SOLVING GEOSPATIAL TASKS IN 2D AND 3D VISUALIZATIONS OF MAPS - Information Visualisation: Techniques, Usability and Evaluation - page 248

Fig. 12.11. Results in the 3D visualization for groups experiencing visual
discomfort or not in 3D (a) and (b). Sub-figures (c) and (d) show results for the
groups of frequent game players and non-players.
Conclusion and discussion
The two experimental studies investigated potential benefits of a 3D map
representation for different types of relative assessment of quantitative
information in a geographic context.
The first experiment looked into a generic map-related visual task - the
identification of shortest distances between geographic positions in a map.
The results of this experiment suggest that accuracy of assessment in a
strong 3D visualization condition, as tested here, is not affected up to a
slant angle of 55 ͼ . This is a clearly higher threshold than the slant angle of
35 ͼ that was reported by Lind et al. [12] as a limit for the slant where
readability of symbols was affected in a similar type of visualization
employing weak 3D. Apparently, larger slant angles are tolerable with
sustained accuracy in a strong 3D visualization condition. It should,
however, be considered that mean accuracy in experiment I was generally
at fairly low levels for the stimuli tested in this experiment (according to
Fig. 12.5(a) about 0.6 to 0.75). Accuracy depends also on the degree of
difficulty. That is, for the task chosen in the first experiment, the ratio
between the shortest and next shortest distance must be less than 1.0, and
lower ratios will result in more reliable discrimination. In the strong 3D
visualization as investigated here, the above ratio can, with linear
regression, be predicted to be approximately 0.7 to assure approximately
100% correct identification of shortest distances. This ratio is somewhat
lower than ratios reported in similar studies. For example, in a study on
comparative assessment of the height of bars in 2D and 3D, Bleisch et al.
[18] found that errors occurred at a proportion of bar heights of 86%. It
should be noted though, that their experiment involved users in a two-
alternative forced choice test, where only two equally oriented bars had to
be compared, whereas the task in experiment I offered 120 potential
alternatives and randomly oriented distances.
The second experiment investigated whether more complex geo-spatial
assessment tasks benefit from visualizations using 3D mappings. The main
results of this experiment indicate neither improved accuracy nor shorter
task solution times for the 3D visualization condition. Although previous
investigations of similar, albeit simpler, geospatial tasks prefigure such
results [18][19], we expected to observe some clear differences in this
study.