No, serif fonts are not reliably easier to read than sans serif fonts. Controlled studies going back to Tinker and Paterson in 1932, and the cleanest modern test (Arditi and Cho, 2005, which added serifs to otherwise identical letters), find no meaningful difference in reading speed at normal sizes. What does change reading speed is print size, x-height, letter spacing and stroke weight, and a 2022 study of 352 readers found that the fastest font differs from person to person.
- No serif effect: letters identical except for serifs read at the same speed (Arditi and Cho, 2005).
- 35% faster reading in each person's fastest font than in their slowest (Wallace et al., 2022).
- 0.2° to 2° of x-height is the range read at full speed (Legge and Bigelow, 2011).
- x-height differences explain most "serif vs sans" results at a matching point size.
- 14% faster in readers' fastest font than in the font they said they preferred.
The serif vs sans serif readability argument has lasted because both sides have a story that sounds right:
- serif fans say the little feet "guide the eye along the line";
- sans fans say serifs turn to mush on screens.
Neither story survives the data well. This post goes through what the studies measured, why the old rules of thumb appeared, and what to actually check when you choose a text face for a book, a website or a 14-point government memo.
What is the difference between serif and sans serif type?
A serif is the short finishing stroke at the end of a letter's main strokes; a sans serif ("without serif") typeface leaves the stroke ends plain.
Serif text faces such as Garamond, Times New Roman (designed for The Times of London in 1932) and Georgia usually also have stroke contrast, meaning thick and thin parts within each letter. Sans faces such as Helvetica (1957), Arial, Calibri (Microsoft's Office default from 2007) and Aptos (its replacement, announced in July 2023) tend to have more even strokes.
That second difference matters for research. Most "serif vs sans" comparisons actually compare two whole typefaces, say Times against Arial, which differ in many ways besides serifs:
- x-height and width;
- letter spacing;
- stroke contrast;
- familiarity.
Any difference the study finds could come from any of those. Only a few experiments isolate the serif itself.
In December 2025 the US Secretary of State, Marco Rubio, told diplomats worldwide to stop using Calibri and set every official document in 14-point Times New Roman. The department had used Times from 2004 until 2023, when Antony Blinken switched to the sans serif Calibri to make paperwork more accessible to readers with disabilities. Rubio's cable called that change part of "misguided" diversity policies, and the department's statement said serif fonts are "more formal and professional". Kristen Shinohara of the Rochester Institute of Technology told NPR that the finishing strokes can make text harder to read for some people with dyslexia or low vision. Two secretaries, two memos, and neither cited a reading-speed study.
NPR via KPBS: The State Department reinstates an old font in a typeface about-face (Dec 2025)Do serifs make text easier to read? The controlled test
The one experiment that changed only the serif found no effect on reading speed. Aries Arditi and Jianna Cho built a parametric family of fonts at Lighthouse International in New York in which the letters were identical except for serifs of 0%, 5% or 10% of cap height, and they compensated for the extra letter spacing serifs bring.
Tested lowercase fonts that differed only in serif size (0%, 5%, 10% of cap height) with four normally sighted readers and two readers with age-related maculopathy. They measured size thresholds for letter recognition, rapid serial visual presentation (RSVP) reading speed and continuous reading speed. Serifs had no effect on either reading speed measure. The 5% serif font had slightly lower size thresholds, a small gain the authors put down to the extra spacing serifs add, and which they judged unlikely to matter at typical print sizes. The limit: a very small sample, and custom fonts rather than real typefaces.
Read the paper (free)A century of null results
That result fits a long line of work. Alex Poole's review of the literature, written up in 2008 from a 2003 master's thesis, lists study after study with no reliable serif effect:
- Tinker and Paterson (1932) found most typefaces about equally legible;
- Moriarty and Scheiner (1984), Tullis and colleagues (1995) and Boyarski and colleagues (1998, testing Times, Georgia and Verdana on screen) found no legibility difference;
- Ole Lund's 1999 doctoral thesis, a critical review of the legibility literature, concluded that serifs are peripheral to reading.
The studies that did find a serif advantage, such as Cyril Burt's 1959 work and Colin Wheildon's 1995 survey, have been widely criticised for their methods.
| Study | Year | What was compared | Result |
|---|---|---|---|
| Tinker & Paterson | 1932 | 10 typefaces, print | Most equally legible; the one sans (Kabel Light) slightly slower |
| Moriarty & Scheiner | 1984 | Serif vs sans, print | No difference |
| Boyarski et al. | 1998 | Times, Georgia, Verdana, screen | No legibility difference |
| Arditi & Cho | 2005 | Same letters with and without serifs | No reading-speed effect |
| Wallace et al. | 2022 | 16 fonts (3 serif, 13 sans), 352 readers | No font fastest for everyone; 35% gain from each person's best font |
| Sawyer et al. (VSS abstract) | 2025 | Print size vs x-height, 60 readers | Size mattered; x-height alone did not |

Where did the "serifs guide the eye" idea come from?
