1 Reading Text in Serif and Sans Serif Typefaces

The work by Gould et al. (1987) was mentioned in Sect. 10.3. In another experiment, Gould et al. followed a number of previous studies comparing screen and paper presentation in a proof-reading task: 18 research workers read through articles of about 1,000 words to try to locate between 1 and 10 misspelled words by saying them aloud to the experimenter. The screen versions were presented as anti-aliased characters on a cathode-ray tube (CRT) screen using a system which had been specifically designed to simulate their appearance on paper, and high-quality printed versions were generated using the same script files. The screen and paper versions of each article were both presented using a serif typeface (Press) and two sans serif typefaces (Letter Gothic and Univers). Each participant read one article on-screen and one article on paper in each of the three typefaces. Overall, the paper versions were read significantly faster than the screen versions, although this might have been partly because the screen system required 1.5 s scrolling time to go from one page of text to the next. There was no significant difference in the participants’ accuracy between the two versions. There was no significant variation across the typefaces and no significant interaction between the effects of display mode and typeface in either speed or accuracy.

Also using a CRT screen, Tullis et al. (1995) compared the legibility of four typefaces available in the Microsoft (MS) Windows operating environment: two serif typefaces, MS Serif and Small Font, and two sans serif typefaces, Arial and MS Sans Serif. Each was used in two, three, or four different sizes, yielding 12 combinations of typeface and size. Each of these was shown in either a bold or regular style and against a white or grey background, resulting in a total of 48 conditions that were administered in a random sequence. All of the participants viewed the same paragraph of text presented in each condition and were asked to count the number of typographical errors that it contained. After they had read the paragraph, they pressed the “Enter” key and reported the number of errors using a dialogue box. Tullis et al. measured the time taken to read the paragraph and whether or not the correct number of errors had been reported. The two sans serif typefaces yielded faster reading times, higher accuracy, and higher ratings than the two serif typefaces, especially with larger type sizes.

Garcia and Caldera (1996) asked ten students to read aloud paragraphs of 30 words from a computer screen (neither the computer nor the monitor was specified). The paragraphs were presented in three different type sizes and in five different typefaces: a serif typeface (Times New Roman), three sans serif typefaces (Arial, System, and MS Sans Serif), and a cursive typeface (i.e., a typeface intended to mimic handwriting) (Lucida Casual). There was significant variation across the 15 conditions, with 10-point Arial yielding the fastest reading time. System yielded the second fastest time, but no further results were reported.

Stone et al. (1999) asked 48 female survey interviewers to read aloud 24 sets of material, each consisting of 30 random words chosen to be at the eighth-grade reading level. Half of the sets were presented in black ink on white paper, whereas the other half were presented in black type against a white background on the liquid crystal display (LCD) screen of a laptop computer to simulate the former. In both cases, each set of words was presented on a single page, and the order of the conditions was counterbalanced. The sets of words were presented in three different typefaces: a sans serif typeface (Helvetica), a serif typeface (Times Roman), and a slab serif typeface (Courier). Stone et al. found that their participants read the sets in serif typefaces faster than the sets in sans serif typeface, but there was no significant variation in the number of errors made. There was no significant difference in either reading speed or accuracy between the sets presented on paper and those presented on-screen, and neither of the interactions with the effects of typeface and mode of presentation was significant.

Josephson (2008) carried out an exploratory study of eye-movements in reading from screens. She presented six participants with four news stories of around 250 words on a high-resolution monitor, and their eye-movements were tracked using a corneal reflection system. The stories were presented in the four different typefaces shown in Fig. 10.1 in Sect. 10.1. Afterwards, the participants were asked to rate each typeface on several 10-point scales and then to say which was easiest to read and which they liked the most. The story presented in Verdana was read the fastest, followed by that presented in Times New Roman. The story presented in Times New Roman yielded the fewest fixations, whereas that presented in Verdana yielded the most. Conversely, the story presented in Verdana yielded the fewest backward movements to reread previously presented words, whereas that presented in Times New Roman yielded the most. Verdana was rated highest whereas Times New Roman was rated lowest of the four typefaces. When asked which typeface was the easiest to read, three of the participants chose Verdana, whereas none chose Times New Roman. Apart from the small number of participants, one limitation of this study was that one story was assigned to each typeface, and so differences among the typefaces were directly confounded with differences among the news stories themselves.

