A beautiful scientific plot can still fail quietly when its colors collapse into the same gray fog on paper. If you have ever printed a manuscript figure and watched your elegant blue-green-orange palette turn into statistical oatmeal, this guide is for you. Today, you will learn how to choose colorblind-safe figure palettes, protect print contrast, and make your plots readable for reviewers, students, clinicians, collaborators, and that one lab printer that behaves like a moody espresso machine.
Why Colorblind-Safe Palettes Matter
Color is not decoration in a scientific figure. It is evidence wearing a small jacket. When the jacket is badly chosen, the evidence may still be true, but readers cannot follow it without squinting, guessing, or silently judging your plot from behind their coffee cup.
Color vision deficiency is common enough that every research figure should assume some readers will not distinguish red from green, purple from blue, or pastel categories that seemed charming on your monitor at midnight. The National Eye Institute describes color blindness as usually inherited and more common in men, which means it will appear in real classrooms, labs, clinics, review panels, and conference rooms.
I once watched a speaker explain three survival curves that looked distinct on screen but identical from the back row. The data had work boots on. The palette had slippers.
- Colorblind-safe colors reduce avoidable misreading.
- Print contrast protects figures in grayscale and photocopies.
- Labels, line styles, and markers should carry meaning too.
Apply in 60 seconds: Convert your figure to grayscale and check whether the main comparison still makes sense.
For adjacent figure workflow problems, it may help to pair this article with a broader data visualization best practices guide and a multi-panel figures checklist. Color decisions rarely travel alone.
Who This Is For / Not For
This is for you if...
You create figures for journal articles, theses, grant applications, conference posters, clinical reports, technical white papers, or teaching slides. You may use R, Python, MATLAB, Prism, Excel, Illustrator, Inkscape, or whatever software your lab adopted during a funding drought and never quite replaced.
This is also for early-career researchers who know their results are solid but worry the plot looks like a small carnival collapsed into a PDF. There is no shame there. Most of us learn figure design by pain, reviewer comment, and one heroic postdoc.
This is not for you if...
This is not a full color science textbook, a journal-specific formatting manual, or a replacement for accessibility testing with real users. It also will not teach you how to rescue a figure with 19 categories, six nested legends, and a title longer than a grant abstract. Some plots need redesign, not a nicer palette.
| Question | Good Sign | Fix First |
|---|---|---|
| Can the reader identify the main comparison in 5 seconds? | Yes, even before reading the caption. | Simplify the message before changing colors. |
| Does color encode necessary information? | Color supports a clear variable or group. | Remove decorative color. |
| Will the figure be printed? | Yes, and grayscale has been checked. | Add line styles, markers, or direct labels. |
| Are there more than eight categories? | No, or categories are grouped logically. | Consider faceting, grouping, or reducing categories. |
The Fast Palette Rule
The fastest practical rule is simple: use hue for categories, lightness for order, and redundancy for meaning. That sentence can save a surprising number of figures from becoming interpretive soup.
Use hue for categories
For separate groups, choose colors that differ in hue and lightness. A blue, orange, and bluish-green palette is usually safer than red, green, and brown. Red and green may look festive on holiday cards, but in a results figure they can stage a tiny readability coup.
When I review student plots, I often ask, “Could I describe this figure over the phone?” If the answer is “the reddish one, not the reddish-reddish one,” the palette has already started a small fire.
Use lightness for ordered values
For heatmaps, density maps, risk gradients, fold changes, or concentration levels, use a sequential palette with steady light-to-dark movement. Readers should feel the order without needing to decode a rainbow staircase.
Use redundancy for important signals
Never make color carry the whole piano by itself. Add direct labels, line types, marker shapes, annotations, or panel titles. This matters especially for line plots, survival curves, ROC curves, dose-response curves, and stacked bars.
Visual Guide: The 4-Step Palette Safety Check
Decide what comparison the reader must see first.
Choose categorical, sequential, or diverging colors.
Check grayscale and small-size readability.
Add labels, markers, line styles, or patterns.
