Readwyn 0.1.0

Serif or sans serif: which is easier to read on a screen?

Open the font menu of almost any reading app and you'll find both kinds: faces with small finishing strokes at the ends of the letters, such as Georgia or Literata, and faces without them, such as Helvetica or Avenir. Choose the one you find comfortable. Controlled experiments have found little or no difference in reading speed between serif and sans serif type, and the old case against serifs on screens was made for monitors that drew each letter with a handful of pixels.

What the experiments found

Comparing one whole typeface with another, say Times with Arial, tells you little about serifs, because the two also differ in letter width, stroke contrast and spacing. Aries Arditi and Jianna Cho, at Lighthouse International in New York, avoided that by building lower-case fonts that were identical apart from the serifs: none, serifs 5% of the cap height long, and serifs 10% long. They measured the smallest size at which people could read each font, and their reading speed both in ordinary text and with words flashed one at a time. Serifs changed neither speed. The 5% font was slightly more legible at the smallest sizes, a gain the authors suspected came from the extra space that serifs push between letters, and they concluded that fonts differing only in serifs showed no difference in legibility (Vision Research, November 2005).

The font you read fastest in is a personal matter. In a 2022 paper in ACM Transactions on Computer-Human Interaction, Shaun Wallace and ten co-authors compared fonts on hundreds of crowdsourced readers, with sizes matched by how big the text looks rather than by point size. On average, each reader was 35% faster in their own fastest font than in their slowest, with comprehension unchanged. Readers' fastest fonts weren't the same, which is why the authors argued that no single font suits everyone.

Familiarity matters as well. Sofie Beier and Kevin Larson had people read fonts with common and with unusual letter shapes, tested them, let them read a story in the font for 20 minutes, and tested them again. The 20 minutes of exposure had an effect on reading speed and the unusual shapes did not, though readers disliked them (Information Design Journal, 2013). So for speed, what counted was the time spent reading a face, not how ordinary its letters looked, and the style you read most easily may be the one you read most.

Since the best face varies from person to person, read a chapter in each and notice which one you stop seeing. (Readwyn's Text panel offers four serifs and two sans serifs, and the choice applies to every book on that Mac.) Two well-known faces drawn for reading on screens happen to be serifs: Literata, which TypeTogether made for Google Play Books, and Bookerly, the Kindle's default.

Where the rule about screens came from

The idea that screens need sans serif comes from monitors whose letters were only a few pixels tall. Microsoft's notes on Verdana, which Matthew Carter designed for screen display and which was first released on 8 July 1996, explain the problem: with so few pixels, letters set close together collide, and an "r" and an "n" can merge into an "m". Verdana's answer was wide letters and generous spacing.

Yet Carter also drew a serif face for the same screens in the same year. Microsoft describes Georgia, released in September 1996, as made for clarity at low resolutions: compared with the Scotch Roman types it descends from, it has lighter capitals and a larger x-height. Microsoft added both to its Core fonts for the Web that year, Verdana in July and Georgia on 1 November.

Pixel density has since climbed well past those monitors. In September 2022 Amazon moved its cheapest Kindle from 167 to 300 pixels per inch. Type set at 10 point has an em of 10/72 of an inch, top to bottom: about 23 pixels at 167 ppi and about 42 at 300 ppi, so every stroke and serif gets nearly twice as many pixels in each direction (more on pixels per inch).