1 Introduction

The last 60 years have seen fundamental changes in the way that people acquire and share information. At the beginning of this period, information was mainly represented in the form of written text that was either read directly by readers themselves or else communicated by teachers or other speakers. Nowadays, much information is available on computer monitors or other screens. (One should, of course, recognise the “digital divide” both between people in the developed world and people in the developing world and also among diverse groups within the developed world in terms of their access to the relevant technologies.) The availability of this technology is particularly apparent in educational settings, where students routinely expect to acquire information through computer systems and also use such systems to submit their academic assignments to be evaluated by their teachers and other assessors.

Users of screen-based material have a choice in how they make use of that material. They can either read it in the form in which it is presented, or they can print it in hard copy. Shaikh (2004; Shaikh & Chaparro, 2004) carried out an online survey concerning reading habits and obtained responses from 330 participants: 120 were students, and the rest were recruited from a variety of e-mail lists. Most respondents were happy to read news items, newsletters, and product information online, but they preferred to scan articles in journals online before printing them off to read in more detail. This was consistent with the findings of previous studies that researchers and other professionals tended to print off articles or other important documents to read.

It used to be popular to assume that the increased use of digital technologies among young adults meant that they constituted a distinct population who thought and learned in qualitatively different ways from older people. They were variously called “Millenials” (Strauss & Howe, 1991), the “Net Generation” (Tapscott, 1998), “Digital Natives” (Prensky, 2001), and “Generation Y” (Jorgensen, 2003). However, these ideas were not supported by research evidence (see, e.g., Pedró, 2009). Surveys found quantitative differences between older and younger people in their attitudes to technology and their uses of technology, but there was no evidence for any qualitative differences in people born since the early 1980s (e.g., Jelfs & Richardson, 2013). In fact, many of today’s students retain a strong preference for print (Baron et al., 2017; Mizrachi, 2015), although some do not (Singer & Alexander, 2017). (This may well depend on mundane factors such as local printing costs.)

Which typefaces should be used in the presentation of text on computer screens? One approach was to employ versions of typefaces that were already well established in conventional printing. The top row of Fig. 10.1 contains two examples of such typefaces. As was mentioned in Sect. 1.2, the serif typeface Times New Roman was devised in 1932 for use in the London newspaper The Times. Versions became popular in printing and publishing more generally, and in the 1980s these were provided by Macintosh and Microsoft in their word-processing software. The typeface known as “Times Roman” was adopted by Apple. The sans serif typeface Arial was devised in 1982 for use on IBM printers. It was adopted by Microsoft in 1990, and it was the default typeface for several years in some of its applications; it is also available for Macintosh users.

Fig. 10.1
figure 1

Examples of common serif typefaces (Times New Roman and Georgia) and common sans serif typefaces (Arial and Verdana) for displaying text on screens

Full size image

An alternative approach was to devise new typefaces, often with the aim of ensuring their legibility on small or low-resolution screens. The bottom row of Fig. 10.1 contains two examples of such typefaces. The serif typeface Georgia and the sans serif typeface Verdana were both developed for Microsoft in the 1990s and were released in 1996. They were subsequently made available for installation on Macintosh computers. Other researchers developed entire families of typefaces. In the United States, Bigelow and Holmes (1986) devised the Lucida family, while in Russia Paratype was devised for material in both Latin and Cyrillic text (Akhmadeeva et al., 2012). Other artificial typefaces were devised by Arditi (2004), Beier and Larson (2010), and Sanocki (1987), although some of these were so radically different from traditional typefaces that they might well have caused difficulties even for experienced readers.

It was mentioned in Sect. 5.1 that actual serif and sans serif typefaces typically differ in a number of characteristics. In principle, it should be possible to devise artificial typefaces in which serif and sans serif differ only in the presence or absence of serifs. In fact, efforts to devise such typefaces disclosed a major confound with the width of the letters and the inter-letter spacing. In particular, Arditi and Cho (2005) found that adding serifs to Arditi’s (2004) sans serif style led to an increase in the mean width of the letters in order to accommodate the serifs. Conversely, Moret-Tatay and Perea (2011) found that removing the serifs from the serif style Lucida Bright led to an increase in the average inter-letter spacing. This confounding means that any results that are obtained using such typefaces are likely to be ambiguous.

