Está en la página 1de 14

This article was downloaded by: [University of Saskatchewan Library]

On: 08 January 2015, At: 02:08


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Visual Communication Quarterly


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/hvcq20

Keeping Your Readers' Eyes on the Screen: An


Eye-Tracking Study Comparing Sans Serif and Serif
Typefaces
a
Sheree Josephson
a
Weber State University , Ogden, Utah
Published online: 20 Jan 2011.

To cite this article: Sheree Josephson (2008) Keeping Your Readers' Eyes on the Screen: An Eye-Tracking Study Comparing Sans
Serif and Serif Typefaces, Visual Communication Quarterly, 15:1-2, 67-79, DOI: 10.1080/15551390801914595

To link to this article: http://dx.doi.org/10.1080/15551390801914595

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 67

Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

Keeping your readers’ eyes on the screen:


An eye-tracking study
comparing sans serif
and serif typefaces

An exploratory eye-tracking study was designed to compare the onscreen


legibility of sans serif and serif typefaces. The typefaces selected for this
study included the serif font Times New Roman and the sans serif font
Arial—both originally designed for printing on paper—and the serif font
Georgia and the sans serif font Verdana—both designed in recent years
especially for reading onscreen. Six participants read four short news stories,
each displayed in a different typeface, while their eye-movement behavior
was recorded. Overall, Verdana performed best. Participants were able
to read more quickly and experienced fewer regressions (backward
movements) when the type was set in Verdana. They also expressed
a strong preference for this font on the computer screen.

By Sheree Josephson

Pages 67–79 67 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 68

T
o do the research for this article, I bitmap—essentially a bunch of rows and columns
printed out numerous articles about of pixels that are either on or off. A digital font
onscreen typography so that I could read consists of thousands of these bitmaps—for every
them off screen. Since I located these letter, number, and symbol and for every point
articles in electronic databases, I could have easily size, style, and resolution (Larson, 2007). For a
read them on my computer screen, but I definitely bitmapped letter, the minimum stroke width is one
prefer to read long, scholarly, and trade journal arti- pixel, while the next possible stroke is double that
cles on paper. Yes, this practice wastes ink and or two pixels wide. Especially in smaller size fonts
paper (even though I print on the back of previously (such as those used for body copy), there is little
used paper), but it saves me from considerable room for nuance. The open space in some letters
eyestrain and nasty headaches. I print and will can fill in and important pixels can be deleted alto-
likely continue to print because the quality of com- gether. In larger sized fonts, jagged edges are
puter text is often terrible. apparent on the enlarged type. Bold-facing or itali-
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

cized type compounds the problems—especially


The biggest problem is the low resolution of com- when the type contains serifs. Bold-faced serifs
puter screens. Color LCD screens on most desktops often run together, while the slanted letters in itali-
and laptops today have a resolution of only about cized type for both serif and sans serif type often
100 pixels per inch. Kevin Larson (2007, p. 27), become especially jagged.
who is a member of Microsoft’s Advanced Reading
Technologies group, points out, “You need at least To deal with the jagged edges, font developers
two or three times that many pixels to begin to learned how to add slightly lighter squares at the
approach the quality of the printed page.” Screen edges of curves and diagonals, a technique known
resolution has not improved in the last few years, as antialiasing. Viewed close up, the lines actually
Larson says, because you’d need a lot more com- appear a little blurry, but at a normal reading dis-
putational power to make a difference you could tance, the shaded pixels trick the eye into seeing
see on the screen. To move from 100 pixels per what it thinks should be there: smooth continuous
inch to 200 ppi means your computer would have lines (Larson, 2007, p. 30).
four times as many pixels to fill, and that would
probably bog down your graphics processor or Forming computer fonts today has become a little
quickly drain the battery of your laptop. Plus, he more streamlined. Rather than forming individual
says, there are costs. Denser screens would be bitmaps for every letter, number, and symbol and
more expensive, and the higher pixel densities for every point size, style, and resolution, the font
would increase the chances of dead pixels from software running on your microprocessor relies on
manufacturing. scalable outlines of each letter. To display a letter
onscreen, the font software forms the letter outline,
The low resolution of computer screens is espe- automatically scales it up or down to the desired
cially problematic for fonts with serifs, those fine size, and then creates a bitmap on the fly.
cross-lines at the extremities of letters. Serif fonts
render poorly onscreen at small sizes because low- Apple and Adobe are focusing their efforts on mak-
resolution screens cannot handle the subtle details ing the onscreen text as faithful to the printed output
of the type. Serif fonts are also likely to have differ- as possible by performing antialiasing on the letter
ences in stroke widths—another challenge for outline, with slight stem-weight adjustments
computer screens with low resolution. Sans serif (Larson, 2007). Larson says, “The result is that
fonts render more cleanly onscreen since they have when you look at a page of text onscreen, the text
less fine detail. They are also likely to have more will have a very smooth, even appearance, much
uniform stroke widths. like the printed page.” But he says, “The tradeoff
is that the individual letters are less crisp and there-
Displaying fonts on a computer screen—whether fore more difficult to read onscreen” (p. 30).
serif or sans serif—is not easy or straightforward. Microsoft, on the other hand, is hinting letters,
It requires that each letter be created with a unique which slightly adjusts the letter’s shape. The result-

Visual Communication Quarterly 68 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 69

ing bitmap is even and crisp, although it is slightly These six fonts began shipping with the new oper-
distorted, which can impact onscreen legibility. ating system in 2007.