Banerjee and Bhattacharyya (2011) presented 40 young postgraduate researchers with 18 passages of roughly the same length on an LCD monitor. The passages were presented in one of three serif typefaces (Times New Roman, Georgia, and the slab serif Courier New) or one of three sans serif typefaces (Arial, Tahoma, and Verdana) in one of three sizes (10, 12, or 14 points). The passages were assigned at random to the 18 conditions for each participant and presented in a different random order to each participant. The time taken by the participants to read each passage was recorded, they then rated the overall ease or difficulty of reading each passage, and finally they completed a short questionnaire on the mental workload involved in reading each passage. Their eye movements were also tracked using a binocular eye movement recorder.

There was a significant interaction between the effects of typeface and type size on the participants’ reading time. The effect of type size was significant for Courier New and Arial, but not for the other four typefaces. The mean reading time was significantly less for the serif typefaces than the sans serif typefaces, but only for 10-point and 14-point sizes. Courier New and Arial also yielded the fastest average reading time. Verdana in 14-point was rated as easier to read than any other combinations of typeface and size, and Arial in 14-point was rated as the second most preferred. Verdana in 14-point was also rated as having the least mental workload, followed by Arial in 14-point and Tahoma in 14-point. Courier New yielded the lowest fixation duration, followed by Verdana and Arial; Courier New also yielded the lowest total gaze duration, followed by Verdana. In contrast, Times New Roman yielded the longest fixation duration and the longest total gaze duration. However, there were no overall differences between serif and sans serif typefaces in eye movement parameters.

Perea (2013) presented 24 Spanish undergraduate students with individual sentences on a computer screen and measured their eye movements using a video-based eye-tracking device. Each sentence appeared when the participant looked at a square on the left-hand side of the screen, and the participant pressed a key when they had finished reading the sentence to themselves. To check on their comprehension, a yes/no question was presented after 20% of the sentences. (Overall, 96% of these questions were answered correctly.) The sentences were presented in four blocks of 30. Half were presented in the serif typeface Lucida, whereas half were presented in the sans serif typeface Lucida Sans. There was no significant difference in the participants’ reading times, in the total number of fixations, or in the mean duration of their fixations between the serif and sans serif sentences. Perea concluded that the presence or absence of serifs did not materially affect the process of normal reading, although different results might have been obtained with longer extracts, as in Josephson’s study.

2 Comprehending Text in Serif and Sans Serif Typefaces

As with research on reading from paper (see Chap. 5), asking the participants to read continuous text provides less opportunity for researchers to impose experimental control over their reading behaviour. Some researchers have therefore focused on their participants’ comprehension of meaningful material read from screens rather than upon its legibility per se.

Williams (1990) carried out an experiment in which 56 students read two passages on a CRT monitor and answered four comprehension questions after each passage. Each passage contained 650–700 words and filled seven screens, so that the students had to scroll down to read the complete passages. The material was presented in either a serif typeface (10-point Times Roman) or a sans serif typeface (9-point Helvetica); the use of different type sizes ensured that the typefaces were of similar x-height. The participants were randomly assigned to five different groups. Four groups saw the two passages in different typefaces, with the order of the two passages and the order of the two typefaces counterbalanced across the four groups. The fifth group saw both passages in the sans serif typeface to check for any difference in difficulty between the two passages. The monitor screen also contained a clock, and the students were asked to record the times when they started and finished reading each passage. The comprehension questions were multiple-choice with five alternatives. The fifth (control) group showed no difference in mean reading rate between the two passages. Across the other four groups, the mean reading rate was 14.75 words/min faster for the sans serif typeface than for the serif typeface, but the difference was not statistically significant.

Lenze (1991) used a personal computer to present 84 students with a brief paragraph and then asked them a question based on its content to which they would have to type a one-word answer. The paragraph was initially presented for 1 s; if the participant answered incorrectly, the paragraph was presented for 1 s longer, and this process was continued until they answered the question correctly. They were then presented with three further paragraphs in the same way. Half the participants chosen at random were shown the paragraph in a serif typeface, and the other half were shown the paragraph in a sans serif typeface. (Neither of the typefaces was specified.) Finally, all the participants were shown examples of text in both serif and sans serif typefaces and were asked which they preferred. Lenze found that there was no sign of any significant difference between the two groups of participants in the time needed to achieve comprehension of the texts, which suggested that serif typefaces and sans serif typefaces were equally effective in supporting reading comprehension. Even so, 77% of the students preferred the sans serif typeface to the serif typeface.

