i
i
i
i
i
i
i
i
684 27. Visualization
Figure 27.6. The Tableau/Polaris system default mappings for four visual channels accord-
ingtodatatype.
Image courtesy Chris Stolte
(Stolte et al., 2008),
c
2008 IEEE. (See also
Plate XLII.)
range necessary to show the data dimension and the dynamic range available in the
channel. For example, encoding with line width uses a one-dimensional mark and
the size channel. There are a limited number of width steps that we can reliably
use to visually encode information: a minimum thinness of one pixel is enforced
by the screen resolution (ignoring antialiasing to simplify this discussion), and
there is a maximum thickness beyond which the object will be perceived as a
polygon rather than a line. Line width can work very well to show three or four
different values in a data dimension, but it would be a poor choice for dozens or
hundreds of values.
Figure 27.7. Color and lo-
cation are separable chan-
nels well suited to encode
different data dimensions,
but the horizontal size and
and vertical size channels
are automatically fused into
an integrated perception of
area.
Redrawn after (Ware,
2000).
Some visual channels are integral, fused together at a pre-conscious level, so
they are not good choices for visually encoding different data dimensions. Others
are separable, without interactions between them during visual processing, and
are safe to use for encoding multiple dimensions. Figure 27.7 shows two channel
pairs. Color and position are highly separable. We can see that horizontal size and
vertical size are not so easy to separate, because our visual system automatically
integrates these together into a unified perception of area. Size interacts with
many channels: as the size of an object grows smaller, it becomes more difficult
to distinguish its shape or color.
We can selectively attend to a channel so that items of a particular type “pop
out” visually, as discussed in Section 22.4.3. An example of visual popout is
when we immediately spot the red item amidst a sea of blue ones, or distinguish
the circle from the squares. Visual popout is powerful and scalable because it
occurs in parallel, without the need for conscious processing of the items one
by one. Many visual channels have this popout property, including not only the
list above but also curvature, flicker, stereoscopic depth, and even the direction
of lighting. However, in general we can only take advantage of popout for one
channel at a time. For example, a white circle does not pop out from a group of
circles and squares that can be white or black, as shown in Figure 22.43. When we