Print Contrast Before Color
Print is where many pretty palettes go to confess. A color set can pass on a laptop but fail in grayscale, low toner, small journal columns, conference handouts, or PDF previews on a tired office printer.
For print-safe figures, contrast is not only about color difference. It is about luminance, line width, marker size, background, grid intensity, and how much ink the figure asks from the universe.
The grayscale test
Before submission, export your figure and convert it to grayscale. Then shrink it to the expected final size. For many journals, a single-column figure may be roughly 3.3 inches wide, while a double-column figure may be about 6.8 inches wide. Always check your target journal’s actual specs.
If two lines merge in grayscale, do not merely pick brighter colors. Change dash patterns, markers, or labels. Contrast is not a beauty contest. It is a delivery system.
The small-size test
View the figure at 50 percent and 100 percent. If axis labels become dust, legends become embroidery, or markers vanish like socks in a dryer, increase size and simplify.
The background test
White or very light backgrounds usually print more predictably than saturated backgrounds. Avoid placing low-contrast colors on gray panels unless you have tested the final PDF. Scientific readers are patient, but not archaeological.
- Check final figure size, not only full-screen previews.
- Use line width, markers, and labels to protect meaning.
- Keep backgrounds clean unless color contrast has been tested.
Apply in 60 seconds: Export one plot as PDF, print it in grayscale, and circle any comparison that becomes unclear.
Recommended Scientific Palettes
The best colorblind-safe palette depends on what your data are doing. Are you comparing separate groups? Showing a gradient? Highlighting deviation around a midpoint? Each job needs a different tool. A violin does not become a hammer because the grant deadline is near.
For categorical data
Use categorical palettes when groups have no natural order: treatment arms, species, algorithms, cohorts, experimental conditions, survey categories, or model types.
Reliable starting combinations include blue and orange, blue and vermillion, bluish-green and orange, purple and gold, or dark gray plus one accent color. The popular Okabe-Ito style palette is widely used because it avoids the worst red-green traps while preserving clear category separation.
For sequential data
Use sequential palettes when values move from low to high: concentration, time, count, probability, intensity, elevation, risk, or expression level. Choose a light-to-dark scale that remains ordered when printed.
Viridis-style palettes are popular because they were designed to be perceptually more uniform and more accessible than many older rainbow scales. They also tend to behave better when converted to grayscale.
For diverging data
Use diverging palettes when values move away from a meaningful center: below versus above baseline, negative versus positive residuals, loss versus gain, pre versus post difference, or cold versus hot anomalies.
For diverging palettes, do not assume red-blue is always the royal road. It can work, but the midpoint must be visually neutral, and both directions should feel balanced in lightness. Otherwise one side looks more important simply because it shouts louder.
| Palette Type | Best For | Print Risk | Safer Add-On |
|---|---|---|---|
| Categorical | Groups without order | Similar colors merge in grayscale | Direct labels and marker shapes |
| Sequential | Low-to-high values | Weak lightness steps flatten | Legend ticks and clear endpoints |
| Diverging | Values around a midpoint | One side may dominate visually | Neutral center and symmetric limits |
| Highlight palette | One key group among many | Accent may vanish if too pale | Use gray context plus bold annotation |
For figure production timing, a palette decision should happen before final export. If your lab uses a lock process, this pairs naturally with a lightweight figure freeze policy. Color revisions after manuscript assembly are tiny paper cuts with administrative shoes on.
Palette Decision Card
When people ask, “What palette should I use?” the honest answer is, “What job is the figure doing?” This decision card gives you a quick route without requiring a design degree, a colorimeter, or a ceremonial candle beside your plotting script.
Decision Card: Choose a Palette in 90 Seconds
- One main group versus background? Use mostly gray with one strong accent color.
- Two to five unrelated groups? Use a colorblind-safe categorical palette with direct labels.
- Six to eight unrelated groups? Add marker shapes, line styles, or facets. Do not rely on hue alone.
- More than eight groups? Reconsider the plot design. Use small multiples, grouping, or interactive supplements.
