When One Isn’t Enough: An Analysis of Virtual Desktop
Usage Strategies and Their Implications for Design

Meredith Ringel
Stanford University, Computer Science Department
353 Serra Mall, Stanford, CA 94305 USA
merrie@cs.stanford.edu

ABSTRACT
Screen space is a limited resource for computer users—
multiple monitors are one means of workspace expansion,
and “virtual desktops” are yet another way to increase
screen real-estate. We present a taxonomy of organization
strategies based on our observations during a series of
interviews with virtual desktop users. Additionally, we
explore causes of varying user preferences for physical
versus virtual means of screen-space expansion. Finally, we
discuss the design implications of our findings.
Keywords
Virtual desktops, multiple monitors, information organization.
INTRODUCTION
Virtual desktops are window management systems that
expand the space available for application windows by
allowing users to switch between different workspaces. We
interviewed twenty people who use virtual desktops on a
regular basis in order to learn how information organization
strategies differ between virtual desktop users and multiple
monitor users, and to learn what guides users’ preferences
for either virtual or physical means of expanding their
workspaces.
Related Work
Grudin’s field study of multiple monitor use [1] explored
current user practices in “multimon” situations. He found
that most people used their extra monitor as a place for
keeping secondary applications, such as email, while their
main monitor (often larger and/or higher resolution) was
devoted to their primary task.
FIELD STUDY
The population we interviewed for our study is
representative of typical virtual desktop adopters—twenty
men aged 20 through 50 who work with computers on a
daily basis. All participants were asked a series of questions
about their experience with virtual desktops and multiple
monitors and about their computer setup. Interviews were
tape recorded, with the participants’ permission.
We observed that the participants’ information organization
strategies could be grouped into five distinct categories, and
that their mappings of windows to particular virtual
desktops were remarkably consistent for individual users
over time. We also learned that many users did not view
virtual desktops as a cheaper and less optimal solution than
multiple monitors, but rather that they preferred to use
virtual desktops in some situations and multi-monitor setups
in others.
Partitioning Strategies
After asking twenty users about how they divided their
windows among the available virtual desktops, five
different organization strategies emerged:
• Tasks (6 users): Examples included having one
desktop devoted to each course that the user was taking,
or having one desktop for each of several different
programming projects, with each desktop containing all
the code, testing, and documentation related to one of the
projects.
• Subtasks (9 users): This group of users tended to be
working on a single large programming project, and
would typically devote one desktop to source code
editing, another desktop for testing, and another desktop
for support documentation.
• Primary/Secondary (2 users): These users had one
desktop for work, and another for email and the web.
Both of the participants who used this strategy also said
that they would prefer to use multiple monitors instead of
virtual desktops; interestingly, Grudin observed that
primary/secondary was the predominant organization
strategy among multimon users [1].
• Systems (2 users): Different operating systems (via
remote shells or windows) were on separate desktops.
• Applications (1 user): Applications were grouped onto
virtual desktops by type, rather than by task. For example,
all text documents were on one desktop, all web browser
windows on another, source code editors on a third, etc.
Consistency of Mappings
There was an unexpected amount of consistency in users’
mappings of groups of windows onto specific virtual
desktops. Here, “consistency” refers to day-to-day
consistency for each individual, rather than to uniformity
across multiple users. Consistency in mappings suggests