Bumping Windows Between Monitors
Bumping Windows between Monitors
Tim Regan
Mary Czerwinski
Brian Meyers
Greg Smith
3/12/2003
Technical Report
MSR-TR-2003-13
Microsoft Research
Microsoft Corporation
One Microsoft Way
Redmond, WA 98052
Bumping Windows between Monitors
Tim Regan, Mary Czerwinski, Brian Meyers, and Greg Smith
Microsoft Research, 1 Microsoft Way, Redmond, WA 98052 USA
timregan@microsoft.com
Abstract: Users’ move from single to multiple monitors so that they can use more screen real estate. This increase enables
them to keep a greater number of windows open and visible at the same time. But there is a cost: arranging a window takes far
longer since there is more screen space to traverse and more relationships between windows to take into account. To address
this we added ‘bumping’ to an application and tested it in a user study. Bumping allowed users to automatically move a
window across multiple monitors. In this paper, we present our experiment contrasting three styles of bumping. We found less
window repositioning required when users were given bumping in contrast to their existing practices. However we also found
that simple implementations of intelligent bumping are not predictable, causing problems for our participants.
Keywords: Windows management, multimon, multiple monitors
1 Introduction
To test the efficacy of these bumping methods against
As increasing numbers of users turn to multiple monitors to
users’ existing practices we conducted a user study, the
results of which are presented in this paper. Firstly we
solve their problems of limited screen real estate, new
present related work and give more detail of our design and
problems with windows management arise, problems not
implementation.
dealt with by the operating system. To quote Grudin
(Grudin, 2001):
“Multimonitor development has focused on getting the
2 Related Work
system display software and the application program
Data visualization techniques often have to deal with
interface to work, with little attention to the human
packing a large number of visible items (data points) into a
computer interface or intelligent exploitation by system or
limited screen space. For example, in Valence (Fry, 2002)
applications”
Fry looks at building representations that explore the
One such problem is the increased effort in arranging
structures and relationships inside very large sets of
windows. Multiple monitor users can (and do) keep more
information in a limited space. But unlike generalized
windows open, but as they drag windows to arrange them
visualizations, windows are simple objects: they are
there is more screen space to traverse. The traditional single
rectangular, they need to be large enough for their contents
monitor method of minimizing the window or moving it
and surrounding tools to be legible, they live on a
further back in the z-order of open windows is often not
rectangular background, and there are not that many of
used on multiple monitors, as there is the space available to
them open at any one time. Hence we can restrict our
keep the window visible. We know from Fitt’s law that the
attention to algorithms for manipulating rectangles. In (Bell
time taken to move the cursor to an item on the screen is
and Feiner, 2000), Bell and Feiner detail the algorithms
proportional to the log of the distance to the item over the
they have used to manage the layout of rectangles, covering
items size.
notions of empty space and its dual, and how to add and
To make windows arrangement faster and easier on
remove rectangles from each.
multiple monitors we added a new feature that we called
There are a number of interaction techniques and
bumping to an application. Bumping took a window and
representational metaphors that work on, and exploit, large
automatically moved it to a new position. We implemented
displays. Example display metaphors include fish-eyes
three different bumping methods:
(Furnas, 1986), zoomable UIs (Bederson, 2000), and focus
1. Faithful bumping maintains the relative position of the
plus context screens (Baudisch et al, 2001). Flow menus
window. If the source and target screen are the same size
(Guimbretière et al, 2001) and gestural interactions (Myers
(in pixels squared) then no resizing will take place.
et al, 2002) are good examples of large screen interaction
2. Dark-space bumping moves and resizes the window to
techniques. Whilst informative, these do not deal directly
cover a currently unobscured region of the desktop.
with the problem of automatic layout of information or
3. Unobscured bumping moves and resizes the window to
windows.
cover only unobscured desktop or windows that are already
partially obscured.