Boyarski et al. (1998) compared the serif typeface Georgia and the sans serif typeface Verdana, both of which had been designed for on-screen display. Sixteen university staff and graduate students read two passages from a standard reading test presented in the two typefaces in the same body size in Microsoft Word without anti-aliasing. The orders of the typefaces and the passages were both counterbalanced. They were allowed 1 min to read each passage and were then asked four questions to test their comprehension of the passage (each answer being scored between 0 and 3). Finally, the participants were asked to compare the two typefaces. A measure of “effective reading speed” was obtained by dividing their comprehension score by their actual reading time on each passage in order to allow for the possibility of a trade-off between speed and accuracy. Their comprehension scores were significantly higher for passages in Georgia than for passages in Verdana. Nevertheless, there was no significant difference between the two conditions either in their actual reading time or in their effective reading speed. The participants judged the passage presented in Verdana to be easier to read than the passage presented in Georgia, but there was no significant difference in their ratings of the passages’ sharpness or legibility.

Hojjati and Muniandy (2014) presented 30 international students at a Malaysian university with four 200-word English texts in different typefaces. Two were presented in the serif typeface Times New Roman, and two were presented in the sans serif typeface Verdana; in each case, one text was presented single-spaced, and the other was presented double-spaced. After each text, the students were presented with questions designed to test their retention of its content. Finally, they were asked to rate the ease with which they had been able to read each text on a 6-point scale. Regardless of spacing, they found text displayed in Verdana easier to read than text displayed in Times New Roman, they read text displayed in Verdana more quickly, and they recalled more of the content of text displayed in Verdana. However, the researchers had used Amazon Kindles (which use microcapsules containing electronic “ink”) rather than LCDs, and at least some of the texts did not fit completely onto the visible area of the screens (p. 167). The material had been taken from Wikipedia and other sources and had been checked by subject experts, but the texts themselves suffered from grammatical and stylistic problems. It is also not clear whether the typefaces used for different texts were counterbalanced across participants or whether each text was always presented in the same typeface.

Csilla et al. (2016) asked 74 Hungarian students to read to themselves four self-contained excerpts from a Hungarian translation of The Nature of Space and Time by Hawking and Penrose (1999). The excerpts ran for four pages in the original book but had been reformatted to fit onto two pages of European A4-sized paper. Each excerpt was prepared in six different typefaces and then saved in Portable Document Format. Two excerpts were presented on paper, and two were presented on an LCD computer monitor. In each case, one excerpt was presented in a serif typeface randomly chosen from Book Antiqua, Garamond, and Times New Roman, and the other was presented in a sans serif typeface randomly chosen from Arial, Calibri, and Verdana, so that each student saw four different typefaces. The order of the four texts was randomised for each student. The students timed themselves reading each excerpt and then answered a number of questions about its content.

Csilla et al. analysed the data for each of the four excerpts separately and carried out independent-groups tests comparing (a) the participants who had read the excerpt on paper with those who had read the excerpt on-screen and (b) the participants who had read the excerpt in a serif typeface with those who had read the excerpt in a sans serif typeface. There were no significant differences in either reading time or comprehension for any of the four excerpts. It is unfortunate that Csilla et al. did not increase the statistical power of their analysis by using the data from all four excerpts and carrying out repeated-measures tests on the variables of presentation medium and typeface within the 74 participants. They also did not examine whether there was any interaction between the effects of these two variables on either reading time or comprehension.

3 Rapid Serial Visual Presentation

The introduction of new technologies that enable images to be presented in a variety of visual forms prompted new research methods to be used for investigating reading. One such method is that of rapid serial visual presentation (RSVP), in which letters, words, or groups of words are presented one at a time at the reader’s point of fixation. This method was first developed by Gilbert (1959a, b), who presented sentences containing eight words using a movie projector to each of 76 participants and asked them to write down immediately afterwards what they had seen. He found that they were much more accurate when successive pairs of words were shown at the point of fixation than when successive pairs of words were presented across the screen as if in a line of text. Gilbert concluded that the former procedure enhanced the identification of stimuli by eliminating the need for eye movements.

Rubin and Turano (1992) digitised characters from the output of a laser printer using a Times Roman typeface and then used them in text displayed one word at a time on a high-resolution screen controlled by an IBM computer. They found that readers who were previously unfamiliar with this procedure could achieve reading speeds that were several times faster than when reading the same material presented on-screen as a single paragraph, typically in excess of 1,000 words/min for reading aloud or as fast as 1,800 words/min for silent reading.

