Why is Helvetica bad

The felt legibility

Contribution by Martin Liebig

Serif versus grotesque: the textbook opinion persistently assumes quality differences that obviously do not exist. A new study shows: Much more important than the often invoked “objective” legibility is what a font exudes - its “look and feel”.

Let's start out banally: writing must be easy to read. Do I hear contradiction? No result, of course.

Of course, writing can also look a little good. Emphasize the thematic context. Radiate value in their visual tonality. Kindly invite you to read. But: First and foremost, it must still be easy to read. Be absolutely legible. Good legibility is the most important thing. Do I hear contradiction?

Probably not. And that's a shame. Because typography, especially web typography, threatens to get under the wheels of technocrats, of self-chosen readability optimizers, of half-informed system administrators, of textbook-loyal media designers who have given us a web world full of "only Verdana" for years. Give pages. Often these are clerks with responsibility for optics, who reduce the appeal of good typography to the “reading speed” factor, because unfortunately aesthetics and topic adequacy cannot be measured in seconds. It's those people who like to tell us: Serifs are difficult to read on the monitor, everyone knows that by now, you'd better use “Verdana”. The addition: Does Spiegel Online do it too? Spiegel Online always comes at some point.

Legibility: a dangerous manslaughter argument

I would like to use the design diary to counter these clerks to the design: They are legible anyway, the pitifully few fonts that we can use in HTML-based web design. Perhaps not as good as their paper equivalents, but in a mutual comparison almost equally good, whether with or without serifs, whether strong in the trunk or finely chiseled, whether generous in the interior or urgent.

So before we assemble the next page again using only “Verdana” variations - always for the sake of legibility, of course - we give the typographic aesthetics a new chance, the harmonious typeface, the exciting mix above all, and the theme-appropriate look . The opportunity has seldom been more favorable: With the foreseeable spread of the new Vista series, the typographic potential on the web will grow immensely.

So if we take the chance, let's focus more on what is wrongly threatening to degenerate more and more into the marginal aspect of web typography: the “look and feel” of font. The character of the font. What a script represents and radiates: its essence. Let's also rediscover on the web what has ennobled good typography for half a millennium: creative, content-oriented, imaginative use of characters, their combination, their mutual tension, their visual mantra. We also use the medium of font in the browser to prove individuality and quality. In short: It is time to also prefer the “soft” font factors to the supposedly “hard” ones online. And more clearly than ever.

An experiment with more than 3000 participants

A daring thesis? I do not think so. I have the privilege of being able to report in the design diary about an online experiment that the Gelsenkirchen University of Applied Sciences carried out under my direction at the end of 2008 and beginning of 2009. Over 3000 people took part in this large-scale experiment : the reading times of 12 different fonts were measured, we tested different font sizes and line spacing, varied the line widths in a targeted manner - and combined all these factors with one another, resulting in 1440 typographical test configurations.

At the very end we asked for sympathy notes for various fonts - and were somewhat surprised to find that this hit the core of the matter. Because apart from the font size, there is apparently no “hard” typographical aspect whose meaning exceeds the simple aesthetic effect of the typeface.

As far as I know, this experiment was the most ambitious in terms of tasks and number of participants that has been started so far. Worldwide, mind you. But now enough boasted.

How do you measure “readability”?

“Easily legible” - what clues is this typographical compliment actually attached to? And if the clues for the quality of a typographical arrangement exist and have been named: how do you then, in the following step, convert them into a comparative unit, make them empirically measurable? A look at earlier studies shows that the chimera “readability” can be captured in different ways.

First of all: In the experiment described here, “readability” was equated with “reading speed”. That means: It was assumed that “good” typography more quickly is readable as "bad". This is a common, if not uncontroversial, method: As an alternative, for example, one could also use the so-called “increase in knowledge” as a measure of “good readability”, measured by the question: Did the reader even take in, process and understand the information in the text? But: how do you want to measure “knowledge growth” without querying the previous level of knowledge? And how can you query a participant's knowledge before the start of the test without them having any idea what to look out for in the following text?

Some researchers have also tried to use the “fatigue” of a reader as an indication of good or bad typography. This fatigue can be measured, for example, by the pulse or the number of winkers while reading. However, both are very difficult to achieve in an online experiment.

