A sans for small text. And another for large.

2 208 words12 min read
roissy frutiger 1970 cropped

Sans-serif typefaces were once considered unfit for small text due to their historical association with larger applications, but contemporary research challenges this belief, revealing that factors like letter spacing and counter openness are crucial for legibility, leading to the creation of new generation of Sans-serif typefaces designed specifically for small sizes.

Over the course of the 20th century, sans- serif typefaces acquired a reputation for being useful only in large-size text. They were a little bit grotesque, not quite fit for serious use. In discussions about typog- raphy, the superiority of serif over sans-serif typefaces (or vice versa) is one of the most disputed topics, a debate coloured by longstanding conventions and beliefs, without much substantiation. Many typographers believe that the purpose of serifs is to emphasise the reading direction, aiding the eyes to lead from left to right, and that a sans is unsuitable for running text. How much truth is there to this belief, and is there any evidence to support it?

Serif fonts are much older than sans fonts. Long admired are the 2,000-year-old stone inscriptions on Roman monuments, which became a major source of inspiration for punch-cutters and type designers. Sans-serif fonts date from the 19th century, first used on coins, but more commonly for larger applications on posters and in advertising. The technology of the time required specific versions of a typeface to be created, optimised for a fixed physical size. A typeface for small sizes would require more white space around the letters, and a bolder and wider body, while for larger sizes the letter spacing should be tighter and the construction more compact. A little later, in the mid-20th century, we witnessed major technological advancements in the printing industry, including the introduction of the phototypesetting machine. This new technology replaced the traditional hot metal typesetting methods, such as Linotype and Monotype machines, by using light to project characters onto photosensitive paper or film, rather than casting physical characters in metal. Phototypesetting allowed faster production times and lower costs, and paved the way for the digital typesetting and desktop publishing revolutions. Unlike the traditional hot metal typesetting methods that required the creation of a specific version of each typeface for each point size, phototypesetting and digital type worked with a single size of the typeface. The same design of letters would be scaled to small and large sizes. Often, the largest available metal versions of typefaces were used as a source to create sharper images for photocomposition, typically around 72 points. Text at such a large size doesn’t need much white space between letters. But when it is scaled down to small sizes, it feels too light, too narrow, and too hard to read. Unfortunately, this is the story of many typefaces that have been produced for photocomposition and later digitised; they are just not as usable, for the simple reason that they are being used in an incorrect size. Let’s look at Helvetica Neue, and compare it to Akzidenz-Grotesk, a structurally similar typeface, but one optimised for small size.

Fig1 bauer akzidenz grotesk clean 5pt web
A five-point version of Akzidenz-Grotesk (1898), from the Bauer type specimen, compared to the digital version of Helvetica Neue that ships with Mac OS. Note how much darker, wider and looser the Akzidenz-Grotesk is compared to a digital interpretation of Helvetica.

In the days of metal type, the production of a typeface was so labour intensive and costly that it was economically impossible to produce a typeface for the sole purpose of empirical legibility research. Research at the time was purely comparative, comparing the legibility of different typefaces, such as Gill Sans versus Times New Roman, which differed in all visual parameters, making any generalised conclusions flawed and highly questionable. It was only with the advent of computer fonts that a relatively inexpensive way came about to create experimental modifications of digital fonts that isolate each design parameter. This way, researchers for the first time could remove serifs from the typefaces while keeping the remaining parameters intact, and study the effects of reading performance. One such study by American scientists Aries Arditi and Jianna Cho from 2005 created rudimentary low-contrast letters that could be made different from each other by simply adding or removing serifs at the end of strokes and then measuring the text size thresholds and reading speed. Their study exhibited no difference in legibility between typefaces that differ only in the presence or absence of serifs. A more recent study in 2022 on behalf of the Centre for Visibility Design in Copenhagen, carried out by Katsumi Minakata and Sofie Beier, aimed to settle the dispute as to whether serif or sans-serif fonts are more legible. The authors created four new fonts, isolating the stylistic features of the serif and letter stroke contrast. Their word recognition research showed that fonts with low stroke contrast could still be read at smaller text sizes, and sans-serif fonts with low stroke contrast yielded better performance.

Fig2 Minakata Beier web
Katsumi Minakata and Sofie Beier developed a new font family for their 2022 experiment, with different variables for serif and stroke contrast, while keeping the other typeface variables identical.

The absence of serifs makes sans-serif letters narrower compared to the sans, or in other words, serif fonts usually have more white space between letters, which may be the most crucial difference between the two types of letter constructions. It has been confirmed by numerous studies that tracking fonts, or adding more white space between letters, is the single most important attribute that contributes to significantly improved ease of reading.

It is not just white space that makes text easy to read. The openness or closure of letters such as ‘c’ and ‘s’ is an essential element for letter differentiation and identification. The argument for this is rooted in the Gestalt principles that emerged in the early 20th century in Germany, describing how we interpret the complex world around us. According to the Gestalt law of closure, when looking at an image with missing parts, our brains fill in the gaps in the information and perceive the whole.

Fig3 gestalt web
The image above is made of three circular sectors. However, rather than seeing three individual shapes, our human perception interprets it as a triangle overlaying three circles, completing an incomplete shape.

Closed letter counters emphasise letter similarities and create a predictable text rhythm in large sizes. The same features, however, can make letters confusable. That’s why a typeface for small sizes requires a different approach to counter openings, to prevent the internal white space being invaded and to increase letter differentiation. This fact is supported by empirical studies, and as early as 1964, the prolific 20th-century legibility researcher Miles Tinker observed that ‘the greater the enclosed white space of a letter, the greater the legibility’. This tenet was confirmed in a study by Sofie Beier and Chiron Oderkerk from 2021, which concluded that open apertures in the letters ‘a’, ‘c’, ‘e’, ‘r’ and ‘s’ increase recognition for glance-like reading.

