Elastic Windows: Improved Spatial Layout and
Rapid Multiple Window Operations
Eser Kandogan Ben Shneiderman
Department of Computer Science & Department of Computer Science,
Human-Computer Interaction Laboratory Human-Computer Interaction Laboratory &
University of Maryland Institute for Systems Research
College Park, MD 20742 University of Maryland
Tel: (301) 405-2725 College Park, MD 20742
kandogan@cs.umd.edu Tel: (301) 405-2680
ben@cs.umd.edu
ABSTRACT
Most windowing systems follow the independent overlap-
ping windows approach, which emerged as an answer to the
needs of the 80s’ applications and technology. Advances
in computers, display technology, and the applications de-
mand more functionalityfrom window management systems.
Based on these changes and the problems of current window-
ing approaches, we have updated the requirements for multi-
window systems to guide new methods of window manage-
ment. We propose elastic windows with improved spatial
layout and rapid multi-window operations. Multi-window
operations are achieved by issuing operations on window
groups hierarchically organized in a space-filling tiled lay-
out. Sophisticated multi-window operations and spatial lay-
out dynamics helps users to handle fast task-switching and
to structure their work environment to their rapidly chang-
ing needs. We claim that these multi-window operations and
the improved spatial layout decrease the cognitive load on
users. Users found our prototype system to be comprehensi-
ble and enjoyable as they playfully explored the way multiple
windows are reshaped.
KEYWORDS: Window Manager, CAD, Task Switching,
Multi-window operations, Personal Role Manager, Program-
ming Environment, Elastic Windows
to appear in ACM AVI’96 Advanced Visual
Interfaces, Gubbio, Italy, May 27-29.
INTRODUCTION
It is widely believed that windowed environments are su-
perior to non-windowed ones. However, an early study by
Bury et al. [5] (1985) comparing users’ performance in win-
dowed systems to non-windowed systems revealed that task-
completion time in windowed systems can be longer due to
window arrangement time. A detailed analysis, however,
showed that actual times spent on solving a task were lower
in windowed environments compared to non-windowed envi-
ronments. Their experiments also showed that the error rates
in windowed environments were significantly lower. Al-
though systems compared in these experiments were rather
old, the results clearly indicate that benefits of windowing
can be overshadowed by the extra time spent on window
housekeeping activities.
Card et al. [6] analyzed window usage according to tasks and
identified seven functional uses of multiple windows. Among
these, independent control of multiple programs, referred to
here as multitasking, is the most significant. Basically, it
is the ability of users to work on different tasks in separate
windows. Analyses of work flow determined that people deal
with many tasks concurrently with frequent switches among
them [2]. For example, a researcher preparing a paper might
draw the figures in one window while writing the text of the
document using an editor in another window. Multitasking
results in improvements on the overall user performance due
to the decreased average task-completion time. Windowing
systems must provide good mechanisms for task-switching
to make multitasking more beneficial.
Windowing allows access to multiple sources of informa-
tion. It is possible to reduce the cognitive load on users by
allowing them to examine other windows for supplementary
information, or multiple representations for the task at hand
or use task-aids like cut-and-paste.
As stated by Card et al. [6], the computer display is used
not only as a communication medium but also as an external
memory for users. Thus having all the necessary informa-
tion on the screen and filtering out unnecessary windows is
a required property of windowing systems. Malone [16] ob-
served that the way people organize papers on their desk helps
them to structure their work and reminds them of unfinished
tasks. As Funke et al. [11] suggested, windowing systems
should support users to integrate, organize, compare, distill,
summarize, and apply the information.
MOTIVATION
Today’s windowing systems do not differ much in their basic
principles of window management. Almost all systems fol-
low the independent overlapping windows approach, where
windows are allowed to overlap each other, operations on

windows are performed one at a time, and size and location
of each window is independent.
With the typical early 80’s display resolution (640  480)
it was not possible to display two page-sized documents on
the screen simultaneously. Overlapping windows came as
a solution to the small-screen problem by allowing more
windows to be open simultaneously.
Resolutions like 1280  1024 are quite common these days,
which is roughly four times the 80’s resolution. Besides
the resolution, graphics processing speed increased as well,
which made sophisticated animations feasible. Animations
in windowing systems help users to understand the result of
operations and decrease the cognitive load.
