Cognition: Human–Computer Communication - Foundations for Designing User-Centered Systems - page 206

When reading text on paper, the print quality is usually quite high, often as high
as 1,000 dpi (dots per inch) for a high-end or photo-quality printer, and usually at
least 600 dpi for most laser and inkjet printers, with the light used to read it being
reflected from the paper. When you read text from a screen, however, the reso-
lution is typically much lower (at best, most laptops and display screens use not
much more than 250 dpi), and the light to read comes from the display. Greater
resolution of the text generally leads to faster, less fatiguing reading. Reflective
light using printed inks generally provides greater contrast between light and dark
than do luminance changes from a screen. So reflected light also tends to lead to
faster, less fatiguing reading.
Reading from the screen is generally slower, more error prone, and more
fatiguing than reading from hardcopies. Researchers have frequently found that
reading from the screen is 10-30% slower, with similar results reported for errors
and for fatigue (Dillon 1992 ; Gould et al. 1987 ; Kurniawan and Zaphiris 2001 ). It
is hard to ascribe the difference between these factors solely to the resolution and
luminance sources. Some think there are other causes as well, such as the func-
tionality of page turning and other aspects related to the text being on paper or on a
screen (Dillon 1992 ).
7.3.4 Scanning Displays and Menus
Users typically scan displays, rather than read them, particularly for menus and
web pages with links. This is an important skill related to reading. If you observe
someone using a web page, particularly a complicated one, you should be able to
observe this scanning behavior. Most (although perhaps not all) will not read every
item on the page, but will scan the page, looking for, and being susceptible to, the
visual factors explained in the chapter on vision.
This scanning behavior is also seen when users read menus. Hornof and Kieras
( 1997 ) found that longer menus take about 120 ms/item longer to use. They also
found that users read more than one menu item at a time. The physiological
structure of the eye allows two items to be read within a single fixation. As users
read, they scan menus in both systematic and random ways: they are systematic in
that they mostly read top to bottom; they also sometimes randomly skip around the
items on the menu.
Further details of human behavior in this area and models to predict behavior in
menu reading are available (Byrne 2001 ; Hornof and Halverson 2003 ; Rieman
et al. 1996 ). The models can help explain parts of behavior in this area that are
predictable (e.g., walking through the menu items) although the reasons for the
random skipping around menu items are not yet well explained. Figure 7.5 shows
two menus for a VPN client. The first menu has many more items than the second.
If the targets are ISPtoPSU and ITS Wireless at University Park, the first menu
definitely makes the items take longer to find.