in figures | Arial, Times or Symbol, 8 to 12 point | Sans-serif, preferably Helvetica or Arial; panel labels 8 pt bold, other text 5 to 7 pt |
Source: PLOS ONE figure guidelines and Nature final-submission instructions, accessed 2026-09-30.
Three things in those pages trip people more than the headline numbers:
- PLOS ONE wants captions in the manuscript, not in the figure file, and it wants the file named to match the citation (Fig1.tif for "Fig 1"). A figure with its own title baked into the image gets sent back.
- Nature asks authors not to rasterize line art or text. A bar chart exported from a plotting tool as a flat PNG and pasted into a slide is the classic miss. It looks fine on screen and blurs at print width.
- Nature's lettering sizes are small. A 5 to 7 point limit means axis labels made for a 12-point slide will be shrunk until they break. Check the figure at 89 mm wide before you check it at full screen.
At initial submission most journals are more relaxed than at acceptance. Our Nature figure guidelines page separates the two stages for that journal, and the Science figure guidelines page does the same for Science.
Image-integrity rules: the ones that end careers
Here the language across publishers converges, because it comes from the same problem. Nature Portfolio's policy is a fair summary of the field:
- Images should be minimally processed, and the final image must correctly represent the original data.
- Brightness and contrast changes are acceptable only when applied equally across the entire image and equally to controls, and never so that data disappear.
- Cloning, healing, or any tool that deliberately obscures a manipulation is unacceptable.
- Images from different times or places should not be combined into one panel unless the legend says so; juxtaposed pieces need a visible border and a legend note.
- For life-science papers, unprocessed gels and western blots go in with the final accepted version, and editors may ask for unprocessed files during review. If those files cannot be produced, the policy says review may be halted until the issue is resolved.
PLOS ONE adds a practical point many authors miss: do not upsample an image to reach a dpi target. Setting more pixels than the camera recorded makes the software invent data that were never there. If the original is too small, say so and supply it at its native size.
A failure pattern we see often is not fraud at all. It is a lab that crops a blot tightly for a clean-looking panel, loses the molecular-weight markers and the neighbouring lanes, and then cannot find the uncropped original eight months later when a reviewer asks. Nature also requires that cropped gels keep all important bands. Save the raw file the day you acquire it. For a full pre-upload screen, see how to check figures for image integrity.
Readability rules: what reviewers notice without a checklist
No journal software flags a figure for being hard to read. Editors and reviewers do. Rougier, Droettboom and Bourne's ten rules (PLOS Computational Biology, 2014) remain the most cited statement of them. Four carry most of the weight in a submission:
- Identify your message. Each figure should make one point, and a reader should be able to say what it is from the figure and legend alone.
- Do not trust the defaults. Default tick density, line weight, and colour maps in plotting software are chosen for screens, not journal columns.
- Use colour effectively. IEEE's author guidance suggests printing a graph in greyscale and checking whether it still reads, then using line style and marker shape as well as colour, because readers with colour vision deficiency cannot separate red from green.
- Do not mislead the reader. Truncated axes, dual axes that suggest a correlation, and bar charts that hide the spread of the data all read as advocacy.
A useful test: cover the Results text and ask a colleague outside your subfield what Figure 1 shows. If they cannot say it in one sentence, the figure is not ready, whatever its dpi.
Readiness check
Run the scan to see how your manuscript scores on these criteria.
See score, top issues, and what to fix before you submit.
Which rule to fix first
If your figure has this problem | Fix it | Why |
|---|---|---|
Uneven contrast, an unmarked splice, or no raw file | Now, before anything else | This is the only category that can stop review or follow the paper after publication |
A number in the figure that disagrees with the text | Now | Reviewers read this as carelessness with the data, not a typo |
Axis labels unreadable at column width | Before submission | An editor judging significance in minutes will not zoom in |
Wrong file format or dpi for the journal | Before upload, or at the production request | Usually recoverable, rarely decisive |
In our pre-submission review work: the figure problems that matter
In our pre-submission review work, the figure problems that change how a manuscript is read are rarely about file format. Three patterns come up again and again:
1. The figure and the Results disagree on a number. The legend says n = 6, the Results say n = 8, or a p value in a panel annotation differs from the one in the text. Referees treat this as a sign the authors did not recheck their own data, and it colours how they read every other figure. The fix is mechanical: list every number that appears in both a figure and the text, and reconcile them before upload.
