Human-Computer Interaction -- INTERACT'03
M. Rauterberg et al. (Eds.)
Published by IOS Press, (c) IFIP, 2003, pp. 57-64
Drag-and-Pop and Drag-and-Pick: techniques for accessing
remote screen content on touch- and pen-operated systems
Patrick Baudisch1, Edward Cutrell1, Dan Robbins1, Mary Czerwinski1,
Peter Tandler2, Benjamin Bederson3, and Alex Zierlinger4
Microsoft Research, Redmond, WA; 2Fraunhofer IPSI, Darmstadt, Germany;
HCIL, University of Maryland, MD; 4Maila Push, Darmstadt, Germany
{baudisch, cutrell,czerwinski, dcr};;;
Abstract: Drag-and-pop and drag-and-pick are interaction techniques designed for users of pen- and touchoperated display systems. They provide users with access to screen content that would otherwise be impossible
or hard to reach, e.g., because it is located behind a bezel or far away from the user. Drag-and-pop is an extension of traditional drag-and-drop. As the user starts dragging an icon towards some target icon, drag-and-pop
responds by temporarily moving potential target icons towards the user’s current cursor location, thereby allowing the user to interact with these icons using comparably small hand movements. Drag-and-Pick extends the
drag-and-pop interaction style such that it allows activating icons, e.g., to open folders or launch applications. In
this paper, we report the results of a user study comparing drag-and-pop with traditional drag-and-drop on a 15’
(4.50m) wide interactive display wall. Participants where able to file icons up to 3.7 times faster when using the
drag-and-pop interface.
Keywords: Drag-and-drop, drag-and-pick, interaction technique, pen input, touchscreen, heterogeneous display.
With the emergence of pen- and touch-operated personal digital assistants (PDAs), tablet computers,
and wall-size displays (e.g., Liveboard, Elrod et al.,
1992; Smartboard,,
touch and pen input have gained popularity. Over
the past years, more complex display systems have
been created by combining multiple such display
units. Wall-size touch displays have been combined
into display walls, such as the DynaWall (Streitz
2001), or the iRoom Smartboard wall (Johanson,
2002b). Recent PDAs and tablet computers allow
connecting additional displays, such as another tablet or a monitor in order to extend the device’s internal display space.
Touch/pen-operated screens that consist of multiple display units bring up a new class of input challenges that cannot always be solved with existing
techniques, because many of the existing techniques
were designed for indirect input devices, such as
mice, track pads, or joysticks. Indirect input devices
can be used on arbitrary display configurations, because they can simply be mapped to the respective
topology (e.g., PointRight, Johanson 2002a). Touch/
pen input, however, is based on the immediate
Figure 1: Drag-and-pop
correspondence between input space and display
space and thus requires users to adapt their input
behavior to the physicality of the display system.
Here are three examples where this can become
Scenario 1: External monitors. One or more display units within a display system may not be
equipped with a touch or pen sensor. Connecting an
external monitor to a tablet computer or PDA, for
example, allows users to see more material, but requires them to use an indirect input device, such as a
mouse, when interacting with content on the external
monitor. Since some tablet-specific tasks, such as
scribbling, are hard to accomplish with a mouse,
users find themselves continuously switching between pen and mouse.
Scenario 2: Interactions across display units.
Some interaction techniques, such as drag-and-drop,
require users to interact with two or more icons in a
single pen-down interaction. If these icons are distributed across physically separate pen/touch input
display units, users first have to bring all involved
icons to the same display unit, a potentially timeconsuming activity (Figure 2a-c).
Scenario 3: Bridging long distances. Accessing
icons located far away from the user, e.g., on the
opposite side of a 15’ DynaWall, requires users to
physically walk over, the time for which may in
some circumstances increase linearly with distance
(Guiard et at, 2001). In addition, drag interactions
get more error-prone with distance, because users
drop objects accidentally when failing to continuously keep the pen tip in contact with the display
surface (Rekimoto 1997).
Drag-and-pop & drag-and-pick
Drag-and-pop and drag-and-pick are interaction
techniques that address these issues. We will begin
by giving an overview; more detailed descriptions of
both techniques can be found in Section 4.