The “stumbling block” principle

But how do you measure reading times online? Our experiment was designed as follows: All participants were presented with 4 texts one after the other, each of which had two words hidden that clearly disrupted the flow of reading, but which definitely matched the topic. Readers should click on these “stumbling blocks”. As soon as the second stumbling block was discovered, time stopped. From the time until the second stumbling block was discovered, we concluded that it was “legible”.

Another very important note

All texts in the experiment comprised around 110 words. This means: Measurements were made under the conditions of classic, linear reading texts. In this respect, caution is required when it comes to transferring our measurement data to other types of text: It cannot be ruled out that the results can also be used, for example, on navigation bar fonts or mini-teasers, the sizes of which tend to be below average. It may just as well be that some of our results can also be applied to the proud commercial type “headline”. However, that is not certain.


The fracture is almost extinct

Passionate typographers will howl at the rough cut of this statement - but the fact is: we live in an Antiqua-dominated type culture. In a world that is dominated by signs that go back in essence to the capitalis of Roman antiquity. And then the next rough cut follows: In essence, these modern antiqua can be assigned to two major classes - namely that of the serif-bearing and that of the sans-serif antiqua. That the rich sphere of typography deserves (and also knows) much finer differentiations should be mentioned and simply ignored in the following.

Because serif versus grotesque: in the microcosm of typographers, this debate was the defining one of the 20th century. With the advance of the screen media, the discussion has shifted a little in terms of reasoning, but it continues.

Serif vs. grotesque - a hundred years of religious warfare

Whereby one has to admit that today we are - fortunately! - are far removed from the relentlessness, the ideological charge, even the militancy with which the writing debate was conducted in the twenties and thirties. The Nazis denigrated sans serif creations as “un-German”, as “Jewish-Bolshevik”, as “anti-popular”.

Whereby they meant less of the written forms as such than the people who created and used these typefaces: the modernists, the representatives of the “new objectivity”, the opponents of the Wilhelmine pomp in architecture, furniture construction and handicrafts. Anyone who was for the Dessau Bauhaus and “De Stijl” was for rationalization, for the uncompromising objectification of the world, including the characters - consequently the serifs had to give way.

This attitude naturally brought on the scene for whom the radical eradication of all ornament went too far. Mind you, they didn't have to be Nazis, protests were also expressed by conservative aesthetes who - quite rightly - resisted throwing overboard everything that had matured over four centuries of typography and proven itself out of ideological uncompromisingness. These people found sans serif fonts “grotesque”. A name was created for the typeface that characterizes an estimated 90 percent of all websites today.

Serif vs. grotesque - objectively it is undecided

In the 1930s and 1940s in particular, Anglo-Saxon researchers set about creating a scientific foundation for the ideological debate. The results are quickly summarized: Virtually no study of this time (and hardly any later) revealed any significant differences between serifs and sans serifs in terms of objective legibility.

In the prevailing opinion, of course, the serifs tended to keep the scepter in their hands: In the alternative, it was argued that the serifs emphasize and facilitate horizontal reading guidance (how often do you read vertically?), The upper halves of serif fonts are easier to identify (I personally only read whole letters) and the readers don't like experiments, it is best to read familiar typefaces (definitely a good argument). The overwhelming majority of the newspapers also stayed with the familiar, with the serif. But nothing helped: Objectively, it was a tie between serif and grotesque.

Screen typography: Unfortunately, the round has to go into the square

And then came the screen media with their angular pixel grids. And raised new, this time rather non-ideological questions. Because the digital processors tended to construct gruesome figures on the monitors. Which brings us to the screen debate, monitor typography.

To anticipate it right away: I think the "Times" on the screen is terrible. The angular monitor grid rapes - especially in smaller font sizes - this font, which is rightly popular in itself, into indignity, degrades fine serifs to club feet, lets curves degenerate into kinking driveways, diagonals into steep stairs and curves into saw teeth. Basically, it is an insult to the “Times” font, finely crafted in the late twenties, that you call your changeling on the screen by the same name in the first place.

The principle of font smoothing only helps a little further - the simulation of curves and fine lines by means of gray gradations. ClearType is the name of this technology, which at least in Windows Vista has meanwhile become the standard for font display. Even with this procedure, however, the original times is only partially approached.

