20
resulted in the screen image moving in the opposite direction, giving the impression that the
view was actually moving over a stationary document.
5.1.3. Task
We adapted Fitts’ (1954) reciprocal tapping task for the purpose of this study. Participants had
to record as many hits as possible for a variable, pre-determined period of time (a trial) by
clicking alternatively on two colored discs. Once the crosshair was judged to be on one target,
validation had to be performed by clicking. A hit caused the current array of concentric circles
to be replaced at once by the alternate array centered on the other target (typically located a
long way away, out of sight). A miss was ineffective, and so an error always had to be
corrected before the next movement could start— therefore error rate was always 0%.
5.2. Experiment 1: Fitts’ Law Beyond the 10-Bit Barrier
In order to test the validity of Fitts’ law in multiscale, pan-and-zoom pointing, we focused in
this experiment on tasks of very high difficulty. Fitts’ law experiments usually have IDs in the
2-10 bit range (Soukoreff & MacKenzie, this issue). We presented our participants with a
selection of IDs ranging from 3.2 bits (D/W = 8) up to 30 bits (D/W = 109).
5.2.1. Procedure
The critical independent variable, manipulated in a within-participant design, was task
difficulty, quantified as ID =log2 (D/W + 1) (MacKenzie, 1992). We chose seven levels: 3.2,
5.1, 8, 12, 17, 23, and 30 bits (for a W of 1 pixel, these IDs corresponded to distances of 8, 32,
257, 4 097, 31 073, 8 388 608, and 1 073 741 825 pixels, respectively). Note that pointing
with an ID of 30 bits is equivalent to reaching and selecting a 1-inch target at a distance of
about half the circumference of our planet. At the first two levels of ID, zooming was not
necessary because at the outset of a trial the zoom could be preset so as to show both targets
on the screen with a size compatible with acquisition. For ID = 8 bits, the zoom began to be
very useful, and beyond this level of difficulty it became indispensable.
5.2.2. Results and Discussion
Zoom Utilization
For the first two levels of difficulty (ID = 3.2 and 5.1 bits), the zoom facility was
systematically ignored by the participants, who were content to carry out the task with just
panning. For higher levels of the ID, all the movements of all participants involved a zooming
component.
Using as our metric the output value of the zoom index (Z, which was made to vary between -
150 and +50), we defined zooming amplitude (AZ) as the difference between the zoom index
reached at hit time and the zoom index reached at the time of peak zoom-out (Zmin). As visible
in Figure 18, which takes into account only the five levels of ID where the zoom was used,
the relation between zooming amplitude and task difficulty was rigorously linear, with AZ =
6.6 ID - 14.8 (r2 = .997). The more difficult the task, the larger the zooming action.