Abstract
A tool for designing experiments, capturing data, and
building Fitts' law models is described. The software runs
on an IBM or compatible computer equipped with an
appropriate graphical display and selection device (e.g.,
mouse, joystick). Features intended for HCI educational
purposes or experimental research are included, making
this a very powerful utility for research in input techniques
or Fitts' law. The software is available via anonymous FTP
through the internet.
Keywords:
Fitts' law, mouse, input techniques, human
performance modeling, HCI education.
Introduction
Fitts' law is a highly successful relation that accurately
models human movement time [3]. It has attracted
increasing attention since its inception in 1954, and has
been widely researched and applied in human-computer
interaction (HCI) [2, 7].
It has been suggested that HCI remains "soft science" [8],
that research is short on mathematical models and theories.
Instead, there is an abundance of anecdotal "research"
(We built this new interface and users liked it!).
It is felt that HCI could benefit from "explicit computer
program tools ... The tools may be of any kind -
simulation, measurement, analysis or system building" [8].
This implies a role for tools to assist in the rapid
development of experimental software to test a variety of
input devices in diverse task settings. This poster presents
a tool for this purpose: the Generalized Fitts' Law Model
Builder. The program is highly extensible and supports
the capture and subsequent analysis of relevant data for
experiments using input devices (e.g., a mouse) in
point/select tasks. The software assists in building Fitts'
law models, or in capturing data for other analyses.
Applications include experimental research and HCI
education.
A Brief Introduction to Fitts' Law
Fitts' law can be stated as
(1)
where MT is the movement time to complete a task, ID is
the index of task difficulty (in bits), and IP is the index of
performance (in bits/s).
The index of difficulty was originally defined by Fitts as
(2)
where A is the movement distance or amplitude, and W is
the width of the target.
Other formulations have been proposed for the index of
task difficulty, including
(3)
proposed by Welford [9] and
(4)
introduced by MacKenzie [4]. See also [5, 6].
Fitts' Law Studies
The first study to apply Fitts' law to the mouse was by
Card, English and Burr in 1978 [1]. Since then, Fitts' law
has been applied to many other selection tasks, all of which
differ in small but significant ways. In all cases, the
researchers developed their own software to present tasks
to the subjects. Typically, for each trial, subjects
manipulate a cursor using a pointing device and select a
target using a button on the device. Data on the speed and
accuracy of the trial are captured and stored in a file. The
utility of the Generalized Fitts' Law Model Builder is its
built-in flexibility and configurability. This software can
be used for many such studies, without requiring the
experimenters to develop new software.
THE GENERALIZED FITTS' LAW MODEL BUILDER
The software allows many types of movement tasks to be
tested in experimental settings. To facilitate this, many
features are tunable through a setup screen and a
configuration file read by the program.
The most common task parameters are amplitude (distance
from starting position to the target), target width (horizontal
size), target height (vertical size), and approach angle (from
the starting position to the target. The experimenter can set
specific values or enable random values to be chosen.
The type of selection task is fully configurable, including
the trial repetition (serial versus discrete), whether the task
should be a pointing or dragging task, and how the user
indicates the beginning or ending of a trial.
The target may be configured as a circle, square, rectangle,
or a word chosen from text. It follows that the set of
tunable parameters is not orthogonal and some exclusions
apply (such as the inability to control target height and
width for square or circular targets).
The behavior of the equipment is configurable, including
C-D gain of the device, device type (mouse, joystick,
other), and type of feedback (e.g., beeps upon an erroneous
movement). The aspect ratio of the display device is
corrected for, and all distances are measured in millimeters.
Lastly, the software supports the recording of external
factors which the experimenters may be interested in, such
as the age, sex, or previous computer experience of the
subjects.
The Setup Screen
All adjustable parameters are controlled by means of a
configuration file. This file uses a simple syntax and can
be edited, created, or modified with any DOS text editor.
Additionally, there is a setup screen through which all
parameters can be set. This allows the configuration file to
be altered using a friendly graphical user interface.
The Results Screen
After a block of trials, summary statistics and Fitts' law
models are displayed in graphic and text forms. The results
are tentative at best, since a single block of trials for one
subject is inadequate to draw conclusions. However, the
immediate results are useful to educators or students
investigating Fitts' law.
Output Data File
Output data files contain the data measured by the software
for each session and subject. The data files are numbered,
and well commented to reduce the chances of confusing the
files during the follow-up analysis.
AVAILABILITY
The Generalized Fitts' Law Model Builder is available
through anonymous FTP at our local site
(snowhite.cis.uoguelph.ca) to anyone with
internet access. The software is in the directory
pub/fitts-law/gflmb. The filename is
gflmb_xx.exe, where xx is the version number. This
directory also contains a file called GFLMB-README
which contains information relevant to the installation and
operation. The file is a self-extracting archive file, which,
when executed on DOS, creates files containing the
software, a sample configuration file, a users' manual, and
any further instructions pertaining to the software. Further
inquiries regarding the software should be directed to the
second author.
CONCLUSION
The Generalized Fitts' Law Model Builder is useful to HCI
educators and students alike. The ease of use and instant
data analysis are valuable features of this teaching aid.
At the same time, the Generalized Fitts' Law Model Builder
is a robust tool with enough flexibility and rigor for use in
serious experimental research in input devices and
interactive techniques for human-computer interfaces.
References
1. Card, S. K., English, W. K., and Burr, B. J. Evaluation of
mouse, rate-controlled isometric joystick, step keys and text
keys for text selection on a CRT. Ergonomics,
21 (1978), 601-613.
2. Card, S. K., Mackinlay, J. D., and Robertson, G. G. The
design space of input devices. Proc. of the CHI '90
Conference on Human Factors in Computing
Systems, 1990, pp. 117-124.
3. Fitts, P. M. The information capacity of the human motor
system in controlling the amplitude of movement.
Journal of Experimental Psychology, 47
(1954), 381-391.
4. MacKenzie, I. S. A note on the information-theoretical
basis for Fitts' law. Journal of Motor Behavior,
21 (1989), 323-330.
5. MacKenzie, I. S. Fitts' law as a research and design tool
in human-computer interaction. Human-Computer
Interaction, 7 (1992), 91-139.
6. MacKenzie, I. S. Movement time prediction in human-
computer interfaces Graphics Interface '92,
1992, pp. 140-150.
7. Marchionini, G., and Sibert, J. An agenda for human-
computer interaction: Science and engineering serving
human needs, SIGCHI Bulletin, 23, 4, (1991),
14-32.
8. Newell, A., and Card S. K. The prospects for
psychological science in human-computer interaction.
Human Computer Interaction, 1 (1985), 209-
242.
9. Welford, A. T. The measurement of sensory-motor
performance: Survey and reappraisal of twelve years'
progress. Ergonomics, 3 (1960), 189-230.