One solution to the windows arrangement problem is to
data manipulation. He may turn to the web to find relevant
swap from overlapping windows to other possibilities, for
newsgroup threads and having read them; move the
example tiling windows (Bly and Rosenberg, 1986). Today,
newsgroup browser aside to return to my spreadsheet.
innovative ways of handling windows are examined to
However he still needs to refer to the suggested solution, so
handle new tasks and new ways of working. For example in
the browser must remain visible. The reference may be
(Kandogan and Schneiderman, 1996), Kandogan and
purely visual or may involve transferring information (e.g.
Shneiderman revisit tiled windows to address tasks like
through cut-and-paste or drag-and-drop) between open
photo sorting, where masses of images need to be addressed
windows.
by the user at once. A categorization of such windows
Move and keep working
coordination actions is contained in (North and
Here the user wants to move the window to a new position
Schneiderman, 1997). Similarly in (Beaudouin-Lafon,
and keep working on it there. For example, a user may turn
2001), Beaudouin-Lafon proposes rotating and peeling back
to his secondary screen and start writing an email, but if the
windows as a way to address the increasing number of open
email becomes complicated he may want to move it to his
windows made possible by today’s powerful PCs. Through
primary work area. This would be a semantic-move in that
our notion of bumping we seek to tackle the problem of
it positions the window following the user’s semantic
window arrangement without abandoning overlapping
designation of the screen space. Alternatively a user may
windows, since they are the established norm.
turn from some code one is writing on her new 21” monitor
The closest work we have found to our notion of bumping
to perform some SQL queries on her old 15” screen. As the
is by Hutchings and Stasko in (Hutchings and Stasko,
data becomes hard to view on the smaller screen she may
2002). Their ‘expand and shove’ techniques allow users to
move the window to the larger screen. This would be a
expand a window in such a way that other windows on the
quality-move in that it positions the window following
screen are not further obscured or shrunk, though they may
qualitative differences within the screen space. There is
be moved. Any movement maintains windows’ relative
overlap between the notions of semantic-move and quality-
positioning. Expand and shove work together to give more
move; users may designate one screen as their secondary
screen space to the window that the user is currently
monitor on which they conduct peripheral tasks purely
focused on, but they result in older windows collecting
because of qualitative differences between that and their
towards the periphery of the display. While this may work
primary screen. These qualitative differences extend
well on large homogeneous display surfaces, multiple
beyond the technical specification of the screen and
monitors are used differently. Multiple monitor users will
graphics card. One of the participants in the experiment
turn from one screen to another, changing their experience
described later in this paper who ran a three screen set-up
of which screen is the focus and which peripheral. Expand
reported using the left hand screen for instant messaging
and shove never make a window (or its visible region)
and emailing friends because the left monitor was not
smaller, thus leading to a number of large windows vying
visible to people walking past him, and thus maintained his
for desktop space. Bumping allows users to take up unused
privacy.
desktop with the bumped window – which may involve
Move and return to later
making it smaller. Bumping is also useful for discarding
Single screen users often minimize windows they wish to
windows in such a way that they may be easily reacquired
stop working on now but return to later. For example a user
later. In summary expand and shove are useful for
may turn from authoring a report to answer incoming email.
increasing the visible area of the window a user is currently
Once the email is dealt with she returns to the report. There
attending to without sacrificing relative positioning of
are disadvantages with minimizing windows, disadvantages
windows and without complex windows arrangement.
associated with the limited size of the Microsoft Windows
Bumping is useful when a user’s attention is swapping from
start-bar or the Mac task-switcher. Because both are small
one window to another.
they quickly get crowded with icons if many applications
are running. Because they are small they cannot effectively
3 Arranging Windows
leverage users’ spatial navigation skills when users seek to
return to an application. Placing a window on another
When people move windows aside, where do they want to
screen allows users to get on with their work and more
put the window? A field study to answer this question
easily go back and get the window when it is required.
accurately is beyond the scope of this paper, but we can
Remove
analyze the possibilities.