Despite the significant advances that have been Despite the fact that both serif and sans serif type-
made in making type more readable onscreen, it faces have been created especially for reading on a
is still not possible to make type as easy to read computer screen, the common wisdom on the Web
on a computer screen as it is on the printed page is that the sans serif type is still more legible than
because of the issues created by low resolution. serif type. An article featured on Webcast1.com
Some typography experts have decided it is (2007) rates the “readability on screen” of the sans
better to start from scratch and design type fonts serif font Verdana as “exceptional,” saying “its
especially for reading onscreen. In 1985, Charles wide body makes it the clearest font for onscreen
Bigelow and Kris Holmes designed the Lucida fam- reading even at small sizes.” Trebuchet is said to
ily of type—a sans serif font—paying particular be “difficult to read in small sizes.” In comparison,
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

attention to its legibility, especially at low resolu- the readability onscreen of the serif font Georgia is
tion. Verdana, released in 1996 and designed listed as “very good.” Webcast1.com also tested
by Matthew Carter, was the first typeface that two popular long-time fonts—the serif type Times
Microsoft created especially for use on computers New Roman and the sans serif type Arial. Times
(Larson, 2007). This sans serif typeface “designed New Roman was said to be “acceptable” onscreen
to be readable at small sizes ...has many features for type 12 point and above, but “terrible for
to enhance legibility on screen: lowercase letters smaller sizes.” Arial’s readability onscreen was
that are proportionately tall compared to uppercase “not the worst or the best, especially at small sizes
letters, stroke widths that are not too thin, and when it becomes too narrow and the spacing
generous spacing both inside the letter and between characters becomes or is too small.” For
between letters. Well over 90 percent of Windows reading onscreen, Webcast1 says Verdana is the
and Macintosh computers now have Verdana “clear winner.”
installed on them, making it one of the most widely
available typefaces in the computer world” (p. 28). Sara Quinn (2005) of the Poynter Institute explains
Trebuchet is another sans serif typeface designed why Verdana, Georgia, and Trebuchet work well
at Microsoft. It is meant to mimic handwriting and onscreen. She says they work because “(1) their
is informal and friendly. Also released in 1996, it lowercase characters are slightly larger than the
was designed by Vincent Connare. average typeface. This larger ‘x-height’ makes the
character look bigger overall. The open spaces are
Serif fonts have also been created specifically slightly larger than average, so they don’t seem
for reading onscreen. Carter designed the Georgia to ‘fill in’; (2) to limit jagged edges the curves are
font for Microsoft in 1993, 3 years before he reduced to a minimum in the open spaces of the
worked on Verdana. Georgia was designed as a letters; (3) the letters are spaced further apart, in
more readable alternative to Times Roman and a more regular way, so they don’t seem to touch;
featured a relatively large x-height. Georgia’s serifs (4) some combinations of letters that might bump
are slightly wider with blunter, flatter ends but is or overlap, like ‘ft,’ and ‘fl,’ are specifically drawn
otherwise indistinguishable from Times Roman for so that they have extra space between them;
many typography novices. For the 2007 launch of (5) Verdana, Georgia, and Trebuchet are installed
Microsoft’s Windows Vista, its typography group both on Windows and Apple operating systems,
wanted to include several new screen-friendly making them universally available for use on any
typefaces, so it staged a competition. Of the 26 web page. Some people call these ‘web-safe’
submissions from the world’s top type designers fonts, because most users have them.”
(Larson, 2007), six Western fonts were selected.
The results were two serif faces called Cambria Numerous scholars and typographers have studied
and Constantia; two sans serif faces called Calibri the legibility and/or readability of type onscreen.
and Corbel; a flared-serif face, Candara; and a Legibility is generally defined as the ability to discern
monospaced face for programmers, Consolas. characters and words, while readability is the

Pages 67–79 69 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 70

ease of reading and understanding the material Riley, Hackler, and Janzen (2002) compared four
(Tinker, 1963). sans serif fonts—Arial, Comic, Tahoma, and
Verdana—and four serif fonts—Courier New,
In a study of the “readability” of fonts and sizes Georgia, Century School Book, and Times New
in the Microsoft Windows environment, Tullis, Roman—at 10-, 12-, and 14-point sizes. No signifi-
Boynton, and Hersh (1995) examined the reading cant differences in reading efficiency (reading
rates for Arial, MS Sans Serif, MS Serif, and Small time/accuracy) were detected between the font
Fonts in sizes ranging from 6.0 to 9.75 points. They types at any size. There were, however, significant
found no difference in reading speed between the differences in reading time irrespective of accu-
serif and sans serif fonts; however, they found a racy. The two fonts used widely in print—Times
greater preference for the sans serif font compared New Roman and Arial—were the fonts read most
to the serif font. They also found that the larger quickly onscreen possibly due to the participants’
9- and 10-point fonts elicited faster reading times. familiarity with these typefaces. In terms of size,
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

12-point type was read faster than 10-point type.


