Visual modalities and data visualisation

In this post we will look at some visual modalities for data visualisation – size, colour, and shape. People use these all the time, but not always well. We will use the following matrix of numbers as an example, encoding the numbers in different ways (large values are highlighted in the image on the right):


Using shape to convey magnitude does not reveal the pattern here, since the triangles do not stand out from the diamonds and squares (the relationship between magnitude and shape is also rather arbitrary):

However, both size and brightness succeed in getting the message across (size works best):


On the other hand, varying only the hue or the saturation of colours does a poor job:


Combining hue and saturation variation with brightness variation, as in the various sequential ColorBrewer palettes, works very well – which is why this is recommended for colouring maps:


See also my three lenses on data visualisation post from last year and this paper by Cynthia Brewer on colour use guidelines.