Having become acquainted with using window moves in
Move and refer to
preference to minimize, multiple monitor users may also
Here the user wants to move a window away and start work
associate the ‘window move’ action with the ‘get window
in another window, while keeping the first visible for
out of the way’ intent, thus making use of fast muscle
reference. Suppose that a user is working on a spreadsheet
memory. Hence some moves may in fact be in place of
and becomes stuck trying to perform a complex sequence of
closing windows.
Figure 1: Faithful Bumping
Figure 2: Dark-space Bumping
Figure 3: Unobscured Bumping
Faithful bumping is the easiest to understand and is shown
4 Bumping Functionality
in Figure 1. The bumped window moves from one screen to
another, and maintains its relative position. In Figure 1 the
In this section we will explain the bumping mechanism and
clear IE window on the left hand screen is moved as
discuss our design choices.
indicated by the grey arrow to the same size and position on
As mentioned in the introduction we implemented three
the right hand screen. We implemented two styles of
bumping methods:
faithful bumping: one that retains the absolute size of the
1. Faithful bumping,
window and another that resizes the window so that it
2. Dark-space bumping, and
occupies the same proportion of the new screen. This is
3. Unobscured bumping.
particularly useful for multiple monitor users who have
used an old discarded monitor as their second screen. This
• Gestural input would be a substantial departure
second screen is often physically smaller and runs at a
from current window manipulation techniques and
lower resolution than the primary monitor. In these
hence less likely to be adopted
conditions a faithful bump of a window, from the primary
A keyboard shortcut to bump the window was also included
to the secondary screen, may result in obscuring all of the
in our prototype.
second screen without displaying the whole window.
Resizing during faithful bumping is also useful for users
5 Experiment
with homogeneous screen who run their start bar as a wide
horizontal bar at the vertical edge of one of the monitors, as
To test our bumping idea, both quantitatively and
this reduces the available desktop space on that monitor.
qualitatively, we ran a user study. In this section we present
Dark-space bumping takes a bumped window and places it
the study and discuss the results.
inside the largest rectangle of unobscured desktop available
in the direction of the bump, resizing if necessary. Figure 2
1.1
Research Questions
The specific questions we wished to address were:
shows the result of dark-space bumping. The clear IE
1) Is the bumping button effective in reducing the amount
window on the left hand screen is bumped to the clear IE
of window dragging required of users?
window on the right hand screen, in order to avoid the two
2) Does the bumping button improve productivity?
open Windows Explorer windows, its size is reduced.
3) Do users prefer a bumping button?
Like dark-space bumping, unobscured bumping searches
With reference to the above questions we wanted to know
for new rectangles of unobscured space in which to position
which bumping algorithm performed best.
the bumped window. In addition to uncovered desktop it
includes windows that are already partly obscured by other
1.2
Experimental Setup
windows. Figure 3 shows how the clear IE window on the
Participants
left hand screen will be bumped to the right hand screen
17 volunteers (5 female and 12 male) from the greater
using unobscured bumping. Because one of the Windows
Puget Sound area were recruited from our company’s
Explorer windows is partially obscured by the other, the
usability database to participate in the study. Unfortunately
bumped IE window assumes its new position over a
3 cancelled and 3 interpreted the screening question “Do
combination of the already obscured window and
you currently use Windows XP with two or more
unobscured desktop space.
monitors?” differently from us, bringing our final number
Note that if there is sufficient space available, a dark-space
of participants to 11 (2 female and 9 male). We wanted to
or an unobscured bump may not result in the window
use only people experienced with multiple monitor use for
switching screens but moving on the current screen instead.
two reasons. Firstly, we have found that the learning effects
Our prototype application was a text editor with two
of multiple monitor use can dominate other factors in
buttons added to a toolbar with arrows signifying ‘bump
studies. Secondly, we wanted to get suggestions from the
left’ and ‘bump right’. In most windows based operating
users on how a bump button should work in practice.
systems (e.g. Microsoft Windows, Mac, and X Windows)
Participants’ jobs were mainly (but not all) technical.