In conventional paper-based printing, individual letters and other symbols are discrete physical entities. As was mentioned in Sect. 2.4, the overall height of typefaces (their body size) is traditionally expressed in terms of points, where one point is approximately equal to 0.35 mm. However, the size of typefaces is also expressed in terms of the dimensions of lowercase letters. The x-height of a typeface is the height of lowercase letters that do not have either ascenders or descenders (such as the letter x itself). When the same characters are presented on computer screens, they are simply fragmented digitised representations. Body size and x-height become relative terms since they depend on the size in which the characters are displayed on-screen. Rendering printed typefaces as digitised letter forms has been an exceedingly complex process involving complex debates and decisions (Bigelow, 2020a, b), although much of this process may well not be apparent to most display-screen users.

Section 2.5 argued that there was no reason to think that serifs and other features had the same consequences when people were reading from paper as when they were reading from computer monitors or other screens. In fact, with regard to reading from computer screens, designers and design educators have typically claimed that the legibility of sans serif typefaces was superior to that of serif typefaces (e.g., Poncelet & Proctor, 1993; Schriver, 1997, p. 508; “Universal Design”, 1999, p. 5), hence my colleague’s assertion, mentioned in Sect. 1.1, that “everybody knew” that sans serif typefaces were easier to read on screens than were serif typefaces. Even so, advocates of this position have usually failed to present any empirical evidence in support of this claim, and accordingly Part II of this book reviews the research literature with regard to the legibility of serif typefaces and sans serif typefaces when they are used to generate material on computer monitors and other screens.

2 Legibility of Serif and Sans Serif Typefaces Using Older Technology

Before considering the legibility of typefaces using modern computers, it should be acknowledged that the use of technology to enable information to be presented in other ways than as material printed on paper is by no means a new phenomenon. From the eighteenth century onwards, speakers used epidiascopes (opaque projectors also known as episcopes) and “magic lanterns” (using transparent plates) to project images of objects and other material onto viewing screens for potentially large audiences. However, from the 1950s these were superseded by overhead projectors and slide projectors. The widespread adoption of the latter technologies led to the development of recommendations for best practice. It was widely asserted, in particular, that sans serif styles rather than serif styles should be used for the projection of textual material. Nevertheless, as Phillips (1976, pp. 18–19) observed, such assertions seem to have been based on personal preferences rather than empirical research.

Adams et al. (1965) compared the legibility of different typefaces in the images produced by an overhead projector. They tested 120 children in Grades 1, 2, 3, 5, and 6 (i.e., aged 6–12) of a university’s laboratory school. The children in each grade rotated among five rows of seats at varying distances from the projection screen. They were shown groups of four uppercase letters produced on an electric typewriter in five different styles and sizes and were asked to list them on a prepared response form. For one style, the letters were in a sans serif typeface (Bulletin). For the other four styles, they were produced in different sizes in a serif typeface (Elite). One of these styles, Elite 6/32 in. (4.76 mm), was the same size as the Bulletin type. Adams et al. noted that letters in the Elite 6/32 in. style were significantly more likely to be reported correctly than were letters in the Bulletin style by children in four of the five grades. Even so, they added that this “may be a phenomenon of the sample and might not similarly be observed in a replication of the study” (p. 427).

Grooters (1972) carried out a similar study in which rows of ten uppercase letters were presented to 60 adult participants in four different sizes at four different distances by means of a Kodak Carousel slide projector. Instead of actual typefaces, he employed the templates that were in use at the time for lettering in technical drawing. The participants were required to read the letters aloud as if in an eye examination. He found that the sans serif style LeRoy Standard was marginally more legible than the slab serif style LeRoy Stymie Medium, but that both of these were significantly more legible than the sans serif style LeRoy Condensed Gothic, in which the letters were 60% of the width of those in LeRoy Standard. These results were found at all distances and in all sizes, except for the closest distance and the largest size where performance approached 100% for all three styles (in other words, there were ceiling effects).

Phillips (1976) similarly compared a slab serif style of lettering (Leroy Stymie) with a sans serif style (Twentieth Century). Using a Kodak Carousel projector, he presented 31 volunteers with six slides, each containing five lines of ten randomly ordered uppercase letters in decreasing size. Three slides contained letters printed in the slab serif style, and three contained letters printed in the sans serif style, in each case using a light, medium, or bold stroke width. Once again, the participants were asked to read the letters in each slide aloud, line by line. Overall, performance was better with the slab serif style than with the sans serif style (pp. 53–54). There was however a significant interaction between the effects of letter style and letter size, such that the difference between the two styles was only significant using one of the smaller letter sizes (pp. 56–57). There was also a significant three-way interaction with stroke width, such that the difference between the two styles was only significant for two of the 15 combinations of size and stroke width (pp. 59–60). Phillips concluded that performance was so poor with the smaller letter sizes that neither style would be acceptable for projected visual materials; but that conversely performance was so good using either style with the larger letter sizes that both serif and sans serif lettering would be acceptable for projected visual materials (p. 70).

