THE MULTIPLE VISUAL ATTRIBUTES OF SHAPE - Information Visualisation: Techniques, Usability and Evaluation - page 44

Since Bertin, many researchers have extended the domain of visual
attributes: either identifying sub-attributes of a particular visual attribute -
for example, colour is now understood to have sub-attributes of hue and
saturation in addition to brightness; or identifying new attributes, such as
volume, connection, containment, motion, and blur [Ber67

[Che12][Cle85][Hea09][Mac95][Mac06][Maz09][War00][Wol04].
Until recently, within information visualizations, shape has been used
traditionally to encode either a single variable or to encode multiple data
variables by re-using a single type of shape attribute (Fig 3.2). Typical
examples include:

x Global Form: These are the familiar simple regular geometric

primitives, such as circles, squares, triangles, stars, etc. that are
accessible in many software charts, such as Excel. While the
number of shapes can be increased by identifying more geometric
primitives (e.g. teardrops, darts, trefoils) the approach is limited to
using shapes simply to differentiate categories and does not
consider any potential sub-attributes of shape.

Fig. 3.2. Excel scatterplot using different shapes - there are only 9 unique shapes
after which shapes begin to repeat with a new colour.

x Icons : One way to enhance the range of potential shapes is to use

icons, i.e. abstracted pictographic representations of real-world
objects (e.g. [Mod76]). The use of icons was effective in the
information graphics of Isotype (e.g. [Neu30]) and other
pictographic information graphics (Fig. 3.3). Challenges with icons
include difficulty designing uniquely identifiable icons; designing
icons that can be unambiguously decoded; and designing icons that
can convey multiple simultaneous data attributes.