Geske (1996) tested the legibility of Helvetica type The most legible typefaces included Arial, Courier,
in three different sizes and in three different faces. and Georgia, indicating that participants did not
He found no significant differences for legibility of distinguish between serif and sans serif in terms of
9-, 10-, and 12-point type in the normal face. He legibility. However, the most attractive fonts were
found that Helvetica Bold significantly increases Georgia and Times, both serif type faces. Verdana,
the legibility in most cases and does not decrease a serif font designed specifically for onscreen
the legibility. Geske also found that Helvetica italics reading, was the most preferred font choice. The
were illegible and recommended avoiding them or researchers conclude that Verdana is the best
at least using “extreme” caution. choice for onscreen type because it was the
most preferred. It was read fairly quickly and it
A study by Boyarski, Neuwirth, Forlizzi, and Regli was perceived as being legible.
(1998) evaluated the reading speed of participants
using the serif Georgia, the serif Times Roman, Wilson (2001) conducted several surveys to ask
and the sans serif Verdana fonts. Georgia per- 1500 respondents their preference of serif and sans
formed well compared to Times Roman, indicating serif typefaces. First, he compared 12-point Times
that this font, designed for computer screen New Roman with 12-point Arial, the default for
design, is perceived by readers to facilitate the many web browsers at the time. By a 2-to-1 mar-
ease of reading online. A comparison of Verdana, gin, respondents rated the sans serif font more
a sans serif font designed for screen display with readable than the serif font. In the second survey,
Georgia, produced mixed results. In tests of aliased 12-point Times New Roman was compared with
v. anti-aliased fonts, they found no performance dif- 12-point Georgia, the serif typeface developed by
ferences in reading speed. In these tests, all text Microsoft specifically for screen readability.
was displayed at 16 points since a character must Wilson’s results found that Georgia was substan-
be at least 14 points for the anti-aliasing to show. tially more readable than Times New Roman. Next
he compared Verdana, the sans serif font developed
Bernard, Mills, Peterson, and Storrer (2001b) by Microsoft, and Arial. Wilson’s results showed
tested a range of 12 fonts—sans serif, serif, and more respondents preferred Arial at 12 point and
ornate styles—for effective reading speed in con- Verdana at 10 point because 12-point Verdana is
junction with accuracy. They also asked partici- perceived as too large for body text. Overall, his
pants about their perception of font legibility. Differ- results show readers prefer sans serif fonts to
ences were found for reading time with Tahoma, a serif fonts for body text on screen.
sans serif font, being read significantly faster than
Corsiva, an ornate font. Perceived font legibility Morrison and Noyes (2003) found that the read-
also showed significant differences across the 12 ing time for serif font Times New Roman was
fonts with sans serif fonts being more legible than faster than for the ornate sans serif font Gigi.
the ornate fonts. A later study by Bernard, Lida, Comprehension was also better for Times New

Visual Communication Quarterly 70 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 71

Roman as was preference for web use. However, of the middle of a word. Not all words are fixated.
Gigi was rated as being more attractive by Short words and particularly function words such
80 percent of the participants. as “the” and “a” are frequently skipped.

Bernard, Liao, and Mills (2001a) examined passages During a single fixation, there is a limit to the
containing two serif and sans serif fonts at 12- and amount of information that can be recognized.
14-point sizes for differences in legibility, reading The fovea, which is the clear center point of vision,
time, and general preference when read by an older extends 1 degree of visual angle to the left and
population. They found that 14-point fonts were right of fixation (Rayner, 1993). In other words, the
more legible, promote faster reading, and were pre- eyes can only see three to four letters to the left
ferred to the 12-point fonts. At 14 point, serif fonts and right of fixation at normal reading distances.
tended to support faster reading, but were generally Visual acuity decreases quickly in the parafovea,
less preferred than sans serif fonts. Bernard et al. which extends out to 5 degrees of visual angle on
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