While most fonts for continuous reading tend to be serif fonts, there are some sans-serif typefaces that were designed primarily for continuous text, using the aforementioned design principles. An early humanist sans Gill Sans (1928), rooted in Edward Johnson’s London Underground typeface (1916), has been successfully used in book settings. Jan van Krimpen’s Romulus Sans (1931) was intended for running text but was never released. Perhaps the most successful sans typeface for small text is Frutiger (1976) by Adrian Frutiger. The starting point is the sans-serif text typeface Concorde (1964), which was adapted for the signage of Charles de Gaulle airport in Paris-Roissy, and named Roissy (1972). Frutiger remarked that ‘What was important was total clarity – I would even call it nudity – an absence of any kind of artistic addition.’ It was a typeface for distance reading; however, typefaces for distance reading typically have the same visual angle as typefaces for small size text and therefore work well for both. The early version of Roissy is rather compact and tightly spaced, something that Frutiger considered a mistake at the time. This was corrected in the adaptation of the typeface when it became a general purpose typeface marketed as Frutiger by Linotype. Frutiger is now a de facto gold standard of signage, yet it has also been used commonly for small size text, even for setting newspapers and countless magazines. Adrian Frutiger made comparisons between his other successful typeface Univers (1957) and Frutiger (1976) and noted which adjustments needed to be made to reduce the visual noise, by means of open counters and by inserting more white space.

Fig4 visual angle web
The visual angle measures the size of an object as it appears from the viewer’s perspective, rather than its actual size. Reading 10 pt text at 30 cm has the same visual angle as reading 11.75 cm text from a distance of 10 m. For this reason, a typeface for distance reading can also perform very well in small text size.
Fig5 frutiger univers web
Recreation of Adrian Frutiger’s illustration from his book Une vie consacrée à l’écriture typographique, 2004, comparing Univers 55 with Frutiger 55.

This is the inspiration behind Zed Text and Display. Rather than creating one typeface, with compromises needed in order to work in both larger and small text, Zed comes in two separate, dramatically different versions. The Text version has higher x-height, higher stroke contrast, simplified letter constructions, open counters, and loose spacing. Zed Display takes the opposite approach, with lower stroke contrast, and featuring fully formed counters that reduce the amount of white space between letters and allow tight spacing that is appealing in larger text sizes.

Fig6 zed design features web.svg
The Text version uses open counters to prevent confusion, allowing a reader to accurately decipher text at a glance. Loose spacing is probably the most important attribute of a text typeface and the Text version not only has open forms, but also plenty of white space around the letters to avoid character crowding. The Display version has unusually low contrast, with horizontals taking 90% of the vertical stem thickness. The text version has increased contrast, 80% of the verticals, which makes text clearer in small sizes. The Text version has slightly higher x-height, and extenders (ascenders and descenders) have increased in size to allow fully formed shapes such as ‘f’ and ‘g’ in the darkest cuts.

The original concept envisioned the creation of an optical design axis that would interpolate between the two separate versions, with separate versions for in-between text sizes. That would require Higher Order Interpolation, a concept proposed by Underware type foundry in 2018, which allows glyph contour points to be controlled by moving along curves as a variation axis changes, rather than in straight lines. We built a non-linear interpolation prototype and tested it in Latin and Devanagari scripts, but found it challenging to obtain the desired outcome. More importantly, we had planned from the outset to expand the range of languages in the Zed family, to a set of scripts that are fundamentally different from Latin. The Thai script, for example, is easier to read in small sizes when it includes traditional loops at the end of strokes, while at a large size, loopless forms are commonly used in advertising and display typefaces. Essentially, in various writing scripts such as Thai, Lao, Khmer, Devanagari, Bengali and many others, the basic structure of letters intended for small text size needs to be very different from that intended for large text. A single design scaled to large and small text would result in a less-than-optimal reading experience. Because of such differences in the forms of letters, mathematical interpolation between the text and display versions is impossible, and therefore we created the separate styles of Zed Text and Zed Display.

thai looped loopless
Many world writing scripts, for example Thai in this sample, use different construction models for small text type compared to larger text in display sizes. The text above is the same, with the top line set in Zed Text and the lower line in Zed Display. For this reason, Zed doesn’t have a continuous variable optical size, but rather two separate optical versions that work without compromises.

In conclusion, the reputation of serif typefaces as being more useful for smaller text, and sans better for larger text, is unjustified. Zed is a typeface family that learns from the existing cognitive psychology research of the past decades, as well as being informed by novel research that we have undertaken during the development of the typeface. It is also a design project that builds upon the best examples of 20th-century sans-serif typefaces and intends to advance the profession by rethinking how to make type as accessible as possible.

  • bilak square 900

    Peter Biľak works in the field of editorial, graphic, and type design. In 1999 he started Typotheque type foundry, in 2000, together with Stuart Bailey he co-founded art & design journal Dot Dot Dot, in 2012 he started Works That Work, a magazine of unexpected creativity, in 2015 together with Andrej Krátky he co-founded Fontstand.com, a font rental platform. He collaborates with the choreographer Lukas Timulak on creation of modern dance performances, and together they started Make-Move-Think.org, a foundation for interdisciplinary artistic collaborations. Peter is teaching at the Type & Media, postgraduate course at the Royal Academy of Arts, The Hague. Member of AGI (Alliance Graphique Internationale).