Geoscience Reference
In-Depth Information
TABLE 5.2
Ranking of Elementary Perceptual Tasks according
to Statistical Scale after Mackinlay (1986)
Quantitative
Ordinal
Nominal
Position
Position
Position
Length
Grey saturation
Colour hue
Angle
Colour saturation
Texture
Slope
Colour hue
Connection
Area
Texture
Containment
Volume
Connection
Grey saturation
Grey saturation
Containment
Colour saturation
Colour saturation
Length
Shape
Colour hue
Angle
Length
Texture
Slope
Angle
Connection
Area
Slope
Containment
Volume
Area
Shape
Shape
Volume
Note:
Items in bold are not applicable.
5% of males are also red-green colour-blind), so using hue to highlight subtle differences between
two similar demographic datasets might be less effective than using positional offset - say by ver-
tical displacement - or movement (animation). To complicate things further, human perception
of visual variables is non-linear but may follow a known activation function (e.g. Robertson and
O'Callaghan, 1988). For example, humans are particularly poor judges of the comparative magni-
tudes of both areas and volumes, though we seem to underestimate them in a known and consistent
manner!
The total bandwidth available across all visual variables sets a theoretical upper bound on the
amount of information that can be conveyed concurrently. However, even within this limit, the use
of a greater number of visual variables to encode more data attributes does not necessarily increase
the effectiveness of a scene, since many combinations of visual variables are known to interfere
with each other (e.g. colour
intensity
and
transparency
tend to cancel each other out so using them
together will likely add confusion, not information). The use of such knowledge in the construction
of a scene is described in the following.
5.5.2 e
xaMPle
of
r
ePeated
V
iSual
e
ncoding
for
e
xPloration
: c
horoPleth
M
aPPing
One of the most widely used techniques in GeoViz is an exploratory form of choropleth mapping,
whereby a succession of different classification schemes are essentially applied to the data, to see
if useful structure appears to have been imposed, or with a view to uncovering some pattern that
was hitherto unknown. For example, an obvious spatial clustering of geographical regions, to which
the same colour has been applied by the classifier, may become apparent. One might regard each
such test as being a kind of hypothesis, which is quickly posited and then withdrawn if it proves
unfruitful (see Section 5.6). Visual classification schemes are imposed on data by using grouping
or binning methods (which most GeoViz systems provide), by amplifying some differences in the
data, while holding others constant. For example, we may choose to organise geographical regions
into five classes based on unemployment levels. These classes are then reflected into the display in
some way that visually differentiates them, by assigning to them different visual variables; colours
would be used in choropleth mapping, but other visual variables such as size could be used instead.