(2001a) found no difference between the fonts either side of fixation. This is an area of about 15
designed for the computer and the print fonts. to 20 letters to the left and right of the fixation
point. The purpose of eye movements in reading,
Horton (1990) points out that the proper size for therefore, is to place the foveal region on that part
text depends on the following factors: reading dis- of the text readers are trying to process (Rayner,
tance, screen resolution, contrast between text and 1993). The area beyond the parafoveal region is
background, visual acuity of the user, and how the known as the peripheral region.
text is read (skimmed lightly, read word for word,
or read character by characters). Other factors that Readers take in information from all three zones—
might impact the proper size for text include the foveal, parafoveal, and peripheral—during the
contrast and color of the characters and the angle length of a fixation. Word recognition takes place
of regard from which it is viewed. in the area closest to the fixation point. In a review
article about word recognition, Larson (2004)
explains, “This zone is usually large enough to cap-
Eye Movements and Reading ture the word being fixated, and often includes
smaller function words directly to the right of the
Measuring eye movements to study reading fixated word. The next zone extends a few letters
onscreen has been done since the invention of the past the word recognition zone, and readers gather
desktop computer. While reading, the eyes do not preliminary information about the next letters in
continuously scan a line. Rather, they fixate for this zone. The final zone extends out to 15 letters
about 200 to 250 milliseconds (Rayner, 1993) and past the fixation point. Information gathered out
then rapidly make a saccade to another place on this far is used to identify the length of upcoming
the line. These ballistic movements usually take words and to identify the best location for the next
only 20 to 35 milliseconds. Most saccades are for- fixation point.” It should also be noted that readers
ward movements from seven to nine letters. About only use information to the right of the fixation
10 to 15 percent of saccades are regressive or point and that they don’t use any letters to the
backward movements (Rayner, 1993). Rayner left of the word that is currently being fixated.
points out that most readers are not aware of the
frequency of regressive saccades while reading.
When the end of the line of type is reached, read- Eye Movements and Reading Onscreen
ers traverse back to read the beginning of the new
line. These return movements are referred to as Various studies have measured the effects of eye
return sweeps. Information is perceived during the movements on the readability and legibility of
fixations, but no useful information is perceived onscreen text in terms of fixation frequency, fixation
while the eyes are moving. The location of a fixa- duration, reading time, regression frequency, and
tion is not random. Fixations almost never occur saccade duration. In particular, issues such as the
between words, and usually occur just to the left number of characters per line, line length,

Pages 67–79 71 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 72

Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

A research participant reads a news story while her eyes are tracked. The ISCAN eye-tracking system,
which does not require attachments to the head, uses a real-time digital image processor to automati-
cally track the center of the pupil and a low-level infrared reflection from the corneal surface. Ron
Hendricks © 2007. Original in color.

line spacing, number of lines, scrolling, and visual was more efficient with the 80-character line. Duch-
angle have been studied using eye-movement data. nicky and Kolers (1983), replicating findings from
Until now, the legibility of sans serif type versus Kolers et al. (1981), supported this conclusion that
serif type onscreen had not been studied using very short lines elicit more and longer fixation
eye-tracking technology. However, before the pauses. With short line length, readers may not be
present study is discussed, a review of the eye- able to make use of much information in each fixa-
tracking studies of onscreen typography is in order. tion and may decrease their saccade length, which
could slow down the reading (Rayner & Pollatsek,
Kolers, Duchnicky, and Ferguson (1981) measured 1989).
the readability of CRT displays based on eye move-
ments. They compared the reading speed for 40- Lynch and Horton (1999) showed that wider lines
versus 80-character lines. Doubling the number of text require readers to move their heads slightly
of characters was achieved by halving the width of or strain their eye muscles to track over the long
the letters. Results showed that doubling the num- text, making readability suffer because readers
ber of characters per line increased the number may lose track of the next line on the long trip back
of fixations per line from 4.82 to 8.00. However, to the left margin. Bernard et al. (2002) observed
the total number of fixations per passage was the same behavior. They tested the upper limits
decreased with 80 characters per line and the num- with text that contained 132 characters per line.
ber of words extracted per fixation was larger, the Viewing this long line required a large visual angle
durations of each fixation was longer, and the total that may have caused significant disruption to the
reading time was shorter. In other words, reading return sweep, often requiring head movements as

Visual Communication Quarterly 72 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 73

well as saccades. In increasing line length with (1983) suggest that it may be the upward move-
more characters per line, the advantages of fewer ment of scrolled text that reduces the difficulties
return sweeps and more fixations per line may associated with longer lines, but Dyson (2004)
be offset by difficulties in locating the next line. points out that this explanation has not been
Bouma (1980) too observed it is difficult for read- explored empirically.
ers to accurately locate the beginnings of new
lines after long lateral eye movement, particularly Another factor that has been suggested in explain-
if there is close vertical spacing. ing the decrement in performance when reading
from computer screens is the flicker of the com-
In a study of the effect of vertical space between lines puter screen caused by the refreshing of the image.
on reading performance, Kolers et al. (1981) found In a study using 1980s technology, Wilkins (1986)
that single spacing produced more fixations per line, found that eye movements are different for text
meaning that fewer words were read per fixation. that is flickering than for text that is not. Gould
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