Drag-and-pop extends traditional drag-and-drop
as illustrated by Figure 1. (a) The user intends to
delete a Word memo by dragging it into the recycle
bin. (b) As the user starts dragging the memo’s icon
towards the recycle bin, icons that are of compatible
type and located in the direction of the user’s drag
motion “pop up”. This means that for each of these
icons a link icon is created (tip icon) that appears in
front of the user’s cursor. Tip icons are connected to
the original icon (base icon) using a rubber band.
(c) The user drags the memo over the recycle bin
and releases the mouse button. The recycle bin accepts the memo. Alternatively, the user could have
dropped the memo over the word processor or the
web browser icon, which would have launched the
respective application with the memo. (d) When the
user drops the icon, all tip icons disappear instantly.
Figure 2d shows how drag-and-pop simplifies
dropping icons onto targets located at the other side
of a bezel that separates display units (scenario 2).
Figure 9 shows a user performing a drag-and-pop
interaction to drop an icon on a distant target.
Figure 2: (a-c) Traditional drag-and-drop: Dragging
an icon across the bezel requires the user to drop the
icon half way across the bezel and pick it up at the
other side (d) Drag-and-pop temporarily brings matching target icons to the current pen location, allowing
the user to file icons without having to cross the bezel.
Drag-and-pick modifies the drag-and-pop interaction concept such that it allows activating icons,
e.g., to open a folder or to launch a program. While
drag-and-pop is initiated by the user dragging an
icon, drag-and-pick starts with the user performing a
drag interaction on empty screen space. The system’s response to this drag interaction is similar to
drag-and-pop, but with two differences. First, all
icons located in the direction of the drag motion will
pop up, not only those of compatible type (Figure 3).
Second, as the user drags the mouse cursor over one
of the targets and releases the mouse button, the
folder, file, or application associated with the icon is
activated as if it had been double clicked.
Figure 4 shows how this allows users to use the
pen for launching an application, the icon of which
is located on a monitor not supporting pen input.
In principle, drag-and-pick can be applied to any
type of widget, e.g., any buttons and menus located
on a non-pen accessible monitor. In this paper, however, we will focus on the manipulation of icons.
Related work
Drag-and-drop is a well-know interaction technique
for transferring or copying information using a
pointing device, while avoiding the use of a hidden
clipboard (Wagner, 1995; Beaudouin-Lafon, 2000).
Hyperdragging (Rekimoto, 1999), allows extending
drag-and-drop across physically separate displays
(Scenario 2), but requires an indirect input device,
such as a mouse. Most techniques compatible with
pen usage are based on point-and-click, e.g., pickand-drop (Rekimoto, 1997) and take-and-put (Streitz
et al., 2001). These techniques, however, cannot be
used to access material on a display unit not providing pen support (Scenario 1).
Figure 3: Drag-and-pick makes all icons in the direction of the mouse motion come to the cursor.
users to accelerate an object with a small gesture; the
object then continues its trajectory based on its inertia (Geißler, 1998). The imprecision of human motor skills has prevented throwing from being used
for reliable target acquisition. Myers et al. (2002)
used laser pointers to acquire targets on a Smartboard, but found them to be slower than touch input.
A variety of mouse-based interaction techniques
use destination prediction to simplify navigation
(e.g., Jul, 2002). Dulberg et al. (1999) proposed a
flying click or flick for snapping the mouse to target
locations. Swaminathan and Sato (1997) proposed
making relevant controls on the screen “sticky”.
As an alternative way of launching applications,
today’s operating systems offer menus containing
lists of available application or documents. A ‘send
to’ option (Microsoft Windows) allows sending an
icon to a target selected from a predefined list.
Compared to 2D desktops, which typically use a
larger amount of screen space than pull-down or
pop-up menus, menus are limited to a smaller selection of choices unless they use a hierarchical menu
organization, which makes their usage less transparent and often less efficient. Furthermore, invoking a
content-menu may require hitting a qualifier key,
which can be problematic on touch-based systems.
Design and algorithms
In this section, we will take a more detailed look at
the design and algorithms behind drag-and-pop/pick.
4.1 Selecting candidates
Figure 4: Drag-and-pick allows users to temporarily
move icons from an external monitor to the tablet
where the user can interact with them using the pen.
