NI Lab All articles
Lab Practice

Readable Without the Color: A Twenty-Minute Accessibility Gate for Lab Figures

NI Lab
Readable Without the Color: A Twenty-Minute Accessibility Gate for Lab Figures

Photo by Photo by Evgeniy Surzhan on Unsplash on Unsplash

A figure is usually judged in the environment where it was made: a calibrated monitor, at 150 percent zoom, with the author's memory of which line is which. None of those conditions travel. The figure will be viewed on a projector in a dim hotel ballroom, printed on a departmental laser printer that renders only grayscale, and read by someone who cannot separate the red series from the green one.

About one in twelve men and a smaller but nonzero share of women have some form of color vision deficiency, most commonly in the red–green range. Publishers have begun asking authors to confirm that figures remain legible without color, and several funding agencies now treat accessibility as part of the accessibility of federally supported research products. The compliance framing, however, undersells the case. A figure that works for only one kind of reader is not yet finished; the check below is as much a quality control on the encoding as on the audience.

Twenty minutes, run before submission, is enough.

Check one: convert and shrink (five minutes)

Flatten the figure to grayscale, then view it at the width it will occupy in print — typically one column, roughly 3.5 inches — rather than full screen. The question is not whether the figure is attractive but whether adjacent categories remain distinguishable by tone alone.

A useful threshold: neighboring categories should differ by a visible step in lightness, not a subtle one. Two mid-tone colors that look elegant side by side in the authoring software often collapse into the same gray. Reading the grayscale version on a phone at arm's length is a reasonable proxy for a bad projector in a large room.

Check two: encode every distinction twice (five minutes)

Color should never be the sole carrier of meaning. Pair it with a second channel:

Red–green pairings are the most frequent offender. A palette designed for perceptual uniformity, such as viridis or the Okabe–Ito set, solves the palette problem but not the encoding problem; two series still need different shapes if the reader is to separate them in grayscale.

Check three: control line style and density (five minutes)

Dashed and dotted lines that read clearly at four inches become indistinguishable at 3.5 inches and vanish entirely on a projector. Use at most three line styles in one panel, and make the differences structural — long dashes versus dots, not two shades of dashes. Thin rules below half a point often disappear in printing; check the figure at final size rather than trusting the screen.

Density deserves the same scrutiny. A legend with eight series is a signal that the panel should be split. Small multiples — the same axes repeated per condition — almost always communicate more than a crowded single plot, and they survive grayscale conversion intact.

Check four: labels and type (five minutes)

Direct labeling, placing the series name beside the line rather than in a separate legend, removes an entire class of reading errors. Where a legend is unavoidable, order its entries to match the visual order of the series in the panel.

Text at final size should not fall below roughly 7 points, and 8 is safer for anything a reader must decode under time pressure. Check that asterisks marking significance, error-bar caps, and axis tick labels all survive the grayscale conversion; significance marks in red are a common casualty.

The pass/fail checklist

Run all eight items. A single failure returns the figure to the author; partial credit is how unreadable panels reach print.

  1. Converted to grayscale, all categories remain distinguishable by tone.
  2. No distinction relies on color alone.
  3. Marker shapes or fill patterns differ across series.
  4. Line styles differ structurally and survive at final print width.
  5. No red–green pairing carries meaning.
  6. Type is at least 7 points at final size, including within panels.
  7. Direct labels replace the legend wherever the panel allows it.
  8. The figure reads correctly at one-column width on a phone screen.

Accessibility as a rigor claim

The argument for this audit is not primarily ethical, though the ethical argument stands. It is methodological. Many figures carry their central claim in the visual comparison between two conditions, and an encoding that merges those conditions in grayscale has obscured the evidence. Reviewers who cannot parse a panel either ask for clarification — costing a revision cycle — or read the figure as ambiguous and discount the result.

In practice, the check also catches ordinary mistakes. Labs running this audit frequently discover that two categories were assigned near-identical colors in the plotting script, that a legend was ordered inconsistently with the data, or that an axis label was clipped at the print width. These are errors no amount of statistical care corrects downstream.

When there is no time to redraw

Sometimes the revision deadline leaves hours, not days. Three changes recover most of the value: add direct labels to the series, switch to a perceptually uniform palette, and increase marker or line differentiation in the two panels that carry the main claim. The remaining panels can be fixed in the next revision, but the ones a reader must decode should not be deferred.

Accessibility checks are cheap. Unreadable figures are not.

All Articles

Related Articles

The Paper Is Already Out: A 45-Minute Rehearsal for Post-Publication Fixes

The Paper Is Already Out: A 45-Minute Rehearsal for Post-Publication Fixes

Where the Protocol Hides: A One-Hour Rescue Audit for Chat and Camera-Roll Methods

Where the Protocol Hides: A One-Hour Rescue Audit for Chat and Camera-Roll Methods

Six Panels to the Bench: A Visual Comprehension Check for Lab Onboarding

Six Panels to the Bench: A Visual Comprehension Check for Lab Onboarding