The claim is a rationalisation of habit, not a measured effect. Books and newspapers were set almost entirely in serif type for centuries, so readers were most practised in it, and Miles Tinker's work in the 1960s already noted that perceived legibility tracks familiarity. Poole's review found no eye-movement evidence that serifs pull fixations along a line.
The idea also mixes up two jobs:
| Job | How it is read | What matters most |
|---|---|---|
| Headlines and signs | At a glance or from far away | Letter shapes and spacing |
| Running text | A practised reader recognises whole words | Size, spacing and weight; stroke ends contribute little |
That is why the US Americans with Disabilities Act standards ask for sans serif on physical signage, while serif body text in books works fine.
Is sans serif better on screens?
It used to be, because of pixels, and it mostly isn't any more. On the 72 to 96 pixels-per-inch monitors of the 1990s, a 10-pixel-high lowercase letter had no room for a serif and a thin hairline at once, so fine serif detail broke up. Microsoft commissioned Georgia (serif) and Verdana (sans) from Matthew Carter in 1996 specifically for those screens, with large x-heights and hinting.
Phones and laptops now commonly run at 300 to 460 pixels per inch, where a serif has several device pixels to work with.
Wallace and colleagues' 2022 crowdsourced study, run on people's own devices, found no consistent population-wide font effect for readers under 35, and among readers aged 35 and over, the serif EB Garamond stood out as a best performer while the sans Open Sans was the slowest. That is the opposite of the old screen rule.
Recruited 500 Amazon Mechanical Turk readers and kept 352 after quality filters (ages 18 to 71, mean 33). Each read short passages in Times, Noto Sans and other fonts from a set of 16, with sizes matched by perceived size rather than point size, plus comprehension questions. Across everyone, font had no consistent effect: no single font was fastest for most people. But each reader's fastest font gave 314 words per minute on average, against 275 in their preferred font and 232 in their slowest, a 35% spread with comprehension unchanged (median 90%). Reading speed fell 1.51 words per minute per year of age. Limits: online, uncontrolled screens, only three serif fonts, and short passages.
Read the paper (free)See the data as a table
| Words per minute (wpm) | |
|---|---|
| Each reader's fastest font | 314 |
| Each reader's preferred font | 275 |
| Each reader's slowest font | 232 |
| EB Garamond (serif), group average | 312 |
| Open Sans (sans), group average | 254 |
Source: Wallace et al., ACM TOCHI 2022 · 352 readers. Group averages per font mix different readers, so the authors warn they partly reflect who was assigned each font.
Two details worth more than the headline
- Preferences were poor guides: people read 14% faster in their fastest font than in the font they said they preferred.
- Self-comparison changed the ranking: the authors' "speed rank" measure, which compares each reader with themselves, put four very different sans faces on top (Franklin Gothic, Calibri, Oswald and Montserrat), while EB Garamond's high group average came with a speed rank of only 48%.
The font effect is real but personal.
Readers read 14% faster in their fastest font than in the font they said they preferred. Wallace et al., ACM TOCHI, 2022
If not serifs, what makes text readable?
Size, measured as x-height, comes first. Gordon Legge and Charles Bigelow's 2011 review shows that normally sighted adults read at maximum speed across a tenfold range of lowercase heights, from about 0.2° to 2° of visual angle; at a 40 cm reading distance that is an x-height of 1.4 mm to 14 mm.
Combines decades of psychophysics with new measurements of printed matter. The fluent range, where reading speed is at its maximum, runs from about 0.2° to 2° of x-height. The authors measured running text in newspapers (mean 0.23°, range 0.20° to 0.26°) and in paperback and hardcover books (mean 0.24°), all sitting just above the critical print size. Karow's survey of 1,049 roman typefaces found x-height fractions from 0.28 to 0.58 of the body size, and their own 51 serif text fonts ranged 0.36 to 0.55 (mean 0.45), which is why two fonts at "12 point" can look very different. Limit: their print survey covers serif text faces, and the speed data come mostly from lab tasks.
Read the paper (free)Why the same point size is not the same size
That x-height spread explains many "serif vs sans" results. Times New Roman was cut to fit more words in a newspaper column and has a small x-height; Verdana, Calibri and Montserrat have large ones. Compare them at the same point size and you are really comparing a small font with a big one.
A 2025 Vision Sciences Society abstract by Ben Sawyer's group (60 readers) found print size drove reading speed while x-height alone did not, with one exception: at small sizes, fonts with larger x-heights were read slightly faster, about 248 against 232 words per minute.
Then spacing and weight
After size come spacing and weight. Pombo, Kim and Pelli (VSS 2025, 74 online readers, 12 fonts) found that visual crowding, the way neighbouring letters interfere with each other, correlated with reading speed (r = -0.58) and strongly with reading comfort (r = -0.93); text fonts generally crowded less than display fonts.
Guidi, Kosovicheva and Wolfe's variable-font work found the thinnest hairline strokes slowed reading regardless of width. So a high-contrast Didone with hairline serifs can hurt at small sizes, not because it has serifs but because its thin strokes and tight fit disappear.