buttons associated with managing the positioning or size of
Participants had been using multimon for an average of 2 ½
a window are placed in the window’s title bar. Most
years (SD = 2 years 5 months). This was not evenly spread:
systems provide alternatives to this. For example, on
7 users had been using multiple monitors for less than 1 ½
Microsoft Windows PC the keyboard shortcut Alt-Space M
years while the remaining 4 had been using multiple
acquires a window for moving, so that pressing the arrow
monitors for more than 4 years. The participants were
keys will move the window until the user presses Enter to
screened to be intermediate to expert Windows and Office
leave the mode. Our bump buttons’ placed in the toolbar
users, as per validated internal screening tools.
just below the menus was a pragmatic choice for ease of
Task & Design
implementation.
We had participants do two tasks four times on a two
Gestural interaction styles have been successfully applied to
monitor PC. The two monitors were identical 21” CRTs
large screens (Guimbretière et al, 2001) and to small
(i.e. not flat LCD screens) each at 1024 by 768 pixel
screens (Perlin, 1998). Instead of a button we could, for
resolution with the start bar along the bottom edge of the
example, have had users acquire a window and then flick
left hand screen. Participants completed a brief
the mouse to send the window in the direction of the flick.
questionnaire after each of the four sessions and a longer
We decided against this in favour of a button push for three
one at the end. Before each pair of tasks the participants had
reasons:
a short practice session to familiarize themselves with the
• Gestural input has not taken off for standard
changes in behaviour of the bump button. The ordering of
personal computing size screens (i.e. 15” to 21”
the conditions was fully balanced across the subjects. The
diagonal screens) which multiple monitors extend
four conditions were the three bumping methods discussed
already (faithful bumping, dark-space bumping, and
unobscured bumping) as well as a condition with the
ANOVAs unless otherwise stated. The discussion therefore
bumping buttons removed.
hangs on trends inferred from the data and should not be
The prototype application we chose was a simple text
interpreted as statistically significant.
editor, but with the minimize and the maximize buttons
disabled.
Task 1: The Reconstruction Task involved opening 9 Rich
Text Format (RTF) files. The contents of 5 of the files
were to be found, jumbled line by line, in 3 of the other
files. The files opened in the same place and with the same
size on left hand screen so that each file is initially obscured
by the previous one. The last file to be opened was the
instructions. Users had to reconstruct the missing contents
of the 5 files using the other 3. So, for example, the line
starting “H-10” is line 10 from file H and the user, having
located it, would cut-and-paste it back into position in file
H. After 3 minutes the users were stopped and asked to start
the second task.
Task 2: The Alphabet Task instructions gave users two
random lines of ten letters and asked them to recreate them
Figure 4: Task 2 in Progress
across the two screens using the letter files provided so that
Preference was measured by the question “If you had to do
the letters were not obscured. Figure 4 shows one screen
the tasks again, which version would you use?” 9 of the 11
during the second task in progress.
participants choose a bumping case (χ2 (1, n=11) = 4.45, p
The Reconstruction Task was designed to involve a lot of
= 0.04). They were also asked “Which task did you enjoy
switching back and forth between windows: 3 files were
the most?” and 7 of the 11 participants choose a bumping
required for repeated searching for lines to paste into the
case (χ2 (1, n=11) = 0.82, p = 0.36). Although the presence
other 5 files. The Alphabet Task was designed to require a
of a bumping button was preferred, the preference is not
lot of careful window positioning and resizing. As we have
entirely explained by enjoyment (perceived performance
argued already, positioning windows so that they are
gains etc. could be other factors considered by participants).
readable (unobscured) and so that they are available are key
Of those who chose the bumping cases there was no
windows management tasks on multiple monitor systems
appreciable difference shown between the faithful, dark-
and so our tasks are typical, if abstract, windows
space, and unobscured bumping conditions. We also asked
management tasks.