As was mentioned in Sect. 11.1, Suen and Komoda (1986) had digitised printed characters in a slab serif typeface (Courier), a sans serif typeface (Letter Gothic), and the output of a dot-matrix printer. They used these characters to present paragraphs of roughly 160 words drawn from magazine articles to 36 participants. Each paragraph was shown on a high-resolution CRT screen, one word at a time for just 100 ms. Multiple-choice questions (how many was not specified) were then employed to assess the participants’ comprehension of the paragraph. The dot-matrix style yielded the worst performance, but there was no difference between the comprehension of material presented in the slab serif typeface and that presented in the sans serif typeface. Suen and Komoda argued that the higher-order skills and strategies involved in reading connected text had overridden any differences in legibility between the serif and sans serif typefaces.

Yager et al. (1998) used the RSVP method to compare the legibility of two typefaces matched for x-height: Dutch, a serif typeface similar to Times; and Swiss, a sans serif typeface similar to Helvetica. Sentences were presented on a CRT monitor in white letters on a black background under conditions of either high luminance or low luminance. The latter was intended to stress the visual system and hence to simulate the situation of readers with visual impairment. The participants were 46 normally sighted college and high-school students who were asked to read each sentence aloud immediately after it had been presented. The presentation rate was calibrated to identify the threshold for correct responding for each participant. Under high luminance, there was no significant difference between the number of sentences read correctly in the two typefaces. Under low luminance, performance was significantly better with the Swiss typeface. Yager et al. noted that in this situation the Dutch typeface was close to the threshold of visual acuity. They speculated that under conditions of low luminance either the serifs or the thinner strokes of the letters of the Dutch typeface tended to be invisible, and that this was responsible for the significantly poorer reading performance.

Also using RSVP, Morris et al. (2002) presented 27 participants with words that made up meaningful but unconnected sentences. The words were presented on a CRT monitor in slab serif and sans serif typefaces taken from Bigelow and Holmes’ (1986) Lucida styles to ensure that the typefaces were matched in all respects except for the presence or absence of serifs. When the words were presented at the equivalent of 14-point type (a normal reading size), there was no difference in the number that were read correctly in the slab serif and sans serif typefaces. However, when they were presented at the equivalent of 4-point type (a small but still tolerable size), performance was better with the sans serif typeface than with the slab serif typeface. Morris et al. suggested that the letters appeared relatively crowded in the latter situation, and that rendering serifs in small sizes might be counterproductive.

As mentioned in Sect. 5.1, Arditi and Cho (2005) devised lowercase typefaces of uniform thickness with slab serifs that extended for 0% (sans serif), 5%, or 10% of their cap height. They presented sentences in these typefaces on a CRT display screen using RSVP and adjusted the presentation speed for each participant to ensure a 50% correct reading rate. Data were obtained from two participants with normal vision and two with impaired vision. The former participants achieved higher reading speeds than the latter participants, but there was no effect of serif size and hence no effect of the presence versus the absence of serifs. Arditi and Cho acknowledged that the small number of participants was a limitation of their study.

One motivation for investigating the RSVP procedure was that it was suspected it might help to compensate for the limitations of handheld devices with small screens, such as cellular phones, palmtop computers, and personal digital assistants (Bernard, Chaparro, & Russell, 2000; De Bruijn et al., 2002). Palmtop computers and personal digital assistants were fairly popular towards the end of the twentieth century, but during the 2000s their functionality was superseded by that of smartphones. It is thus not surprising that the RSVP procedure also became less popular as a research tool. In any case, a major criticism of the procedure is that it lacks ecological validity, insofar as it does not represent a situation that is characteristic of normal reading in real-life settings (Perea, 2013).

4 Reading Material on Handheld Devices and Smartphones

It has tended to be assumed that sans serif typefaces are more legible than serif typefaces when used on handheld devices or smartphones. The sans serif typefaces Droid and Roboto were developed for Android mobile phones, while the sans serif typeface San Francisco was devised for Apple products, although all three typefaces are available in serif styles. The limited amount of research comparing the legibility of serif and sans serif typefaces when employed on handheld devices and smartphones has been carried out by Korean researchers.