Pixelated = illegible?

But what follows from this statement? If you ask around and read a bit, you will often come across a causal conclusion: Because the Times in particular, serif fonts in general, are so devastated on the monitor, they are difficult to read, claim many authors. And should therefore not be used. At least not in reading font sizes. Rather, sans serif fonts are recommended: that is, “grotesque” like “Arial” or “Verdana”, because their rapes on the monitor were comparatively limited. And this thesis is just as persistent as that of the superiority of serifs on paper. Allow me to insert: I consider both to be wrong.

12 fonts in comparison

Which fonts should we now put into the running for measuring the browser world? It was clear early on in the conception phase of the experiment: At least the six “classic” web fonts had to be included - those six that are presumed to be installed on almost every receiving computer worldwide and that 98 percent of all reputable websites are Estimated to be used precisely because of this mass distribution of the year 2009 anno domini: “Times”, “Arial”, “Verdana”, “Georgia”, “Tahoma” and “Trebuchet”.

We also took the "C" series from Windows Vista. Because even if the market penetration of the operating system has not yet worked out as usual at Microsoft headquarters in Redmond - it can be assumed that the XP successors Vista and 7 will establish themselves on PCs worldwide in the next two or three years. And with them the fonts “Calibri”, “Cambria”, “Candara”, “Constantia” and “Corbel”.

We have only excluded the “Consolas” from the “C series”, by the way, as well as the much older “Courier” - these two Monotype fonts, whose letters are all cut exactly the same width, are at least that way on the web at the moment beyond the fashion that an investigation didn't seem worth the effort.

Finally, “Segoe” was added as the twelfth font - the new system font from “Vista”, which, not without reason, is assumed to have an astonishing resemblance to one of the greatest typographic creations of the past 100 years: the “Frutiger ”Font. When designing the experiment, however, we underestimated the potential of the “Lucida Grande”, which has seen a huge surge in popularity, not least thanks to Twitter, and has been used on more and more websites in recent months. That is why it was unfortunately not included in the experiment. Nobody is more angry about it than myself.

Differences that are hardly worth mentioning

First of all, the most important result of the font comparison: The mentioned causal conclusion “font defaced = font is difficult to read” simply does not seem tenable. A look at the results of the study shows that the presence or absence of serifs does not play a role in the objective reading speed. Even more: From a legibility point of view, at least the 12 tested fonts hardly differ at all.

It is true that the participants read texts in the “Arial” font (a grotesque) the fastest. The worst times, however, also came about with a sans serif (“Corbel”). And what is even more important: the difference between the supposedly “best” and the supposedly “worst” font was just 3.9 percent in reading speed. Extrapolated, this means: If you read through a text in the “Arial” font in ten minutes, you need ten minutes and twenty-three seconds for the same text in “Corbel”. Can one seriously derive binding recommendations from such values?

Results of previous studies: definitely confirmed

Anyone who researches looks for laws. Anyone who collects empirically wants to derive forecasts from data. So the question is: To what extent can these first results of our experiment be generalized? Are the results applicable to all Antiqua that are used on the web, including those that we have not tested? Are they, above all, applicable to those fonts that will be created in the years to come? At this point we enter the speculative sphere. In my opinion, however, there are several arguments in favor of generalizing the font results of our study. Especially since they turned out to be highly significant in stochastic terms and on the basis of a large number of participants - and in addition, they are anything but isolated in the overall view of previous studies.

The analysis of variance of the data showed that the “font” factor as a whole has no so-called “main statistical influence” on reading performance. The choice of font is therefore not statistically significant as an influencing factor for the reading speed. Our results thus coincided with various earlier studies in which all of these were “Georgia” “won”, sometimes “Arial” and sometimes “Times” - but in which there was never a really statistically significant difference between the compared Fonts (Weisenmiller; Bernard et al.; Boyarski et al.; Hill / Scharff; Redelius; Bernard / Mills; Bernard et al.; Tullis et al.; Liebig).

The study by Bernard et al., Which showed that the test winners “Times” and “Arial” (which won “Times”!) Were significantly more readable than “Century Schoolbook”, “Courier” and “Verdana”. However, it should be noted that a total of seven fonts performed just as well as the winning letters - and this study is very isolated in its results.