SOLVING GEOSPATIAL TASKS IN 2D AND 3D VISUALIZATIONS OF MAPS - Information Visualisation: Techniques, Usability and Evaluation - page 242

In 3D, maps were virtually slanted out, off the plane of the display (55 ͼ
slant angle) and rendered using stereoscopic off-axis projections based on
current observer positions, as acquired with a camera-based tracking
system. This allowed the map to be perceived correctly according to the
head movement of the participants, as well as allowing for the
visualization of maps with different perspectives.
A visual stimulus consisted of 10 bars with varying positions within
the map, from which the subjects had to search and identify the pair with
the largest height difference in relation to their distance in the map. Fig.
12.8 shows one stimulus configuration visualized in 2D and 3D. Upon
deciding on a candidate pair of bars, participants registered their responses
by clicking with the mouse on the bottom of the two bars that they
identified as the best candidates (i.e., biggest height difference in relation
to distance). To avoid incorrect parallax in the 3D visualization, the 2D
mouse cursor was replaced with a 3D cross-hair moving within the slanted
map.
Experimental design
The experiment followed a within-subject design, meaning that all
participants solved the task in both visualization conditions (2D and strong
3D). One trial, i.e., one repetition of the experiment, comprised the
assessment of 25 stimuli. Subjects performed two trials and hence they
viewed 25 stimuli in the 2D condition, and viewed the same 25 stimuli in
the 3D condition. To avoid carry-over effects between repeated trials, the
order of visualization conditions (2D-3D or 3D-2D) was determined
randomly and balanced so that 50% of the test population started with each
of the two conditions.
To evaluate potential learning effects within each trial (and visualization
condition), stimuli for each trial were grouped into two blocks, with block
1 comprising the first 13 stimuli, and block 2 containing the last 12
stimuli. Since the 25 different stimuli had varying degrees of difficulty,
their order of appearance within each trial was randomly assigned to
counterbalance potential differences in difficulty between the blocks. This
implies that subjects saw the stimuli in different orders.
Test procedure
The same experimental apparatus as for experiment 1 was used; that is, the
3D condition in this experiment provided both stereoscopic and dynamic
perspective cues (strong 3D). A different group of participants was