How big is your x-height, really?
Use this to check what you just read. Enter a font size in CSS pixels, the font's x-height as a fraction of the em (from the font file or its specimen; 0.45 is the serif average Legge and Bigelow measured) and your viewing distance.
x-height in degrees of visual angle
At the defaults, 16-pixel text with a 0.45 x-height viewed at 40 cm gives an x-height of about 1.9 mm, or 0.27°, comfortably fluent.
Drop to 12 pixels with a small-x-height font such as Garamond at 0.40 and you are at about 0.18°, below the critical size. Laptop screens viewed at 60 cm need proportionally larger text: the same 16 pixels shrinks to about 0.18° at that distance.
So which should you use for body text?
Choose by size, spacing, language support and voice, not by serif. In practice, that means a short list of checks:
- Set the size by x-height. Aim for at least 0.24° at the expected reading distance, and test the font at that size, not at a matching point size. Our guide to kerning, tracking and leading covers the spacing that goes with it.
- Check the thin strokes. At small sizes or on low-quality print, avoid hairline serifs and high contrast. A sturdy serif such as Merriweather or Source Serif reads as easily as a humanist sans.
- Check every script you set. In India, the Latin face also has to sit next to Devanagari, Bengali or Tamil. A pairing that matches x-height to the Devanagari headline matters more than serifs; see bilingual Hindi-English layout and how Devanagari fonts work.
- Let readers change it. Wallace's 35% spread is the strongest argument for reader settings: an e-reader font menu or a site that respects browser font-size preferences helps more than any single "most readable" choice.
- Use contrast, not just shape. Grey 300-weight text fails readers whatever its serifs. Check it with the methods in WCAG and APCA contrast.

Why did the US State Department switch fonts twice?
For reasons of image and accessibility policy, not reading-speed data, as the story near the top of this post shows. The department used Times New Roman from 2004 to 2023, when Secretary Antony Blinken moved it to Calibri to make documents more accessible; in December 2025 Secretary Marco Rubio ordered a return to Times New Roman at 14 point, calling serifs more formal and professional. Accessibility researchers quoted by NPR warned that serifs can be harder for some readers with dyslexia or low vision, but the controlled evidence above suggests the bigger lever in both memos was the 14-point size.
The lesson for a designer is that typeface choices carry meaning (formal, modern, friendly) well before they carry any legibility effect. That is a legitimate reason to pick a serif or a sans. Just don't justify it with a reading-speed claim the research doesn't support. For the brand side of that argument, see why logos look dated, and for the layout that holds the type, the typographic grid guide. More of my studio notes are on the Kala page.

What's new
- US State Department drops Calibri and returns to 14-point Times New Roman NPR via KPBS, Dec 2025
- Readability Consortium publishes VSS 2025 abstracts on x-height, crowding and variable fonts Readability Matters, Sep 2025
- Posters report younger readers faster in sans, older readers faster in bold serif Readability Matters, Oct 2024
Questions people ask
Is serif or sans serif better for dyslexia?
Neither has strong evidence on its own. The controlled studies above found no serif effect in general readers, while accessibility researchers such as RIT's Kristen Shinohara caution that decorative strokes can add difficulty for some dyslexic and low-vision readers. The safe levers are the measured ones: larger x-height, looser letter spacing and sturdy strokes.
Why do books use serif fonts and websites use sans serif?
Mostly history and technology. Book typography inherited centuries of serif text faces, while early screens at under 100 pixels per inch rendered serifs poorly, so web design adopted sans. On today's high-density screens that technical reason has largely gone, and many reading apps default to serifs.
Is Helvetica a good font for long text?
It is fine at generous sizes but not ideal. Its tight spacing and closed apertures (the openings in letters like c, e and s) make similar letters harder to tell apart at small sizes. Humanist sans faces with open apertures, such as Frutiger or Source Sans, usually work better for long passages.
What font size is most readable for body text?
For screens read at 40 to 60 cm, 16 to 18 CSS pixels with an x-height around 0.45 to 0.5 of the em keeps text well inside the fluent range. In print, 9.5 to 12 point is typical, adjusted for the font's x-height. Use the calculator above for your own distance.
Sources
- Arditi, A. and Cho, J.: Serifs and font legibility, Vision Research (2005)
- Wallace, S. et al.: Towards Individuated Reading Experiences, ACM TOCHI (2022)
- Legge, G. E. and Bigelow, C. A.: Does print size matter for reading? Journal of Vision (2011)
- Alex Poole: Which are more legible, serif or sans serif typefaces? (literature review) (2008, updated 2012)
- Readability Matters: VSS 2025 abstracts (Sawyer et al.; Pombo, Kim and Pelli; Guidi et al.) (2025)
- Readability Matters: VSS 2024 readability posters (2024)
- NPR via KPBS: The State Department reinstates an old font (2025)
- Microsoft Design: A change of typeface, Microsoft's new default font has arrived (2023)
- Graphic Design Stack Exchange: Serifs improve readability for long text, myth or truth?