- Continuous values? Use a sequential palette with clear lightness movement.
- Positive and negative values around zero? Use a diverging palette with a neutral midpoint.
Risk scorecard
| Risk Factor | Low Risk | High Risk |
|---|---|---|
| Number of categories | 2–5 | 9 or more |
| Line width | Visible at final size | Hairline strokes |
| Legend dependence | Direct labels included | Reader must constantly cross-check legend |
| Color pair choice | Blue-orange or blue-vermillion style pairs | Red-green, pink-gray, pale yellow-white |
| Grayscale survival | Main message remains clear | Groups become indistinguishable |
Score your figure honestly. One high-risk item is a warning. Three high-risk items mean the plot is standing on a chair changing a lightbulb during an earthquake.
How to Test Your Figure
Testing is where color confidence becomes reader confidence. You do not need fancy equipment to catch most problems. A few practical checks will reveal whether your figure is ready for peer review, print, and tired humans.
The 5-minute test
- Export the figure at final intended size.
- Convert it to grayscale.
- View it at 100 percent and 50 percent.
- Ask whether the main conclusion is still visible.
- Send it to one person who has not stared at the data for six weeks.
I once sent a grayscale plot to a collaborator who replied, “Are there supposed to be four lines?” There were seven. That email did more good than an afternoon of font polishing.
The legend removal test
Hide the legend and see whether the reader can still infer the main comparison from labels, annotations, or plot structure. If the figure becomes a scavenger hunt, add direct labels. Legends are useful, but they should not be the entire navigation system.
The photocopy test
Print the figure in grayscale or save it using a black-and-white PDF setting. If your institution still has a copier that groans like an ancient boat, use it. That machine knows things your monitor hides.
Mini Calculator: Print-Failure Risk
Use this tiny scoring tool as a rough pre-submission check. Lower is better.
Score: Not calculated yet.
Show me the nerdy details
Color difference is not only hue difference. Two colors may have different hues but similar luminance, which means they can collapse into nearly identical grays when printed. For ordered data, perceptual uniformity matters because equal steps in the data should feel like equal visual steps. For categorical data, separability matters more than smoothness. For print, line weight, marker size, and background contrast often rescue meaning when hue alone fails.
Common Mistakes
Most palette mistakes come from good intentions moving too quickly. The figure looks polished on screen, the deadline breathes warmly on your neck, and suddenly the color choices become permanent. This is how small design gremlins sneak into respectable manuscripts.
Mistake 1: Using rainbow palettes for everything
Rainbow palettes can create false boundaries and uneven emphasis. They often imply sharp transitions where the data are continuous. They also tend to behave poorly in grayscale. Use them only when you have a defensible reason, not because the plot menu offered a tiny weather map.
Mistake 2: Red versus green as the main contrast
Red-green comparisons are risky for colorblind readers and often weak in print. If your figure depends on that pair, change it or add strong redundancy through labels, shapes, or line types.
Mistake 3: Too many categories
A palette cannot solve category overload. If you need 14 colors, the real problem may be plot structure. Use facets, grouping, sorting, or a table. Your figure should not feel like a jar of spilled buttons.
Mistake 4: Pale colors on pale backgrounds
Pastel colors can look elegant in a slide deck and disappear in a printed PDF. Pale yellow on white is especially fragile. It has the confidence of a whisper in a train station.
Mistake 5: Legend-only decoding
Readers should not bounce between the plot and legend 30 times. Direct labels reduce cognitive load, especially in line charts. If space is tight, label the most important groups and reduce the rest to context.
- Avoid rainbow defaults unless the data truly require them.
- Replace red-green contrasts with safer pairings.
- Use layout and labeling to reduce color dependence.
Apply in 60 seconds: Pick one plot and add direct labels to the two most important lines.
If your figure contains microscopy, western blots, gels, or images, palette issues can overlap with image integrity expectations. A separate image manipulation red flags guide can help you avoid accidental trouble during enhancement and export.
