Nature Formatting Requirements 2026: Figures & Limits

Nature applies two sets of figure requirements: a relaxed one at submission, a strict one after acceptance.

That single fact explains most of the confusion around Nature's scientific figure specifications, including why you will find three different column widths on the journal's own pages. It also means the honest answer to "what size should my figure be?" depends on which stage you are at — and this page gives both, with the source for each.

Everything here is read from Nature's own author pages and the artwork guides they link to. Where those sources disagree — and on two significant points they do — both versions appear, with the stage each applies to.

Nature figure specifications at a glance

Every value in that table is stated on a Nature page or in an artwork guide those pages link to. The rest of this page explains what each one means in practice, and flags the two places where the journal's own sources do not agree with each other.

How wide can a Nature figure be?

The wording, in each journal's own words:

"For guidance, Nature's standard figure sizes are 90 mm (single column) and 180 mm (double column) and the full depth of the page is 170 mm." — formatting guide

"For guidance, Nature's standard figure sizes are 89 mm wide (single column) and 183 mm wide (double column). The full depth of a Nature page is 247 mm. Figures can also be a column-and-a-half where necessary (120–136 mm)." — final submission

"The widths of printed figures are: 89 mm (single column) / 183 mm (double column) — The maximum height for a Nature figure is 170 mm, to allow space for the figure legend to fit underneath." — Nature's figure guide

• Build to 89 mm and 183 mm. Those are the numbers on the production pages and in the artwork guides the art department works from, and every source that describes the printed page agrees on them.
• Treat 170 mm as your height ceiling, not 247 mm. The legend has to fit underneath.
• The 1–3 mm difference only costs you something if your figure is flattened. In editable vector artwork, a rescale from 90 mm to 89 mm changes nothing you can see. In a bitmap sized to exactly 90 mm with baked-in 5 pt labels, the same rescale pushes your smallest text below the minimum. That is the real reason to keep everything vector, and it is a stronger argument than any of the individual numbers.
• If you build at 90 / 180 during writing and revision, that is fine — the initial submission pages say so. Just do not flatten it.

What is the minimum font size inside a figure?

"Minimum text size: 5 pt." · "Maximum text size for all other text: 7 pt."

The one exception is panel labels:

"Separate panels in multi-part figures should each be labelled with 8 pt bold, upright (not italic) a, b, c."

Three related rules come from the same pages and are easy to satisfy once you know them:

Which file formats and resolutions does Nature accept?

1. Every main figure that contains line art, axes, labels or a schematic: supply .ai, .eps or .pdf with editable layers. Every source above agrees on this, and it is the only delivery that satisfies all five simultaneously.
2. Photographs: layered PSD, or TIFF at 300 dpi or above. TIFF is refused by the artwork PDFs but explicitly preferred for photographs on the current final submission page. Layered PSD is the format both sides accept, so it is the one to reach for when you have a choice.
3. Do not let a .png or .jpg be the only file you hold for any panel with text in it. Even where it would be accepted, it eliminates the edits the art department needs to make.
4. Extended Data is the exception, in the other direction. There, .jpeg is the preferred format — see below.

Resolution

Two traps in that table.

How many display items can a Nature paper have?

"The typical length of a 6-page article with 4 modest display items (figures and tables) is 2500 words… The typical length of an 8-page article with 5-6 modest display items is 4300 words."

The definition of "modest" is the part that makes this usable:

"A 'modest' display item is one that, with its legend, occupies about a quarter of a page (equivalent to ~270 words)."

"A composite figure (with several panels) usually needs to take about half a page, equivalent to about 600 words."

So a multi-panel mechanism figure spends roughly two modest items' worth of your budget. An uninterrupted page of text is "about 1250 words," and 50 references take up approximately 900 words of space.

Nature is explicit about what happens if you overrun:

"Authors of contributions that significantly exceed the limits stated here… will have to shorten their papers before acceptance, inevitably delaying publication."

Physical sciences papers "do not normally exceed 6 pages on average," biological, clinical and social-sciences papers "do not normally exceed 8 pages on average" — with final print length at the editor's discretion.

Extended Data and Supplementary Information

This is where Nature differs most from journals that simply have a "supplementary" bucket, and it is worth getting right because the two containers have opposite rules.

Extended Data

"A maximum of ten Extended Data display items (figures and tables) is typically permitted."

Two rules that surprise people:

"Extended Data figures are not edited or styled by Nature's art department; for this reason, authors are requested to follow Nature style as closely as possible."

If space allows, Nature "encourages authors to include a simple schematic, as a panel in an Extended Data figure, that summarizes the main finding of the paper" — the closest thing in Nature's format to a graphical summary.

Supplementary Information

The rule that catches people is the first one:

"Supplementary Information should not contain figures (any figures additional to those appearing in print should be formatted as Extended Data figures)."

So the ten-item Extended Data cap is a real cap on your additional figures. You cannot overflow into Supplementary Information.

Word limits and the summary paragraph

"Articles start with a fully referenced summary paragraph, ideally of no more than 200 words, which is separate from the main text and avoids numbers, abbreviations, acronyms or measurements unless essential."

Nature also prescribes the shape of those 200 words: "2-3 sentences of basic-level introduction to the field; a brief account of the background and rationale of the work; a statement of the main conclusions (introduced by the phrase 'Here we show' or its equivalent); and finally, 2-3 sentences putting the main findings into general context."

Formatting quirks specific to Nature

If you read only one section here, read this one. These are the requirements that do not follow from general good practice and that no publisher-wide guideline would tell you.