if participants felt a bumping button to move a window
These tasks abstract the two main behaviours encountered
automatically was useful. On a seven point scale, with 1
when users move and refer to windows, as discussed in
representing “not at all” and 7 representing “yes very
Section 3. Task 2 abstracts visual reference (i.e. the task of
much”, the average answer was 5.73 (STD 1.85).
looking at the information in multiple windows
simultaneously). Task 1 abstracts transferring information
30000
between open windows (e.g. cut-and-paste). We could
instead have chosen more realistic, domain specific, tasks
25000
instead of abstract ones (e.g. building a stock report in
s)
Word from a PowerPoint deck and a number of company
i
xel 20000
p
(
and financial websites, or predict tomorrow’s weather from
a number of current and recent weather charts). We chose
15000
ment
not to for two reasons, one pragmatic and one theoretical.
10000
Pragmatically our bumping button was implemented as part
of an application (not added to an existing application) and
ag Move
Dr
5000
so sticking to a simple WordPad kept the programming
required manageable. Theoretically we believed that the
0
results from abstract tasks could more easily be generalised,
Faithful
Dark-space
Unobscuring
None
precisely because of the abstraction.
Figure 5: Reconstruction Task, Average Total Windows Dragging
1.3
Experimental Results
Partly due to our small sample size the differences and
distinctions given in this section are mostly not significant
as shown by the test values reported. The tests are one-way
40000
40000
35000
35000
s) 30000
s) 30000
i
xel
i
xel
p 25000
p
(
25000
(
20000
20000
ment
ment
15000
15000
10000
ag Move
10000
ag Move
Dr
Dr
5000
5000
0
0
Faithful
Dark-space
Unobscuring
None
Faithful
Dark-space
Unobscuring
None
Figure 6: Alphabet Task, Average Total Windows Dragging
Figure 8: Alphabet Task, Average Adjusted Total Windows
The basic purpose of the bumping button was to reduce the
Dragging
amount of window dragging required. Figure 5 and Figure 6
Figure 7 and Figure 8 show a similar measure: the average
show the average total number of pixels that windows are
total number of pixels windows are dragged for the
dragged for the Reconstruction Task and the Alphabet
Reconstruction Task and the Alphabet Task, but this time
Task. The values for the Reconstruction Task are averages
with position tweaking taken into account. Bumped window
of 7391, 7729, 6351, and 19039 pixels for the faithful,
moves that require immediate repositioning are added to the
dark-space, unobscured, and no bumping conditions
total window dragging number. The values for the
respectively (F(3,30)=1.82, p=0.17). The values for the
Reconstruction Task are averages of 14498, 15120, 12511,
Alphabet Task are 9914, 9549, 8502, and 19039 pixels for
and 25877 pixels for the faithful, dark-space, unobscured,
the faithful, dark-space, unobscured, and no bumping
and no bumping conditions respectively (F(3,30)=1.21,
conditions respectively (F(3,30)=1.97, p=0.14). In both
p=0.32). The values for the Alphabet Task are 22311,
cases we see a difference between the case without a
29600, 24398, and 25877 pixels for the faithful, dark-space,
bumping button and the cases with: there is more windows
unobscured, and no bumping conditions respectively
dragging required without the bumping button. But this is
(F(3,30)=0.52, p=0.67). The prominent difference, between
misleading. The reduction in window dragging was only of
the case without a bumping button and the cases with, is
benefit if the bumped window alighted in a position the user
retained in the Reconstruction Task as we go from the drag
was happy with. If the user had to immediately move their
figures to these adjusted drag figures but lost in the
cursor to the window and correct its position, then the
Alphabet Task. It is replaced with a slight increase in the
bumping did not reduce mouse movement.
dark-space dragging over the other conditions. In the
Reconstruction Task users needed to swap between
30000
windows, but the windows positioning was left to the users’
own preferences. In this case the bumping buttons reduced
25000
the amount of dragging required. The Alphabet Task was
s)
about moving and resizing windows where their exact
i
xel 20000
p
(
placement was important and largely prescribed. Hence any
initial advantage gained by bumping for this task is lost as a
15000
ment
window’s positions and size are adjusted until perfect.