A different set of interaction techniques have
been proposed to help users overcome large distances (Scenario 3). Manual And Gaze Input Cascaded (MAGIC) pointing (Zhai et al., 1999) uses eye
tracking to move the cursor to the target area, from
where the user guides the cursor manually (which
requires an indirect input device). Gesture input
techniques allow selecting a target and a command
in a single interaction and are generally compatible
with pen input (Rubine, 1991). ‘Throwing’ allows
In order to reduce clutter, drag-and-pop creates tip
icons only for a subset of the icons on the screen.
Drag-and-pop’s candidate selection algorithm is
initialized with the entire set of icons on the screen;
it then successively eliminates candidates using the
following four rules.
First, icons of incompatible type are eliminated.
If the user drags a text file, the icon of a text processor can create a tip icon; the recycle bin icon can
create a tip icon; the icon of another text file, however, cannot, because dragging two text files onto
each other is usually not associated with any behavior. Drag-and-pick bypasses this selection step in
order to allow users to activate any type of icon.
Second, icons located between the cursor and the
location where the tip icons cluster will appear (see
following section) are eliminated. This rule avoids
creating tip icons that move away from the cursor.
Third, only icons that are located within a certain
angle from the initial drag direction (the target sector) are considered. The initial drag direction is determined the moment the user drags an icon further
than a given threshold (default 15 pixels). During
preliminary testing on a Smartboard, we got good
results with first-time users when using sector sizes
of ±30 to ±45 degrees. The sector size could be reduced to sector sizes of ±20 degrees as users gained
more experience.
Forth, if the number of qualifying icons is above
some hard limit, drag-and-pop eliminates tip icon
candidates until the hard limit is met. Icons are removed in an order starting at the outside of the target
sector moving inwards. This rule assures the scalability of drag-and-pop to densely populated displays, but requires drag-and-pop users working with
densely populated screens to aim more precisely.
We typically use hard limits between 5 and 10.
4.2 Computing the tip icon layout
Once tip icon candidates have been selected, dragand-pop determines where on the screen to place the
tip icons. In order to avoid interference between tip
icons, the location of all tip icons is computed in a
centralized fashion.
Our drag-and-pop prototype uses the following
algorithm that is illustrated by Figure 5: (1) Snap
icons to a grid and store them in a two-dimensional
array, with each array element representing one cell
of the grid. If two or more icons fall into the same
cell, refine the grid. (2) Shrink the icon layout by
eliminating all array columns and rows that contain
no icons. (3) Translate icon positions back to 2D
space by mapping the array onto a regular grid. By
default, the output grid is chosen to be slightly
tighter than the input grid, which gives extra compression.
Figure 5: Drag-and-pop computes tip icon layouts
(a) by snapping icons to a grid and then (b) removing
empty rows and columns.
We chose this algorithm, because it preserves
alignment, proximity, and spatial arrangement between icons, which allows users to use their spatial
memory when identifying the desired target within
the tip icon cluster. This is especially useful when
tip icons look alike (e.g., a folder in a cluster of
folders). In order to help users distinguish local icon
clusters from surrounding icons more easily, the
algorithm may be adjusted to shrink empty rows and
columns during layout computation instead of removing them entirely.
After the tip icon layout has been computed,
drag-and-pop positions it on the screen such that the
center of the layout’s bounding box is located at the
direct extension of the user’s current mouse motion.
The distance of the tip icon cluster to the user’s current cursor position is configurable. For inexperienced users, we got best results with distances of
around 100 pixels; shorter distances made these users likely to overshoot the cluster. For more experienced users, we were able to reduce the distance to
values around 30 pixels, which allowed these users
to operate drag-and-pop with less effort, in a more
“menu-like” fashion. In order to reduce visual interference between tip icons and icons on the desktop,
drag-and-pop diminishes desktop icons while tip
icons are visible.
4.3 The rubber band
When the tip icon cluster is displayed, users need to
re-identify their targets within the tip icon cluster in
order to be able to successfully acquire them.