Park et al. (2008) presented texts to four Korean students by means of a personal digital assistant (a Pocket PC 2002). They were asked to read the texts silently, and their eye movements were monitored using an eyeball-tracking camera which reflected infrared rays on their corneas. The texts were shown in three type sizes and in three typefaces: two serif typefaces (Batang and Gungseo) and one sans serif typeface (Gulim). The participants’ fatigue was measured both by monitoring their blink rate and by asking them to rate how easy or difficult it had been to read each text on a 7-point scale. There was no significant variation across the three typefaces in their reading speed, their error rate, their eye-blinking, or their subjective ratings.

Kim et al. (2015) presented 14 Korean students with pairs of two-syllable words using an Apple iPhone with a display of 90 mm by 50 mm using one of two serif typefaces (Batang or Gungseo) or one of two sans serif typefaces (Dodum or Gulim). There were 10 trials for each typeface, and the order of the typefaces was counterbalanced across the participants. On each trial, one member of the word pair had been designated as the target, whereas the other was a distractor, and the participants’ task was to read aloud the target in each pair. Their mean reading time was marginally faster for the words shown in serif typefaces than for words shown in sans serif typefaces, but the difference between the two means was not at all statistically significant.

5 Connotations of Serif and Sans Serif Typefaces

Several of the studies mentioned asked participants to express a preference between serif and sans serif typefaces. Misanchuk (1989) argued that, in the absence of evidence for objective differences in legibility between typefaces, designers might be guided by readers’ preferences or satisfaction ratings. In fact, several studies found a significant preference for sans serif typefaces over serif typefaces (Boyarski et al., 1998; Hojjati & Muniandy, 2014; Lenze, 1991; Tullis et al., 1995), some found no significant difference (Garcia & Caldera, 1996; Holleran, 1992; Muter & Mauretto, 1991), but none found a significant preference for serif typefaces over sans serif typefaces.

Savory et al. (2012) found a clear preference for sans serif typefaces among military experts in the highly specialised area of the design of radiation detector screens. They used a focus group to discuss other aspects of these devices, a methodology where responses can be vulnerable to peer pressure from the other participants. However, the participants expressed their preferences among the different typefaces in an individual written survey, which should have avoided this problem.

Section 5.4 noted research that had focused on the connotations of different typefaces when used for print-based text, and analogous research has been carried out on the connotations of different typefaces when presented on computer monitors. Shaikh et al. (2006) examined the connotations of 20 different typefaces; they included four serif and six sans serif typefaces, among which were the six typefaces introduced by Microsoft to make use of ClearType software (see Sect. 10.3). Using an online survey, more than 500 participants rated the characteristics of each typeface using 15 bipolar scales and then indicated whether they would use the typeface for each of 25 online purposes.

The serif typefaces were regarded as being stable, practical, mature, and formal, but the sans serif typefaces were not regarded as being especially high or low on any of the 15 traits. The serif typefaces were regarded as being appropriate for business documents, website text, or online magazines but not for digital scrapbooking, children’s documents, or e-greetings. The sans serif typefaces were regarded as being appropriate for website text, e-mail, or online magazines but not for scrapbooking, computer programming, or mathematical documents. Shaikh et al. concluded that computer users consistently attributed personalities to typefaces displayed on-screen and that both serif typefaces and sans serif typefaces were seen as being appropriate for the kinds of material that were typically read on-screen.

Shaikh (2007, pp. 44–100) carried out another online survey in which 379 participants rated text samples of 40 typefaces on 15 bipolar constructs. The typefaces included ten examples of each of four categories: serif, sans serif, display (used in advertisements or logos), and cursive (akin to handwriting). Each participant was shown 20 typefaces including five serif typefaces, five sans serif typefaces, five display typefaces, and five cursive typefaces; these were randomly selected and presented in a random order. The participants also rated the legibility of each typeface using a 7-point scale. There were significant differences between the serif and the sans serif typefaces on five of the scales: the serif typefaces were regarded as more delicate, beautiful, expensive, warm, and old, whereas the sans serif typefaces were regarded as more rugged, ugly, cheap, cool, and young. However, these differences were small in magnitude, and in general the serif and sans serif typefaces tended to be regarded as similar. The participants’ ratings of legibility showed no significant difference between the serif and the sans serif typefaces, but both were rated as being significantly more legible than the display typefaces or the cursive typefaces.

