A Comparison of the Affordances of a
Digital Desk and Tablet for
Architectural Image Tasks
Ame Elliott and Marti A. Hearst
102 South Hall
University of California, Berkeley
Berkeley, CA 94170
{ame,hearst}@sims.berkley.edu
The physical context of architectural design includes large workspaces, typically draft-
ing tables covered with piles of images and sketches. We are investigating if and how a
large computerized workspace can be integrated usefully into such a design environment.
To this end, we compared a large computerized desktop (digital desk) to a standard desk-
top computer and a small tablet environment for two typical architecture design tasks:
sketching and image sorting. For the sketching task, the participants’ preferences were
evenly divided between the digital desk and the tablet. For the image sorting task, the desk
was the least preferred environment, and produced significantly higher sorting times and
more mistakes. Investigation into the causes of this difference yielded several interesting
findings, including: the height of the participant was significantly associated with their
speed on the sorting task, the larger image size available on the desk compensated for its
poorer resolution in subjective preferences, and the quality of the alignment of the pen was
an important factor both for preference and scoring results in the sketching task. Highly
responsive pen input devices seem critical for user satisfaction not only for sketching,
but also for image sorting; the effects of large display spaces are difficult to isolate from
the limitations of input device. This paper elaborates on these findings and considers
the implications for the design of user interfaces for image manipulation, in particular
interaction techniques appropriate to using pen-input with large display surfaces.
1. Introduction
The physical context of architectural design includes large workspaces, typically
drafting tables covered with piles of images and sketches. We are investigat-
ing if and how a large computerized workspace can be integrated usefully into
such a design environment. Current computer support for architectural design
focuses primarily on the last stages of the design process, that is, in aiding the
production of drawings describing fully-designed buildings (Kalay and Carrara,
1994). Previous studies of architectural practice explore the organizational and
economic aspects of running an office without addressing architects’ activities
during the design process (Larson, Leon, and Bolick, 1983; Blau, 1984; Cuff,
1991; Stevens, 1998). The early phase of design in architectural work is generally
not well-understood, but recent results show that the process is creative, open-
Volume 0, Numb er 0
ended, and dynamic, consisting of retrieving and looking at images of precedents
and making sketches (Elliott, 2000). Making sketches while looking at an image
is one way architects determine the relevance of that particular image to a task
at hand. Sorting images into groups is another way architects gather design in-
formation from images. The categorization process helps clarify design intent and
allows relevant images to emerge via their relationship to members of a group. A
recent overview of visual databases in architecture reiterates the importance of
looking at images and sketching, but falls short of attempts to explore how these
databases could be accessed in the context of the professional practice of design
(Koutamanis, Timmermans, and Vermeulen, 1995).
Our ultimate goal is to devise computerized tools to better support architects
in their work, particularly in their use of images. Before evaluating new image
search interfaces in such an environment, we set out to determine if a large desk-
like computerized display is useful for simpler tasks. To help determine this, we
conducted a user study comparing the use of three different environments:
• A digital desk, a 35”x 26” display housed in a table the size of a drafting
table using pen-based input
• a tablet, an 8.5” x 11” LCD display with pen-based input; and
• a standard monitor with mouse-based input.
Several physical properties of the digital desk suggest it would be better than
the standard keyboard-mouse-monitor environment for the sketching and image
browsing tasks common to the early phases of architectural design. The desk
has a pen input device, which allows architects to take notes, select objects, and
sketch in a way that should be natural and comfortable. The large display size
lets users work with many images simultaneously, and the upright orientation of
the display allows architects’ to work standing, or seated on a high stool, just as
they often work in non-computerized environments.
Numerous research projects use large displays as part of an attempt to sup-
port non-computerized work practices in electronic environments, including the
DigitalDesk (Newman and Wellner, 1992), Tivoli (Pedersen et al., 1993), In-
foWall (Rekimoto and Saitoh, 1999), HoloWall (Rekimoto and Matsushita, 1997),
metaDESK (Ullmer and Ishii, 1997), i-LAND (Streitz et al., 1999), Flatland (My-
natt et al., 1999), and the Interaction Mural (Guimbretiere, Stone, and Winograd,
2001). The advantages of using a large display for direct manipulation are well
documented. Fitzmaurice and Buxton (1997) advocate using specialized input
devices for specialized tasks, suggesting that a configuration designed for archi-
tects’ image manipulation needs would be more effective than a one-size-fits all
setup of mouse and keyboard. These systems are intended to serve as either
relatively static artifacts of handwritten notes, or as dynamic capturing media
designed to support collaborative work. Our intention is to support tasks relating
to browsing, searching, and using large collections of information.
In this paper we report the results of a comparative usability study of ar-
chitects performing two tasks: image sorting and sketching, in each of the three
2
Elliott and Hearst Affordances of a Digital Desk
different environments. The results of the user study disprove our initial hypothe-
sis that the desk environment would be unilaterally more helpful for architectural
tasks than the tablet or the monitor. Instead, for qualitative measures, the desk
did only as well as the tablet for the sketching task and much worse than ei-
ther the tablet or the monitor for image sorting. Quantitative results show more
sorting mistakes on the desk than other environments and significantly longer
times to complete the image sorting task on the desk. Interestingly, we found
that the height of the participant was significantly associated with their speed
on the sorting task; shorter participants were significantly slower on the desk,
but not so on the other environments. The monitor and mouse were preferred
over both devices for the sorting task. This may be the result of a combination
of factors: the sorting application is designed to be used with the pointing acuity
of the mouse, and the monitor screen size strikes an acceptable balance between
having images large enough to view easily but not so large as to require large
head movements to see all the images.