Woods et al. (2005) showed pairs of lowercase letters to groups of children from kindergarten to fourth grade using a tachistoscope attached to a slide projector. The children had to say whether the letters in each pair were the same or different and to write them down. Each pair was presented in a serif typeface (Times New Roman) or in a sans serif typeface (Arial), and the two typefaces were presented either in separate blocks of slides or in the same block. The children’s scores on both discrimination and identification were higher for letters that were presented in the sans serif typeface than for those presented in the serif typeface. This was true regardless of the children’s age or the size of the typeface used. One problem with this study is that the children were tested in small groups, and some cheating had taken place.

Overhead projectors and slide projectors have in turn been superseded by computer-based projection software, most obviously by Microsoft PowerPoint. (This was developed in the 1980s by an independent software company to produce both overhead transparencies and slides. However, the company was acquired by Microsoft in 1987, and it was then developed to display presentations through digital projectors on both Windows and Macintosh systems.) PowerPoint has been in general use since the early 1990s, and in practice its applications in education and in other fields have tended to borrow techniques adopted with the older technology of slide projectors. In particular, presenters are still being advised to use sans serif typefaces rather than serif typefaces in their PowerPoint presentations (e.g., Garon, 1999). More recently, Ing et al. (2017) found that 14 out of 17 speakers at an ophthalmology conference had used sans serif typefaces in their presentations. They suggested that “serif fonts may be harder to read in digital slides” (p. 172). Phillips’ (1976) suggestion that such advice is based more upon personal preference than upon empirical research probably still applies, but one study has evaluated this directly.

Earnest (2003) compared the legibility of serif and sans serif typefaces for material presented using PowerPoint. He assigned 138 students to five different groups. Four groups viewed a recording of a speech given by their university’s president incorporating a slide presentation, whereas the fifth group only viewed the speech. Two of the first four groups viewed slides using a serif typeface (Times New Roman), and the other two groups viewed slides using a sans serif typeface (Verdana). Immediately afterwards, the students answered nine multiple-choice questions on factual points mentioned in the speech. Nine days later, they were asked to complete the test for a second time. Earnest found that the groups who had viewed slides obtained higher scores than the group who had only viewed the speech. However, there were no significant differences between the groups who had viewed the slides in a serif typeface and the groups who had viewed the slides in a sans serif typeface.

The limited number of studies using these older technologies have failed to produce unequivocal evidence favouring either serif or sans serif typefaces. Consistent with this idea, participants tend to give similar qualitative ratings of serif and sans serif typefaces presented using either slide projectors (Kastl & Child, 1968) or PowerPoint (Mackiewicz, 2007). There is certainly no support for the notion that sans serif styles should be routinely preferred for the projection of textual material.

3 Issues with Screen Technology

Early computers typically lacked the facility for generating visual displays; instead, they produced text that was generated using rudimentary printers. In the 1960s, however, it was realised that visual displays might be useful, and appropriate technology was at hand to facilitate this in the form of cathode-ray tubes (CRTs), which were widely used in scientific research (as oscilloscopes) and most obviously in television sets. In CRTs, an electron gun stimulates pixels arranged in a checkerboard pattern on a phosphorescent screen, and this process is carried out repeatedly in a systematic manner to generate a visible display. Early monitors tended to be very low resolution, meaning that the details of images were lost. Indeed, this is probably the origin of the idea that sans serif typefaces should be used, because serifs would have been among the lost details. Schriver (1997, p. 403) noted that designers for television had mainly used sans serif typefaces for many years (see also McVey, 1985). Even so, technology improved, and by the year 2000 high-resolution colour CRT monitors had been developed.

In the 1990s, however, an alternative form of technology became available through liquid crystal displays (LCDs). These use liquid crystals to modulate the light emitted from a background to generate images on a computer screen, again using a checkerboard pattern of pixels that is repeatedly scanned. LCD monitors were initially developed for use with laptop computers because of their reduced size, weight, and power consumption; however, during the 2000s they became available more generally, and their resolution typically exceeded that of CRT monitors. As a result, since the late 2000s LCD monitors have generally superseded CRT monitors in computer-based applications.