With advances in computer technology, more demanding ap-
plications come into existence. The amount and variety of
information that users have to deal with increased a lot with
advances in networks and the Internet. The information that
is facing the users is usually unorganized and dynamically
changing, thus users themselves need to do the organization.
Typically when exploring information users want to keep
both detail and overview.
Computer-Aided Design (CAD), Computer-Aided Engineer-
ing, Object-Oriented Development Environments, and Geo-
graphic Information Systems (GIS) are typical multi-window
applications. In these applications, it is typically necessary
to open many windows displaying simultaneously different
parts or representations. Also opening separate windows for
toolboxes, commands, and options is becoming the practice
in complex applications.
With the increase in the number of windows, visualizing si-
multaneously all the necessary information for a task became
difficult. As the number of windows per task increases, task-
switching becomes more time-consuming since more win-
dows need to be opened/closed or moved/resized under the
independent overlapping windows approach. Due to the in-
dependence of windows, each window must be handled sep-
arately. Longer delays due to housekeeping further increase
task-completion time because of the loss of users’ mental task
context, kept in short-term memory. Increase in the number
of windows also prevents users to see the overview of their
desktop due to overlapping windows. This might delay users
to switch to unfinished tasks.
Contents of short-term memory are not only affected by the
time that passes, but also by the type of work carried out
during that time period. Since window housekeeping is an
activity related to the computer domain and not to the users’
task [21], the time spent on window management substan-
tially increases the disruptive effect on the short-term mem-
ory, thus implies a non-linear cost curve as the number of
windows per task increases.
Gaylin [12] observed that the number of window operations
that are used to switch the active window set constitutes 63%
of all the operations in an independent overlapped window
manager. This result supports the findings by Bannon et
al. [2] that people switch among tasks frequently forcing
them to change the visible set of windows on the screen.
Bederson and Hollan [3] observed that in traditional window-
based systems there is no graphical depiction of the relation-
ship among windows even when there is a strong semantic
relationship. This problem is most apparent in hyper-text
browsers and CAD systems, where each subwindow is either
a link followed or part of the system under design. In current
approaches, users have to deal with each window separately
when organizing their desktop.
Kahn et al. [14] observed a similar phenomenon and called the
presence of too many open windows “Windowitis”. They ob-
served that in Windowitis situations the users become quickly
disoriented, lose the relationships that exist between windows
due to loss of spatial cues, and become unproductive in com-
pleting their tasks.
Bly and Rosenberg [4] characterized the requirements of
multi-window systems as the ability of the windows to con-
form to their contents and the ability of the system to relieve
the user of window management.
On the basis of the problems discussed, we have updated
these requirements:
 support users to promote organization and coordination of
windows according to tasks.
 allow fast task-switching and resumption.
 free users’ cognitive resources to work on task related op-
erations rather than to window management operations.
 use screen space efficiently and productively for the tasks.
 provide a spatial layout that indicates the relationship be-
tween windows.
Earlier research that addresses some of these is described in
the Related Work section at the end of the paper.
PROPOSED SOLUTION: ELASTIC WINDOWS
Our method is based on three principles: hierarchical window
organization, space-filling tiled layout, and multi-window
operations.
Hierarchical Window Organization:
Hierarchical window organization supports users structuring
their work environment according to tasks. The hierarchi-
cal organization of windows allows users to map their task
hierarchy onto the nested rectangle tree structure.
Figure 1 displays the mail-tool application written using
elastic windows principles. The new messages are shown
iconized in the left window. Old messages are displayed
as icons, grouped hierarchically in separate windows on the
right. The hierarchical layout clearly indicates the semantic
relationship between the contents of the windows by the spa-
tial cues in the organization of windows. The layout provides
the user with an overview of all correspondence, where users
can pick any category and work on it.
Multi-window operations:
Typically, people organize papers on their desk as piles, and
move all of them simultaneously. Malone [16] found out that
users like to group items spatially. We claim that providing
multi-windowoperations on groups of windows can decrease
the cognitive load on users by decreasing the number of
window operations.