Our first implementation of drag-and-pop created
tip icons on top of their bases and used slow-inslow-out animation (Shneiderman 1998) to move tip
icons to their final location. While this approach
allowed users to locate the final position of the desired tip icon by visually tracking it on its way from
basis to final position, it also required users to either
wait for the animation to complete or to acquire a
moving target. We therefore chose to abandon the
animation and immediately display tip icons at their
final destinations.
In lieu of the animation, we provided tip icons
with rubber bands. The design prototype of the rubber band is shown in Figure 6. For performance reasons, our prototype, which is shown in all other
screenshots, uses rubber bands of a lower level of
graphical detail, i.e., a tape and three lines in the
color scheme of the corresponding icon.
The purpose of the rubber band is to offer the
functionality of the animation, but without the problems alluded to above. The rubber band, decorated
with the respective icon’s texture, can be thought of
as having been created by taking a photograph of the
tip icon animation with a very long shutter speed
(so-called motion blur, e.g., Dachille and Kaufman,
2000). Like the animation, the rubber band allows
users to trace the path from base to tip icon. However, users can do this at their own pace and the customized texturing of the rubber band allows users to
start tracing it anywhere, not only at the base.
The rubber band is provided with a narrow midriff section, suggesting that the rubber band is elastic. This design was chosen to help users understand
that tip icons have retracted to their bases when they
disappear at the end of the interaction. This feature
may also help users find their way to the tip icon
faster, because it provides users with a visual cue
about how far away the tip icon is located. A thick
rubber band section implies that the tip icon (or
base) is close; a thin rubber band section indicates
that the target is further away.
motion segment moving away from the tip icons,
thus terminating tip icons as soon as they appear.
The algorithm: the tip icon cluster is kept alive as
long as at least one of the following three rules is
successful. The first rule checks whether the mouse
cursor has moved closer to the center of at least one
of the icons in the tip icon cluster. This rule makes
sure that the cluster does not disappear while users
approach their targets. The second rule checks if the
cursor is in the direct vicinity of an icon. This rule
provides tolerance against users overshooting a tip
icon while acquiring it. The third and last rule keeps
the cluster alive if the cursor is stationary or if it is
moving backwards very slowly (up to 5 pxl/frame).
This rule makes drag-and-pop insensitive to jitter.
Figure 7 illustrates the resulting behavior.
Figure 6: The motion blur textures on the rubber
bands that connect tip icons with their bases are made
by overlaying skewed copies of that icon.
Figure 7: The tip icon cluster is kept alive as long as
the user moves towards the cluster (arrows) or inside
the convex hull surrounding the cluster (dashed).
4.4 Aborting drag-and-pop interactions
As soon as tip icons and rubber bands are shown on
the screen, drag-and-pop waits for the user to acquire one of the tip icons to complete the ongoing
drag-and-pop or drag-and-pick interaction. There are
two cases, however, in which users will want to
abort the interaction without acquiring a tip icon.
The first case is when the user dragged the
mouse at a wrong angle so that the desired target
icon did not pop up. In this case, the user may either
drop the icon and try again or complete the interaction as a regular drag-and-drop interaction, i.e., by
dropping the icon onto the target icon’s base instead.
The other case occurs if the user is intending to
perform a regular mouse drag operation, for example
to rearrange icons on the desktop or to capture a set
of icons using a lasso operation. For these cases,
drag-and-pop allows users to terminate tip icons onthe-fly and to complete the interaction without dragand-pop/pick. To abort, users have to move the
mouse cursor away from the tip icon cluster while
still keeping the mouse depressed. This can be done
by overshooting the cluster or by changing mouse
direction. In particular, this allows users to access
the underlying drag-and-drop and lasso-select functionality by introducing a simple zigzag gesture into
their cursor path. The zigzag contains at least one
User study
In this section, we report the results of a user study
comparing drag-and-pop with the traditional dragand-drop technique. To examine the effects of bezelcrossing as well as distance, as described in Scenarios 2 and 3, we chose to run the study on a tiled
wall-size display. During the study, in which participants filed icons into folders or dragged them
onto the icons of matching applications, we recorded
the time and accuracy of these movements. Our
main hypothesis was that participants would perform
faster when using the drag-and-pop interface, primarily because it would avoid the need for crossing
the bezels, but also because it would bridge the
space to very distant icons more efficiently.