10000
ag Move
Dr
5000
0
Faithful
Dark-space
Unobscuring
None
Figure 7: Reconstruction Task, Average Adjusted Total Windows
Dragging
many people wanted it to retain its size (of the 8
14
participants whose comments included sizing information,
12
7 of them suggested no resizing).
ed
We observed some unexpected uses of the bumping
10
buttons, especially in the Alphabet task. One was the use of
8
the ‘bump left’ button. Windows were fixed to initially
i
nes Copi
open towards the left of the left hand screen. So whilst in
L
6
the conditions which allowed automatic resizing clicking
of
the bump left button had the effect of making the window
4
smaller. Although this did not save participants time,
Number
2
several tended to prefer using it over manual resizing.
Another unexpected use was the sizing and positioning of a
0
window before clicking the bump button. Because of its
Faithful
Dark-space
Unobscuring
None
easy predictability, some users positioned and sized a
Figure 9: Reconstruction Task, Average Number of Lines
window on the left hand screen before bumping it over to
Completed in 3 Minutes
the right hand screen using in the faithful bumping
condition.
4
1.4
Discussion of Experimental Results
The bumping button was clearly a popular addition for
es)
multimon users: our participants chose it above the no
3
bumping condition and enjoyed using it more. They
i
nut
m
declared that the addition of a bumping button was useful.
2
It reduced the amount of dragging users needed to perform,
but the total amount of window management related cursor
me Taken (
movement was only reduced in the Reconstruction Task.
1
Ti
The Reconstruction Task was typical of tasks where
multiple windows provide the sources and targets for
0
content. The fine-grained windows positioning required in
Faithful
Dark-space
Unobscuring
None
the Alphabet Task exceeded our bump methods’ abilities.
Performance was better with the simple behaviours: without
Figure 10: Alphabet Task, Average Time Taken
bumping or using faithful bumping. Each version of the
Figure 9 and Figure 10 show measures of task performance.
bumping had been explained to participants and practiced
Figure 9 shows the number of lines cut-and-paste in the 3
by them before the tasks, but dark-space bumping and
minutes of the Reconstruction Task. The values are
unobscured bumping was too complex for participants to
averages of 10, 8, 8, and 12 lines for the faithful, dark-
predict. It may be that longer term usage of the bump button
space, unobscured, and no bumping conditions respectively
would allow users time to develop an effective mental
(F(3,30)=3.53, p=0.03). Figure 10 shows the time taken to
model of the more complex bumping algorithms. It also
complete the Alphabet Task in minutes. The values are
seems that the single aspect of the more complex bumping
averages of 3.45, 3.82, 3.82, and 3.27 minutes for the
that participants found least useful was the resizing. One
faithful, dark-space, unobscured, and no bumping
participant suggested that we keep our algorithms for
conditions respectively (F(3,30)=1.32 p=0.29). Both charts
intelligent placement of windows but just remove the
show an increase in productivity for the faithful and the no
resizing element.
bumping conditions over the dark-space and the unobscured
However, the best bumping button may be the most
bumping conditions.
complex. One participant’s advice to us on the best
This distinction between the more complex conditions
bumping method to encode was that it should place the
(dark-space and unobscured) and the simple conditions
window “where I want it to go”.