The journal goes further and actively discourages doing the work early:

"Please note that print-publication quality figures are large and it is not helpful to upload them at the submission stage."

"Do not submit publication-quality figures until asked to do so by an editor."

"Line weights and strokes should be set between 0.25 and 1 pt at the final size (lines thinner than 0.25 pt may vanish in print). Do not rasterize or outline these lines if possible."

Hairline defaults in most plotting libraries sit below 0.25 pt. This is the single most common way a figure that looks perfect on screen loses its axes in print.

"Our art team need to make edits to text, lines, shapes, etc and therefore require everything on separate layers, including scale bar text in images."

"For photographic images with scale bars, please do not flatten the scale into the image. Keep the scale bar and any text on a separate, editable layer."

And use scale bars at all: "Scale bars should be used rather than magnification factors."

"In order to help cover some of the additional cost of four-colour reproduction, Nature Portfolio charges our authors a fee for the printing of their colour figures. Please contact our offices for exact pricing and details. Inability to pay this charge will not prevent publication of colour figures judged essential by the editors, but this must be agreed with the editor prior to acceptance."

• The Methods section "cannot contain figures or tables." Everything visual belongs in the main display items or Extended Data.
• Nature "discourages the use or adaptation of previously published display items" — reusing your own published figure is a conversation with the editor, not a formality.
• Chemical structures go in at "100% as .cdx files," built from the Nature Portfolio ChemDraw template.

Figure problems that delay publication at Nature

Ranked by how much time they cost once production finds them:

1. Line art or text rasterized. The art team cannot edit it, and the fix is rebuilding from source, not re-exporting. Explicitly forbidden: "do not rasterize line art or text in submitted figures."
2. Text converted to outlines, or fonts embedded as TrueType 3. Visually identical to a correct file, and equally uneditable.
3. Hairlines below 0.25 pt. They survive your screen and your laser printer and disappear in print.
4. Scale bar flattened into the image. Named specifically in the guidance, and it forces a redraw of the panel.
5. A flattened bitmap composed to exactly 90 mm. Rescaling it to the production 89 mm drags baked-in 5 pt labels under the minimum text size. Vector artwork absorbs the same rescale without a change.
6. Extended Data in the wrong container. Over 10 MB, over 300 dpi, in a format other than .jpeg / .tiff / .eps, or a table body supplied as editable Word rather than as an image.
7. Additional figures placed in Supplementary Information. They belong in Extended Data; SI "should not contain figures."
8. Significantly exceeding the length budget, which Nature says will require shortening "before acceptance, inevitably delaying publication."

Can you use AI to generate a Nature figure?

"Use of AI is governed according to risk level, not by tool type. The same framework applies across all roles: authors, reviewers and editors."

If you have seen the sentence "Springer Nature journals are unable to permit its use for publication" quoted in guidance about AI-generated images, that wording is no longer on the publisher's policy page, and neither are the three exceptions that used to accompany it. The current page carries no version number and no effective date — which is the reason to read it directly before you submit rather than trusting any summary of it, this one included.

"Creating photorealistic images (deepfakes)"

The other Red entries that can catch a figure are about substance rather than medium: "Fabricating data, citations or results," and presenting AI-generated conclusions "as human-derived." A pathway schematic is neither, so it is not excluded by the medium it was drawn in. A simulated micrograph presented as a result is both, and always was excluded.

"Human accountability is non-transferable — Accountability for scholarly content, evaluation, and editorial decisions cannot be delegated to AI systems."

And for authors specifically: "Disclose AI use for trust and transparency." Green-tier use — the lowest-risk tier — is characterised as "likely to be reversible and verifiable," which is a workable self-test for any figure: can you show where each element came from, and change any of them?

What this means for using SciFig

Three boundaries for this journal:

• Disclose it. If a generative step contributed to a manuscript figure, document it in Methods. The current policy asks for disclosure even in its lowest-risk tier, and under-declaring is the failure mode that costs you a correction.
• Check every element yourself. Accountability is non-transferable, and no export step transfers it. A mislabelled axis or a reversed arrow is your error regardless of what drew it.
• Photorealistic imagery is out. That entry sits in Red without qualification, and it is the one line here with no "with disclosure" route around it.

None of this reaches the figures you make outside a manuscript — grant applications, conference posters, slide decks, theses, lab websites. A journal's AI policy governs what that journal publishes; it says nothing about the rest of your year, which for most researchers is the larger share of the figures they make.

Build a Nature-ready figure

Three things to settle before you start drawing, not after:

• Width and height, because relabelling a rescaled figure is the expensive rework
• Colour choices, because avoiding red/green contrast and the rainbow scale is free at the start and a re-render later
• Layer discipline, because a flattened scale bar or an outlined label cannot be recovered from the export

Frequently asked questions

Sources

Every specification here was read from Nature's own author pages and the artwork guides linked from them. Where two official sources disagree, both are shown above with the stage each applies to. Reviewed 2026.

• Nature — Formatting guide — display item budget, word and reference limits, summary paragraph, colour charge
• Nature — Initial submission — first-stage figure sizes, accepted formats, file size, colour accessibility
• Nature — Final submission — production column widths, resolution, formats, line weights, Extended Data
• Nature — Final guide to authors (PDF) — artwork specification linked from the final submission page
• Nature — Final artwork guide (PDF) — older artwork specification, still online
• Nature research figure guide — panel construction, export settings, colour palette, format list
• Nature — Extended Data guide (PDF) — Extended Data dimensions, formats and limits
• Nature — Supplementary Information — what SI may contain and its size limits