5.1 Desktop layout
To obtain a representative set of icon arrangements
for the study, we gathered desktop screenshots from
25 coworkers who volunteered their participation
(15 single, 6 dual, and 4 triple monitor users). Overall resolutions ranged from 800,000 pixels to
3,900,000 pixels (66% more than the display wall
used in the experiment).
We clustered the obtained desktops by number
of icons and arrangement pattern. Then we chose
representatives from each of the three resulting main
clusters for the study (Figure 8). The “sparse” desktop reflected the desktops of roughly two thirds of
the participants. It contained only 11 icons, most of
which were lined up in the top left corner of the
screen. The “frame” desktop reflected the desktops
of three of the participants. It contained 28 icons
arranged around the top, left, and right edge of the
screen. The “cluttered” desktop, finally, contained
35 icons that were spread primarily across the top
and left half of the screen. Five participants had chosen this style of arranging their icons.
Icon layouts were stretched to fit the aspect ratio
of the display wall used in the experiment. An area
at the bottom right of the screen was reserved for the
starting locations of the icons to be filed during the
study (dashed shape in Figure 8).
(1.12m) high. Display units could be operated by
touching the display, but for easier handling participants were provided with color-free felt pens. Each
of the three display units ran at a resolution of
1024x768 pixels, offering an overall resolution of
3072x768 pixels. The three display units were connected to a single PC equipped with two Matrox
Millennium graphics cards and running WindowsXP. During the experiment, the DynaWall ran
a simulated Windows desktop. We compared dragand-pop to a control condition of drag-and-drop.
Since our preliminary Windows-based version of
drag-and-pop did not support the full functionality
required for the study, we implemented a simulation
using Macromedia Flash (
The drag-and-pop interface used in the experiment
was configured to a ±30 degree target sector, 35
pixel target distance, and a maximum number of 5
tip icons.
Figure 8: The (a) sparse, (b) frame, and (c) cluttered
desktop layouts used in the study. The dashed line indicates the space reserved for the icons users had to file.
Boxes around icons indicate icon to be filed and target.
5.2 Participants
Eight colleagues with no experience using drag-andpop were recruited for this experiment. Due to technical problems, the data from one of these participants had to be dropped leaving us with 7. There
were 2 female and 5 male participants ranging in age
between 18 and 35. All were right handed with normal or corrected-to-normal vision.
5.3 Method
The test was run on the DynaWall (Streitz, 2001), a
display wall consisting of three Smartboard units
(Figure 9). Each Smartboard consisted of a backprojected 72”display with resistive touch input, so
that the entire display was 15’ (4.50m) long and 45”
Figure 9: DynaWall setup used in user study
To each desktop layout we added 10 document
icons in the lower right quadrant of the screen.
These appeared in six different arrangements (Figure
8 shows 2 of them). The participants’ task was to
drag these icons into a given target folder or application. Icons of image files, for example, were to be
filed in a folder labeled “My Pictures” and all Word
documents should be dropped onto the Word application. To counterbalance for order effects, we required participants to file the documents in a randomized order. That is, for each movement, the item
to be filed was highlighted along with the target
icon. All other document icons were frozen, so that
participants could only move the highlighted icon.
As soon as participants began moving an item, all
highlighting was removed, forcing participants to
remember the destination item. We did this to assure
that participants would have to re-identify tip icons
when using the drag-and-pop interface, just as they
would in a real-world task.
Participants were allowed several minutes to
practice moving and filing icons in the prototype to
get them accustomed to both the DynaWall display
and the drag-and-pop interface. Once it was clear
that users understood how to use the display and the
interfaces, they were allowed to go on to the study.
Participants filed 2 sets of icons for each interface
(drag-and-pop and control), for each of the three
desktops. Thus participants filed 2 x 10 icons x 2
interface x 3 desktops for a total of 120 movements.
To mitigate learning effects associated with new
desktop arrangements or a new interface, we omitted
the first 5 trials for any desktop-interface combination from our analyses, yielding ~15 correct trials
per cell or 90 movements per participant.