Koch (2012) asked 100 volunteers to evaluate six typefaces in an online survey: five were variants of a sans serif typeface, Helvetica, and one was a slab serif typeface, Glypha Medium. For each typeface, participants were presented with the uppercase and lowercase alphabets, together with a grid showing 12 cartoon characters, each representing a different emotion (although the emotions themselves were not named). When they clicked on each character, they were shown a short animation in which the character enacted the relevant emotion; they were also shown a 5-point scale and indicated how much the relevant typeface had aroused the emotion in them (from “I do not feel this” to “I feel this strongly”). Out of the 100 volunteers, 42 completed the survey (yielding 12 ratings for each of six typefaces), of whom 32 had had some prior training in design. Comparisons between the overall ratings given to the slab serif typeface and the sans serif typefaces yielded just one significant comparison, in that Glypha Medium yielded higher ratings of satisfaction than the average across the different variants of Helvetica. However, Koch had carried out 48 different comparisons in total, suggesting that even this effect might well have been a spurious result due to chance variation (i.e., a Type I error).

Kaspar et al. (2015) carried out an online survey in which 188 students evaluated three scientific abstracts on six dimensions. The students were randomly assigned to receive the abstracts and the rating scales either in the serif typeface Lucida Bright or in the sans serif typeface Lucida Sans. (As mentioned in Sect. 11.2, these typefaces are identical except for the presence or absence of serifs.) There was no significant difference in the time taken to read the abstracts or in the time taken to complete the rating scales between the groups who received material in the two typefaces. The students who read the abstracts in the serif typeface rated them significantly more positively than the students who read them in the sans serif typeface: they rated them as significantly more comprehensible and more appealing overall, and they were significantly more interested in reading the full paper from which the abstract had been taken; they also rated the topicality, the quality, and the importance of the research question significantly more highly. Kaspar et al. noted that reports of scientific work were more often presented in a serif typeface, and they concluded that the use of such a typeface increased the impression of a work’s scientific character.

Kaspar et al. (2015) carried out a similar study in which 187 students were randomly assigned to receive abstracts and rating scales either in the serif typeface Times New Roman or in the sans serif typeface Arial. (These typefaces differ on several other characteristics as well as in the present or absence of serifs.) Once again, there was no significant difference in reading speed between the groups who received material in the two typefaces. However, in this case, the participants who read the abstracts in the sans serif typeface rated them significantly more positively than the students who read them in the serif typeface: they rated them as significantly more comprehensible and more appealing overall, and they were significantly more interested in reading the full paper from which the abstract had been taken; they also rated the quality of the research question significantly more highly, although there was no significant difference between the students who read the abstracts in different typefaces in their ratings of either the topicality or the importance of the research question. Kaspar et al. concluded that there was no simple rule of thumb that favoured the presence or absence of serifs under all circumstances and that sans serif typefaces could lead to more desirable text evaluations when other features of the text compensated for the missing serifs.

6 Conclusions

Studies of the legibility of connected sentences that have measured readers’ speed and accuracy have proved inconclusive. Research on reading text presented on computer screens has enabled investigators to use other forms of technology such as eye-tracking equipment. However, research into readers’ eye movements has not yielded conclusive findings with regard to the presence or absence of serifs. As with research on reading from paper, asking participants to read continuous text provides less opportunity for researchers to impose experimental control over their reading behaviour. Some researchers have instead focused on their participants’ comprehension of material. Such studies have not yielded consistent findings with regard to the presence or absence of serifs on readers’ speed or accuracy, and at least one study suffered from serious methodological problems.

A particular device that has been investigated is the presentation of letters, words, or groups of words one at a time at the reader’s point of fixation: RSVP. This was originally thought to compensate for the limitations of handheld devices. Five studies have compared readers’ comprehension, speed, or accuracy and found no significant differences except that serif styles were less legible with very small type or under conditions of low luminance when, of course, serifs are likely to have been faint or even completely invisible. It has tended to be assumed that sans serif typefaces are more legible than serif typefaces when used on handheld devices or smartphones. However, two studies failed to find any difference between serif and sans serif typefaces in terms of the participants’ reading performance when using such devices.

Finally, this chapter described research on the connotations of different typefaces when presented on computer screens. Some studies, but not all, have found that readers express a preference for sans serif typefaces when reading on computer screens, but in general both serif and sans serif typefaces are regarded by users as appropriate for online purposes. There is some evidence that serif and sans serif typefaces differ in their connotations or “personality”. These differences seem to reflect variations in readers’ expectations, which in turn depend on their prior experience and familiarity with different typefaces.