One issue with the presentation of text using both CRTs and LCDs is that of aliasing. In general terms, this is the under-sampling of the information needed to produce an accurate reproduction of a particular character. More specifically, in both CRT and LCD technology, text is displayed in the form of arrays of square pixels. If a stroke in a character is oblique rather than horizontal or vertical, its contour will receive only an approximate representation as an array of pixels. As a result, it will appear irregular and ragged rather than continuous (an effect sometimes known as “staircasing” or “the jaggies”). In theory, such “aliased” text should be less legible than text printed on paper, and the effect should be more pronounced with low-resolution monitors than with high-resolution monitors.

This issue was originally handled by means of anti-aliasing software. This smoothed the edges of characters by averaging the surrounding pixels to yield varying levels of grey scale in the contours of the displayed text. Gould et al. (1987) found that text presented on paper was read significantly faster than aliased text presented in the same typeface on a CRT; however, the difference became nonsignificant when text presented on paper was compared with anti-aliased text presented in the same typeface. An analogous process known as “scale-to-grey” was used when displaying images scanned from printed text on computer monitors. Sheedy and McCarthy (1994) found that scale-to-grey led to enhanced reading performance compared to simple black-and-white text, and participants reported fewer symptoms of eye strain as a result of reading scale-to-grey text; as expected, both differences were greater with low-resolution CRTs than with high-resolution CRTs.

The increasing use of LCDs in the 1990s enabled a different approach to be taken to the aliasing issue. In LCDs, each pixel consists of three vertical bars representing the colours red, green, and blue. In 1998, Microsoft introduced ClearType software which used sub-pixel rendering with the aim of enhancing the legibility of text presented in LCDs. However, later research using a variety of tasks failed to show an unequivocal advantage of ClearType text over aliased text (Aten et al., 2002; Dillon et al., 2004, 2006; Gugerty et al., 2004; Slattery & Rayner, 2010; Tyrrell et al., 2001). Gugerty et al. (2004) also compared ClearType text with anti-aliased text using 10-point Verdana. They found that words presented in anti-aliased text were read significantly more slowly and less accurately than words presented in ClearType or aliased text, and they argued that anti-aliasing software should not be employed with smaller type sizes.

One problem with ClearType software was that the resulting characters had coloured borders that were generally thought to be distracting for readers. Microsoft therefore offered ClearType with five levels of sub-pixel rendering that varied from greyscale with no colour filtering to a high level of colour contrast. Sheedy et al. (2008) evaluated both objective performance and subjective preference for material presented in a sans serif typeface (Verdana) with all five levels of sub-pixel rendering. They found that readers preferred a moderate level of ClearType rendering to higher levels or to greyscale, but that ClearType rendering did not improve text legibility, reading speed, or reading comfort.

Another issue with screen technology is the rate at which images are refreshed: with CRTs, too slow a refresh rate leads to flicker. Wilkins (1986) showed that flicker tended to disrupt readers’ saccadic eye movements, even when the refresh rate was sufficiently high to render the flicker imperceptible. (Wilkins showed that eye movements were also disrupted when readers read a printed page that was illuminated by conventional fluorescent lighting.) With LCDs, the screen itself does not flicker, but in many models the screen is backlit using pulse width modulation. The flicker may not be perceptible, but readers’ eye movements may be disrupted, and they may complain of discomfort (Brown et al., 2020; Wilkins, 2021). Thus, any findings regarding participants’ eye movements obtained when reading from either CRTs or LCDs need to be interpreted with caution.

4 Conclusions

This chapter has discussed the increased use of screen-based reading in education and in daily life generally. Readers usually have the option of printing off screen-based material to be read on paper, and this seems to be popular when researchers have to read more serious material. Both the use of computer technology and attitudes to such technology seem to vary with the user’s age, but there is no support for the so-called “digital natives” hypothesis. Some existing typefaces were taken over to use in computer systems, while other typefaces were developed specifically for on-screen use. The chapter discussed the legibility of serif and sans serif typefaces when projected by means of older technology such as slide projectors, overhead projectors, and PowerPoint, but this did not show any consistent difference in their legibility. Finally, the chapter described some of the technical issues concerning the way that images are displayed using CRTs and LCDs.