5.4 Results
Task performance
Task performance was evaluated through speed and
accuracy measurements. Error rates were considerably larger for drag-and-pop than for the control
(6.7% vs. 1%). We observed two things that made
this type of error more likely in the drag-and-pop
condition. First, in the drag-and-pop condition candidate targets were brought closer together, making
it easier to accidentally drop an item on the wrong
target. Second, because drag-and-pop targets had
been translated away from their “home” location,
participants would sometimes forget which item was
in fact the target, especially if visually similar icons
(e.g., other folders) had created tip icons as well.
All data analyses for movement times were performed on the median movement times for each participant in each condition to normalize the typical
skewing associated with response time data. Summary statistics report the means of these times.
Target icons could be located in the same display
unit as the icon to be filed, in a neighbor display
unit, or in the display unit at the other end of the
display wall, requiring users to cross 0, 1, or 2 bezels in order to file the icon. To test the effect of
bezel crossing on performance, we ran a 2 (Condition) x 3 (Bezels Crossed) within subjects ANOVA
on the median movement data. This revealed a significant main effect for condition, F(1,6) = 18.2,
p<0.01. Collapsed across all distances, drag-and-pop
was significantly faster than the control. There was
also a significant main effect of bezels crossed,
F(2,12) = 19.5, p<0.01; movement time increased as
the number of bezels participants had to cross to get
to the target icon increased. As hypothesized, we
also saw a significant interaction between condition
and number of bezels crossed, F(2,12) = 15.2,
p<0.01. As seen in Figure 10, an increase in the
number of crossed bezels resulted in only a small
increase in movement time for drag-and-pop,
whereas it had a huge effect for the control interface.
When no bezels had to be crossed, drag-and-pop
appeared to be slightly slower than control, although
follow-up t-tests showed that this difference was not
significant, t(6)=1.73, ns. When 1 or 2 bezels had to
be crossed, drag-and-pop was significantly faster
than drag-and-drop (t(6)=4.02, p<0.01 & t(6)=4.12,
p<0.01, respectively). With 1 bezel crossed, dragand-pop was twice as fast as the control and with 2
bezels it was 3.7 times as fast.
Number of Bezels Crossed
Figure 10: Mean movement time for control and dragand-pop interfaces (± SEM).
Figure 11 shows a scatter plot of movement time
versus target distance for both conditions. The best
linear fit for drag-and-drop was f(x)=0.007x-1.76,
r2=0.23. The linear fit for drag-and-pop was
f(x)=4.19, r2<0.0001. This reinforces what can be
seen in Figure 10—movement time increases with
distance for the control interface, but stays relatively
constant for the drag-and-pop interface.
Target Distance (pixels)
Figure 11: Movement time vs. target distance.
Questionnaire and subjective feedback
At the end of the study, participants answered a
short questionnaire about their experience using the
DynaWall and drag-and-pop. Participants were very
enthusiastic about drag-and-pop. On a 7 point Likert
scale (where 7=strongly agree and 1=strongly disagree), there was a mean > 6 for questions such as,
“I liked using drag-and-pop”, “I always understood
what was happening when drag-and-pop was on,”
and “I would use drag-and-pop for large displays.”
There was a mean of less than 3 for “It took a long
time to get used to drag-and-pop” and “It was hard
to control what the targets did when drag-and-pop
was on.” Participants reported the drag-and-pop
interface to cause less manual stress and fatigue than
the control interface.
The most common problem with drag-and-pop
was in getting the right group of targets to pop up,
and several participants requested a wider angle for
destination targets. This relates to an observation we
made about how people interact with touch-sensitive
wall-displays. On the wall display, participants had
to employ their whole arm to make a movement,
resulting in targeting motions in the shape of arcs.
This means that the initial direction of the movement
was not in the direction of the target. To accommodate such arcs in the future, we have adapted the
target selection algorithm of drag-and-pop by giving
the target sector extra tolerance for movements towards the top of the screen.