Total reading time was therefore slightly longer with et al. (1987) studied the effects of flicker on CRTs
single spacing than with double spacing. However, by taking photographs of displays and having sub-
when fixations were analyzed, Kolers et al. (1981) jects read the photographs that were flicker-free.
found that the duration of individual fixations and the The decrement in reading performance with com-
rate of fixation did not vary in a single-spaced versus puter screens persisted even when text was read
a double-spaced document. Duchnicky and Kolers from the photographs. Gould et al. (1987) con-
(1983) also studied how the number of lines displayed cluded that of the factors that he and his associ-
on a computer screen impacted reading times. They ates have studied—angle of regard, visual angle,
displayed either one, two, three, four, or 20 lines of flicker—no single factor produces the decrement
type. They found that reading speed increased 9 per- when reading from computer screens, but rather it
cent when the number of lines displayed was is some combination of the factors.
increased from 1 to 20 lines and that 4 and 20 lines of
text were read faster than 1 and 2 lines of type. Over-
all, it appears that the denser the text onscreen, the
Research Questions
faster it is read. Dyson (2004, p. 378) points out that there are a
“limited number of published studies on typefaces
We tend to sit farther away from the screen than on screen, and no clear evidence of differences in
from printed matter when reading (Gould et al., their legibility.” A small, exploratory eye-tracking
1987), so that a longer physical line length sub- study to look at the legibility of Times New
tends a smaller visual angle compared to the Roman and Arial—a serif font and a sans serif
same physical line in print. Reading of long lines font traditionally designed for reading in print—
on screen does not appear to be slowed down, so and Georgia and Verdana—a serif font and a sans
it is possible that reading from the screen presents serif font designed specifically for onscreen read-
less of a problem with return sweeps to the next ing—was designed to add to the limited number
line, as these cover a smaller visual angle. Citing of published studies on typefaces onscreen. The
her earlier study (Dyson & Kipping, 1998), Dyson research questions for this exploratory study are
(2004, p. 390) said, “Without this difficulty, there as follows:
may be a saving in time spent in eye movements
of long lines as the total time in return sweeps is 1. Are there eye-movement behavioral differences in
theoretically decreased.” She further explains that the onscreen viewing of Times New Roman, Arial,
the greater number of return sweeps apparently Georgia, and Verdana typefaces?
takes longer than the additional time required to 2. Are there legibility differences onscreen for Times
read a long line of text. New Roman, Arial, Georgia, and Verdana fonts?
3. Are there differences in onscreen reading speed
Scrolling through the text may provide a cue for for readers of Times New Roman, Arial, Georgia,
the eye to locate the next line, especially if the and Verdana fonts?
line is not overly long. Duchnicky and Kolers 4. What fonts do readers prefer onscreen?

Pages 67–79 73 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 74

Methodology
This study was designed to investigate eye-movement
behavior, legibility, reading speed, and preference of
four commonly used typefaces on the screen—two
long-time fonts—Times New Roman and Arial—and
two fonts recently designed for reading onscreen—
Georgia and Verdana. These fonts were selected
because they are widely used in the design of web
pages, and appear to be the most widely used fonts
on the Internet.

The participants were six students recruited from


Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

a mid-sized university in the western United States.


The participants included four women and two
men. Their average age was 24.2 years. They were
rewarded with extra credit points in a college
course.

Participants were asked to read four timely Asso-


ciated Press wire stories on a computer screen.
Actual news stories published the week of the
study were selected for use in the study. The
news stories were about science and technology
topics and were chosen because the participants
would not likely have heard or seen the
information.

Four stories were displayed in one of the four differ-


ent typefaces. The stories contained approximately
the same number of words. The average number of
words in each story was 245.5. The story displayed
in Times New Roman contained 252, words, while
the story shown in Arial had 239. The story dis-
played in Georgia contained 245 words, while the
story shown in Verdana had 246 words. The stories
were assigned to a particular font in a random fash-
ion. Because of the small scope of this study, the
typefaces were not rotated between stories.

Because Tinker (1963) found that the optimal line


length is 12-point text with 80 characters for consec-
utive reading of text, the type for this study was dis-
played in 12 point with approximately 80 characters
per line, approximately 700 pixels wide. The leading
was set at a 16-point line height, which is found on
popular websites such as ESPN.com. Because each
of the four kinds of type occupies slightly different
amounts of space, the number of lines varied per
story. The story displayed in Times New Roman

Visual Communication Quarterly 74 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 75

contained 18 lines, while the story in Georgia had 17 on a 10-point scale on four items: not pleasing to the
and the story in Arial had 19. The biggest variation eye/pleasing to the eye, difficult to read/easy to read,
was the news story displayed in Verdana, which fuzzy/sharp, and too big or small/size about right. At the
required 22 lines to display its 246 words. end of the survey, they were asked two questions:

Participants were
instructed to “carefully”
read the stories as though
they would need to
remember the information
for a current events quiz.
The order of each
font/story was counter-
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

balanced. The eye-move-


ment data were collected using an ISCAN RK-426PC Above. Participants’ eye movements regressed
Pupil/Corneal Reflection Tracking System (ISCAN, more while reading the type set in Times New
Inc., Burlington, MA), which uses a corneal reflec- Roman. Three regressions or backward move-
tion system to measure the precise location of a ments are noted on the first line, while none
person’s eye fixations when looking at a visual dis- were measured in the rest of the paragraph. Used
with permission of the Associated Press, 2007.
play—in this case a news story—on a computer
Original in color.
monitor. The eye-tracking system does not require
attachments to the head.
It uses a real-time digital
image processor to auto-
matically track the center
of the pupil and a low-
level infrared reflection
from the corneal surface
(ISCAN, Inc. Operating Manual, 1998). The system What font is easiest to read? What font do you like
collects data at 60 Hz or about every 16.7 millisec- the most?
onds. For the purposes of this study, the default set-
ting was used to record saccades, which Rayner
(1995) has found to take 20 to 35 milliseconds. Results
Participants were positioned at a fixed distance of Even with a small sample size, a few interesting
about 18 to 24 inches from the computer screen, results emerged. Because the study included only
which had a resolution of 1024 ⫻ 600 pixels with six participants, none of the differences between
a 32-bit color depth. the four typefaces were statistically significant,
but the descriptive statistics are still interesting
Before the testing proceeded, the eye-tracking and important. For several of the variables, the
apparatus was calibrated by having participants differences are large enough to lead to speculation
fixate on a series of points positioned at various that statistical significance may be found in a
locations around the screen. Calibration was also study with a larger sample size.
checked after the showing of each story and in
most cases the equipment was recalibrated in In an attempt to answer the research question
order to be as accurate as possible. about which font is easiest to read onscreen, the
number of seconds it took to read the short news
After the eye-tracking test, participants filled out a stories was recorded. The results are as follows:
short survey. While looking at examples of the four Times New Roman, 59.17 seconds; Arial,
fonts onscreen, they were asked to rate the typography 62.09 seconds; Georgia, 60.01 seconds; and

Pages 67–79 75 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 76

Verdana, 58.11 seconds. Interestingly, in this lim- story set in Times New Roman. The other sans
ited exploratory study, Verdana—the sans serif font serif type, Arial, had the second lowest number
designed especially for computer use—was the of average regressions—20.33—while the other
font read most quickly, followed by Times New serif type, Georgia, had the second highest number
of average regressions at
21.17.

Regressions were also ana-


lyzed according to their
overall percentage of eye
movements. When the
data are examined in this
way, the typography
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

designed especially for the


Below. Participants were able to read the story computer screen performed better than the typography
displayed in Verdana typography more quickly designed for print use. Regressions amounted to 6.87%
and with relatively few regressions. This partici- of the eye movements during the reading of the
pant’s eyes did not regress while reading this story set in Verdana and 9.29% for the story set in
particular paragraph. Used with permission of the Georgia. The percentage of regressions is higher for
Associated Press, 2007. Original in color.
the traditional print fonts.
Some 11.07% of the eye
movements were regres-
sions for Times New
Roman and 9.83% of the
eye movements were
regressions for Arial.

Roman, which may be the font people are most The average number of regressions and the per-
accustomed to reading. centage of regressions were also analyzed for
sans serif versus serif typefaces. For sans serif
In addition, to examine the question about which fonts onscreen, the average number of regres-
font is easiest to read onscreen, the number of sions was 18.58 compared to 22.17 for serif
fixations was recorded. Despite its smallish fonts, although this difference of almost four
appearance on screen, the traditional serif font regressions was not statistically significant. The
Times New Roman outperformed the other three average percentage of the backward regressions
fonts. On average, participants fixated just 187.83 compared to the average percentage of the for-
times to read the news story, compared to 211.50 ward movements known as saccades equaled
times for Arial, the traditional sans serif font. Partici- 8.35% for sans serif typefaces and 10.18% for
pants fixated 218.83 times to read the story set in serif typefaces. The statistics became almost the
Georgia type and 231.83 to read the story set in same when looked at for the two typefaces
Verdana type. designed especially for viewing onscreen versus
the two typefaces designed for viewing in print.
Probably the most interesting results from this The number of regressions in the computer fonts
exploratory study are those involving regressions was 20.33 versus 20.04 for print fonts. The per-
or the backward movements. Even with just six centage of regressions for the computer fonts
participants, some interesting findings are noted. was 9.52% versus 9.01% for print fonts. While the
Participants experienced the fewest regressions— results were not widely spread for these two vari-
16.83 on average—while reading the story set in ables, there is greater variation in the statistics
Verdana type. They experienced the greatest num- when the variable is serif versus sans serif type
ber of regressions—23.17—while reading the fonts onscreen.

Visual Communication Quarterly 76 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 77

The difference in the pixel structure of


12-point Verdana (Top) versus 12-point
Times New Roman is shown when
magnified by 1000 percent.

each picked Verdana, Arial, and Georgia.


Times New Roman was not selected by even
one subject in answer to either question.