(faithful and no bumping) are reflected in the participants’
comments. Typical comments include:
• “made it quicker to move them, especially when I
6 Conclusions
knew where they would go”,
•
We have explained how the benefits afforded as users adopt
“it is useful, but the bump button needs to place
multiple monitors come with an associated cost in terms of
the window in an expected place”, and
windows management. On single monitor PCs complex
• “[it needs to be] simple to guess what it does”.
window arrangements are not desirable since the resulting
When asked where a bumping button should place the
windows are too small to work with. In multiple monitor
window, the participants gave a variety of responses, but
systems users may lay windows side by side, and allow
each window ample space to read or work in. This
8 References
arrangement involves users dragging and resizing windows
across large screen distances. We showed how automating
Baudisch, P., Good, N., and Stewart, P. (2001). Focus Plus
this arrangement by bumping windows can be an
Context Screens: Combining Display Technology with
advantage. We implemented three simple methods of
Visualization Techniques, in Proceedings of UIST2001.
bumping and tested them in a user experiment. Bumping
reduced the amount of windows dragging required, though
Beaudouin-Lafon, M. (2001). Novel Interaction Techniques for
Overlapping Windows, in Proceedings of UIST2001.
not for all tasks: some tasks requiring exact windows
placement still need to be accomplished manually.
Bederson, B. (2000). Jazz: an extensible zoomable user interface
Bumping may be added to the multiple monitor UI in a
graphics toolkit in Java, in Proceedings of UIST2000.
variety of ways. It could be implemented within an
application, in an OS, or as part of a set of windows
Bell, A.B. and Feiner, S.K. (2000). Dynamic Space Management
management functions (e.g. Ultramon
for User Interfaces, in Proceedings of UIST2000.
http://www.realtimesoft.com/ultramon/) has a notion
similar to our faithful bumping in their multiple monitor
Bly, S.A. and Rosenberg, J.K. (1986). A Comparison of Tiled and
management software). For those working on such
Overlapping Windows, in Proceedings of CHI86.
enhancements our recommendations are:
• Include bumping – we have shown that it can be
Furnas, G. (1986). Generalized Fisheye Views, in Proceedings of
effective in reducing the windows drag required
CHI86.
for windows management.
• Keep it simple – we have shown that the semantics
Fry, B. (2002). Valence,
of the bumping button must be transparent to users
http://acg.media.mit.edu/people/fry/valence/
for the benefits to emerge.
Grudin, J. (2001). Partitioning Digital Worlds: Focal and
Peripheral Awareness in Multiple Monitor Use, in
7 Next Steps
Proceedings of CHI2001.
Although each bumping method we implemented was
Guimbretière, F., Stone, M., and Winograd, T. (2001). Off the
obviously algorithmic, and hence predictable, two of the
wall: Fluid interaction with high-resolution wall-size
algorithms proved too complex for the user to accurately
displays, in Proceedings of UIST2001.
and quickly predict. To address this we intend to re-
implement similar algorithms but without the windows
Hutchings D.R. and Stasko, J. (2002). QuickSpace: New
resizing element. This implies that the bumped window
Operations for the Desktop Metaphor, in Proceedings of
would obscure more of the screen (since we cannot make it
CHI2002.
smaller) and so we will try new methods of determining
valid areas to obscure. For example we can analyze
Kandogan, E. and Shneiderman, B. (1996). Elastic Windows:
windows bitmaps to find large areas of white-space
Improved Spatial Layout and Rapid Multiple Window
adjacent to a windows edge.
Operations, in Proceedings of AVI96.
We will also add animation to the window bump to test if
Myers, B.A., Bhatnagar, R., Nichols, J., Peck, CH., Kong, D.,
that helps users gain an understanding of the underlying
Miller, R., and Long, C.A. (2002). Input Devices:
mechanism and hence find the bump more predictable.
Interacting at a distance, in Proceedings of CHI2002.
We intend to add the bumping button to a greater array of
applications (or all applications) so that we can study a
North, C. and Shneiderman, B. (1997). A Taxonomy of Multiple
richer media mix of tasks in our experiment, for example
Window Coordinations, Technical Report, University of
pasting pictures into reports or presentations or data
Maryland.
between spreadsheets. We can then install the enhancement
in participants’ workspaces to gain a longitudinal
Perlin, K. (1998). Quikwriting: Continuous Stylus-Based Text
understanding of the usage of bumping.
Entry, in Proceedings of UIST99