2. The text cites a figure panel that is not there, or is not what it says. After two rounds of internal revision, "Fig. 3E" in the Results can point to a panel that was moved to the supplement or relabelled. A reviewer hunting for a panel that does not exist loses patience fast, and a mislabelled panel can make a correct result look wrong.
3. The figure shows less than the claim says. A representative image carries a sentence that implies quantification, or a bar chart with no individual points supports a claim about consistency across samples. This is where readability and integrity meet: the figure is honest, but the text asks it to prove more than it can.
The Manusights AI review is built around these checks. It cross-references the quantitative values that appear in both the figures and the text (n numbers, p values, r values, fold changes), confirms each cited panel exists, comments on presentation when it gets in the way (unreadable fonts, missing annotations, unclear axis labels), and asks whether the data shown support what the text claims. It does not certify dpi or run forensic image-manipulation analysis; those belong to the journal's own checks and your raw files. The engine behind it was trained by 35+ current top-tier-journal reviewers, which is why it reads a figure against the claim rather than against a style sheet.
If you want that read on your own manuscript before upload, start with the free manuscript readiness scan. The figure quality check before submission page covers what a fuller figure review should deliver, and the figure legend guide covers the text that sits under each figure.
Pre-upload figure checklist
- Every figure makes one point you can state in a sentence.
- Every number shared by a figure and the text matches exactly.
- Brightness and contrast changes were applied to whole images and to controls alike.
- Any spliced or rearranged lanes are marked in the image and described in the legend.
- Raw, uncropped originals are saved and findable.
- Line art is vector; photographs meet the journal's dpi at the printed width.
- Text is legible at the journal's single-column width.
- Graphs still read when printed in greyscale.
- File names, formats, and caption placement match the target journal's instructions.
Sources accessed September 30, 2026.
- Figures, PLOS ONE submission guidelines.
- Final submission, Nature.
- Image integrity and standards, Nature Portfolio editorial policies.
- Create graphics for your article, IEEE Author Center.
- Rougier NP, Droettboom M, Bourne PE. Ten Simple Rules for Better Figures. PLOS Computational Biology, 2014.
Frequently asked questions
Most journals ask for photographic images at 300 to 600 dpi at the size they will print. PLOS ONE states 300 to 600 dpi with a width between 789 and 2,250 pixels at 300 dpi, and Nature asks for TIFF or layered Photoshop files at 300 to 600 dpi for photographs at final submission. Line art and graphs should stay as vector files rather than being converted to pixels.
Yes, within limits. Nature Portfolio's image-integrity policy allows brightness and contrast changes only when they are applied equally across the whole image and equally to controls, and never so that data disappear. Cloning or healing tools that remove or add features are not acceptable.
Follow the target journal. PLOS ONE allows Arial, Times, or Symbol in 8 to 12 point. Nature's final-artwork guidance asks for a sans-serif typeface, preferably Helvetica or Arial, with panel labels in 8 point bold and other text between 5 and 7 point.
Image-integrity problems are the serious ones: journals can halt review when unprocessed data are unavailable and can correct or retract papers later. File-specification problems usually cost a production query or a resubmission request. Readability problems rarely trigger a formal rule, but they shape how an editor and reviewers read the result.
No. Rougier and colleagues' 2014 PLOS Computational Biology article gives design principles such as knowing your audience, not trusting software defaults, and not misleading the reader. Journals do not enforce those principles as file checks, but reviewers notice when a figure breaks them.
Final step
Find out if this manuscript is ready to submit.
Run the Free Readiness Scan. See score, top issues, and journal-fit signals before you submit.
Private API processing. Your manuscript is not used to train models.
See example reportsPut the guidance to work
Move from draft advice to a practical manuscript pass.
Choose the reporting framework first, then use the final checklist when the manuscript and submission files are ready to travel together.