Journal Submission7 min read•Updated Sep 7, 2026

Science Figure Guidelines: A Pre-Upload Decision Matrix

A practical figure audit for Science submissions that distinguishes editorial readability, file integrity, and production readiness.

By Manusights Editorial Team•September 7, 2026

Editorial processThe Manusights editorial team researches and maintains these guides using source review, field-specific analysis, and our documented editorial process.How we work

Next step

Choose the next useful decision step first.

Use the guide or checklist that matches this page's intent before you ask for a manuscript-level diagnostic.

Open Journal Fit ChecklistPrivate API processing. Your manuscript is not used to train models.Run Free Readiness Scan

Quick answer: Prepare a Science figure for two gates: an editor and reviewer must be able to interpret it now, and the underlying file must remain production-ready later. Verify the current stage-specific instructions before upload; do not turn an old pixel or dimension rule into a timeless requirement.

Evidence basis: We reviewed Science's initial-manuscript instructions, information for authors, and author resources on September 7, 2026. Those pages control; the matrix below is a Manusights preflight, not a substitute for the submission system.

Official sources define the stage-specific rules. Manusights judgment is not an editorial-outcome forecast; this preflight tests whether the figures and manuscript tell the same bounded scientific story.

Gate	Pass condition	Stop condition
Scientific claim	Every panel supports a named result	Decorative or unreferenced panel
Visual encoding	Symbols, groups, scale, and units are unambiguous	Meaning depends on color alone
Reduction	Labels and key features survive likely print/web size	Readable only while zoomed in
File integrity	Fonts, vectors, images, and layers export predictably	Clipping, substitution, or compression artifacts
Reporting	Legend and methods expose necessary interpretation details	Essential methods exist only in the graphic

Audit the scientific job first

Write one sentence for each figure: “This figure lets the reader decide whether ___.” If the sentence cannot be completed, the figure may be combining unrelated claims. Check that panel order follows the reasoning in the Results and that the text does not claim more than the panels show.

Then inspect axes, denominators, uncertainty displays, sample definitions, statistical annotations, image adjustments, and scale bars. A polished graphic with a missing unit is still a failed figure. Use the reporting guideline appropriate to the study design and retain the raw data and analysis record.

Test at realistic size

Export the figure, place it in a manuscript-sized page, and read it without zoom. Test grayscale and a color-vision simulation when color carries meaning. Use direct labels where they reduce legend hunting. Keep line weights and type hierarchy consistent across the set.

This is a functional check, not a universal typography prescription. Science can change technical specifications, and different content types can have different constraints. Record the instruction URL and access date in the submission ledger.

Separate figure, legend, and source data

The figure should carry the visual comparison. The legend should define panels, symbols, conditions, and only the interpretation details needed to read it. Methods should carry reproducible procedures. Source data should preserve the values behind the display when the journal requests them.

Information	Best home	Failure to catch
What each panel shows	Legend	Reader must infer panel purpose
Acquisition and analysis	Methods	Procedure buried in tiny legend text
Exact plotted values	Source-data route	Values unrecoverable from the chart
Main result	Results narrative	Caption makes a claim the text never evaluates

Before upload, compare the figure callouts, file names, panel labels, legends, supplement, and data availability statement. Use the figure-quality check for general design work and the pre-submission checklist for the complete package. A manuscript figure check can help find claim-to-panel mismatches before submission.

Preserve a production-safe handoff

Keep the editable source, linked assets, raw-data export, analysis code, and final rendered file together. Open the exported file on a second machine or in a clean viewer when possible. Check that fonts were embedded or converted as the current instructions permit, transparency behaves predictably, and no panel was rasterized unintentionally.

Name files and panels consistently enough that an editor can reconcile the manuscript callout with the upload list. If a production request arrives later, make the change from the retained source rather than editing a compressed submission image. Record what changed so the accepted figure and the underlying analysis do not silently diverge.

Think twice if polish is hiding an evidence defect

Stop before upload when a visual redesign changes the apparent denominator, removes a null result, crops a relevant image region, or makes uncertainty harder to see. The strongest-looking version is not necessarily the most faithful one. Compare every transformed panel with the raw data, analysis output, and Results sentence it supports.

Three figure failures routinely survive a cosmetic review. The scale mismatch places panels with different ranges side by side without making the difference visible. The selective-label pattern annotates favorable comparisons while leaving equivalent null comparisons unlabeled. The legend dependency forces the reader to reconstruct group identity from dense prose because the encoding itself is ambiguous. Repair the evidence path first, then refine spacing and style.

Use a panel ledger with figure number, panel, claim, source-data file, analysis output, manuscript callout, and final verification status. That ledger is the durable artifact when a late revision changes a panel after the text has already been edited.

Readiness check

Run the scan while the topic is in front of you.

See score, top issues, and journal-fit signals before you submit.

Get free manuscript previewPrivate API processing. Your manuscript is not used to train models.See example reports

In our editorial work: read the figure without the Results text

In our editorial work, we first inspect each figure with the Results narrative hidden. The title, panel labels, axes, units, group definitions, sample information, uncertainty display, and legend should let a careful reader reconstruct the comparison without guessing. We then reveal the Results and test every directional and quantitative claim against the panels.

Next, we reverse the test. Every major visual result should have a clear manuscript callout, and every callout should identify the correct figure and panel. Supplementary panels should not carry evidence essential to a main-text claim unless the narrative makes that dependency explicit. Repeated displays must use consistent group labels and denominators.

Finally, we compare the rendered PDF with the editable source and analysis output. That catches clipped labels, font substitutions, accidental panel swaps, stale plots, and exports made before the final analysis. The method does not predict Science's editorial decision. It creates a traceable chain from data to display to manuscript claim, which is the part authors can verify before upload.

Sources accessed September 7, 2026.

Instructions for preparing an initial manuscript, Science.
Science information for authors, Science.
Science author resources, Science.

Frequently asked questions

Use Science current author instructions and initial-manuscript guidance. Initial editorial review and final production can have different file requirements, so verify the rule for the current stage.

Yes. Judge the figure at its likely published size, not only while zoomed in on a large design canvas.

A legend can define symbols and conditions, but it cannot repair missing units, ambiguous encodings, or an unreadable visual hierarchy.

Put the guidance to work