Short Story: The Plot That Survived Peer Review
Short Story: The Gray Printer Trial
The lab had a bar chart that everyone loved. Four treatment groups, a neat legend, soft colors, and error bars standing upright like well-behaved fence posts. On screen, it looked calm and expensive. Then someone printed it for a meeting.
The blue and purple bars became nearly identical. The pale green bar faded into the background. The key result, the one sentence the manuscript needed most, vanished into gray paper. For a minute, the room was quiet in that special academic way where nobody wants to accuse the figure of being the problem.
A postdoc changed the palette to blue, orange, dark gray, and bluish green. Then she added direct labels above each group and made the key comparison slightly darker. The figure lost a little glamour but gained a spine. Two weeks later, a reviewer praised the clarity of the results. Nobody mentioned color. That was the victory.
The lesson is practical: a scientific figure does not need to look flashy. It needs to survive the journey from data to reader.
Tool-Specific Tips
Every plotting tool can make accessible figures, and every plotting tool can make a small disaster wearing a legend. The trick is to set defaults intentionally and test exports, not just previews.
R and ggplot2
In R, consider using viridis-style scales for continuous data and colorblind-safe manual palettes for categorical plots. Keep your palette choices in reusable variables so every figure in a paper speaks the same visual dialect.
For manuscripts with multiple figures, build a small palette object early. This helps prevent “Figure 2 blue” from meaning a different group than “Figure 4 blue.” Reviewers notice those little continuity goblins.
Python and Matplotlib or Seaborn
In Python, Matplotlib supports several perceptually friendlier colormaps, including viridis, plasma, inferno, and cividis. For categorical plots, define a stable color dictionary rather than letting categories change color when data order shifts.
I have seen one notebook assign the control group three different colors across three plots. The notebook was not evil. It was merely unsupervised.
GraphPad Prism, Excel, and point-and-click tools
Point-and-click tools are fine when handled with discipline. Replace default red-green palettes, increase line widths, and export at final size before judging. In Excel, avoid relying on automatic chart colors, which may look cheerful but often ignore accessibility.
Illustrator and Inkscape cleanup
Vector editing can help with labels, alignment, and final polish. But do not manually recolor figures in a way that breaks consistency with the data source. Keep a record of final palette values. A small figure style note in your project README can prevent future detective work.
For computational papers, document plotting packages and palette choices in your environment notes. A computational environment appendix or README-first research workflow can make figure reproduction less mysterious.
When to Seek Help
This topic is not medically or legally high-risk in the usual sense, but it can become professionally important. Seek help when figure readability affects clinical interpretation, public health communication, regulatory submissions, safety instructions, or major publication decisions.
Ask a statistician or data visualization specialist when...
- The plot contains many groups, nested variables, or uncertainty intervals.
- The color scale may change the perceived conclusion.
- The figure is central to a manuscript, grant, policy memo, or technical report.
- The audience includes non-specialists who need fast, accurate interpretation.
Ask an accessibility specialist when...
- The figure will be used in public-facing education or training.
- The plot appears in a web dashboard, report portal, or online supplement.
- You need compliance with institutional accessibility standards.
- You are designing figures for patients, students, or community stakeholders.
For web-published figures, the Web Content Accessibility Guidelines from W3C are often used as a practical reference point for contrast and accessible communication. For eye-health basics, the National Eye Institute is a clear public resource. In academic settings, journal instructions remain the final boss, complete with oddly specific file requirements.
- Complex statistical figures need design and analysis alignment.
- Public-facing graphics need accessibility review.
- Journal requirements should be checked before final export.
Apply in 60 seconds: Identify the one figure in your project that most needs an outside readability check.
Accessibility and Ethics Note
Accessible scientific color is not a decorative kindness. It is part of clear communication. When a figure excludes readers who cannot distinguish certain colors, the problem is not with those readers. The problem is with the design.
Colorblind-safe palettes also help readers without color vision deficiency. They help people reading on dim screens, projectors, photocopies, mobile devices, grayscale printouts, and PDFs compressed into strange afterlives by submission systems.
