Spatial Cues in 3D Visualization - Ambient Intelligence for Scientific Discovery - page 116

Results. The data was again analyzed by fitting a repeated measures MANOVA
model to the experimental observations. The MANOVA model tested each of
the four main effects (shadows on/off, viewing mode, number of shadow-casting
lights, and background) on the error magnitude and response time dependent
variables. The data from the positioning and resizing tasks were segregated and
analyzed separately. In both the positioning and resizing tasks performance data,
there were significant differences (at the 95% confidence level) in the mean values
of the dependent variables (error magnitude and response time) as a function of
all four main effects: use of shadows; viewing mode; number of shadowing lights;
and scene background. Complete quantitative results for both the positioning
and resizing tasks will be provided by requesting same from the lead author.
In the positioning task with shadows on, subjects were more accurate, al-
though they took longer to position the misplaced vertex object than when
there was no shadow present. Moreover, subjects viewing the vertex objects in
stereo were more accurate and faster performing the positioning task than they
were when viewing the virtual worlds in mono. Also, subjects viewing the vertex
objects shadowed by two lights sources were less accurate and took more time
than they were when viewing the vertex objects shadowed by one light source.
Subjects viewing the vertex objects in a room, or over a flat plane background
were more accurate than were subjects viewing the vertex objects over the stair-
step background. The error magnitude difference between the room and plane
background conditions was not significant. Furthermore, subjects viewing the
objects over a stair-step background took more time performing the positioning
task than did subjects viewing the objects in a room or over a flat plane back-
ground. The difference in the response times between the room and flat plane
scene background conditions was not significant.
In the resizing task with shadows on, there was no difference in resizing
accuracy, but subjects took longer resizing the vertex objects using shadows
than they did without the shadows. Furthermore, subjects viewing the scenes in
stereo were more accurate and faster than they were viewing the scenes in mono.
There was no difference in accuracy with shadows created by two lights compared
to one light source, although subjects viewing the objects shadowed by two light
sources took more time performing the resizing task than they did when viewing
the objects shadowed by one light source. Finally, the scene background also
significantly impacted subjects' resizing task performances, but only with respect
to resizing accuracy, not response time. Subjects viewing the vertex objects in a
room were more accurate than they were when viewing the objects over either
a flat plane or stair-step background condition. The differences in the response
times among the three scene background conditions were not significant.

3.3 Experiment #3:

Object Positioning and Resizing in Virtual Rooms

Method. The third experiment involved object positioning and resizing tasks
in 2D and 3D 'virtual rooms' [20]. These virtual rooms were bounded by four
surfaces: a floor, and left, right and rear walls. The objects to be manipulated