Discussion
This study is important because it appears
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

to be the first eye-tracking experiment to


address the question of the legibility of serif
versus sans serif fonts on a computer
screen. Specifically, this study compared
the legibility of two traditional fonts initially
designed for reading in print and two fonts
specifically designed to be read onscreen.
The rest of the variables were also analyzed for sans Although no statistical differences were found on
serif versus serif type, and for fonts designed for a number of variables, some interesting results
print use versus fonts designed for onscreen use. emerged.
The averages varied little in the two treatments. The
survey results to gauge preference show a strong Probably the most interesting finding revolves
preference for Verdana type onscreen. Overall, around regressions, those backward eye move-
Verdana was the most preferred typeface, followed ments that occur when readers need to refixate
by Georgia, Arial, and Times New Roman. Verdana because a forward saccade landed in a position
scored high on being pleasing to the eye (9.2), ease that’s not ideal or when readers need to revisit
in reading (9.8), legibility (9.2), and size (7.7), for an the content because of a failure to understand
overall score of 9.0. Verdana tends to look large at the material. Verdana, which has many features
12-point, probably leading to the lower score on to enhance legibility on screen such as lowercase
size. Georgia was rated a little lower with an overall letters that are proportionately tall compared to
score of 7.6. It received an 8.3 on being pleasing to uppercase letters, stroke widths that are not too
the eye, 8.5 for ease of reading, 5.8 in legibility, and thin, and generous spacing both inside the letter
7.8 for size. Arial scored 7.0 overall with a score of and between letters, required almost 6.5 fewer
7.8 on being pleasing to the eye, 6.0 for ease of regressions on average than Times New Roman,
reading, 8.3 in legibility, and 6.0 in size. Times New probably the most popular typeface used in printed
Roman scored relatively low overall with a compos- materials. Arial, possibly the most widely used
ite rating of 4.1, less than half of that for Verdana. sans serif font in print, had the second lowest
Specifically, it scored 4.7 on being pleasing to the average number of regressions. Georgia and Times
eye, 4.0 for ease of reading, 4.5 for legibility, and 3.2 New Roman, both serif typefaces, had the most
for size. The results are not surprising since Times regressions. That makes sense. It has long been
New Roman’s serifs appear to be swallowed up known by both creators and users of reading mate-
onscreen, while overall it looks too small at 12 point. rial onscreen that serif fonts render poorly at small
sizes because low-resolution computer screens
When participants were asked what font is easiest cannot handle the subtle details of the type.
to read onscreen, three selected Verdana, two
selected Arial, and one picked Georgia. When partic- When the regression data are analyzed according
ipants were asked what font they like most, two to the percentage of backward saccades to forward

Pages 67–79 77 Visual Communication Quarterly


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 78

saccades, a slightly different pattern emerges. emerge if longer reading periods were used.
Again, Verdana had the lowest percentage of Eyestrain might be more problematic for one
regressions at 6.87%. However, the second-lowest font over another.
percentage in this exploratory study belonged to the
other typeface designed specifically for onscreen Warde (1955) said type should be a “nonintrusive
use—Georgia—at 9.29%. Arial outperformed Times servant,” improving effortless reading and not inter-
New Roman with 9.83% and 11.07%, respectively. rupting the reading process. According to this
It would be interesting to see if the findings regard- approach, the form of typographic text should
ing regressions hold in a larger, more controlled be invisible or transparent, making it so the reader
study of reading onscreen. is not aware of the formal aspects of a typeface.
Computer fonts such as Verdana and Georgia are
Interestingly, in this limited study, the news story helping to make it easier to read onscreen. In this
displayed in Times New Roman was read almost information age, where we are reading more and
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

as quickly as the story presented in Verdana type. more information on a computer screen, more
When the number of fixations is analyzed, partici- legible onscreen type can save both trees and
pants fixated far fewer times on the story in Times headaches.
New Roman (187.83) as they did on the story in
Verdana (231.83). This occurred despite the fact
that the Times New Roman story contained 252 References
words, while the Verdana story contained fewer
words—246. However, the Verdana story may Bernard, M. L., Lida, B., Riley, S., Hackler, T., &
have taken longer to read because of content Janzen, K. (2002). A comparison of popular
online fonts: Which size and type is best?
and/or because it required 22 lines to display its
Usability News, 4.1 [online]../
content compared to the 18 lines it took to display usabilitynews/41/onlinetext.htm.
the story set in Times New Roman. Bernard, M., Liao, C. H., & Mills, M. (2001a). The
effects of font type and size on the legibility
However, it is the preference survey data that most and reading time of online text by older
strongly support the use of Verdana onscreen. The adults. Proceedings of CHI ‘01, 175–176.
six participants most preferred Verdana, followed Bernard, M. L., Mills, M. M., Peterson, M., & Storrer,
by Georgia, Arial, and Times New Roman for read- K. (2001b). A comparison of popular online
ing on a computer screen. The preference for Verdana fonts: Which is best and when? Usability
was twice as strong as it was for Times New Roman News 3.2 [online] ../usabilitynews/3S/font.htm.
Bouma, H. (1980). Visual reading processes and
for reading onscreen.
the quality of text displays. In E. Grandjean
& E. Vigliani (Eds.), Ergonomic aspects of
Much more eye-tracking research needs to be con- visual display terminals (pp. 101–114).
ducted comparing sans serif fonts to serif fonts London: Taylor & Francis.
before any concrete conclusions are reached, but Boyarski, D., Neuwirth, C., Forlizzi, J., & Regli,
this preliminary study is a good precursor to what S. H. (1998). A study of fonts designed for
may be found. In these future stories, more control screen display. In Proceedings of CHI ’98
needs to be exhibited, and more research partici- (pp. 87–94). Los Angeles, CA: ACM Press.
pants need to be used. In future studies, it is espe- Duchnicky, R. L., & Kolers, P. A. (1983). Readabil-
cially important to rotate the content between the ity of text scrolled on visual display termi-
different typefaces to control for the effects of the nals as a function of window size. Human
Factors, 25, 683–692.
content. It would be especially interesting to vary
Dyson, M. C., & Kipping, G. J. (1998). Exploring
the reading periods. The length of time it took for the effect of layout on reading from screen.
a participant to read one story in this exploratory In R. D. Hersch, J. Andre, & H. Brown (Eds.),
examination was about 60 seconds, which is a Electronic documents, artistic imaging and
fairly long period in an eye-tracking study of read- digital typography (pp. 294–304). Berlin:
ing. However, perhaps significant differences would Springer-Verlag.