Conclusions and future work
The substantial time-savings found in the user study
confirm our expectations. Although when used
within a single screen unit drag-and-pop does not
seem to by faster than traditional drag and drop (first
pair of bars in Figure 10; drag-and-pop’s capability
of bridging distance to the target seems to be nullified by the need for re-orientation), its advantages
on very large screens and its capability of bridging
across display units are apparent. On the usability
side, we were glad to see that participants had no
trouble learning how to use the technique and that
they described the technique as understandable and
predictable. The single biggest shortcoming, the
target selection, is the subjects of current work. In
addition to the changes described above, we consider dropping the notion of a fixed target sector size
and replace it with a mechanism that adjusts the sector size dynamically based on the number of matching targets.
Given the recent advent of commercially available tablet computers, our next step will be to explore how drag-and-pop and especially drag-andpick can help tablet computer users work with external monitors. While this paper focused on icons, we
plan to explore ways of operating menus, sliders,
and entire applications using the techniques described in this article.
Thanks to Diane Kelly, Dieter Böcker, Lance Good,
Amanda Williams, and LDUX. Work done at IPSI
was supported by a grant from Microsoft Research.
Beaudouin-Lafon, M. (2000) Instrumental Interaction: An
Interaction Model for Designing Post-WIMP Interfaces. In Proc. CHI ’00. pp. 446–453.
Dachille, F. and Kaufman, A. (2000) High-Degree Temporal Antialiasing. In Proc. Computer Animation
‘00, pp. 49-54.
Dulberg, M.S., St. Amant, R., and Zettlemoyer, L.S.
(1999), An Imprecise Mouse Gesture for the Fast
Activation of Controls. In Proc. Interact ’99, pp.
Elrod, S., et. al. (1992) Liveboard: a large interactive display supporting group meetings, presentations, and
remote collaboration. In Proc. CHI'92, pp. 599-607.
Geißler, J. (1998) Shuffle, Throw or Take It! Working
Efficiently with an Interactive Wall. In CHI '98
Late–Breaking Results, pp. 265–266.
Guiard, Y., Bourgeois, F., Mottet, D. & Beaudouin-Lafon,
M. (2001) Beyond the 10-Bit Barrier: Fitts’ Law in
Multiscale Electronic Worlds. In Proc IHM-HCI
2001, pp. 573-587.
Johanson B., Hutchins, G., Winograd, T., and Stone, M.
(2002a) PointRight: Experience with Flexible Input
Redirection in Interactive Workspaces. In Proc.
UIST ’02, pp. 227–234.
Johanson, B., Fox, A., and Winograd, T. (2002b) The
Interactive Workspaces Project: Experiences with
Ubiquitous Computing Rooms, IEEE Pervasive
Computing, 1 (2), 67-74.
Jul, S. (2002) Predictive targeted movement in electronic
spaces. In CHI’02 Extended Abstracts, pp.626-627.
Myers, B., Bhatnagar, R., Nichols, J., Peck, C., Kong, D.,
Miller, R., and Long, C. (2002) Interacting At a Distance: Measuring the Performance of Laser Pointers
and Other Devices. In Proc CHI’02, pp 33-40.
Rekimoto, J. and Saitoh, M. (1999) Augmented Surfaces:
A Spatially Continuous Work Space for Hybrid
Computing Environments. In CHI'99, pp. 378-385.
Rekimoto, J. Pick–and–Drop: A Direct Manipulation
Technique for Multiple Computer Environments. In
Proc. UIST'97, pp. 31–39, 1997.
Rubine, D. (1991) Specifying Gestures by Example. In
Proc. Siggraph’91, pp. 329-337.
Shneiderman, B. (1998) Designing the User Interface:
Strategies for Effective Human-Computer Interaction. Third edition. Reading MA: Addison-Wesley.
Streitz, N.A., Tandler, P. Müller-Tomfelde, C., Konomi, S. (2001) Roomware. In: Carroll, J.A. (Ed.),
Human-Computer Interaction in the New Millennium, Addison Wesley, pp. 553-578.
Swaminathan, K. and Sato, S. (1997) Interaction design
for large displays. In Interactions 4(1):15 – 24.
Wagner, A., Curran, P., O'Brien, R. (1995) Drag me, drop
me, treat me like an object. In Proc CHI ’95. pp.
Zhai, S., Morimoto, C. Ihde, S. (1999) Manual And Gaze
Input Cascaded (MAGIC) Pointing. In Proc. CHI
’99, pp. 246-253.
Download PDF