Ethical figure design is not about perfection
You do not need to become a color scientist overnight. You need a reliable workflow: choose safer palettes, test grayscale, add redundant encodings, and ask whether the figure still tells the truth when color is weakened.
Be careful with medical and safety graphics
If a plot supports clinical decisions, medication guidance, environmental risk, engineering safety, or public health advice, test it more carefully. In those cases, unclear color can become more than inconvenience. It can become misunderstood action.
Buyer Checklist: Tools and Services for Better Figures
If you are choosing software, plugins, templates, or a figure design service, look for practical accessibility support rather than vague promises.
- Built-in colorblind-safe palettes for categorical and continuous data.
- Grayscale preview or contrast checking.
- Export controls for PDF, SVG, PNG, TIFF, and journal-size output.
- Reusable style templates for multi-figure manuscripts.
- Clear documentation of color values and plotting settings.
- Support for direct labels, patterns, line styles, and marker shapes.
FAQ
What is the best colorblind-safe palette for scientific plots?
There is no single best palette for every plot. For categories, use a tested colorblind-safe categorical palette such as blue, orange, bluish green, vermillion, purple, and dark gray. For continuous values, use a perceptually ordered sequential palette such as viridis-style options. For centered differences, use a balanced diverging palette with a neutral midpoint.
Is blue and orange colorblind-safe?
Blue and orange is usually much safer than red and green because it remains more distinguishable for many readers with common color vision deficiencies. It is also often stronger in grayscale because the colors can be chosen with different lightness levels. Still, test the final figure rather than trusting the pair blindly.
Why should I avoid red and green in scientific figures?
Red and green are difficult for many colorblind readers to distinguish. They can also collapse into similar tones when printed in grayscale. If red and green are scientifically meaningful in your field, add labels, symbols, line styles, or patterns so readers do not rely on color alone.
Are rainbow colormaps bad for scientific data?
Rainbow colormaps are often risky because they can introduce artificial visual boundaries and uneven emphasis. They may make some value ranges appear more important than others. For ordered data, a sequential palette with steady lightness changes usually communicates values more honestly.
How do I make a figure readable in black and white?
Convert the figure to grayscale, shrink it to final publication size, and check whether the main message survives. Use different line styles, marker shapes, direct labels, thicker strokes, and adequate spacing. A black-and-white figure should not require heroic concentration.
Should I use patterns instead of colors?
Patterns can help in bar charts, maps, and filled areas, especially when printed. Use them carefully because dense patterns can become noisy. For line plots, line type and marker shape are usually cleaner than patterns.
How many colors can I safely use in one scientific figure?
For most static figures, two to five colors are easiest to read. Six to eight can work with careful labeling and distinct markers. More than eight often suggests the figure needs grouping, faceting, or a different design rather than a larger palette.
Do journals require colorblind-safe figures?
Requirements vary by journal, but many journals expect clear, accessible figures that remain readable in print and online. Even when not explicitly required, colorblind-safe figure design reduces reviewer friction and improves reader trust.
What is the easiest way to test a palette?
Export the figure, convert it to grayscale, view it at final size, and ask someone unfamiliar with the project to identify the main comparison. If they hesitate, the figure needs stronger contrast, simpler design, or redundant encoding.
Can I use colorblind simulators for final approval?
Colorblind simulators are useful screening tools, but they are not perfect substitutes for real accessibility review. Use them alongside grayscale testing, direct labels, and feedback from actual readers whenever the figure is important.
Conclusion
The quiet tragedy from the opening is avoidable: your figure does not have to become gray oatmeal in print. A strong colorblind-safe palette starts with the reader’s task, not the software default. Choose the right palette type, protect lightness contrast, test in grayscale, and make color share the workload with labels, markers, and line styles.
Your next step is simple and useful within 15 minutes: take one important plot, export it at final size, convert it to grayscale, and add one redundant cue where the meaning gets weak. That small act can turn a pretty figure into a trustworthy one.
Last reviewed: 2026-06
