Drafted with AI assistance, edited by Yahya Hafez.
Loading post…
Monthly notes on comprehension, accessibility, and AI tutoring. No spam, no upsells, unsubscribe anytime.
Accessibility guidelines have a reputation problem. The Web Content Accessibility Guidelines read like a rulebook written for someone else: a minimum contrast ratio for people with low vision, a maximum line length for people with dyslexia, resizable text for people who are getting older. If none of those labels fits you, it is easy to assume the rules do not either. The research the rules are built on says otherwise. Every threshold in the text-presentation criteria comes from laboratory measurements of how fast people read, and most of those measurements were taken on readers with ordinary eyesight. The numbers mark the point where reading starts to degrade for anyone. People with low vision just reach that point sooner.
The guidelines ask for a contrast ratio of at least 4.5:1 between body text and its background at the baseline level, and 7:1 at the stricter level. A disclosure first, since this blog belongs to a reading app: We do not yet claim full conformance with WCAG 2.2 at any level, and this post is not a scorecard. The W3C's own explanation of the stricter criterion is unusually candid about the reasoning: 7:1 was chosen to compensate for the loss of contrast sensitivity typical of a person with roughly 20/80 visual acuity, the point at which low vision usually begins (W3C, Understanding SC 1.4.6 (opens in new tab)). That sounds like a rule for a specific population. The underlying measurements are not.
Gordon Legge and colleagues at the University of Minnesota spent the 1980s measuring reading speed under controlled changes to the text itself. In the contrast study, readers with normal vision read text at falling contrast levels while the researchers tracked their speed (Legge, Rubin & Luebker, 1987 (opens in new tab)). Two findings matter here. First, normal readers tolerate a surprising amount of contrast loss on large letters; for letters a full degree of visual angle tall, contrast had to fall to about 6 percent before speed halved. Second, that tolerance shrinks fast as letters get smaller. For letters a quarter of a degree tall, about the size of ordinary body text, the critical contrast rose to 33 percent. Contrast and size trade off against each other, and the corner where both are low is exactly where fashionable light-grey, small-print web typography lives.
The population that corner catches is not small, and it grows. Contrast sensitivity declines with age beginning in the forties and fifties, a pattern first mapped across 91 adults aged 19 to 87 by Cynthia Owsley and colleagues in 1983, and the Minnesota group later showed that those age-related losses translate directly into slower reading of small and low-contrast text (Akutsu, Legge, Ross & Schuebel, 1991 (opens in new tab)).
The same laboratory produced the most useful single fact about print size. Plot reading speed against letter size and you get a plateau and a cliff. Across a range spanning a factor of ten, from an x-height of about 0.2 degrees of visual angle up to about 2 degrees, people read at their maximum speed. Below the low end of that fluent range, speed collapses toward the acuity limit (Legge & Bigelow, 2011 (opens in new tab)). Legge and the type designer Charles Bigelow surveyed five centuries of printed books and found that typographers had converged on sizes inside the fluent range long before anyone measured it.
Two things follow. The low end of the plateau is a cliff, not a slope, so text set near the critical size has no margin: a dim screen, a longer viewing distance, or a reader past fifty pushes it over the edge. And the headroom above the floor is enormous. Going larger costs nothing in speed until letters get very large indeed.
Modern screen studies agree, and they add comprehension to the picture. Luz Rello, Martin Pielot, and Mari-Carmen Marcos tested six font sizes from 10 to 26 points on ordinary web readers and found that readability and comprehension improved continuously up to 18 points, while line spacing at either extreme, very tight or very loose, made reading worse (Rello, Pielot & Marcos, 2016 (opens in new tab)). That was the general public, not a clinical group.
Debates about accessible typography tend to fixate on fonts: serif or sans, this face or that one. Aries Arditi and Jianna Cho built fonts that differed only in serif size and measured legibility by size thresholds and reading speed, in people with normal vision and in two with macular degeneration. Serifs had no measurable effect on reading, and the tiny advantage of light serifs was explained by the extra letter spacing serifs happen to add (Arditi & Cho, 2005 (opens in new tab)). In a follow-up on letter case, all capitals read faster than mixed case only when the text was near the acuity limit, because capitals are physically bigger; at comfortable sizes the advantage vanished (Arditi & Cho, 2007 (opens in new tab)). Both results point the same way as the dyslexia font research: what looks like a letterform effect is usually a size or spacing effect in disguise.
Fonts do still matter, just not in the way a single recommendation can capture. In 2022 Shaun Wallace and colleagues had 352 adults aged 18 to 71 read on their own devices in 16 common fonts, with sizes normalised so that no font was simply bigger. Comparing each person's fastest font with their slowest, reading speed differed by about 35 percent with no loss of comprehension. The fastest font was different for different people, and the font a reader said they preferred did not predict the one they read fastest (Wallace et al., 2022 (opens in new tab)). There is no best font. There is a best font for you, and you cannot find it by looking.
This is why the most important criterion in the guidelines' text-presentation section is the one almost nobody implements. Success Criterion 1.4.8, Visual Presentation, asks that readers be able to choose their own text and background colours, that lines run no wider than about 80 characters, that text is not fully justified, that line spacing is at least one and a half times the font size with wider paragraph spacing, and that text can be enlarged to 200 percent without scrolling sideways. Read it closely and it is not a list of values. It is a list of adjustments the reader must be able to make.
That matches the evidence. The fluent range varies with eyesight, age, screen, and lighting. The fastest font varies by person. Line length and spacing have a comfortable middle and two bad extremes. A fixed design can only be right for one imagined reader. A design that exposes the dials can be right for each real one. There is a cognitive dividend too: every bit of effort spent decoding faint or cramped letters is effort taken from working memory, the capacity you need to follow an argument rather than merely see the words.
The accessibility rules were never a special case. They are the measured edges of ordinary reading, and the people who need them most are simply standing closer to the edge. Move away from it and everyone reads faster.