Another interesting result was that the desk’s poorer screen resolution did
not register negatively with the participants. For subjective measures, the larger
image size apparently made up for the relatively poorer resolution.
Finally, the quality of the alignment of the pen was an important factor both
for preference and scoring results in the sketching task. The alignment of the pen
to the desk is worse than that of the pen to the tablet, due in part to parallax
errors increasing with display size. Architects who hold the pen at an angle while
drawing are particularly hampered by the alignment problem.
The remainder of this paper, after discussing related work, describes the user
study methodology, analyzes the results in more detail, highlights relevant related
work, and outlines the implications of the findings for the design of user interfaces
for image retrieval on large displays.
2. Methodology
2.1 Overview
The study consisted of two tasks: sorting sets of images and sketching a copy of
a line drawing. These tasks were designed to be more constrained, and thus more
easily evaluated, than the open-ended tasks common to the early phases of the
design process. We assumed that if one tool dominated the others in these tasks,
it would then make sense to evaluate it further with more open-ended tasks.
The sorting problem was a timed task in which participants were encouraged to
arrange images into a pre-determined rectangular layout as quickly as possible.
The sketching task was an untimed task in which participants were asked to copy
an image from a physical piece of paper. Three different computer environments
were used in a within-subjects design: a desktop PC with a mouse, a desktop PC
with an LCD tablet and stylus, and a digital desk with a stylus.
2.1.1 Equipment. The hardware specifications of each of three devices appear
in Table 1. For the image sorting task, participants used the slide sorter view of
3
Volume 0, Numb er 0
Monitor Tablet Desk
Input Device Logitech 3-button mouse Pen type 1 Pen Type 2
Display Monitor LCD Rear-Projected
Display Size 11” x 12” (43 x 30cm) 11” x 8.5” (28 x 21.6 cm) 31” x 26” (79 x 66 cm)
Resolution 1280 x 1024 1024 x 768 1280 x 1024
Pixels/Inch
2
6425 8411 1626
Images/Inch
2
.17 images .37 images .04 images
Image Size 2” x 1.4” 1.43” x 1.06” 4” x 2.9”
Figure 1
Comparison of devices used in the study.
Microsoft PowerPoint to order the images. Each display showed the same number
of images at a time (5 rows and 7 columns), and every set contained exactly 35
images.
1
For the sketching task, participants used Adobe Photoshop with the drawing
window open to full screen size. The cursor was set to draw as a thick black pen
with width equivalent to the line weight of the example sketch. No tool palettes or
other windows were visible, and participants were instructed to react as though
they were given a permanent black magic marker and a piece of blank paper.
They could start over as many times as they wanted, but could not erase any
marks.
Participants were given a workbook containing informed consent paperwork,
surveys, written versions of the oral instructions for the tutorial on how to use
the equipment, examples of the sketches to be copied, written versions of the oral
instructions for how to sort the images, and examples of correctly sorted sets of
images. Participants controlled their own workbooks throughout the study and
could refer to them at any point. During each sorting task participants were given
a large template for easy reference showing the correct order for the images as a
supplement to the workbook.
1 We increased the double-click speed to maximum in all conditions to avoid the undesirable effect
that occurs when a user inadvertently double-clicks an image (it changes the display instead of
selecting an image). During the tutorial participants were shown this possibility and instructed
about what to do if it should occur.
4
Elliott and Hearst Affordances of a Digital Desk
Device Order Number of Participants
777
First Tablet Desk Monitor
Second Monitor Tablet Desk
Third Desk Monitor Tablet
Tabl e 1
The experiment design.
2.2 Experiment Design
Twenty-one paid participants, 13 male and 8 female graduate students in architec-
ture, completed a within-subjects study. The procedure consisted of completing
informed consent paperwork, a background questionnaire, tutorial and ergonomic
adjustment on each device, the sketching task, a written survey and oral follow-up
questions about the sketching and image sorting tasks. The participants’ inter-
actions with the systems were recorded on videotape and with still photos. The
participants spent an hour and a half working one-on-one with an experimenter.
Ergonomic adjustment on the digital desk and tablet included offering the
participant three chairs of different heights and the option of standing, adjusting
the angle and height of the display surface so that the participant understood the
range of possible positions, and demonstrating two lighting options to reduce glare
and eye-strain. The adjustments for the monitor included only two chair options
and lighting options. The device order, derived from a Latin square design, was
the same for both sketching and sorting tasks. Table 1 below shows the experiment
design.
For the sketching tasks participants spent a few minutes getting familiar with
the environment and using the input device for sketching. After completing two
trial runs that were not used in analysis, participants sketched a set of six abstract
figures. The participants then moved to the other two devices to complete addi-
tional trial runs and sketch different groupings of abstract figures. The motivation
for sketching abstract figures is to limit the semantic implications of copying a
sketch of a famous building or landscape and isolating aspects of the mechanics
of sketching. Isolating the architects’ responses to the mechanics of each of the
pens is important because little is known about how the mechanics of pen input
for professional design tasks work; the bulk of the work with pen computing is as
a general purpose input device, e.g., (Greenstein and Arnaut, 1988; Kurtenbach
and Buxton, 1991; MacKenzie, Sellen, and Buxton, 1991; Accot and Zhai, 1999).
For the image sorting tasks, participants completed a tutorial consisting of
sorting numbers into ascending order (to allow them to become familiar with the
slide sorter view of PowerPoint for each display and input device). After complet-
ing the tutorial, the participant sorted three sets of images in each environment.
Each consecutive image set increased in complexity. The first image set consisted
of colorful shapes that the participants were asked to group into a pre-determined
order (for example, red triangles first, followed by yellow circles). The second set
of images consisted of sorting a mixed group of photographs by placing people
5