Visual Communication Quarterly 78 Volume 15 January–June 2008


HVCQ_A_291625.qxp 3/26/08 7:12 PM Page 79

Dyson, M. C. (2004). How physical text logout Quinn, S. (2005, March 4). In search of:
affects reading from screen. Behavior & The best online reading experience.
Information Technology, 23, 377–393. http://www. poynter.org/content/content_
Geske, J. (1996). Legibility of sans serif type for view.asp?id=78569. Last accessed
use as body copy in computer mediated com- October 6, 2007.
munication. Presented at Association for Edu- Rayner, K., & Pollatsek, A. (1989). The psychology
cation in Journalism and Mass Communica- of reading. Englewood Cliffs, NJ:
tion (AEJMC) August, Visual Communication Prentice-Hall Inc.
Division, Anaheim, CA. Rayner, K. (1993). Eye movements in reading:
Gould, J. D., Alfaro, L., Barnes, V., Finn, R., Recent developments. Current Directions
Grischkowsky, N., & Minuto, A. (1987). Read- in Psychological Science, 2, 81–85.
ing is slower from CRT displays than from Tinker, M. A. (1963). Legibility of print. Ames,
paper: Attempts to isolate a single-variable IA: Iowa State University Press.
explanation. Human Factors, 29, 269–299. Tullis, T. S., Boynton, J. L., & Hersh, H. (1995).
Downloaded by [University of Saskatchewan Library] at 02:08 08 January 2015

Horton, W. (1990). Designing and writing online Readability of fonts in the windows environ-
documentation: Help files to hypertext. ment. In Proceedings of CHI ’95 (pp. 127–128).
New York: John Wiley & Sons. Denver, CO: ACM Press.
ISCAN, Inc. (1998). RK-726PCI Pupil/Corneal Warde, B. (1955). The crystal goblet, sixteen essays
Reflection Tracking System PCI Card Version on typography. London: Sylvan Press.
Operating Instructions. Rikki Razdan, Webcast1.com. Web-safe fonts for your site.
Burlington, MA. http://www.webcast1.com/webfriendly-
Kolers, P. A., Duchnicky, R. L., & Ferguson, D. C. fonts.html. Last accessed October 7, 2007
(1981). Eye movement measurement of Wilkins, A. (1986). Intermittent illumination from
readability of CRT displays. Human Factors, visual display units and fluorescent lighting
23, 517–527. affects movement of eyes across text.
Larson, K. (2004). The science of word recogni- Human Factors, 28, 75–81.
tion or how I learned to stop worrying and Wilson, R. F. (2001). HTML e-mail: text font read-
love the bouma. Microsoft typography, ability study. http://www.wilsonweb.com/
pp. 26–31. http://www.microsoft.com/ wmt6/html-email-fonts.htm. Last accessed
typograpy/ctfonts/WordRecognition.aspx. October 6, 2007.
Last accessed October 27, 2007.
Larson, K. (2007). The technology of text: Type
designers, psychologists, and engineers are Sheree Josephson is a professor of
joining forces to improve reading onscreen. journalism and visual communication
IEEE Spectrum, May, 28–31. at Weber State University in Ogden,
Lynch, P., & Horton, S. (1999). Web style guide: Utah.
Basic design principles for creating web
sites. New Haven, CT: Yale University Press.
Morrison, S., & Noyes, J. (2003). A comparison of Correspondence should be addressed to
two computer fonts: Serif versus ornate sans Sheree Josephson, Weber State University,
serif. Usability News 5.2 [online] http:// Department of Communication, 1605
psychology.wichita.edu/surl/usabilitynews/52/ University Circle, Ogden, UT 84408-1605.
UK_font.htm. Last accessed October 16, 2007. E-mail:sjosephson@weber.edu

Pages 67–79 79 Visual Communication Quarterly

También podría gustarte